Upload
trandiep
View
215
Download
1
Embed Size (px)
Citation preview
1
Making Up PeoplemdashThe Effect of Identity on Performance in a Modernizing
Society
Karla Hoff and Priyanka Pandey
The World Bank
Forthcoming in the Journal of Development Economics
Abstractmdash It is typically assumed that being hard-working or clever is a trait of the person in the
sense that it is always there in a fixed manner However in an experiment with 288 high-caste
and 294 low-caste students in India cues to onersquos place in the caste system turned out to starkly
influence the expression of these traits The experiment allows us to discriminate between two
classes of models that give different answers to the question of how someonersquos identity affects
his behavior Models of the fixed self assume that identity is a set of preferences Models of the
frame-dependent self assume that identity entails a set of mental models that are situationally
evoked and that mediate information processing Our findings suggest that the effect of identity
on intellectual performance depends sensitively on the social setting This perspective opens up
new policy options for enhancing human capital formation and development
Key words Randomized experiment Culture Identity Mental model Frame India Caste Stereotype threat
Corresponding author Hoff (khoffworldbankorg) 1818 H StWashington DC 20433 USA
Acknowledgements Over the course of many years we have benefitted from very helpful comments from many
people We would particularly like to thank Rachel Croson Allison Demeritt Anjini Kochar Leigh Linden
Tauhidur Rahman Vijayendra Rao Joe Stiglitz and Ann Swidler We also thank the referees We owe a special
debt to Anaka Narayanan Ram Pratap and Mayuresh Kshetramade for assistance with data collection and to
Shweta Arya Sonal Vats and Sam Zhongxia Zhang for research assistance This work was made possible by grants
from the World Bank-Netherlands Partnership Program the World Bank Research Support Budget and the
MacArthur Foundation Research Network on Inequality and Economic Performance
2
Making Up PeoplemdashThe Effect of Identity on Performance in a Modernizing
Society
1 Introduction
A number of models in economics give different answers to the question of how someonersquos
identitymdashan individualrsquos sense of the social categories to which he belongsmdashmight affect his
behavior We present an experiment that allows us to discriminate between some of these
models We show that situational cues to identity can alter intellectual performance Our
findings suggest that identity can have a first-order effect on human capital formation and
development
A central goal in many disciplines is to understand how identity affects behavior
Historians have documented that societies all over the world have invented social identities and
used symbols etiquette rituals dress codes and segregation to impress on people the notion that
certain individuals constituted significantly different categories and were subject to different
constraints For example in Growing up Jim Crow How Black and White Southern Children
Learned Race Jennifer Ritterhouse (2006 p 4) describes how the unwritten rules that governed
interactions across race lines were used ldquonot only as a form of social control but also as a script
for the performative creation ofhelliplsquoracersquo itselfrdquo The Victorian clothing system with its frock
coats for men and tight-laced corsets for women ldquoproduced the existence of [certain] categories
of behaviour and kept them habitually in being by moulding bodily configuration and
movementrdquo The role of men was to be active and strong (their clothes allowed them movement
and emphasized broad chests and shoulders) the role of women was to be inactive and
submissive (their clothes inhibited movement and were constricting) (Connerton 1989 pp 34)
Our title alludes to and amplifies the title of an essay ldquoMaking Up Peoplerdquo in which the
philosopher Ian Hacking (1986) argues that the creation of new slots in which to fit and
3
categorize people eg the perverted the suicidal and the heterosexual or homosexual person
moulds individualsrsquo sense of themselves and produces behavior that would not occur in the
absence of these labels
Our concern in this paper is to measure how labels affect learning and intellectual
performance A large body of work supports the view that identity matters for human capital
achievement Social norms and sometimes unknowingly schools and teachers can encourage or
discourage students from given social backgrounds Akerlof and Kranton (2002) present
evidence that whether students invest in schooling depends in part on their cultural identity
wherein payoffs differ among ldquojocksrdquo ldquonerdsrdquo and ldquoburnoutsrdquo Anthropological studies
document that school routines and curricula can convey to black students that there is something
wrong with them and their background (Ogbu 1999) In schools in which students are from
one background and teachers are from a different background the tension between students and
teachers Ogbu suggests may give rise to an oppositional culture among students Studies of
adolescents find some evidence of racial differences in the relationship between social status
among peers and academic achievement blacks for example may face a tradeoff between
acceptance and academic achievement that whites do not face Not wanting to be rejected by
their peer group for ldquoacting whiterdquo they may put less effort into their schoolwork (see Fryer
2011 for a review) Ferguson (2003) presents evidence that teachersrsquo perceptions expectations
and behavior differ across students of different social groups and that the interaction between the
expectations of teachers and those of students contributes to the black-white test score gap On
the brighter side all these findings suggest that schools have room to reduce the disparity in
educational outcomes by ethnicity and race by addressing at least two sources for this
differencemdashteachers and students Minority groups such as African-American students are
4
more likely to report discrimination by teachers In a sample of ethnically diverse US schools a
large percentage of students reported being bullied by peers based on their ethnicity (Bellmore
and Tomonaga 2009) Students who reported ethnicity-based discrimination were more likely to
experience depressive symptoms
The sociologists Ann Swidler (1986 2001 especially p 161) and Paul DiMaggio (1997)
argue that culture as a matter of self-conscious orientation or identity is not a set of values or
preferences but instead is a fragmented set of mental models understandings worldviews and
guides to action and these elements may not be consistent Culture affects action when it is
given force by contexts that organize cultural meanings and bring them to bear on action
Culture shapes behavior through cognitive frames (mental models) that are situationally evoked
and that determine how an individual construes a situation and which actions seem possible and
desirable in that situation given a personrsquos values ldquoIf one asked a slum youth why he did not
take steps to pursue a middle-class path to success hellipthe answer might well be not lsquoI donrsquot want
that lifersquo but instead lsquoWho mersquordquo (Swidler 1986 p 275)
This perspective suggests that the effect of identity on performance may depend on which
of a set of possible mental models is evoked in a given situation To test the hypothesis we draw
on our experiment in rural India that manipulates the classroom setting in which high-caste and
low-caste students are asked to learn and perform a new skill under incentives (Hoff and Pandey
2006) A feature of the caste system that makes it well-suited to identifying the effect of identity
is that an individualrsquos caste is determined by the accident of birth Although caste boundaries
and caste rankings change over long periods the status in north India of the specific high castes
(primarily Thakur and Brahmin) and the specific low caste (Chamar) from which we draw our
participants goes back millennia (Gupta 2000)
5
A second reason that caste is well-suited for the study of the effect of identity on
performance is that the meaning of the caste categories is not in doubt Two central
characteristics of the caste system are hierarchy and repulsion (or difference) between castes
(Gupta 2000) Hierarchy reflects beliefs in essential differences between the castes
For centuries it was believed that a manrsquos social capacities were known from the caste or
the lineage into which he was born and that no further test was necessary to determine
what these capacities were (Beacuteteille p 99)
Repulsion is expressed in endogamy and restrictions on contact between castes in particular
between those distant in rank A body of work in anthropology supports the view that vertical
mobility of the individual within the caste hierarchy was nearly impossible for someone who
remained in or near his village where his caste identity would be known (eg Gupta 2000 and
Srinivas 2009) A low-caste boy could not move up and a high-caste boy could not move
down1 At the bottom of the traditional caste hierarchy are the castes whose members were
marked as polluted They were called ldquountouchablesrdquo and are today called Dalits
Untouchability is the imposition of social disabilities on persons by reason of their birth in the
certain low castes Dalits were not allowed to own land or use public wells were made to use
separate crockery in food stalls and were not allowed to sit inside a schoolhouse but instead
forced to remain outside
Untouchability is illegal under the Constitution of India and attitudes in Uttar Pradesh
towards Dalits are radically different today from what they were in the recent past (Kapur et al
2010) Evidence of a new social order is visible to every schoolchild in the stipends publicly
distributed to Dalits to encourage school enrollment and in the broad participation of Dalits in
1 Through a political process social mobility was possible on the part of the community but not by individuals (see
Rao and Ban 2007 for a measure of the process of changes in caste groups) For an exception in which individuals
were able to change their status for purposes of British law in India see Cassan (2011)
6
the political process Yet children are also likely to encounter the traditional order of caste
segregation and untouchability in their own experiences through the fables they learn and in
the continued insults and atrocities against upwardly mobile Dalits2 Two surveys give some
indication of how untouchability plays out in schools in humiliation and segregation
One common example of social prejudice in the classroom is the disparaging attitude of
upper caste teachers towards Dalit children This can take various forms such as telling
Dalit children that they are lsquostupidrsquo making them feel inferior using them for menial
chores and giving them liberal physical punishment (PROBE 1999 p 51)
In one out of four primary schools in rural India Dalit children are forced by their
teachers or by convention to sit apart from non-Dalits As many as 40 percent of schools
practice untouchability while serving mid-day meals making Dalit children sit in a
separate row while eating (Shah et al 2006 p168 based on a 2001-02 national survey)
The two social orders coexist in uneasy tensionmdashone in which ldquowe are all equal nowrdquo as some
villagers remarked to us and another in which upward mobility by low-caste individuals is not
liked and could be dangerous
Our experiment assesses the effect on childrenrsquos intellectual performance of making caste
identity public and of segregating children by caste Participants in our experiment were junior
high school boys from the top three castes and from the lowest castes3 In groups of six the
participants were asked to solve mazes under incentives and were randomly assigned to one of
three conditions
(1) In the control condition caste identity which is not discernible from natural physical
2Two examples are illustrative (i) In July 1998 in the state of Uttar Pradesh a High Court judge had his chambers
ldquopurified with Gang jarsquo (water from the River Ganges) because it had earlier been occupied by a Dalit judgerdquo Times
of India (Bombay) July 23 1998 (ii) Increases in wealth among the low-caste community of Pallars from
remittances from the Persian Gulf created tensions that led the entire Dalit population of the village to be driven out
by the marginally higher caste Thevas who set fire to their homes and fields (Human Rights Watch 1999 p 85) a
similar episode occurred elsewhere in 2012 (ldquoWhen development triggers caste violencerdquo The Hindu May 8 2013)
A recent Indian TV show provided one of the first public settings where viewers heard low-caste individuals speak
first-hand about their harsh and horrifying experiences of growing up as untouchables (wwwsatyamevjayatein
episode 10)
3High-caste boys were drawn from Thakurs (62) Brahmins (32) Vaishayas (4) and others (2) Low-caste
boys were drawn only from Chamars In Hardoi district Chamar is considered the lowest caste The set of high
castes and the Chamar caste each makes up somewhat less than 20 of the population of the district
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
2
Making Up PeoplemdashThe Effect of Identity on Performance in a Modernizing
Society
1 Introduction
A number of models in economics give different answers to the question of how someonersquos
identitymdashan individualrsquos sense of the social categories to which he belongsmdashmight affect his
behavior We present an experiment that allows us to discriminate between some of these
models We show that situational cues to identity can alter intellectual performance Our
findings suggest that identity can have a first-order effect on human capital formation and
development
A central goal in many disciplines is to understand how identity affects behavior
Historians have documented that societies all over the world have invented social identities and
used symbols etiquette rituals dress codes and segregation to impress on people the notion that
certain individuals constituted significantly different categories and were subject to different
constraints For example in Growing up Jim Crow How Black and White Southern Children
Learned Race Jennifer Ritterhouse (2006 p 4) describes how the unwritten rules that governed
interactions across race lines were used ldquonot only as a form of social control but also as a script
for the performative creation ofhelliplsquoracersquo itselfrdquo The Victorian clothing system with its frock
coats for men and tight-laced corsets for women ldquoproduced the existence of [certain] categories
of behaviour and kept them habitually in being by moulding bodily configuration and
movementrdquo The role of men was to be active and strong (their clothes allowed them movement
and emphasized broad chests and shoulders) the role of women was to be inactive and
submissive (their clothes inhibited movement and were constricting) (Connerton 1989 pp 34)
Our title alludes to and amplifies the title of an essay ldquoMaking Up Peoplerdquo in which the
philosopher Ian Hacking (1986) argues that the creation of new slots in which to fit and
3
categorize people eg the perverted the suicidal and the heterosexual or homosexual person
moulds individualsrsquo sense of themselves and produces behavior that would not occur in the
absence of these labels
Our concern in this paper is to measure how labels affect learning and intellectual
performance A large body of work supports the view that identity matters for human capital
achievement Social norms and sometimes unknowingly schools and teachers can encourage or
discourage students from given social backgrounds Akerlof and Kranton (2002) present
evidence that whether students invest in schooling depends in part on their cultural identity
wherein payoffs differ among ldquojocksrdquo ldquonerdsrdquo and ldquoburnoutsrdquo Anthropological studies
document that school routines and curricula can convey to black students that there is something
wrong with them and their background (Ogbu 1999) In schools in which students are from
one background and teachers are from a different background the tension between students and
teachers Ogbu suggests may give rise to an oppositional culture among students Studies of
adolescents find some evidence of racial differences in the relationship between social status
among peers and academic achievement blacks for example may face a tradeoff between
acceptance and academic achievement that whites do not face Not wanting to be rejected by
their peer group for ldquoacting whiterdquo they may put less effort into their schoolwork (see Fryer
2011 for a review) Ferguson (2003) presents evidence that teachersrsquo perceptions expectations
and behavior differ across students of different social groups and that the interaction between the
expectations of teachers and those of students contributes to the black-white test score gap On
the brighter side all these findings suggest that schools have room to reduce the disparity in
educational outcomes by ethnicity and race by addressing at least two sources for this
differencemdashteachers and students Minority groups such as African-American students are
4
more likely to report discrimination by teachers In a sample of ethnically diverse US schools a
large percentage of students reported being bullied by peers based on their ethnicity (Bellmore
and Tomonaga 2009) Students who reported ethnicity-based discrimination were more likely to
experience depressive symptoms
The sociologists Ann Swidler (1986 2001 especially p 161) and Paul DiMaggio (1997)
argue that culture as a matter of self-conscious orientation or identity is not a set of values or
preferences but instead is a fragmented set of mental models understandings worldviews and
guides to action and these elements may not be consistent Culture affects action when it is
given force by contexts that organize cultural meanings and bring them to bear on action
Culture shapes behavior through cognitive frames (mental models) that are situationally evoked
and that determine how an individual construes a situation and which actions seem possible and
desirable in that situation given a personrsquos values ldquoIf one asked a slum youth why he did not
take steps to pursue a middle-class path to success hellipthe answer might well be not lsquoI donrsquot want
that lifersquo but instead lsquoWho mersquordquo (Swidler 1986 p 275)
This perspective suggests that the effect of identity on performance may depend on which
of a set of possible mental models is evoked in a given situation To test the hypothesis we draw
on our experiment in rural India that manipulates the classroom setting in which high-caste and
low-caste students are asked to learn and perform a new skill under incentives (Hoff and Pandey
2006) A feature of the caste system that makes it well-suited to identifying the effect of identity
is that an individualrsquos caste is determined by the accident of birth Although caste boundaries
and caste rankings change over long periods the status in north India of the specific high castes
(primarily Thakur and Brahmin) and the specific low caste (Chamar) from which we draw our
participants goes back millennia (Gupta 2000)
5
A second reason that caste is well-suited for the study of the effect of identity on
performance is that the meaning of the caste categories is not in doubt Two central
characteristics of the caste system are hierarchy and repulsion (or difference) between castes
(Gupta 2000) Hierarchy reflects beliefs in essential differences between the castes
For centuries it was believed that a manrsquos social capacities were known from the caste or
the lineage into which he was born and that no further test was necessary to determine
what these capacities were (Beacuteteille p 99)
Repulsion is expressed in endogamy and restrictions on contact between castes in particular
between those distant in rank A body of work in anthropology supports the view that vertical
mobility of the individual within the caste hierarchy was nearly impossible for someone who
remained in or near his village where his caste identity would be known (eg Gupta 2000 and
Srinivas 2009) A low-caste boy could not move up and a high-caste boy could not move
down1 At the bottom of the traditional caste hierarchy are the castes whose members were
marked as polluted They were called ldquountouchablesrdquo and are today called Dalits
Untouchability is the imposition of social disabilities on persons by reason of their birth in the
certain low castes Dalits were not allowed to own land or use public wells were made to use
separate crockery in food stalls and were not allowed to sit inside a schoolhouse but instead
forced to remain outside
Untouchability is illegal under the Constitution of India and attitudes in Uttar Pradesh
towards Dalits are radically different today from what they were in the recent past (Kapur et al
2010) Evidence of a new social order is visible to every schoolchild in the stipends publicly
distributed to Dalits to encourage school enrollment and in the broad participation of Dalits in
1 Through a political process social mobility was possible on the part of the community but not by individuals (see
Rao and Ban 2007 for a measure of the process of changes in caste groups) For an exception in which individuals
were able to change their status for purposes of British law in India see Cassan (2011)
6
the political process Yet children are also likely to encounter the traditional order of caste
segregation and untouchability in their own experiences through the fables they learn and in
the continued insults and atrocities against upwardly mobile Dalits2 Two surveys give some
indication of how untouchability plays out in schools in humiliation and segregation
One common example of social prejudice in the classroom is the disparaging attitude of
upper caste teachers towards Dalit children This can take various forms such as telling
Dalit children that they are lsquostupidrsquo making them feel inferior using them for menial
chores and giving them liberal physical punishment (PROBE 1999 p 51)
In one out of four primary schools in rural India Dalit children are forced by their
teachers or by convention to sit apart from non-Dalits As many as 40 percent of schools
practice untouchability while serving mid-day meals making Dalit children sit in a
separate row while eating (Shah et al 2006 p168 based on a 2001-02 national survey)
The two social orders coexist in uneasy tensionmdashone in which ldquowe are all equal nowrdquo as some
villagers remarked to us and another in which upward mobility by low-caste individuals is not
liked and could be dangerous
Our experiment assesses the effect on childrenrsquos intellectual performance of making caste
identity public and of segregating children by caste Participants in our experiment were junior
high school boys from the top three castes and from the lowest castes3 In groups of six the
participants were asked to solve mazes under incentives and were randomly assigned to one of
three conditions
(1) In the control condition caste identity which is not discernible from natural physical
2Two examples are illustrative (i) In July 1998 in the state of Uttar Pradesh a High Court judge had his chambers
ldquopurified with Gang jarsquo (water from the River Ganges) because it had earlier been occupied by a Dalit judgerdquo Times
of India (Bombay) July 23 1998 (ii) Increases in wealth among the low-caste community of Pallars from
remittances from the Persian Gulf created tensions that led the entire Dalit population of the village to be driven out
by the marginally higher caste Thevas who set fire to their homes and fields (Human Rights Watch 1999 p 85) a
similar episode occurred elsewhere in 2012 (ldquoWhen development triggers caste violencerdquo The Hindu May 8 2013)
A recent Indian TV show provided one of the first public settings where viewers heard low-caste individuals speak
first-hand about their harsh and horrifying experiences of growing up as untouchables (wwwsatyamevjayatein
episode 10)
3High-caste boys were drawn from Thakurs (62) Brahmins (32) Vaishayas (4) and others (2) Low-caste
boys were drawn only from Chamars In Hardoi district Chamar is considered the lowest caste The set of high
castes and the Chamar caste each makes up somewhat less than 20 of the population of the district
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
3
categorize people eg the perverted the suicidal and the heterosexual or homosexual person
moulds individualsrsquo sense of themselves and produces behavior that would not occur in the
absence of these labels
Our concern in this paper is to measure how labels affect learning and intellectual
performance A large body of work supports the view that identity matters for human capital
achievement Social norms and sometimes unknowingly schools and teachers can encourage or
discourage students from given social backgrounds Akerlof and Kranton (2002) present
evidence that whether students invest in schooling depends in part on their cultural identity
wherein payoffs differ among ldquojocksrdquo ldquonerdsrdquo and ldquoburnoutsrdquo Anthropological studies
document that school routines and curricula can convey to black students that there is something
wrong with them and their background (Ogbu 1999) In schools in which students are from
one background and teachers are from a different background the tension between students and
teachers Ogbu suggests may give rise to an oppositional culture among students Studies of
adolescents find some evidence of racial differences in the relationship between social status
among peers and academic achievement blacks for example may face a tradeoff between
acceptance and academic achievement that whites do not face Not wanting to be rejected by
their peer group for ldquoacting whiterdquo they may put less effort into their schoolwork (see Fryer
2011 for a review) Ferguson (2003) presents evidence that teachersrsquo perceptions expectations
and behavior differ across students of different social groups and that the interaction between the
expectations of teachers and those of students contributes to the black-white test score gap On
the brighter side all these findings suggest that schools have room to reduce the disparity in
educational outcomes by ethnicity and race by addressing at least two sources for this
differencemdashteachers and students Minority groups such as African-American students are
4
more likely to report discrimination by teachers In a sample of ethnically diverse US schools a
large percentage of students reported being bullied by peers based on their ethnicity (Bellmore
and Tomonaga 2009) Students who reported ethnicity-based discrimination were more likely to
experience depressive symptoms
The sociologists Ann Swidler (1986 2001 especially p 161) and Paul DiMaggio (1997)
argue that culture as a matter of self-conscious orientation or identity is not a set of values or
preferences but instead is a fragmented set of mental models understandings worldviews and
guides to action and these elements may not be consistent Culture affects action when it is
given force by contexts that organize cultural meanings and bring them to bear on action
Culture shapes behavior through cognitive frames (mental models) that are situationally evoked
and that determine how an individual construes a situation and which actions seem possible and
desirable in that situation given a personrsquos values ldquoIf one asked a slum youth why he did not
take steps to pursue a middle-class path to success hellipthe answer might well be not lsquoI donrsquot want
that lifersquo but instead lsquoWho mersquordquo (Swidler 1986 p 275)
This perspective suggests that the effect of identity on performance may depend on which
of a set of possible mental models is evoked in a given situation To test the hypothesis we draw
on our experiment in rural India that manipulates the classroom setting in which high-caste and
low-caste students are asked to learn and perform a new skill under incentives (Hoff and Pandey
2006) A feature of the caste system that makes it well-suited to identifying the effect of identity
is that an individualrsquos caste is determined by the accident of birth Although caste boundaries
and caste rankings change over long periods the status in north India of the specific high castes
(primarily Thakur and Brahmin) and the specific low caste (Chamar) from which we draw our
participants goes back millennia (Gupta 2000)
5
A second reason that caste is well-suited for the study of the effect of identity on
performance is that the meaning of the caste categories is not in doubt Two central
characteristics of the caste system are hierarchy and repulsion (or difference) between castes
(Gupta 2000) Hierarchy reflects beliefs in essential differences between the castes
For centuries it was believed that a manrsquos social capacities were known from the caste or
the lineage into which he was born and that no further test was necessary to determine
what these capacities were (Beacuteteille p 99)
Repulsion is expressed in endogamy and restrictions on contact between castes in particular
between those distant in rank A body of work in anthropology supports the view that vertical
mobility of the individual within the caste hierarchy was nearly impossible for someone who
remained in or near his village where his caste identity would be known (eg Gupta 2000 and
Srinivas 2009) A low-caste boy could not move up and a high-caste boy could not move
down1 At the bottom of the traditional caste hierarchy are the castes whose members were
marked as polluted They were called ldquountouchablesrdquo and are today called Dalits
Untouchability is the imposition of social disabilities on persons by reason of their birth in the
certain low castes Dalits were not allowed to own land or use public wells were made to use
separate crockery in food stalls and were not allowed to sit inside a schoolhouse but instead
forced to remain outside
Untouchability is illegal under the Constitution of India and attitudes in Uttar Pradesh
towards Dalits are radically different today from what they were in the recent past (Kapur et al
2010) Evidence of a new social order is visible to every schoolchild in the stipends publicly
distributed to Dalits to encourage school enrollment and in the broad participation of Dalits in
1 Through a political process social mobility was possible on the part of the community but not by individuals (see
Rao and Ban 2007 for a measure of the process of changes in caste groups) For an exception in which individuals
were able to change their status for purposes of British law in India see Cassan (2011)
6
the political process Yet children are also likely to encounter the traditional order of caste
segregation and untouchability in their own experiences through the fables they learn and in
the continued insults and atrocities against upwardly mobile Dalits2 Two surveys give some
indication of how untouchability plays out in schools in humiliation and segregation
One common example of social prejudice in the classroom is the disparaging attitude of
upper caste teachers towards Dalit children This can take various forms such as telling
Dalit children that they are lsquostupidrsquo making them feel inferior using them for menial
chores and giving them liberal physical punishment (PROBE 1999 p 51)
In one out of four primary schools in rural India Dalit children are forced by their
teachers or by convention to sit apart from non-Dalits As many as 40 percent of schools
practice untouchability while serving mid-day meals making Dalit children sit in a
separate row while eating (Shah et al 2006 p168 based on a 2001-02 national survey)
The two social orders coexist in uneasy tensionmdashone in which ldquowe are all equal nowrdquo as some
villagers remarked to us and another in which upward mobility by low-caste individuals is not
liked and could be dangerous
Our experiment assesses the effect on childrenrsquos intellectual performance of making caste
identity public and of segregating children by caste Participants in our experiment were junior
high school boys from the top three castes and from the lowest castes3 In groups of six the
participants were asked to solve mazes under incentives and were randomly assigned to one of
three conditions
(1) In the control condition caste identity which is not discernible from natural physical
2Two examples are illustrative (i) In July 1998 in the state of Uttar Pradesh a High Court judge had his chambers
ldquopurified with Gang jarsquo (water from the River Ganges) because it had earlier been occupied by a Dalit judgerdquo Times
of India (Bombay) July 23 1998 (ii) Increases in wealth among the low-caste community of Pallars from
remittances from the Persian Gulf created tensions that led the entire Dalit population of the village to be driven out
by the marginally higher caste Thevas who set fire to their homes and fields (Human Rights Watch 1999 p 85) a
similar episode occurred elsewhere in 2012 (ldquoWhen development triggers caste violencerdquo The Hindu May 8 2013)
A recent Indian TV show provided one of the first public settings where viewers heard low-caste individuals speak
first-hand about their harsh and horrifying experiences of growing up as untouchables (wwwsatyamevjayatein
episode 10)
3High-caste boys were drawn from Thakurs (62) Brahmins (32) Vaishayas (4) and others (2) Low-caste
boys were drawn only from Chamars In Hardoi district Chamar is considered the lowest caste The set of high
castes and the Chamar caste each makes up somewhat less than 20 of the population of the district
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
4
more likely to report discrimination by teachers In a sample of ethnically diverse US schools a
large percentage of students reported being bullied by peers based on their ethnicity (Bellmore
and Tomonaga 2009) Students who reported ethnicity-based discrimination were more likely to
experience depressive symptoms
The sociologists Ann Swidler (1986 2001 especially p 161) and Paul DiMaggio (1997)
argue that culture as a matter of self-conscious orientation or identity is not a set of values or
preferences but instead is a fragmented set of mental models understandings worldviews and
guides to action and these elements may not be consistent Culture affects action when it is
given force by contexts that organize cultural meanings and bring them to bear on action
Culture shapes behavior through cognitive frames (mental models) that are situationally evoked
and that determine how an individual construes a situation and which actions seem possible and
desirable in that situation given a personrsquos values ldquoIf one asked a slum youth why he did not
take steps to pursue a middle-class path to success hellipthe answer might well be not lsquoI donrsquot want
that lifersquo but instead lsquoWho mersquordquo (Swidler 1986 p 275)
This perspective suggests that the effect of identity on performance may depend on which
of a set of possible mental models is evoked in a given situation To test the hypothesis we draw
on our experiment in rural India that manipulates the classroom setting in which high-caste and
low-caste students are asked to learn and perform a new skill under incentives (Hoff and Pandey
2006) A feature of the caste system that makes it well-suited to identifying the effect of identity
is that an individualrsquos caste is determined by the accident of birth Although caste boundaries
and caste rankings change over long periods the status in north India of the specific high castes
(primarily Thakur and Brahmin) and the specific low caste (Chamar) from which we draw our
participants goes back millennia (Gupta 2000)
5
A second reason that caste is well-suited for the study of the effect of identity on
performance is that the meaning of the caste categories is not in doubt Two central
characteristics of the caste system are hierarchy and repulsion (or difference) between castes
(Gupta 2000) Hierarchy reflects beliefs in essential differences between the castes
For centuries it was believed that a manrsquos social capacities were known from the caste or
the lineage into which he was born and that no further test was necessary to determine
what these capacities were (Beacuteteille p 99)
Repulsion is expressed in endogamy and restrictions on contact between castes in particular
between those distant in rank A body of work in anthropology supports the view that vertical
mobility of the individual within the caste hierarchy was nearly impossible for someone who
remained in or near his village where his caste identity would be known (eg Gupta 2000 and
Srinivas 2009) A low-caste boy could not move up and a high-caste boy could not move
down1 At the bottom of the traditional caste hierarchy are the castes whose members were
marked as polluted They were called ldquountouchablesrdquo and are today called Dalits
Untouchability is the imposition of social disabilities on persons by reason of their birth in the
certain low castes Dalits were not allowed to own land or use public wells were made to use
separate crockery in food stalls and were not allowed to sit inside a schoolhouse but instead
forced to remain outside
Untouchability is illegal under the Constitution of India and attitudes in Uttar Pradesh
towards Dalits are radically different today from what they were in the recent past (Kapur et al
2010) Evidence of a new social order is visible to every schoolchild in the stipends publicly
distributed to Dalits to encourage school enrollment and in the broad participation of Dalits in
1 Through a political process social mobility was possible on the part of the community but not by individuals (see
Rao and Ban 2007 for a measure of the process of changes in caste groups) For an exception in which individuals
were able to change their status for purposes of British law in India see Cassan (2011)
6
the political process Yet children are also likely to encounter the traditional order of caste
segregation and untouchability in their own experiences through the fables they learn and in
the continued insults and atrocities against upwardly mobile Dalits2 Two surveys give some
indication of how untouchability plays out in schools in humiliation and segregation
One common example of social prejudice in the classroom is the disparaging attitude of
upper caste teachers towards Dalit children This can take various forms such as telling
Dalit children that they are lsquostupidrsquo making them feel inferior using them for menial
chores and giving them liberal physical punishment (PROBE 1999 p 51)
In one out of four primary schools in rural India Dalit children are forced by their
teachers or by convention to sit apart from non-Dalits As many as 40 percent of schools
practice untouchability while serving mid-day meals making Dalit children sit in a
separate row while eating (Shah et al 2006 p168 based on a 2001-02 national survey)
The two social orders coexist in uneasy tensionmdashone in which ldquowe are all equal nowrdquo as some
villagers remarked to us and another in which upward mobility by low-caste individuals is not
liked and could be dangerous
Our experiment assesses the effect on childrenrsquos intellectual performance of making caste
identity public and of segregating children by caste Participants in our experiment were junior
high school boys from the top three castes and from the lowest castes3 In groups of six the
participants were asked to solve mazes under incentives and were randomly assigned to one of
three conditions
(1) In the control condition caste identity which is not discernible from natural physical
2Two examples are illustrative (i) In July 1998 in the state of Uttar Pradesh a High Court judge had his chambers
ldquopurified with Gang jarsquo (water from the River Ganges) because it had earlier been occupied by a Dalit judgerdquo Times
of India (Bombay) July 23 1998 (ii) Increases in wealth among the low-caste community of Pallars from
remittances from the Persian Gulf created tensions that led the entire Dalit population of the village to be driven out
by the marginally higher caste Thevas who set fire to their homes and fields (Human Rights Watch 1999 p 85) a
similar episode occurred elsewhere in 2012 (ldquoWhen development triggers caste violencerdquo The Hindu May 8 2013)
A recent Indian TV show provided one of the first public settings where viewers heard low-caste individuals speak
first-hand about their harsh and horrifying experiences of growing up as untouchables (wwwsatyamevjayatein
episode 10)
3High-caste boys were drawn from Thakurs (62) Brahmins (32) Vaishayas (4) and others (2) Low-caste
boys were drawn only from Chamars In Hardoi district Chamar is considered the lowest caste The set of high
castes and the Chamar caste each makes up somewhat less than 20 of the population of the district
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
5
A second reason that caste is well-suited for the study of the effect of identity on
performance is that the meaning of the caste categories is not in doubt Two central
characteristics of the caste system are hierarchy and repulsion (or difference) between castes
(Gupta 2000) Hierarchy reflects beliefs in essential differences between the castes
For centuries it was believed that a manrsquos social capacities were known from the caste or
the lineage into which he was born and that no further test was necessary to determine
what these capacities were (Beacuteteille p 99)
Repulsion is expressed in endogamy and restrictions on contact between castes in particular
between those distant in rank A body of work in anthropology supports the view that vertical
mobility of the individual within the caste hierarchy was nearly impossible for someone who
remained in or near his village where his caste identity would be known (eg Gupta 2000 and
Srinivas 2009) A low-caste boy could not move up and a high-caste boy could not move
down1 At the bottom of the traditional caste hierarchy are the castes whose members were
marked as polluted They were called ldquountouchablesrdquo and are today called Dalits
Untouchability is the imposition of social disabilities on persons by reason of their birth in the
certain low castes Dalits were not allowed to own land or use public wells were made to use
separate crockery in food stalls and were not allowed to sit inside a schoolhouse but instead
forced to remain outside
Untouchability is illegal under the Constitution of India and attitudes in Uttar Pradesh
towards Dalits are radically different today from what they were in the recent past (Kapur et al
2010) Evidence of a new social order is visible to every schoolchild in the stipends publicly
distributed to Dalits to encourage school enrollment and in the broad participation of Dalits in
1 Through a political process social mobility was possible on the part of the community but not by individuals (see
Rao and Ban 2007 for a measure of the process of changes in caste groups) For an exception in which individuals
were able to change their status for purposes of British law in India see Cassan (2011)
6
the political process Yet children are also likely to encounter the traditional order of caste
segregation and untouchability in their own experiences through the fables they learn and in
the continued insults and atrocities against upwardly mobile Dalits2 Two surveys give some
indication of how untouchability plays out in schools in humiliation and segregation
One common example of social prejudice in the classroom is the disparaging attitude of
upper caste teachers towards Dalit children This can take various forms such as telling
Dalit children that they are lsquostupidrsquo making them feel inferior using them for menial
chores and giving them liberal physical punishment (PROBE 1999 p 51)
In one out of four primary schools in rural India Dalit children are forced by their
teachers or by convention to sit apart from non-Dalits As many as 40 percent of schools
practice untouchability while serving mid-day meals making Dalit children sit in a
separate row while eating (Shah et al 2006 p168 based on a 2001-02 national survey)
The two social orders coexist in uneasy tensionmdashone in which ldquowe are all equal nowrdquo as some
villagers remarked to us and another in which upward mobility by low-caste individuals is not
liked and could be dangerous
Our experiment assesses the effect on childrenrsquos intellectual performance of making caste
identity public and of segregating children by caste Participants in our experiment were junior
high school boys from the top three castes and from the lowest castes3 In groups of six the
participants were asked to solve mazes under incentives and were randomly assigned to one of
three conditions
(1) In the control condition caste identity which is not discernible from natural physical
2Two examples are illustrative (i) In July 1998 in the state of Uttar Pradesh a High Court judge had his chambers
ldquopurified with Gang jarsquo (water from the River Ganges) because it had earlier been occupied by a Dalit judgerdquo Times
of India (Bombay) July 23 1998 (ii) Increases in wealth among the low-caste community of Pallars from
remittances from the Persian Gulf created tensions that led the entire Dalit population of the village to be driven out
by the marginally higher caste Thevas who set fire to their homes and fields (Human Rights Watch 1999 p 85) a
similar episode occurred elsewhere in 2012 (ldquoWhen development triggers caste violencerdquo The Hindu May 8 2013)
A recent Indian TV show provided one of the first public settings where viewers heard low-caste individuals speak
first-hand about their harsh and horrifying experiences of growing up as untouchables (wwwsatyamevjayatein
episode 10)
3High-caste boys were drawn from Thakurs (62) Brahmins (32) Vaishayas (4) and others (2) Low-caste
boys were drawn only from Chamars In Hardoi district Chamar is considered the lowest caste The set of high
castes and the Chamar caste each makes up somewhat less than 20 of the population of the district
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
6
the political process Yet children are also likely to encounter the traditional order of caste
segregation and untouchability in their own experiences through the fables they learn and in
the continued insults and atrocities against upwardly mobile Dalits2 Two surveys give some
indication of how untouchability plays out in schools in humiliation and segregation
One common example of social prejudice in the classroom is the disparaging attitude of
upper caste teachers towards Dalit children This can take various forms such as telling
Dalit children that they are lsquostupidrsquo making them feel inferior using them for menial
chores and giving them liberal physical punishment (PROBE 1999 p 51)
In one out of four primary schools in rural India Dalit children are forced by their
teachers or by convention to sit apart from non-Dalits As many as 40 percent of schools
practice untouchability while serving mid-day meals making Dalit children sit in a
separate row while eating (Shah et al 2006 p168 based on a 2001-02 national survey)
The two social orders coexist in uneasy tensionmdashone in which ldquowe are all equal nowrdquo as some
villagers remarked to us and another in which upward mobility by low-caste individuals is not
liked and could be dangerous
Our experiment assesses the effect on childrenrsquos intellectual performance of making caste
identity public and of segregating children by caste Participants in our experiment were junior
high school boys from the top three castes and from the lowest castes3 In groups of six the
participants were asked to solve mazes under incentives and were randomly assigned to one of
three conditions
(1) In the control condition caste identity which is not discernible from natural physical
2Two examples are illustrative (i) In July 1998 in the state of Uttar Pradesh a High Court judge had his chambers
ldquopurified with Gang jarsquo (water from the River Ganges) because it had earlier been occupied by a Dalit judgerdquo Times
of India (Bombay) July 23 1998 (ii) Increases in wealth among the low-caste community of Pallars from
remittances from the Persian Gulf created tensions that led the entire Dalit population of the village to be driven out
by the marginally higher caste Thevas who set fire to their homes and fields (Human Rights Watch 1999 p 85) a
similar episode occurred elsewhere in 2012 (ldquoWhen development triggers caste violencerdquo The Hindu May 8 2013)
A recent Indian TV show provided one of the first public settings where viewers heard low-caste individuals speak
first-hand about their harsh and horrifying experiences of growing up as untouchables (wwwsatyamevjayatein
episode 10)
3High-caste boys were drawn from Thakurs (62) Brahmins (32) Vaishayas (4) and others (2) Low-caste
boys were drawn only from Chamars In Hardoi district Chamar is considered the lowest caste The set of high
castes and the Chamar caste each makes up somewhat less than 20 of the population of the district
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
7
markers4 was not made public in a session of three high-caste and three low-caste boys Since
there is a large overlap in poverty between high- and low-caste households in the local
population quality of dress is not an automatic sign of caste status We call the control condition
Caste Not Revealed
(2) The second condition (Revealed Mixed) was the same as the control except that caste
identity was made public
(3) The third condition (Revealed Segregated) was the same as the second except that a
session was composed of only high-caste boys or only low-caste boys Participants would likely
have been aware that the composition of their session reflected deliberate segregation by caste
status This is so because participants were driven to the experimental site with other boys from
their village whose caste they would know in groups equally divided between high and low
castes After arriving at the school that was the site of the experiment individuals were assigned
to new groups of six boys to participate in an experimental session The probability that a boy
would find himself in a group with five other boys of his same caste status as a result of a
random draw of the entire local school population was very smallmdashless than (02)5
= 000032 If
children perceived the segregation as deliberate it is likely that they would also perceive it as
meaningful in the context of the traditional caste system in which high castes are not supposed to
mingle with Dalits The practice of segregation remains in many villages an issue over which
high and low castes struggle (Human Rights Watch 1999)
The control condition shows that low-caste boys solve mazes just as well as high-caste
boys This is true under piece-rate incentives It is also true under the tournament incentive We
4 See Deliege (1999) and Gupta (2000) In contrast Hindu surnames do mark caste For that reason some low-caste
groups have sought a constitutional amendment to abolish Hindu surnames (The Telegraph October 15 2005
ldquoSlash Surname to Kill Casterdquo) Thanks for Pranab Bardhan for this reference
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
8
have three main results for the effects of revealing caste (All effects reported here are
significant and control for individual characteristics)
Result 1 Under piece-rate incentives publicly revealing caste in mixed-caste groups
creates a 23 caste gap in total mazes solved in favor of the high castes Since there was no
caste gap in the control we infer that in other possible worlds the low castes could have been an
equal or dominant group Here a social identity has affected behavior Underlying the caste gap
is an increase when caste is revealed in the proportion of low-caste boys who fail to learn to
solve mazes Our first result extends to a new social groupmdashlow-caste individuals of Indiamdasha
body of work in psychology that finds that cues to a personrsquos identity if it is stereotyped as
intellectually inferior may undermine the personrsquos ability to perform cognitive tasks (stereotype
threat)
Result 2 Under piece-rate incentives both high- and low-caste boys underperform by
over 20 when caste is publicly revealed in segregated groups Stereotype susceptibility would
predict this result for the low caste but it would not predict this result for the high caste
Furthermore all available evidence ndashfrom failure rates to learn how to solve a maze in this
experiment and from a direct test of self-confidence in another study (Hoff and Pandey 2005)mdash
show that the Revealed Segregated condition does not lower the self-confidence of high-caste
boys To us the most plausible interpretation of the decline in high-caste performance is that
segregation is a marker of high-caste dominance and evokes a sense of entitlement in which the
high-caste boys feel less need to achieve As the Indian sociologist Andreacute Beacuteteille notes in the
caste system social preeminence is assigned by birth rather than by competition (Beacuteteille (2011
II [2003] p 11)
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
9
Result 3 Under tournament incentives both high-caste and low-caste boys
underperform when caste is revealed but in the segregated condition the effect is much larger
for low-caste boys In a tournament strategic rather than psychological factors could explain
why revealing identity impairs performance Because we cannot disentangle these factors we
relegate the analysis of the tournaments to the appendix
2 The Fixed Self and the Frame-Dependent Self
Our experiment will allow us to discriminate between two broad sets of theories about how
social identity might affect performance We will call the two sets of theories the fixed self and
the frame-dependent self
The Fixed Self
A broad class of theories takes the view that an individual at a moment in time has fixed
well-defined preferences and abilities They provide all the information that is relevant for
describing the individualrsquos choices and outcomes for a given set of opportunities In the
textbook model in economics an individual has preferences in which a sense of identity with
others has no influence This theory is one of the fundamental differences between the standard
model of economics and the concept of the individual that other social sciences find useful in
which socially defined variables such as conformity affect preferences
The theory that an individual has at any moment in time a well-defined set of
preferences and that they are always salient is maintained in recent work that substantially
broadens the notion of preferences by incorporating onersquos sense of group membership In
Akerlof and Kranton (2000) a social category constitutes part of an individualrsquos identity
Associated with the category are a set of norms or ideals for how someone in the category should
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
10
behave The individual likes conforming to the ideals of that category and dislikes actions by
others that deviate from the ideals
An individual may be able to choose his social identities ie he can define himself and
his relationships to others at a categorical level (Akerlof and Kranton 2002 Fang and Loury
2005 Hoff and Sen 2006 and Munshi and Rosenzweig 2006 Sen 2006 Munshi and Wilson
2008) For example a descendant of Irish immigrants to the US can define himself as Irish-
American or not The individualrsquos choice problem makes sense only under the assumption that
an individual has a meta-utility function However just as in the two theories above an
individual has well-defined preferences that provide all the information that is relevant for
describing his choices and outcomes for a given set of opportunities
The Frame-Dependent Self
The other broad class of theories draws on a certain view of how the mind works
namely that information is processed in relation to mental models (references are in Tversky
and Kahneman 1983 and an elaboration is in DiMaggio 1997) Individuals have multiple mental
models they may not be consistent and they are situationally evoked
An experiment from Tversky and Kahneman illustrates the powerful effect of cues on
cognition A group of subjects are asked How many seven-letter words of which the sixth letter
is ldquoNrdquo (_ _ _ _ _ N _) would you expect to find in four pages of a novel (about 2000 words)
The same group of subjects are also asked How many seven-letter words of which the last three
letters are ldquoINGrdquo (_ _ _ _ ING) would you expect to find in four pages of a novel (about 2000
words) The median estimate is several times greater for ING words than for _N_ words The
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
11
responses violate logic but can be explained this way _ING words are a standard category (a
very simple mental structure) _N_ words are not and categories shape how we think5
There is substantial evidence that the setting in which alternatives are offered affects
choices and that one mechanism underlying this influence is that the setting evokes a particular
mental model or aspect of the self For example in the studies of Benjamin et al (2010)
LeBeouf et al (2010) and Cohn et al (2013) individuals were randomly assigned to fill out
background questionnaires that either included questions related to an aspect of their identities or
that did not A questionnaire that primed Asian identity made Asian-Americans more
cooperative and patient a questionnaire that primed a family-oriented identity strengthened
values related to family obligations and a questionnaire that primed the prisoner identity of
prison inmates led them to cheat more Even the language of thought affects behavior studies
with bilingual students find that randomly directing them to use one or the other of their two
languages shifts their implicit attitudes and behavior (Danziger and Ward 2010 and Ogunnaike et
al 2010) In the study of Lambarraa et al (2012) which also uses bilingual subjects the
language of thought in the period before the subject has to make a choice influences his
behavior Since the decision-making situation is exactly the same but only the prior language
has changed it is clear that context has changed not what it is objectively appropriate to do but
instead the way that the subject evaluates his choices The effect of context on behavior is
analogous to the effect of the environment on the expression of an individualrsquos genes DNA are
the instructions for making an individual but as yet poorly understood features of the
environment determine the on-and-off states of genes (See Salant and Rubinstein 2008 for a
model of frame-dependent utility)
5 A second example is that the distinctions between colors in the language one speaks influence the speed with
which one can perceive distinct colors See Alter (2013 ch 2) for a popular lively account of the influence of
categories on cognition
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
12
The idea of an extended utility function is interesting when it leads to the observation of
inconsistent choices Of course if we knew all the stimuli to the individual then the theory of
rationality (ie consistency) would be trivial Since we do not observe all stimuli and our
understanding of how individuals process information is limited it becomes a useful construct to
posit multiple preferences one for each self-construal or worldview
Useful for what purpose It may be useful for understanding long-run social change
which entails changes in the set of possible identities the salience of particular identities and the
possible ways of understanding a situation In the process of economic development the stimuli
to an individual can change in a way that leads to the expression of one set of preferences rather
than another that is preferences depend on context and frames
Finally an influential body of evidence in psychology on stereotype threat (and to a
lesser extent stereotype boost) relates to the influence of context on human productivity In
experiments over many domains (eg SATs and sports) priming a negatively or positively
stereotyped aspect of an individualrsquos identity shifts performance in the direction of the
stereotype for instance African-Americans do worse on scholastic aptitude tests if before the
test they are asked to check a box for their race (Steele and Aronson 1995) Asian-American
women if the Asian aspect of identity is made salient do better on math tests than women in the
no-prime condition but if their gender is made salient do worse than women in the no-prime
condition (Shih Pittinsky and Ambady 1999) Children in both lower elementary grades and
middle school grades (but not those in upper elementary grades) show shifts in performance
consistent with the patterns of stereotype threat and stereotype boost (Ambady et al 2001 and
Afridi Li and Ren 2010) However other studies do not find evidence of stereotype threat (eg
Fryer Levitt and List 2008) When stereotype threat occurs results in neuroscience support the
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
13
view that the stereotype is a powerful distractor in the task that subjects are trying to accomplish
In Krendl et al (2008) women taking a math test in conditions of stereotype threat did not
recruit the neural regions associated with mathematical learning but instead showed heightened
activation in a neural region associated with social and emotional processing Neuroscientists
also find that early environmental exposure can affect the chemistry of the DNA with a long-
term effect on onersquos responses to stress (Begley 2007 and Pollak 2008) Children who have a
history of exposure to stressful events react more strongly to stress We conjecture that for a
low-caste child who has encountered stressful events related to his caste identity increasing the
salience of caste is a source of stress In contrast a high-caste child because of his early
experiences may find comfort and power in contexts in which his caste identity is salient
3 Participants and Design
288 high-caste (hereafter H) and 294 low-caste (hereafter L) in 6th
or 7th
grade in the district of
Hardoi in the Indian state of Uttar Pradesh participated in the study Hardoi was under feudal
rule in the 19th
century and the feudal elite was made up of the high castes A legacy of feudal
rule is greater high-caste dominance compared to regions of Uttar Pradesh not under such rule
(Pandey 2010) Thus the site of our experiment is a relatively high-caste dominant district of a
region of India (the North) in which caste divisions run deep
In the experiment participants in groups of six solved mazes in a classroom The six
boys in a session were generally recruited from six different villages but since this was not
always the case we will control for the number of other participants that a participant knew
Participants were brought to the experiment site by car Just before entering the car each
participant was asked in private his name village name fatherrsquos name grandfatherrsquos name and
caste On arriving at the site we verified in private with each participant his name and caste
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
14
before randomly assigning him to a treatment
Three conditions varied the salience of caste in the session which was always led by a
high-caste young female experimenter
Caste Not Revealed (the control condition) A session was composed of 3 H and 3 L No
personal information about the participants was revealed
Revealed Mixed (ie caste revealed in a mixed-caste session) The composition of a
session was the same as in the preceding condition but now the experimenter began a
session by saying that she would like to confirm some information with each participant
who should nod if it is correct Then the experimenter turned to each participant and
stated his name village name fatherrsquos name grandfatherrsquos name and caste
Revealed Segregated (ie caste revealed in a segregated session) This was the same as
the preceding condition except that a session was composed of either 6 H or 6 L
The priming mechanism reflects a way in which caste identity is actually made salient in
classroom settings This increases the external validity of our results Although an individualrsquos
caste is widely known in a village publicly referring to a childrsquos caste is not uncommon in rural
schools There is anecdotal evidence of teachers telling low-caste children to not drink from the
tap at the school lest it pollute the water for others While implementing this study we came
across some such instances Caste is commonly recorded in school enrollment books often
using different colors for high and low castes to identify caste-targeted entitlements such as
stipends and uniforms provided by state governments In villages people are frequently called
by their caste names Following the common usage in this area and the way that caste is
recorded in school enrollment books we used the traditional name for each caste (Thakur
Chamar etc) in revealing caste identity6
6 In the 1998-99 Indian National Family Health Survey households were asked to name their caste Most low-caste
respondents gave their actual caste name (eg Chamar) but a few used the more generic and politically correct
names Dalit harijan or Scheduled Caste (Marriott 2003)
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
15
We next describe the incentive schemes Participants were given a packet of 15 mazes to
solve in each of two 15-minute rounds7 Some participants had piece-rate incentives in both
rounds (the ldquoPP treatmentsrdquo) others had piece-rate incentives in round 1 and tournament
incentives in round 2 (the ldquoPT treatmentsrdquo) Under the piece-rate scheme a participant earned
one rupee per maze solved Under the tournament scheme he earned six rupees per maze solved
if he solved the most mazes in his session otherwise he earned nothing In case of a tie both
winners received the prize The tournament provided high-powered incentives a winner could
(and some did) earn 15 x 6 rupees equivalent to almost two daysrsquo unskilled adult wages
Figure 1 gives the organization of the experiment Experimental conditions were
identical in the first round of treatments (1) and (4) (2) and (5) and (3) and (6) and so we will
pool them when reporting first-round results
Figure 1 Experiment Design
Note PP means that the piece rate incentive applies in both rounds of maze-solving PT means that the piece rate
incentive applies in round 1 and the tournament incentive applies in round 2
7 The mazes are Xerox copies from httpgamesyahoocomgamesmazehtml level 3 Gneezy Niederle and
Rustichini (2003) showed that individuals do not solve mazes just for fun they respond to incentives
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
16
Recruitment We conducted the experiment in January and March 2003 and in March
2005 In January 2003 on days that schools were open we went to public schools near the site
of the experiment and chose high- and low-caste children for each day after pooling the
enrollment data for all nearby public schools A letter from the District Magistrate instructed the
teachers to cooperate with our team On days that schools were closed we visited homes in
nearby villages each evening to ask parentsrsquo permission to pick up their children the next day to
drive them to the junior high school that served as the site of the experiment In only rare
instances did parents refuse to let their children participate In March 2003 and March 2005 to
choose the subjects every day our team went to six randomly selected villages within a 20-
kilometer radius of the experiment site From each village we drew an equal number of high-
caste and low-caste children At most ten participants came from a single village nearly always
an equal number of H and L On each day we recruited participants from a new set of villages
Implementation On arrival at the experiment site participants waited in silence in a
large common room A research assistant provided each child a snack When the cars bringing
the participants for the morning or afternoon sessions had all arrived the participants were
directed in groups of six to a new set of classrooms where they remained for the rest of the
experiment They were not told anything about how or why the particular groups were formed
We next describe what took place during an experimental session which lasted about 70
minutes Under the Revealed Mixed and Revealed Segregated conditions but not in the control
the experimenter began a session by making public the identity of the participants as described
above After that all sessions proceeded in the same way The experimenter told the
participants that they would ldquotake part in two games of solving puzzlesrdquo She gave participants
the show-up fee of 10 rupees and described how to solve a maze in this way
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
17
hellipthere is one child The child has to go to the ball The solution is a path that takes the
child to the ball The black lines are walls The child cannot cross a wall
She gave participants five minutes to practice with an additional maze She explained that for
each maze they solved participants would receive an additional one rupee She checked to make
sure each child understood the incentive scheme She said that the incentive payments ldquowill be
given to every child in an envelope after the games are overrdquo Then she told the participants that
they would have 15 minutes to solve a packet of mazes and the first round of maze-solving
began
After the first round and without giving feedback on performance she explained that
there would be one more round of solving mazes explained the incentive scheme (piece rate or
tournament) and checked that each child understood it In a post-play survey participants gave
information about their background in private Mazes were graded blind Participants received
their earnings in sealed envelopes and were then taken home
Predictions Under piece rates the payoff to a participant depends on only his own
actions Therefore the theories of the fixed self would predict that increasing the salience of
caste would have no effect on behavior In contrast the theories of the frame-dependent self
would predict that increasing the salience of caste could evoke for a low-caste individual the
mental model in which Dalits are accepted only so long as they stay ldquoin their placerdquo which
would reduce the utility from high achievement and thus reduce performance For a high-caste
individual in contrast the prediction of the theories of the frame-dependent self is ambiguous
On the one hand making him more aware of his caste should if anything enhance his desire to
conform to the ideal of a high-caste person which is to be superior On the other hand making
caste more salient could evoke a mental model in which he has less need to achieve because he
has an entitlement to status In addition under the theory of stereotype threat or boost making
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
18
caste more salient may entail a negative productivity shock to L and a positive productivity
shock to Hmdashsee Figure 2
Figure 2 Predicted Effects of Increasing the Salience of Caste under Piece Rate Incentives
Theory Predicted effect of increasing caste salience on the performance of
High caste Low caste
The Fixed Self
Individuals have well-defined preferences
that are always salient
None
None
The Frame-Dependent Self
Increasing an individualrsquos awareness of an
aspect of his identity may cue a mental model
Individuals have multiple sets of preferences
one for each mental model Cues to a
negatively stereotyped identity can also
impair the ability to perform
Ambiguousmdash
Cueing an identity whose norm
is to be superior increases the
utility from achievement which
improves performance but
evoking a worldview in which
life chances depend less on
effort than on caste impairs
performance
Declinesmdash
Making a low-caste person
more aware of his caste (i)
evokes a worldview in which it
is a norm violation for him to
excel and (ii) may trigger
stereotype threat
4 Descriptive Statistics
In this section we describe the participantsrsquo characteristics and broadly summarize the results8
Table 1 shows that parents of H have much greater education than parents of L For simplicity
the table groups together Revealed Mixed and Revealed Segregated as the ldquoidentity conditionsrdquo
The table shows that 45 of all H compared to 12 of all L have a mother with at least six years
of schooling (These are weighted averages across conditions calculated using Figure 1) Both
parents are illiterate among only 5 of H compared to 28 of L Only 8 of H have fathers
8 In each time period in which we conducted the experiment (January and March 2003 and March 2005) we held at
least six sessions under PP incentives in the control condition As shown in Web Appendix Table S1 there were no
significant differences in output by time period Therefore we pool the data across the three time periods We also
found no experimenter effects on the number of mazes solved per round
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
19
who are day laborers compared to 18 in the case of L These differences highlight the need to
examine whether the correlates of caste can explain the differences between H and L in our
results We can do that because the distribution of parentsrsquo characteristics for H shares a
common support with that for L For example there are not only L who have mothers with no
schooling there are also H whose mothers have no schooling We collected data on two other
variables in the post-play survey exposure to mazes and the number of other participants in a
session that a subject knows
Table 1 shows that the randomization between the control and identity conditions was
largely successful However in the identity conditions participants have parents with a
significantly higher level of education and are significantly more likely to have had some
exposure to mazes These differences should if anything improve performance in the identity
conditions compared to the control An effect that goes the other way is that the low caste is on
average slightly more likely to be in 6th
than 7th
grade in the identity conditions9 We control for
these factors in the analysis and all results described in Section 1 are robust to these controls In
9 Unlike for L the randomization in terms of grade in school is perfect for H across control and identity treatments
For L the mean grade in school is 653 for control (Caste Not Revealed) and 634 for the identity treatments and this
difference is significant as Table 1 shows Further disaggregating the data by treatment reveals that the problem of
imbalance in grade for L lies with the control versus all other treatments and that this is not so in the case of H For
H for piece rate conditions the means for grade in school (with standard deviations in parentheses) are 653 (050)
for Caste Not Revealed 653 (050) for Revealed Mixed and 669 (047) for Revealed Segregated For the
tournament conditions the means are 647 (050) for Caste Not Revealed 644 (050) for Revealed Mixed and 643
(050) for Revealed Segregated On the other hand for L while there is little difference in the mean grade in school
among the four identity treatments the mean grade for Caste Not Revealed is higher For L for piece rate
conditions the means are 657 (050) for Caste Not Revealed 637 (048) for Revealed Mixed and 630 (046) for
Revealed Segregated For the tournament conditions the means are 643 (050) for Caste Not Revealed 630 (046)
for Revealed Mixed and 639 (049) for Revealed Segregated
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
20
particular the results that the low caste underperforms when caste is revealed are robust to
controlling for grade in school
Table 1 Descriptive Statistics for Participants
High caste
Low caste
Caste Not
Revealed
Identity
conditions
Caste Not
Revealed
Identity
conditions
Motherrsquos education
None 32 25
75 68
Years ϵ (06) 26 29
17 17
At least 6 years 42 46
8 15
Fatherrsquos education
None 6 6
26 31
Years ϵ (06) 7 13
22 19
At least 6 years 86 81
52 50
Both parents illiterate 7 4
26 29
4 7
7 5
Mother works outside
the home
8 9 16 19 Father is a day laborer
Grade in school 651 651 653 634
Previous exposure to
mazes 7 15
4 16
Mean number of other
participants known 055 114 056 103
Note This table looks at the balance between treatments in which caste identity is revealed (ldquoIdentity
conditionsrdquo) and those in which it is not Except for the last row the characteristics reported in this table
are binary For example ldquoboth parents illiteraterdquo =1 if both parents have no formal education and
otherwise it is zero ldquoprevious exposure to mazes beforerdquo =1 if the subject had seen mazes before and
otherwise it is zero and grade in school is equal to either 6th or 7th For binary variables the tests of
equality of means across conditions for the high caste are based on logit regressions one for each
characteristic and similarly for the low caste For ldquomean number of other participants knownrdquo the test of
equality of means is based on a t-test denotes rejection of the equality of means for the control and
identity conditions at p lt 005
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
21
Figure 3 shows the average number of mazes solved by H The figure is divided into
three blocks Block 1 is round 1 block 2 is round 2-piece rate and block 3 is round 2-
tournament In each block H output is lowest in Revealed Segregated Under the Mann-
Whitney U-test the differences between Revealed Segregated and the control are significant at p
lt 05 in all blocks In block 2 average output is higher in Revealed Mixed than in the control
but the difference is not significant
Figure 3 Average Output of High-Caste Participants
Note Brackets indicate differences between treatments with 95 confidence based on the Mann-
Whitney U-test
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
22
Figure 4 superimposes on Figure 3 the average outputs of L All three blocks show that
when caste is not revealed the average output of H is almost the same as that of L However
when caste is made public the performance declines for L are generally steeper than those for H
A significant caste gap emerges in Revealed Mixed in both rounds
Figure 4 Average Output of High-Caste and Low-Caste Participants
Note Vertical lines indicate caste gaps that are statistically significant based on the Mann-Whitney test
with 95 confidence
Figure 5 shows how the identity conditions impair L relative to H performance at the top
of the performance distribution The figure reports the ratio of L participants to all participants
with output at or above each decile in round 2 (If H and L were equally represented in each
0
1
2
3
4
5
6
7
8
Ave
rag
e o
utp
ut
High Caste
Low Caste
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
ROUND 1 ROUND 2
Piece Rate Piece Rate Tournament
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
23
decile and if varying caste salience had the same effect on H and L then all points in the figure
would lie along the horizontal line at 05 ie in any cut of the distribution the proportion of L
participants would be one-half) The figure shows that if the top 10 percent of participants were
selected based on their performance in the control (Caste Not Revealed) L would be in the
majority But if selection was based on performance in Piece rate or Tournament Revealed
Mixed L would be in the minority And if selection was based on performance in Piece rate
Revealed Segregated it would result in an equal representation of H and L Thus whether H or
L are overrepresented in the top decile depends on the context in which the boys perform
Figure 5 Proportion of the Low Caste above each Performance Decile in Round 2
(Cumulative)
Note There is in general more than one participant whose performance ranks him at the border between
two deciles To explain the adjustment we make for this case we use an example Let n(20) = the
number of L in decile 20 n(30) = the number of L in decile 30 and n(b) = the number of L at the border
0
01
02
03
04
05
06
07
10 20 30 40 50 60 70 80 90 100
Pro
po
rtio
n
Decile (10 is top decile)
Tournament Caste Not Revealed
Piece rate Caste Not Revealed
Piece rate Revealed Segregated
Tournament Revealed Mixed
Piece rate Revealed Mixed
Tournament Revealed Segregated
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
24
between deciles 20 and 30 Finally denote an L whose performance ranks him at the border between
these two deciles as a border person Lb Consider a case where n(20) = 2 n(30) = 1 and n(b) = 3 How
do we allocate the three border persons We allocate them so that the proportion of border persons in
decile 20 is equal to the ratio n(b)[n(20)+n(30) + n(b)] = 05 and the proportion of border persons in
decile 30 is also equal to this ratio Thus we allocate 2 border persons to decile 20 and the remaining
border person to decile 30 After the allocation decile 20 has 2L + 2Lb the proportion of boundary Lrsquos
in the decile is 5 as required And decile 30 has 1L + 1Lb the proportion of boundary Lrsquos in the decile is
also 5 as required Intuitively since there are twice as many Lrsquos in decile 20 as in decile 30 we allocate
twice as many border persons to decile 20 as to decile 30 For H the procedure is analogous
_____________________
5 Measuring Treatment Effects
51 Number of Mazes SolvedmdashFull Sample
We find patterns similar to those in Figure 4 in regressions that control for individual and family
characteristics We pool all the observations and allow for interactions among caste context
and incentives Table 2 columns (1)-(4) report OLS estimates with robust standard errors
clustered at the individual level for the following specification
Mazes solved in a round = α + ω(round is 2) + β(subject is H) + γ(session cues identity) (1)
+ (subject is Hsession cues identity) + τ(Tournament) + λ (Tournamentsubject is H) +
ξ(Tournamentsession cues identity) + θ(Tournamentsubject is H session cues identity) + μΖ + error
where Z is a vector of individual and family characteristics The term α measures the predicted
output in the omitted case an L in Piece rate control in round 1 The next eight coefficients
(from ω to θ) measure the effects of round caste and treatments and the two-way and three-way
interactions10
10 For example γ is a vector that measures the difference for L between an identity condition (Revealed Mixed or
Revealed Segregated) and the control under piece rate incentives Using a subscript s for Revealed Segregated α +
ω + γs is the predicted output of L in round 2 of Revealed Segregated under piece rate incentives The predicted
output of H in Revealed Segregated under tournament incentives is α + ω + β + γs + s + τ + λ + ξs + θs
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
25
One result is immediate Low-caste boys solve mazes just as well as high-caste boys
when caste is not revealed The estimated coefficients on H and TH are always very small and
insignificant We get sensible results for round 2 and grade in school both factors significantly
improve round performance in every specification
Specification (1) uses only caste and treatment indicators Specification (2) adds controls
for individual characteristics grade in school previous exposure to mazes and number of other
participants known in a session Specification (3) adds controls for family characteristics In
this section we will analyze the results under the piece rate incentive we will analyze the results
under the tournament incentive in the Appendix
Between specifications (1) and (2) there is only one change in the set of significant
treatment effects the output decline by L in Revealed Mixed is no longer significant Table 3
reports the treatment effects in the first two columns The treatment effects of Revealed Mixed
for each caste (016 for H and -058 for L) are individually insignificant but jointly produce a
significant caste gap in favor of H The treatment effect of moving from the control to Revealed
Segregated is a significant decline of -093 mazes for both H and L This effect is roughly
double the effect on output of being in 6th
instead of 7th
grade (= -043 as shown in Table 2)
In order to check whether the channel through which social identity influences
behavior is classmdasha poor-versus-rich effect on performance as in Croizet and Claire (1998)mdash
rather than caste we next consider the effect of controls for family characteristics In Table 2
column (3) we control for parentsrsquo education motherrsquos employment outside the home and
fatherrsquos employment as a day laborer Because stigma is associated with daily wage-labor
our post-play survey did not ask ldquoIs your father a day laborerrdquo Instead the survey asked
about the fatherrsquos occupation We formed a binary variable for daily wage labor based on the
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
26
responses Adding controls for class reduces the declines in performance in Revealed
Segregated The treatment effect of moving from the control to Revealed Segregated is
marginally significant for H (-090 p lt 010) not for L (-074 p = 011) However two key
results are robust to adding controls for class First merely making caste public without
segregating H from L does not impair H performance (the effect is 014= -51 + 65 and
insignificant) Second H and L are equally good at solving mazes when caste is not
revealed whereas in Revealed Mixed a significant caste gap favoring H emerges (= 035 +
065=10 p = 001)
We have emphasized specification (2) in Table 2 more than specification (3) because
we cannot reject the hypothesis that parental variables have no effect on performance
F(6486)= 158 p-value=011 The only proxy for class that is individually significant is
fatherrsquos education and its effect is not in the direction that is predicted by the hypothesis that
class stigma impedes performance
It might be however that parental variables matter for L but not H because having
educated parents alleviates low-caste stigma Therefore in unreported regressions we reran
specification (3) separately for H and L participants We still find that parental variables have
little explanatory power and are insignificant by an F-test We also checked for the effect of
having two illiterate parents We find that this is not significant (result not shown) In these and
all other regressions that we have run we find no evidence that class is the channel through
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
27
Notes Standard errors in parentheses are robust to heteroskedasticity and observations are clustered at the level of the individual The omitted case is L in Caste Not
Revealed under piece rate incentives Column (4) excludes participants who have zero output in both rounds Round 2 = 1 for round 2 and zero for round 1 Grade in
school = 1 if the participant is in grade 7 0 if he is in grade 6 Previous exposure to mazes = 1 if some time before the experiment the participant had seen mazes 0
otherwise Number of other participants known is the number of others in the experimental session known to a given participant plt001 plt005 plt010
Table 2 OLS Estimates of the Determinants of Output per Round and Output Change between Rounds
Dependent variable Output per round Output change
between rounds
Without
individual and
family
characteristics
With
individual
characteristics
With
individual
and family
characteristics
Excluding
participants
who solved
zero mazes
With individual
characteristics
(1) (2) (3) (4) (5)
High caste (H) 029 016 035 056
025
(035) (036) (039) (034)
(042)
Round 2 214 217 227 233
(015) (016) (016) (016)
Revealed Mixed -070 -058 -051 -007
-054
(034) (037) (038) (035)
(039)
Revealed Segregated -097 -093 -074 -070
-086
(037) (040) (046) (040)
(043)
Tournament (T) 140 145 144 128
106
(065) (066) (066) (066)
(055)
Revealed Mixed H 075 073 065 -012
064
(048) (050) (053) (047)
(060)
Revealed Segregated H 002 -001 -016 -052
-064
(054) (058) (065) (056)
(064)
TH -026 -012 -014 -004
-044
(089) (090) (096) (086)
(077)
Revealed Mixed T -135 -159 -202 -148
-102
(076) (078) (078) (077)
(069)
Revealed Segregated T -277 -305 -302 -282
-138
(076) (077) (082) (077)
(076)
Revealed Mixed T H -007 002 067 -016
-026
(108) (111) (120) (108)
(100)
Revealed SegregatedT H 173 173 191 192
256
(114) (121) (133) (116)
(105)
Grade in school
043 051 045
034
(021) (023) (021)
(021)
Previous exposure to mazes
037 051 035
-019
(030) (033) (029)
(036)
Number of participants
006 010 001
002
known
(009) (009) (008)
(009)
Mothers education Є(06)
028
(030)
Mothers education ge 6
044
(033)
Fathers education Є(06)
-064
(039)
Fathers education ge 6
-091
(034)
Mother employed outside
005
home
(053)
Father not a day
055
laborer
(035)
Constant 326 297 276 298
216
(024) (028) (050) (028)
(032)
R2 0189 0197 0221 0223 0080
N 1164 1076 928 1008 538
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
28
Table 3 Effects of Making Identity Public in Mixed and Segregated Sessions under the Piece-rate Incentive
Output per round full sample
Output per round excluding
participants who solved zero mazes
Output change between rounds full
sample
H L
Caste gap
significant
H L
Caste gap
significant H L
Caste gap
significant
(1) (2) (3) (4) (5) (6)
(7) (8) (9)
Estimated treatment
effect of
Revealed Mixed 016 -058
-019 -007
010 -054
(036) (037)
(034) (035)
-046 -039
Revealed Segregated -093 -093
-123 -070
-150 -086
(042) (040) (041) (040) (049) -043
Mean outcome in
Caste Not Revealed 422 406
471 415
241 216
Effect of Revealed
Segregated as of
mean
in Caste Not Revealed -22 -23 -26 -17 -62 -40
Notes Cluster-robust standard errors are in parentheses The mean output per round pools both rounds The treatment effects can be derived
from the regressions in Table 2 Effects in columns (1)-(3) are from regression (2) those in columns (4)-(6) are from regression (4) those in
columns (7)-(9) are from regression (5) However it is easier to obtain the effects and standard errors for H by running regressions in which H is
the benchmark caste plt001 plt005 plt01
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
29
which caste influences behavior However since we do not have measures of income and
wealth the concern that unobserved class variables may matter remains
52 Between-Round Change in the Number of Mazes Solved
As an additional check on our results we consider the treatment effects on the change in
output between rounds This is shown in the last three columns of Table 3 In Piece-rate control
H and L are significantly improving their skills across rounds (217 mazes p lt 01) Revealed
Segregated reduces output change between rounds for H by 62 (p lt 01) and for L by 40 (p lt
05)
53 Success or Failure in Learning How to Solve a Maze
In the remainder of this section we decompose performance into two stages
Stage 1 Learning a new skill The participant learns what it means to solve a maze
The outcome is binarymdashsuccess or failure We measure failure by zero output by a
participant during the 30 minutes of maze-solving
Stage 2 Applying the skill In this stage the outcome is the number of mazes solved
conditional on success in learning how to solve a maze
Table 4 reports the failure rate by identity condition treatment and caste For H in PP
failure is greater in the control (9) than in the two identity conditions (2 and 3) For H in
PT however the effect of revealing caste is not consistent across the PP and PT treatments
7 (230) fail in the control 0 (060) fail in Revealed Mixed and 10 (330) fail in Revealed
Segregated Since the experimental conditions in PP and PT were identical until round 2 and
since if caste was revealed it happened at the beginning of a session it is reasonable to pool PP
and PT within the Revealed Mixed condition and within the Revealed Segregated condition
When we do this a pattern emerges in which revealing caste decreases the failure rate
among H failure is greater in the control (8 or 9108) than in either Revealed Mixed (2 or
2120) or Revealed Segregated (7 or 460) For L the pattern is the reverse failure is smaller
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
30
in the control (2 or 2108) than in either Revealed Mixed (13 or 15120) or Revealed
Segregated (9 or 666)
To fit a logit model it is necessary to collapse the two identity conditions and also the
two incentive conditions11
We estimate
Failure = α + β(subject is H) + γ(session cues identity) (2)
+ (subject is H session cues identity) + μΖ + error
where Z are individual characteristics The benchmark case is L Caste Not Revealed We use
the logit results reported in Supporting Table S2 to predict the probability of failure Figure 6
presents the results The figure shows that revealing caste reduces failure among H from 8 to
2 and increases failure among L from 1 to 11 controlling for individual characteristics
The treatment effects are significant and robust to the addition of controls for household
characteristics The effects are consistent with the predictions of stereotype susceptibility when
participants are made more aware of caste H are less likely and L are more likely to fail to learn
how to solve a maze
54 Number of Mazes Solved by the Subsample Excluding Non-Learners
An advantage of decomposing performance into stages is that we can consider treatment effects
on performance conditional on knowing how to solve a maze We report the effects in Table 3
columns (4)-(5) Are the qualitative results for the full sample robust in the subsample
11
Otherwise the estimates are unbounded since some cells in Table 4 are empty
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
31
Table 4 Proportion of Participants with Zero Output
Treatment
Number of participants with zero
output Total number of
participants of the respective
caste in the treatment
Proportion
High caste Low caste
High caste Low caste
PP- Caste Not Revealed 778 278 009 003
PP-Revealed Mixed 160 960 002 015
PP- Revealed Segregated 130 230 003 007
PT -Caste Not Revealed 230 030 007 0
PT- Revealed Mixed 060 660 0 010
PT -Revealed Segregated 330 436 010 011
Figure 6 Predicted Probability of Failure
Note Based on the logit regression in Supporting Table S2 column (1) The control variables are grade in school
prevision exposure to mazes and number of other participants known in the session The predicted probabilities are
estimated at the means of the control variables
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
32
The high caste The treatment effects of making caste public are stronger in the
subsample than in the full sample The reason is that in the subsample we are not capturing the
stage 1 effect in which making caste public reduces the probability that H will fail to learn how
to solve mazes In the full sample the treatment effect of Revealed Segregated is -093 (p lt
005) compared to -123 (p lt 001) in the subsample We view this latter figure (-123) which is
a 26 decline in performance as our best estimate of the entitlement effect By this we mean the
effect on a high-caste boyrsquos output conditional on his knowing how to solve a maze of moving
from the control condition (where the authority figure is silent about caste) to Revealed
Segregated (where segregation cues high-caste boysrsquo place in the social order) We call it the
entitlement effect because we conjecture that the treatment effect reflects a reduced need to
achieve in a situation that cues onersquos place in the traditional ascriptive social order The
entitlement effect on H is thus about the same size as the performance decline by L that could be
attributed to stereotype threat (= -23 see column 2) These treatment effects are large They
are the same order of magnitude ndashndashbut of opposite signmdashas the effect of switching from the
piece rate to the winner-take-all tournament in the control condition (25 for H and 28 for L)
Besides the entitlement effect other hypotheses might explain the decline in H
performance in Revealed Segregated The first is a perverse kind of stereotype susceptibility
Such an effect occurred in one of the two experiments in Shih et al (2002) They compared a
mild versus blatant activation of a positive stereotype They found in one experiment that
priming Asian identity had no effect on Asian-Americansrsquo performance on a math test and in the
second experiment that it significantly impaired performance perhaps by creating anxiety in
participantsrsquo minds that their performance would not confirm the high expectations for Asians in
the US But our finding that the identity conditions in aggregate reduced the proportion of
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
33
individuals who failed to learn how to solve a maze does not support the view that the identity
conditions increased anxiety
Second because the caste order is always contested it could be that in Revealed Mixed
H feel a need to demonstrate (even if only to themselves) their superiority and that they do not
feel this need in Revealed Segregated Ultimately the second hypothesis comes down to largely
the same thing as our preferred one namely that contexts that reinforce the complacency of the
high caste in their superior status induce them to value less the rewards from individual
achievement
The low caste Next consider the treatment effects for L in the subsample compared to
the full sample (Table 3 columns (2) and (5)) We find that making caste public reduces output
less in the subsample because in the subsample we are not capturing the stage 1 effect in which
making caste identity public increase the probability that L will fail to learn how to solve a maze
In the subsample the treatment effect of Revealed Segregated is no longer significant under
piece rates This suggests that under piece rate incentives the identity conditions impair L
performance primarily by reducing their success at learning the new task (maze-solving)
6 Further Evidence
In this section we discuss evidence that bears on the mental frames that Revealed Segregated
evokes in high-caste and low-caste boys
Revealed Segregated reduces the self-confidence of the low-caste boys In an earlier
experiment (Hoff and Pandey 2005) we used random assignment of participants to the three
conditions of caste salience (Caste Not Revealed Revealed Mixed and Revealed Segregated) to
assess the relationship between caste salience and self-confidence In a six-person session H
and L were taught how to solve a wooden puzzle that we had constructed along the lines of the
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
34
game Rush-Hour Traffic Jam At the end of the session participants had to make a choice
between a sure payoff and a lottery with a high payoff if the individual solved a new puzzle
successfully and zero otherwise In choosing the lottery a participant was betting on his own
success The outcome is thus a test of self-confidence The results showed no significant caste
gap in the acceptance rate of the lottery in both the Caste Not Revealed and Revealed Mixed
conditions In contrast in Revealed Segregated there was a large and significant caste gap in the
proportion that accepted the lottery when the puzzle was difficult and the judge had some
discretion in evaluating a playerrsquos success 70 of H compared to only 33 of L accepted the
lottery (p = 004) This evidence supports the hypothesis that Revealed Segregated evokes a
mental model in which a low-caste may feel ldquoI canrsquot or donrsquot dare to excelrdquo
Observational studies of school enrollment point to a causal effect of caste salience on
low school enrollment by both high- and low-caste students Consider again our conjecture
that when the social division between high and low castes is salient a high-caste boy may think
ldquoI donrsquot need to excelrdquo If this is correct it would predict that when caste boundaries are eroded
a high-caste boy (or his family on his behalf) may think ldquoI need to work harder to better
myselfrdquo This prediction is consistent with the findings in Kochar (2004) She analyzes an
Indian government policy to construct schools in low-caste hamlets The policy led to an
increase in low-caste enrollment The increased school enrollment of low-caste children had
however an unintended effect it increased the enrollment rate of the upper castes Thus aid
targeted to Dalits did not as policy-makers had expected narrow the schooling gap between the
Dalits and the rest of society The increase in high-caste enrollment maintained the relative
superiority of the high caste in years of education12
12
We thank Anjini Kochar for bringing her work to our attention
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
35
Finally an observational study in Pakistan identifies a causal impact of high-caste
dominance on low enrollment of low-caste children in school This finding is consistent with our
conjecture that cues to the caste order evoke a mental frame that impairs low-caste individualsrsquo
ability or desire (or both) to achieve in the classroom Jacoby and Mansuri (2011) analyze data
from a survey of over 3000 households and 1000 elementary schools in rural Pakistan They
define a settlement as high-caste-dominant if a high caste owns the majority of land in the
settlement giving them the power to enforce the traditional caste order They show that low-
caste children are deterred from enrolling in schools in high-caste dominant settlements They
find that the very low enrollment of low-caste children13
for whom the closest available school is
in a high-caste-dominant hamlet can account for the entire enrollment gap favoring high-caste
over low-caste children The following responses from low-caste women to the question ldquoDo
children receive the same treatment from teachersrdquo illustrate the kinds of exclusionary norms
imposed on low-caste individuals
ldquoThey let the daughters of [high castes] use the latrines but tell our daughters to use the
fields because you stinkrdquo ldquoThe teachers make the daughters of Zamindar Zaats [high
castes] sit inside the rooms under the fans Our poor children are outside under the sun
and dustrdquo (Jacoby and Mansuri p 7)
These two observational studies mdashthe only studies of which we are aware that examine
the effect on achievement of increasing or decreasing the salience of the caste ordermdashindicate
that when onersquos traditional status in the social order seems more fixed both high- and low-caste
individuals are less likely to enroll in school
13
Especially girls for whom honor considerations restrict the freedom to travel outside their own hamlet
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
36
7 Conclusion
The experimental data show that being hard-working or clever is not a trait of the person in the
sense that it is always there in a fixed manner Instead situational cues to onersquos place in the
social order influence the expression of these traits Cues to caste identity in mixed-caste groups
depressed the ability of the low-caste to learn a new skill Segregation by caste status did not
impair the ability of high-caste boys to learn a new skill but sharply reduced their performance
in using the new skill The influence of identity on performance suggests the usefulness of the
theories that posit a frame-dependent self In this view context may affect performance by
changing not what it is objectively appropriate to do but instead by changing the set of
associations that are elicited and the mental model through which the individual interprets the
situation Borrowing from the title of a recent book Framed by Gender (Ridgeway 2011) one
might say that our subjects were framed by caste Our findings provide evidence of one possible
source of the caste gap in school performance in north India Pandey et al (2010) find that after
controlling for observed family characteristics and school fixed effects high-caste students have
significantly higher test scores than low-caste students in rural public primary schools in the
Indian states of Uttar Pradesh and Madhya Pradesh where caste and class hierarchies run deep
Our findings contribute to a richer view of institutional change than one that follows from
the standard definition of institutions-as-rules Besides rules institutions also entail social
identities and worldviews (see Greif 2006 section 216) Social identities made salient by
situational cuesmdashby frockcoats and tight-laced corsets race names or caste namesmdashhave an
independent influence on ldquomaking up peoplerdquo After the rules of an institution have been
abolished and the structure of power that underpinned them has changed the identities founded
on the superseded rules will continue to affect chronic ways that people think about themselves
and interpret the world unless prevailing situations make those identities less salient For
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
37
example when towns emerged in medieval Europeldquo[i]t was not only that the serf having
escaped from the countryside found legal freedom in the town but that the while social
atmosphere there was open to ambition and talentrdquo (Cipolla 1994 p 119) The new possibilities
for upward mobility depended on change in the social as well as the legal environment No one
would argue with this although these ideas rarely enter into economic models
Rural India is in transition between a feudal and a capitalist economic system Our
findings suggest that boys in India hold in uneasy tension a traditional worldview in which
caste is destiny and a modern worldview in which each person shapes his own destiny We
measured the effects on childrenrsquos performance of manipulating the publicness and salience of
the traditional caste identities Our findings are consistent with the hypothesis that changing the
publicness and salience of caste affects which of the two worldviews is uppermost in the
childrenrsquos minds
We have argued that it is a useful construct to posit that an individual has multiple
preferences one for each of his mental models or worldviews because this perspective opens up
a new set of policy options for enhancing human capital formation and development It is also
useful for understanding long-run social change which entails changes in the salience of
particular identities the set of possible identities and the possible ways of understanding a
situation This perspective highlights the relevance to economics of understanding how the set
of possible identities (and not merely the possible technologies and the stock of resources)
evolves in a society Recent work in economics takes up this exciting question (eg Hoff and
Sen 2006 Munshi and Rosenzweig 2006 Hoff and Stiglitz 2010 Greif and Laitin 2004 and
Greif and Tabellini 2012)
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
38
References
Afridi Farzana Sherry Xin Li and Yufei Ren 2010 Social Identity and Inequality The Impact
of Chinarsquos Hukou System Manuscript University of Texas-Dallas
Akerlof George and Rachel Kranton 2000 The Economics of Identity Quarterly Journal of
Economics 115(3) 715-753
Akerlof George and Rachel Kranton 2002 Identity and Schooling Some Lessons for the
Economics of Education Journal of Economic Literature 40(4) 1167-1201
Alter Adam 2013 Drunk Tank Pink and Other Unexpected Forces that Shape how We Think Feel
and Behave New York Penguin Press
Ambady Nalini Margaret Shih Amy Kim and Todd Pittinsky 2001 Stereotype Susceptibility in
Children Effects of Identity Activation on Quantitative Performance Psychological Science
12(5) 385-390
Bandiera Oriana Iwan Barankay and Imran Rasul Social Preferences and the Response to Incentives
Evidence from Personnel Data 2005 Quarterly Journal of Economics 120(3) 917-962
Begley Sharon 2007 Train Your Mind Change Your Brain Ballantine Books New York
Bellmore Amy and Ayako Tomonaga 2009 When Kids Use Ethnicity and Gender to Bully
Web httpwwweducationcomreferencearticleethnicity-gender-bullying
Benjamin Daniel J James J Choi and A Joshua Strickland 2010 Social Identity and
Preferences American Economic Review 100(4) 1913-1928
Beacuteteille Andreacute 2011 The Andreacute Beacuteteille Omnibus Oxford University Press Delhi
Cassan Guilhem 2011 Law and Identity Manipulation Evidence from Colonial Punjab Paris
School of Economics manuscript
Cohn Alain Michel Andreacute Mareacutechal and Thomas Noll 2013 Saliency of Criminal Identity
Causes Dishonest Behaviour An Experiment Behind Bars University of Zurich
manuscript
Connerton Paul 1989 How Societies Remember Cambridge Cambridge University Press
Cipolla Carlo M 1994 Before the Industrial Revolution European Society and Economy
1000-1700 WW Norton amp Company New York 3rd
edition
Croizet Jean-Claude and Theresa Claire 1998 Extending the Concept of Stereotype Threat to
Social Class The Intellectual Underperformance of Students from Low Socioeconomic
Backgrounds Personality and Social Psychology Bulletin 24(6) 588-594
Danziger Shai and Robert Ward 2010 Language Changes Implicit Associations between Ethnic
Groups and Evaluation in Bilinguals Psychological Science 21(6) 799-800
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
39
Deacuteliege Robert 1999 The Untouchables of India Berg Publishers Oxford UK
Deshpande Ashwini 2011 The Grammar of Caste Economic Discrimination in Contemporary
India Oxford Oxford University Press
DiMaggio Paul 1997 Culture and Cognition Annual Review of Sociology 23(1) 263-287
Fang Hanming and Loury Glenn C 2005 Dysfunctional Identities Can Be Rational American
Economic Review 95(2) 104-111
Fehr Ernst Karla Hoff and Mayuresh Kshetramade 2008 Spite and Development American
Economic Review Papers amp Proceedings 98(2) 494-499
Ferguson Ronald F 2003 Teachers Perceptions and Expectations and the Black-White Test
Score Gap Urban Education 38(4) 460-507
Fryer Roland G Jr 2011 The importance of segregation discrimination peer dynamics and
identity in explaining trends in the racial achievement gap in Handbook of Social
Economics eds Jess Benhabib Alberto Bisin and Matthew O Jackson Amsterdam
North-Holland 165-1192
Fryer Roland G Jr Steven D Levitt and John A List 2008 Exploring the Impact of Financial
Incentives on Stereotype Threat Evidence from a Pilot Studyrdquo American Economic
Review Papers and Proceedings 98(2) 370-375
Gneezy Uri Muriel Niederle and Aldo Rustichini 2003 Performance in competitive
environments Gender differences Quarterly Journal of Economics 118(3) 1049-1074
Greif Avner and David Laitin 2004 A Theory of Endogenous Institutional Change American
Political Science Review 98(4) 14-48
Greif Avner and Guido Tabellini 2012 The Clan and the City Sustaining Cooperation in
China and Europe Working paper 445 IGIER Bocconi University
Greif Avner 2006 Institutions and the Path to the Modern Economy Lessons from Medieval
Trade Cambridge Cambridge University Press
Gupta Dipankar 2000 Interrogating Caste Understanding Hierarchy and Difference in Indian
Society Penguin Books New Delhi
Hacking Ian 1986 Making up People In TC Heller M Sosna and D E Wellbery (Eds)
Reconstructing Individualism Stanford University Press Stanford
Hoff Karla Mayuresh Kshetremade and Ernst Fehr 2011 Caste and Punishment The Legacy
of Caste Culture in Norm Enforcement Economic Journal 121(556) F449-F475
Hoff Karla and Priyanka Pandey 2005 Opportunity is Not Everything How Belief Systems
and Mistrust Shape Responses to Economic Incentives Economics of Transition 13(2)
445-472
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
40
Hoff Karla and Priyanka Pandey 2006 Discrimination Social Identity and Durable
Inequalities American Economic Review Papers and Proceedings 96(2) 206-211
Hoff Karla and Arijit Sen The Kin System as a Poverty Trap in Poverty Traps Samuel
Bowles Steven Durlauf and Karla Hoff (eds) Princeton University Press 2006 95-115
Hoff Karla and Joseph E Stiglitz 2010 Equilibrium Fictions A Cognitive Approach to Societal
Rigidity American Economic Review Papers and Proceedings 100(2) 141-146
Human Rights Watch 1999 Broken People Caste Violence against Indiarsquos ldquoUntouchablesrdquo
New York Human Rights Watch
Jacoby Hanan and Ghazala Mansuri 2011 Crossing Boundaries Caste Stigma and Schooling
in Rural Pakistan WPS 5710 The World Bank Washington DC
Kapur Devesh Chandra Bhan Prasad Lant Pritchett and D Shyam Babu 2010 Rethinking
Inequality Dalits in Uttar Pradesh in the Market Reform Era Economic and Political
Weekly 45(35) 39-49
Kochar Anjini 2004 Reducing Social Gaps in Schooling Caste and the Differential Effect of
School Construction Programs in Rural India Manuscript Stanford University
Krendl Anne C Jennifer A Richeson William M Kelley and Todd F Heatherton 2008 The
Negative Consequence of Threat A Functional Magnetic Resonance Imaging Investigation of
the Neural Mechanism Underlying Womenrsquos Underperformance in Math Psychological
Science 19(2) 168-175
Lambarraa Fatima and Gerhard Riener 2012 ldquoOn the norms of charitable giving in Islam A
field experimentrdquo DICE Discussion Paper 59 Heinrich-Heine-Universitat Dusseldorf
LeBoeuf Robyn A Eldar Shafir and Julia B Bayuk 2010 The Conflicting Choices of
Alternating Selves Organizational Behavior and Human Decision Processes 111(1) 48-
61
Marriott Alan 2003 Dalit or Harijan Self-Naming by Scheduled Caste Interviewees
Economic and Political Weekly 38(36) 3751-3752
Munshi Kaivan and Mark Rosenzweig 2006 Traditional Institutions Meet the Modern World
Caste Gender and Schooling Choice in a Globalizing Economy American Economic
Review 96(4) 1225-1252
Munshi Kaivan and Nicholas Wilson 2008 Identity Parochial Institutions and Career
Decisions Linking the Past to the Present in the American Midwest Manuscript Brown
University
Ogbu John U 1999 Beyond Language Ebonics Proper English and Identity in a Black-
American Speech Community American Educational Research Journal 36(2) 147-184
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
41
Ogunnaike Oludamini Yarrow Dunham and Mahzarin R Banaji 2010 The Language of
Implicit Preferences Journal of Experimental Social Psychology 46 999-1003
Pandey Priyanka 2010 Service Delivery and Corruption in Public Services Does History
Matter American Economic Journal Applied Economics 2 190-204
Pandey Priyanka Sangeeta Goyal and Venkatesh Sundararaman 2010 Public Participation
Teacher Accountability and School Outcomes in Three States Economic and Political
Weekly 45(24) 75-83
Pollak Seth D 2008 Mechanisms Linking Early Experience and the Emergence of Emotions
Current Directions in Psychological Science 17(6) 370-375
Public Report on Basic Education in India (PROBE Team) 1999 Public Report on Basic
Education in India Oxford University Press New Delhi
Rao Vijayendra and Radu Ban 2007 Political Construction of Caste in South India
manuscript
Ridgeway Cecilia L 2011 Framed by Gender Oxford University Press Oxford
Ritterhouse Jennifer 2006 Growing up Jim Crow How Black and White Southern Children
Learned Race University of North Carolina Press Chapel Hill
Salant Yuval and Ariel Rubinstein 2008 (Af) Choice with Frames Review of Economic
Studies 75(4) 1287-1296
Schmader Toni Michael Johns and Chad Forbes 2008 An Integrated Process Model of
Stereotype Threat Effects on Performance Psychological Review 115(2) 336-356
Sen Amartya 2006 Identity and Violence The Illusion of Destiny WWNorton ampCo NY
Shah Ghanshyam Harsh Mander Sukhadeo Thorat Satish Deshpande and Amita Baviskar
2006 Untouchability in Rural India Sage Publications New Delhi
Shih Margaret Todd Pittinsky and Nalini Ambady 1999 Stereotype Susceptibility Identity
Salience and Shifts in Quantitative Performance Psychological Science 10(1) 80-83
Shih Margaret Nalini Ambady Jennifer Richeson Kentaro Fujita and Heather Gray 2002
Stereo-type Performance Boosts The Impact of Self-Relevance and the Manner of
Stereotype Activation Journal of Personality and Social Psychology 83(3) 638-647
Srinivas M N 2009 The Oxford India Srinivas Oxford University Press Oxford
Steele Claude and Joshua Aronson 1995 Stereotype Threat and the Intellectual Test
Performance of African-Americans Journal of Personality and Social Psychology 69(5)
797-811
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
42
Swidler Ann 1986 Culture in Action Symbols and Strategies American Sociological Review
51(2) 273-286
Swidler Ann 2001 Talk of Love How Culture Matters University of Chicago Press Chicago
Tverksy Amos and Daniel Kahneman 1983 Extensional versus Intuitive Reasoning The
Conjunction Fallacy in Probabilistic Reasoning Psychological Review 90 293-315 1983
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
43
Appendix Analysis of the tournaments
Recall from Figure 1 that the experiment used two different financial incentives Some
treatments used piece rate incentives in both rounds (ldquoPP treatmentsrdquo) and others used piece
rate incentives in round 1 and tournament incentives in round 2 (ldquoPT treatmentsrdquo) Under the
tournament incentive a participant earned six rupees per maze solved if he solved the most
mazes in his session otherwise he earned nothing A winner could (and some did) earn 15 x 6
rupees equivalent to almost two daysrsquo unskilled adult wages In case of a tie both winners
received the prize
In piece rates payoffs depend only on own performance In contrast revealing the caste
identity of the other participants in a tournament might affect beliefs about the individualrsquos own
chances of winning the tournament In addition if a participant knows that he is competing
against members of his own community he might not wish to best them in the game Thus in
analyzing the results in this section there are many more possible influences on behavior and
we cannot disentangle them We first report all treatment effects and then discuss the possible
explanations
Results
When caste is not revealed the shift from the piece-rate to the tournament incentive
significantly increases performance In Table 2 columns (1)-(4) all coefficients on tournament
are significantly positive The table also shows that in all specifications the coefficient on TH
is insignificant This means that the response of H to tournament incentives is statistically
indistinguishable from that of L
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
44
Revealing caste whether in mixed or segregated groups eliminates the performance
boost from tournament incentives Figure A1 graphs the predicted output in round 2 under all
conditions (two incentive conditions X three identity manipulations) The left panel shows
results for H and the right panel shows results for L As indicated by the dotted lines when
caste is not revealed the tournament boosts output by 25 for H and 28 for L above
performance under piece rates (p lt 05) As shown by both the dark and light solid lines when
caste is made public there is no performance boost for H or L In fact in Revealed Segregated
the tournament incentive reduces L output The decline is 16 mazes (p-value lt 001) which
corresponds to a 38 decline from the predicted level under Piece-rate control These results are
robust to controls for classmdashsee Table 2 column 3
Figure A1 Predicted Output in Round 2 Piece Rate versus Tournament Incentives
Note Predicted output is based on Table 4 column (3) which controls for the participantrsquos grade in
school previous exposure to mazes and number of other participants known in the session Bars show
standard errors
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
45
Table A1 shows the effect of the identity manipulations holding constant the tournament
incentive The first three columns show the results for the full sample and columns 4-6 shows
the results for the subsample that excludes participants who solved zero mazes It is easy to see
that in the full sample the treatment effect of revealing caste is always negative The declines
are much steeper for L than for H For example compared to the control Revealed Segregated
decreases output by 34 for H and by 60 for L The treatment effects are only slightly
reduced in the subsample
Discussion
In the tournaments we cannot disentangle context effects on performance from strategic
responses to equilibrium play The finding that making caste public eliminates the positive
response to tournament incentives is consistent with an interpretation based on context effects
ie when caste is salient H may feel ldquoWhy tryrdquo and L may feel ldquoI canrsquot or donrsquot dare to
excelrdquo But there are three additional possible explanations which we discuss next
Strategic responses to equilibrium play As an example suppose we make the plausible
assumption that H believe that low-caste boys will choose not to compete because for an L to
win against an H would violate the traditional caste order Or suppose we assume that H believe
that low-caste boys cannot compete effectively since they are less able than high-caste boys
Under either assumption the H-segregated tournaments would have low expected returns to
effort compared to the control tournaments because the participants know they are competing
against five H boys
Social preferences The 30 high-caste boys in Revealed Segregated were 20 Thakurs 8
Brahmins and 2 Vaishyas The overall distribution in the treatment was mirrored in the
individual sessions Thus a high-caste boy in Revealed Segregated did not find himself among
boys of only his own specific caste Earlier experiments in the state of Uttar Pradesh reveal
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
46
solidarity only among men of the same specific high-status castes (Hoff et al 2011) not among
men of different specific high-status castes The experiments also reveal no solidarity among
low-caste men even when they are from the same caste Evidence from Fehr et al (2008)
supports the view that many high-caste men are averse to disadvantageous inequality In a
tournament someone must win and we have no evidence from any study of social preferences
that if someone must win a high-caste man would not wish to be the person that won
Work norms Bandiera et al (2005) find evidence from personnel data that workersrsquo
productivity is much lower under a relative incentive scheme than under piece rates In the
relative incentive scheme in their study workersrsquo daily pay depended on the ratio of individual
productivity to average productivity among all coworkers on the same field and day whereas
under piece rates individual pay depended only on individual productivity Thus the harder a
person worked the lower the pay of his coworkers The authors find that workers internalize the
externality of high productivity by working less only when they can monitor others and be
monitored by them Thus the finding reflects work norms enforced by sanctions not social
preferences
It seems to us unlikely that work norms explain the declines in performance in
tournaments when caste is revealed First the tournament incentive is a one-shot winner-take-
all incentive The person with the best performance wins and all others lose By working less a
participant does not increase the earnings of his coworkers as would occur under the scheme
analyzed in Bandiera et al unless he completely forgoes a payoff Further Bandiera et al find
that the decline occurs only when workers can monitor each other In our experiment payoffs
were not public each participant was given his payment such that others did not know the
amount at the time of giving and instructions prior to the game (ldquoThis [the earnings from the
piece rate incentive in round 1 and from tournament in round 2] will be given to every child in an
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
47
envelope after the games are overrdquo) tried to make this clear However it is possible we did not
make it clear enough We cannot rule out that participants feared social sanctions from others in
their group if they won the tournament
We comment finally on the question of work norms in Hardoi The high castes in the
villages shun positions in which they work for others this is considered a violation of their
honor High-caste individuals work for others only if they leave the village to work in the towns
in other districts there is essentially no industry in Hardoi district The issue of work norms in
firms thus does not arise for either high or low castes in Hardoi
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
48
Table A1 Effects of Making Identity Public in Mixed and Segregated Sessions under the Tournament Incentive
Output per round full sample
Output per round excluding
participants who solved zero
mazes
H L
Caste gap
significant
H L
Caste gap
significant
(1) (2) (3)
(4) (5) (6)
Estimated treatment effect of
Revealed Mixed -142 -217
-182 -154
(079) (077)
(075) (077)
Revealed Segregated -225 -397
-213 -352
(092) (075) (086) (075)
Mean outcome
in Caste Not Revealed
663 659
711 659
Effect of Revealed Mixed as
of mean outcome
in Caste Not Revealed
-21 -33
-26 -23
Effect of Revealed Segregated
as of mean outcome
in Caste Not Revealed -34 -60 -30 -53
Note Cluster-robust standard errors are in parentheses The mean output per round pools both rounds All values reported here can be
derived from the regressions in Table 2 effects in columns (1)-(3) from regression (2) and effects in columns (4)-(6) from regression (4)
However it is easier to obtain the treatment effects and their standard errors by running regressions in which the tournament is the
incentive in the baseline case plt001 plt005 plt01
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
49
Supporting Table--Not for Publication
Table S1 Test of Year and Month Effects on Output in the Piece Rate-Control Treatment
Variables
Round
Score
High 031
(041)
Dummy_March03 -017
(057)
Dummy_March05 016
(046)
Constant 428
(042)
R2 0004
N 312
Notes The table reports data for the 156 subjects under the control condition with piece rate incentives in
both rounds Dummy_March03 indicates that a session was held in March 2003 and dummy_March05
indicates that a session was held in March 2005 The omitted category are sessions held in January 2003
60 subjects participated in January 2003 36 subjects in March 2003 and 60 subjects in March 2005
Standard errors clustered at the level of the individual are in parentheses plt001 plt005 plt01
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
50
Supporting Table--Not for Publication
Table S2 Determinants of Failure to Learn How to Solve a Maze
Additional regressors
With individual
characteristics With individual and family characteristics
(1) (2)
High 228 218
(106) (113)
Identity condition 264 275
(102) (103)
Identity conditionH -389 -434
(120) (124)
Grade in school 002 007
(038) (038)
Previous exposure to mazes -005 -097
(057) (076)
Number of other participants known -020 -015
(021) (021)
Mothers education Є(06)
-004
(056)
Mothers education ge 6
004
(052)
Fathers education Є(06)
080
(077)
Fathers education ge 6
130
(065)
Mother employed outside home
-033
(105)
Father not a day laborer
-018
(056)
Constant -451 -524
(100) (118)
Chi square 1492 3057
N 538 464 Notes This table is based on a logit regression The dependent variable is the log odds of failure Failure=1 if a
participant has zero output in both rounds 0 otherwise Robust standard errors are in parentheses plt001
plt005 plt01
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options
51
HIGHLIGHTS
In a randomized experiment high- and low- caste boys solve mazes under incentives
When caste is not revealed low-caste boys solve mazes as well as high-caste boys
Revealing caste in mixed-caste groups creates a caste gap favoring the high caste
Segregationmdasha marker of high-caste dominancemdashreduces even high-caste performance
Productivity is sensitive to social setting which opens up new policy options