Modern Family: Female Breadwinners and the Intergenerational

Transmission of Gender Norms

⇤

Panos Mavrokonstantis

†

April 2016

Abstract

The literature on gender norms has focused on the relation between mothers’ labour supplyand children’s outcomes. The overall finding is that higher maternal labour supply is associatedwith more modern attitudes regarding gender roles among children, and higher labour supplyamong daughters. Instead, I look at the relation between having a breadwinner mother andchildren’s views, and uncover some striking results. Boys raised in modern families (i.e. wherethe mother is the breadwinner) are less likely to develop traditional norms. However, comparedto those in traditional families, girls raised in modern families are actually more likely to betraditional; in opposition to their family’s norm but in line with society’s. I develop a model ofgender identity and show that this unusual finding can be explained by girls’ weaker preferencefor conforming to the family. I provide evidence that girls in modern families are, indeed,less conformist to their families. Then, using quasi-experimental variation and a regressiondiscontinuity design, I demonstrate that this finding is caused by living in a norm-minorityfamily, which pushes girls to adopt more traditional views in order to align themselves moreclosely to the social norm.

Keywords: intergenerational transmission, gender norms, gender inequalityJEL Classification: D10, J16, Z13

⇤I am grateful to Sebastian Camarero-Garcia, Sarah Clifford, Frank Cowell, Anders Jensen, Alan Manning, ArthurSeibold, Viktorie Sevcenko, Johannes Spinnewijn, and seminar participants at the LSE, the 6th ECINEQ meetingin Luxembourg, and the 11th Winter School on Inequality and Social Welfare Theory for helpful suggestions anddiscussions. I acknowledge financial support from the Onassis Foundation.

†STICERD and Department of Economics, London School of Economics (e-mail: p.mavrokonstantis@lse.ac.uk).

1

1 Introduction

Significant attention has been given to the recent rise of female breadwinners in the UK (IPPR2015). The proportion of mothers earning more than 50% of the family income has doubled between1997-2014, increasing from 10% to 20%1. The increase in this modern family structure is challengingthe age-old traditional model of the male breadwinner and female caregiver. An unexplored questionis whether this recent phenomenon, currently driven by a minority of the population, can changepeople’s views about the appropriate roles of men and women in society.

In this paper, I examine the extent to which gender norms are passed on from one generationto the next when a family’s norms oppose those of the society in which the family is embedded.Do children growing up in modern families (i.e. where the mother is the breadwinner) adopt theirfamily’s values or those of society? Using data from England, I find unusual results for girls. Whileboys raised in modern families are less likely to develop traditional norms, girls raised in modernfamilies are actually more likely to do so, in opposition to their family’s norm but in line withsociety’s. Examining further outcomes associated with gender norms, I also find that girls raisedin modern families are less likely to state that being able to earn high wages is important for themand are less likely to want to study science at university. Employing a model of gender identityto examine the gender-socialisation process, I argue that my results can be explained by a weakerpreference among girls for conforming to the family. Using quasi-experimental variation, I then showthat this weaker preference is, in fact, caused by the treatment of living in a modern family, makinggirls react to their modern family’s norm-minority status and adopt more traditional views.

These findings have strong policy implications given the research showing that the gender at-titudes an individual develops by the end of the teenage years become significant determinants oflabour supply in adulthood (Johnston et al. 2014). Moreover, the prevailing gender norm of societyis a strong predictor of gender inequality in the labour market. Consider, for instance, Figure 1,which shows the relation between the gender pay gap and the prevalence of traditional gender normsacross OECD countries. There is a strong positive relation between the proportion of a country’spopulation that believes women should not work full-time, and the percentage difference in the me-dian wage between men and women. Similar conclusions can be drawn from Figures 2 and 3. Themore traditional a country is, the lower female labour supply is, both along the extensive (Figure2) and intensive margins (Figure 3). These stylised facts elucidate how important gender normsare to the analysis of gender inequality. Understanding their intergenerational transmission processcan therefore help identify policies aimed at promoting more egalitarian social norms. Such under-standing can function as an underutilised mechanism through which to reduce gender inequality, thebenefits of which go beyond ameliorating inequalities in the labour market. Promoting more equalgender norms can aid in removing barriers to human development (UNDP 2014) and mitigate theadverse consequences of gender inequality traps (particular gender gaps in education) on economicdevelopment (Dollar and Gatti 1999; Klasen 1999; Knowles et al. 2002; World Bank 2006). Themodernisation of gender norms and the associated increase in female labour-market participationmay also act as a key mechanism to fight poverty and promote rural development, measures thatthe Food and Agriculture Organisation (2009) has strongly advocated.

This study makes several contributions to the literature. First, it sheds light on an importantaspect of the gender inequality debate that is often overlooked in the inequality literature: the‘fundamentals’ of gender inequality. Economists usually attend to other factors. Some studies, forexample, analyse the evolution of gender inequality over time by focusing on the dynamics of femalelabour-force participation (Eckstein and Lifshitz 2011; Goldin 1995; Mammen and Paxson 2000;etc.). Others focus on differences in human capital (Mincer and Polachek 1974), on-the-job training

1Author’s calculations based on the Family Resources Survey.

2

(Blau and Kahn 2000), family-planning preferences (Goldin and Katz 2002), child-care subsidies(OECD 2004), and so forth. These studies’ proposed explanations are only proximate causes, whichare themselves outcomes of more fundamental factors, particularly society’s underlying culturalbeliefs about gender roles. For instance, Vella (1994) finds that between-sex differences in human-capital investment are driven by conservative attitudes towards female labour-force participation.Similarly, Fortin (2015) shows that changes in gender attitudes are the strongest predictor of changesin female labour supply. Understanding the role and propagation mechanism of these fundamentalsis therefore germane to the gender inequality debate.

Second and most important, this study is the first to formulate its conceptual and empiricalframework in a way that assesses socialisation and the intergenerational transmission of gendernorms explicitly, to precisely identify the type of family norm (modern versus traditional) underconsideration. The main approach in the literature is to examine the relation between mothers’ anddaughters’ labour supply, along both the extensive and intensive margins (Del Boca et al. 2000;Fernández et al. 2004; Morrill and Morrill 2013; Olivetti et al. 2016; etc.). The finding of a positivecorrelation between mothers’ and daughters’ labour supply has been interpreted as evidence thatmodern gender attitudes are successfully transmitted. I argue that this interpretation is problematic.Daughters of mothers who worked may well be working more, compared to daughters of motherswho did not work. However, the daughters of working mothers may still be working and earningless than their husbands. Hence, what the literature describes as the transmission of modern normsmay still very well propagate traditional, male-breadwinner family norms. In fact, no studies havelooked explicitly at the effect of living in a modern versus a traditional family on socialisation andthe development of gender norms. The only related study is by Bertrand et al. (2015), which doesnot have an intergenerational approach but examines only between-spouse outcomes. My studyattempts to fill this gap in the literature.

Lastly, this study not only looks at whether parental gender norms are transmitted to childrenbut also seeks to explain why or why not. In particular, it investigates the importance of underlyingchild preferences for conformity to both the family and society. This study is therefore the first tolook at such fundamental factors affecting gender norms, and contributes to our understanding ofhow and why children develop these norms. In addition, by investigating whether the transmission ofgender norms is heterogeneous between boys and girls, this study goes beyond the existing literature,uncovers striking differences, and tries to explain them.

The rest of the paper proceeds as follows. Section 2 reviews the literature. Section 3 describesthe theoretical framework, presenting a model of gender-identity formation, and Section 4 discussesthe main empirical approach. Section 5 describes the data and presents evidence on the social normin England. Section 6 discusses the results, while Section 7 presents an alternative identificationstrategy (a regression discontinuity design), to examine the robustness of the main findings. Section8 concludes.

3

Figure 1: Gender Norms and the Gender Gap

AT

AU

CH

CZ

DE

DK

FI

FR

GB

HU

IL

JP

NL

NO

SE

SK

US

b = 0.409** (0.188)

10

15

20

25

30

35

Male

−F

em

ale

Diff

ere

nce

in W

ages

(%)

40 50 60 70% Against Women Working Full−time

Notes: This figure shows, for various OECD countries, the relationship between the proportion stating thatwomen should not work full-time (horizontal axis) and the gender gap (vertical axis). The gap is measuredas the % difference between the median wage of males and females. The figure is based on data from the2002 International Social Survey Programme and OECD statistics.

Figure 2: Gender Norms and Female Labour Supply: Extensive Margin

AT

AU

BE

BR

CH

CL

CZ

DE

DK

ES

FI

FR

GB

HU

IEIL JP

MX

NL

NO

NZ

PL

PT RU

SE

SI

SK

US

b = −0.305** (0.126)

35

40

45

50

55

60

Fem

ale

Labour

Forc

e P

art

icip

atio

n (

%)

40 50 60 70 80% Against Women Working Full−time

Notes: This figure shows, for various OECD countries, the relationship between the proportion stating thatwomen should not work full-time (horizontal axis) and female labour-force participation (vertical axis). Thefigure is based on data from the 2002 International Social Survey Programme and OECD statistics.

4

Figure 3: Gender Norms and Female Labour Supply: Intensive Margin

AT

AU

BE

DE

DKES

FI

FR

GB

HU

JP

MX

NL

NO

NZ

PL

PT

SE

SIUS

b = −0.812*** (0.195)

40

50

60

70

80

Fem

ale

Rela

tive L

abour

Supply

(%

)

40 50 60 70 80% Against Women Working Full−time

Notes: This figure shows, for various OECD countries, the relationship between the proportion stating thatwomen should not work full-time (horizontal axis) and women’s hours worked (per day) as a proportion ofmen’s (vertical axis). The figure is based on data from the 2002 International Social Survey Programme andOECD statistics.

5

2 Literature Review

A vast literature in sociology and social psychology exists on theories of socialisation and socialidentity (for example Epstein and Ward 2011; Lorber 1994; Lytton and Romney 1991; Tajfel 1978).While different versions of the socialisation process (or its stages) have been proposed, the consensusis that socialisation, the procedure through which individuals learn through human interaction howto be functioning members of society, is the main way in which cultural norms are developed, adoptedand transmitted.

Although studies on the role of socialisation in the intergenerational transmission of culturalvalues exist in the literature of other social sciences, economists have only recently begun to explorethe topic. The seminal work bridging this gap between economics and other disciplines is that ofAkerlof and Kranton (2000; 2002; 2010), who translated theories of social identity into an economicsframework, giving birth to what is now known as Identity Economics. Various approaches to mod-elling identity have ensued2. Among others, Bénabou and Tirole (2007) propose a model whereindividuals hold a range of individual beliefs that they both value and can invest in. Klor and Shayo(2010) model identity as status, while Bisin et al. (2011) introduce the concept of oppositionalidentities. Although the models differ, the common element in this literature is the introductionof identity considerations into a neoclassic framework in which a person’s self-image is valued andbecomes a crucial element of her utility function.

A burgeoning empirical literature has developed that examines the effect of culture and itstransmission. However, little work exists on the intergenerational transmission of explicit gendernorms. Most studies that do consider gender norms focus rather on the effect of norms on someother outcomes (predominantly labour supply), without examining how gender norms are formed inthe first place. For instance, Bertrand et al. (2015) examine the effect of traditional gender norms(defined as aversion to the wife earning more than the husband) on marriage and labour-marketoutcomes. While they argue that this aversion is induced by gender-identity norms, they do notassess how these norms are formed or transmitted intergenerationally, but simply take them as given.They show that women are less likely to work if their potential income exceeds their husband’s; ifthey do work, women are more likely to earn less than their potential income. Moreover, families inwhich the wife earns more face a higher likelihood of divorce and lower marriage satisfaction.

Using a different approach, Alesina and Giuliano (2010) assess family culture by studying theimportance of family ties for economic behaviour. They find that stronger family ties are associatedwith lower female labour-force participation. Their explanation is that strong family ties requirean adult family member to stay home and ‘manage’ the family institution, and this burden fallson women, who are consequently excluded from the formal labour market. In a related paper onculture, Alesina et al. (2013) look at the historical origins of gender norms. They provide evidencefor intergenerational cultural persistence, showing that attitudes towards women are more traditionalamong descendants of societies that practiced plough agriculture. Plough agriculture, in contrast toshifting cultivation, was much more capital intensive and therefore required brawn-intensive labour,favouring men and leading to gender-based division of labour. As a result, the authors find historicalplough use to be negatively related to modern-day attitudes towards gender inequality, female labour-force participation, and female participation in politics and firm ownership. In a somewhat differentsetting, Fernández and Fogli (2009) examine how ancestral culture is related to fertility and labour-market outcomes of second-generation American women. They find that women work more (havemore children) in cases in which their country of ancestry had higher historic female labour-forceparticipation (total fertility rate).

Another strand of the literature looks at the relation between the labour supply of individuals2For a comprehensive survey of identity models, see Costa-Font and Cowell (2015).

6

and of their parents. The overall findings indicate a positive association, implying that gendernorms are transmitted from one generation to the next. For instance, Fernández et al. (2004) findthat women are more likely to work if their mother-in-law also worked, while Morrill and Morrill(2013) show that there is a positive relationship between mothers’ and daughters’ labour supply. Inline with these findings, Del Boca et al. (2000) also find that women’s labour supply is related tothat of both their mothers and mothers-in-law, while Olivetti et al. (2016) find a positive relationbetween daughters’ number of hours worked and the hours worked by both their own mothers andthe mothers of their childhood peers.

Other studies move away from labour-market outcomes and analyse gender norms by exploringmarital satisfaction among males. Butikofer (2013) finds marital satisfaction to be lower when aman’s wife works and contributes to household income, but only for men raised in a traditionalfamily where the mother did not work. Similar findings are reported by Bonke (2008) and Bonkeand Browning (2009). This result provides evidence that gender socialisation at the family level,and at a young age, is crucial in determining lifelong gender norms.

The relation between gender norms and educational outcomes has also been studied recently.For instance, Gonzales de San Roman and de la Rica Goiricelaya (2012) examine the cross-countrygender gap in test scores revealed by PISA data. They find gender norms to be an importantdeterminant; in particular, mothers’ labour-force participation is positively related to daughters’test performance. In a different setting, Blunch and Das (2014) show that increased access toeducation for girls explains much of the rise in egalitarian views towards female access to educationin Bangladesh. Similarly, employing data from 157 countries, Cooray and Potrafke (2010) findconservative culture and religion to be the primary obstacle towards gender equality in education.

Last but not least, some papers investigate the relationship between parental and child genderattitudes. Both Farre and Vella (2013) and Johnston et al. (2014) find strong correlations in theseattitudes in cases in the USA and UK, respectively. In line with previous findings, these two studiesalso find a positive association between boys’ attitudes during childhood and their future wives’labour supply in adulthood.

While these studies shed light on the relation between culture or education and labour-marketoutcomes, they all share one significant limitation: they do not fully capture the essence of thegender norm under consideration (for the reasons explained in the Introduction). The literaturelacks an explanation of how these beliefs are developed and propagated; by focusing on the gender-socialisation process, this paper seeks to fill that gap.

3 The Theoretical Framework

This section describes the process through which children develop their beliefs. Socialisation is firstexplained, followed by a simple model that formalises this process through a utility-maximisingframework. Some theoretical predictions follow, which will be useful in interpreting the empiricalresults in subsequent sections.

3.1 The socialisation process

Children develop gender norms through socialisation (Epstein and Ward 2011; etc.). Social psy-chology identifies two main sources of socialisation: the family (vertical socialisation) and societyat large (horizontal socialisation). Children acquire norms by interacting with and observing theparticular behaviours of these two social institutions.

The family is, of course, the first point of contact with the outside world. Therefore, parentsplay a key role in socialisation. Parents are assumed to be altruistic towards their children, and have

7

preferences regarding the norms their children develop. In particular, they aim to socialise theirchildren to their own values. Parents choose how much effort to exert to increase the probabilitythat vertical socialisation is successful. Effort in this setting can manifest in two main ways. Thefirst is that parents can express their beliefs through direct discussion with their children, whichrequires investment in ‘family time’. Effort can also take the form of parents’ actions, particularlythe role adopted by each parent. Children thereby receive signals about appropriate gender roles byobserving their parents’ household responsibilities, in particular who the breadwinner is and who isresponsible for household production and child care.

As children grow older and start to interact with a social circle beyond their families, they arealso exposed to what society at large deems to be the appropriate role of women. The primarysources of horizontal socialisation are the child’s school, peer group, and exposure to social normsthrough mass media.

Each child has some preference for conforming to the family and society. This preference willdepend not only on individual characteristics, but also on the probability that family socialisationwas successful. This probability in turn, will depend on family and social norms; families followingthe social norm will have a higher chance of transmitting their values because they will not have toovercome an opposing social norm.

Having learnt what the family and society believe to be the appropriate gender roles, the childchooses her own gender values. In doing so, the child takes into account the cost of deviating fromthese norms. This can be thought of as a psychological cost of interacting with others who do notshare the same beliefs, and arises from self-image concerns, that is, concerns about how personalbeliefs will be judged by others. The stronger the preference for conformity is, the higher is the costof deviating from the prescribed norms.

3.2 Formalising the socialisation process

To formalise this process, I build on the Georgiadis and Manning (2013) model of national identity.In my setting, each child has the following utility function3:

U = �1

2

⇥c

F

(xF

, x

S

, Z)(x� x

F

)2 + c

S

(xF

, x

S

, Z)(x� x

S

)2⇤

(1)

where:x 2 [0, 1] represents the child’s choice; larger values indicate more traditional beliefs

x

F

2 [0, 1] is the child’s family norm; larger values indicate more traditional beliefsx

S

2 [0, 1] is society’s norm; larger values indicate more traditional beliefsc

F

2 (0, 1] is how strongly the child wants to conform to the familyc

S

2 (0, 1] is how strongly the child wants to conform to the societyZ is a vector of family and child characteristics affecting preferences for conformity

The intuition of the model is simple. The larger the distance between the child’s belief (x)and that of the family’s (x

F

) and society’s (xS

), the larger the psychological cost is. This cost isincreasing in the preference for conformity to the family (c

F

) and society (cS

). The stronger thispreference is, the more costly it is to deviate from a norm to which the child wants to conform.Taking this into account, the child chooses her optimal belief according to the following rule:

x

⇤ = argmax

x

U = x(cF

, c

S

, x

F,

x

S

, Z) =c

F

(xF

, x

S

, Z)xF

+ c

S

(xF

, x

S

, Z)xS

c

F

(xF

, x

S

, Z) + c

S

(xF

, x

S

, Z)(2)

3Subscript i for each individual has been dropped to avoid cluttering.

8

Hence, the child chooses how traditional her gender view will be by weighting the family andsocial norms by her preferences for conforming to each.

3.3 Comparative Statics

I now use the model to predict how a change in conformity preferences will affect x⇤, as this will beuseful for interpreting my empirical results. The comparative statics are as follows:

@x

⇤

@c

F

=c

S

(xF

, x

S

, Z)(xF

� x

S

)

(cF

(xF

, x

S

, Z) + c

S

(xF

, x

S

, Z))2(3)

(3) shows the response of x⇤ to a stronger preference for conformity to the family4. The resultis ambiguous as it depends on which institution is more traditional, society or the family. As I willshow in Section 5.1, the social norm in England regarding the role of mothers with young childrenis very traditional, while in the empirical framework (Section 4), the family norm I will examinewill represent very modern views. Thus, without loss of generality, these norms can be modelled bysetting x

F

= 0 and x

S

= 1. Given these restrictions, the prediction is that a stronger preference forconforming to the family’s norm leads to a decrease in x

⇤, that is to a less traditional norm. Theintuition is that the more strongly a child wants to conform to her family’s relatively more modernnorm, the more she has to differentiate herself from the traditional social norm and adopt a moremodern belief. Otherwise, the psychological cost of deviating from the more modern norm, to whichshe now wants to conform more strongly, will increase. I will revisit this prediction in Sections 6and 7 where I try to rationalise my finding that girls growing up in modern families develop moretraditional views5.

4 The Empirical Framework

My aim is to estimate x

⇤ = x(xF

, x

S

, Z). The Next Steps survey provides data on children’sgender norms through responses to the following question: “Women should never work full-timewhen they have young children. Do you agree with this statement? ” Stating that a woman shouldnever work full-time if she has young children reflects a very traditional view of gender roles; withinthe language of the model, this would mean x

⇤ = 1. Based on this, I define the outcome variablecapturing children’s norms as follows:

Traditional Norm =

(1 if Agree (x⇤ = 1)

0 if Disagree (x⇤ < 1)

As the aim is to examine the intergenerational transmission of gender norms, I have to assesshow successful vertical and horizontal gender socialisation are. Since the children in the survey areall from England, they are all exposed to the same social norm6. The variation of interest willtherefore come from differences in family norms. Because I am particularly interested in examiningthe development of children’s norms when the family and social norm are oppositional, I define thefamily norm in a way that translates into the model as x

F

= 0. I do so using the following definition:4An analogous result holds for the case of preferences for conformity to society.5In appendix A, I also discuss the prediction @x

⇤

@xFand show how it matches my empirical results.

6The social norm may of course still vary by region. To account for this, regional fixed effects are included in allempirical specifications.

9

ModernFamily =

(1 if Mother Earns More0 otherwise

I thus want to discover how children’s gender norms develop when living in a modern family buta traditional society (evidence on the English society’s gender norm is shown in Section 5.1). To doso, I can estimate the following baseline model:

Pr(Traditional Norm)i

= ↵0 + ↵1ModernFamily

i

+ "

i

(4)

However, as I am particularly interested in testing for between-sex heterogeneity in gender so-cialisation, (while also accounting for a range of possible confounding factors), I will focus on thefollowing extension of the baseline case:

Pr(Traditional Norm)i

= �0 + �1ModernFamilyi

+ �2Femalei

+ �3ModernFamilyi

⇥ Femalei

+ Z0i

⇣ + ui

(5)

This will be estimated using linear probability, probit and logit models to ensure that the findingsare robust to alternative specifications of this binary reponse model7. The identifying assumptionhere is selection on observables, that is E(u

i

|ModernFamily

i

, F emale

i

, Z

i

) = 0. While this is nota weak assumption, it is not implausible given the vast array of characteristics that Z

i

will include.Nevertheless, Section 7 will present another identification strategy - a regression discontinuity design- to provide stronger evidence that the findings have a causal interpretation.

7To avoid excessively long appendices, I have not included the probit and logit estimates in this manuscript.However, every equation shown in the main text that includes a binary dependent variable has been re-estimatedusing a probit and logit specification, and all estimates are robust. The probit and logit estimates are available uponrequest.

10

5 Data

5.1 The Social Norm

What is the social norm to which the children in my sample are exposed? I answer this using infor-mation from the International Social Survey Programme (ISSP), which collects data internationallyon a broad range of social issues. Questions related to gender norms are included in the 1988, 1994,2002, and 2012 surveys. Responses from the 2002 survey were chosen, since this is the period thatcoincides with the Next Steps survey, with children in the sample aged between 15 and 16 at thetime. The ISSP 2002 contains data from a representative sample of 1,960 observations from the UK.Because I study views on the appropriate roles of men and women in childrearing and providingincome, I focus on heterosexual couples with children and analyse the labour supply of both thefather and the mother. In particular, I look at their responses to the following four questions:

“Did you work full-time, part-time, or not at all when. . . ”

1. “. . . you had no children”

2. “. . . you had a child under school age”

3. “. . . your youngest child was still in school”

4. “. . . your children left the home”

These four cases track the parents’ labour-market activity from the time before the child is born(case 1), up to adulthood (case 4). Cases 1 and 4 focus on parents with no dependent children. Case2 concerns parents with children under five years old, and case 3 focuses on parents of school-agechildren. Figure 4 below shows the responses to each question for both the fathers and mothers.Figure 4a reflects the first case, showing that there is no prescription against women working full-timewhen they have no children. When women have no child-care responsibilities, over 80% of womenwork full-time, and only a small minority stay at home (less than 9%). Compared to the fathers’labour supply, the changes in the mothers’ labour supply when they have children are striking.Figure 4b reveals that when women have children below the age of five, the majority (53%) donot work at all, while the percentange of those working full-time drops from 82% to a mere 14%.In the same circumstance, the proportion of fathers working full-time remains steadily above 90%.These figures reveal how traditional the social norm in the UK is. There is a strong prescriptionthat the role of mothers with young children is not at the workplace. It is the role of the father towork full-time and provide for the family, and the role of the mother to reduce her labour supplyin order to provide child care. This prescription seems to hold for all women who have dependentchildren. Figure 4c shows that the majority of women (nearly 80%) still do not work full-time, eventhough their children are older but still of school age. Working full-time only becomes the modeagain when the children have grown up and left the home (Figure 4d), although the proportion ofwomen in this category is still low (46%) compared to the first case. It is interesting to note howstable men’s labour supply is. While the age of a child strongly predicts the mother’s labour supply,fathers always work full-time regardless of the presence and age of children. Overall, these figuresreveal that the children in my sample are growing up in a society that follows very traditional gendernorms.

11

Figure 4: The Social Norm

(a) No children

98.08

1.31 0.61

020

40

60

80

100

Full−Time Part−Time Home

Men

81.91

9.44 8.65

020

40

60

80

100

Full−Time Part−Time Home

Women

(b) Have child under school age

94.53

2.84 2.63

020

40

60

80

100

Full−Time Part−Time Home

Men

14.48

32.79

52.73

020

40

60

80

100

Full−Time Part−Time Home

Women

(c) Youngest child still in school

95.19

2.86 1.95

020

40

60

80

100

Full−Time Part−Time Home

Men

21.30

52.86

25.84

020

40

60

80

100

Full−Time Part−Time Home

Women

(d) Children left the home

94.52

2.40 3.08

020

40

60

80

100

Full−Time Part−Time Home

Men

46.18

38.89

14.930

20

40

60

80

100

Full−Time Part−Time Home

Women

Notes: This figure shows the labour supply of men and women in heterosexual couples, using data from the2002 International Social Surveys Programme. It plots the proportion of men and women working full-time,part-time or not at all in each of the following cases: (a) when they have no children, (b) when they have achild under school age (i.e. below 5), (c) when their youngest child is above 5 but still within schooling age,and (d) when their children have grown up and left the home.

12

5.2 The Next Steps Survey

I will estimate (5) using data from the Next Steps survey. Next Steps follows a nationally repre-sentative cohort of 15,500 individuals born in England in 1989-90. The first wave of data collectiontook place in 2004, when cohort members were attending year nine in school and were aged 13-14.The survey consists of a total of seven waves, with data collected annually between 2004-2010. Dueto the nature of my question, I focus on dual- and heterosexual-parent families (in which parentsare married or cohabiting). This criterion drop approximately 30% of the sample (the vast majoritybeing single-parent families). As is the case with all surveys, Next Steps suffers from unit and itemnon-response. As a consequence, the estimation sample is restricted to the 1,640 cohort-membersfor whom there is no missing information on the variables of interest.

5.3 Choice of Variables

5.3.1 Main Variables of Interest

The outcome variable of interest is ‘Traditional Norm.’ It is a binary variable derived from responsesto the question “Women should never work full-time when they have young children. Do you agreewith this statement?” as explained in Section 4. Responses to this question are taken from the2010 wave. The timing is ideal as it is asked at an age by which children’s socialisation has beencompleted; I thus avoid ascribing norms to possibly ‘transitory’ beliefs. The timing is also supportedby UK evidence showing that beliefs at this age strongly predict individuals’ future labour supply(Johnston et al. 2014). The main independent variable will be a dummy named ‘Modern Family,’ asdefined in Section 4. In the sample, 13.2% of the children live in modern families (i.e., the mothersare the breadwinners), while 32.7% of them hold traditional gender norms. The ‘Modern Family’status is derived from data on gross earnings of the father and mother of each child for the year2004. While data on incomes exists for years 2004-2007, the 2004 wave was chosen to minimise theamount of dropped observations due to item non-response.

5.3.2 Control Variables

A vast range of control variables is included in the regressions. The aim is to control for all variablesthat are potentially correlated with both the child expressing traditional norms and living in amodern family. A detailed list of all variables used in the regression analysis, with the associateddescriptions and summary statistics, is presented in Table B.1 of appendix B. The variables canbe grouped into four categories: family, child and geographical characteristics, as well as proxiesfor parental-socialisation effort. In what follows, I explain which variables I control for and why Iconsider it important to do so.

The family characteristics of immediate concern are cultural and socio-economic attributes. Akey variable controlled for is parental religion. As religion(s) offer(s) various prescriptions on theappropriate behaviour of individuals in society, including gender roles, parents’ religious views canaffect not only how much mothers work and earn, but also the way parents socialise their children(Guiso et al. 2003; Lehrer 1995). Moreover, this effect may depend on how important religion is tothe parents’ way of life; religious parents are more likely than non-practising individuals to conformto the gender roles prescribed by their religion (Guiso et al. 2003; Heineck 2004). I therefore alsocontrol for the strength of parental religiosity.

To account for family structure, I control for whether the child lives with her natural parentsand whether they are married. Evidence (Parker and Wang 2013) shows that marital status affectsparents’ beliefs about the ideal relative labour supply, which may determine who the breadwinneris. Growing up in a ‘non-conventional’ family of unmarried parents may, however, also affect how

13

children view social norms, thereby affecting their preference for conformity to the norms. Forinstance, living in a ‘non-conventional’ family can affect the extent to which children consider socialnorms (which the parents are not following) desirable. Indeed, a positive relation between living in‘non-conventional’ families and adopting ‘non-conventional’ norms is found by Hognas and Carlson(2012), who show that children raised by unmarried parents have a higher likelihood of non-maritalchildbearing in adulthood.

To further account for family structure, I also control for household size and the number ofdependent children. Larger households and more dependents imply a higher financial burden and,therefore, a greater likelihood that both parents are working. Nonetheless, larger households (and,hence, more siblings, chores, and so forth.) to take care of may increase the need for the child’scontribution to household tasks, thereby affecting the child’s belief about her gender’s role withinthe family.

Among parent-specific characteristics, parental age is a crucial factor that needs to be accountedfor. As age increases, so does the probability that a parent is working. Moreover, due to the stylisedfact of a concave earnings profile over the life cycle (Polachek 2008), earnings also increase with agefor working adults. I therefore control for parental age because it directly affects the relative familylabour supply and income.

I also control for parental education, social class and total family income. Education not onlyincreases earnings through higher human capital (Becker 1964) but can also directly affect socialattitudes; higher education is associated with more liberal social views (Dee 2004; Kanazawa 2010;Stankov 2009). The same holds for social class. Although higher class status has historically been as-sociated with more conservative beliefs, social class in modern times (as measured by the Goldthorpescale using occupation types) captures job prestige and is highly correlated with education. For in-stance, Guiso et al. (2003) find that higher income (which is associated with social class, as thisis defined by occupation) leads to more liberal views on gender equality. Thus, I expect income tohave similar effects as education.

To better capture between-family income heterogeneity, I include two additional variables: theIncome Deprivation Affecting Children Index (IDACI) and the Index of Multiple Deprivation (IMD).Both indices are defined at the Super Output Area (SOA) level. SOAs divide England into 32,482local geographical districts - each the size of a neighbourhood - and have been devised to improve thereporting of small-area statistics by the UK Office of National Statistics. The IDACI measures eachSOA’s proportion of children under the age of 16 living in income-deprived households8; a higherIDACI indicates higher deprivation. The IMD is a more composite measure. Besides income, it alsoaccounts for deprivation in the following dimensions: employment, health and disability, educationskills and training, barriers to housing and services, crime, and living environment9. The mostdeprived SOA is given an IMD (rank) of 1, and the least deprived a rank of 32,482. In contrast tothe IDACI, a higher IMD thus indicates a less deprived area.

Geographic characteristics may also confound my results and are therefore controlled for. Theregion and type of area (urban versus rural) may affect the employment opportunities and thusrelative earnings of mothers. In addition, areas may vary in how traditional or modern the prevailinglocal norms are. For instance, larger and more urbanised areas are usually associated with moresecular and liberal views (Maneschiold and Haraldsson 2007).

Besides family and geographic characteristics, I also account for between-child heterogeneity.Research shows that ethnicity has an important effect on attitudes towards gender roles (Kane

8Income-deprived households are those in receipt of income support, income-based jobseeker’s allowance or pensioncredit, or those in receipt of child tax credit with an equivalised income below 60% of the national median beforehousing costs.

9For a more detailed explanation of each index, see Neighbourhood Renewal Unit (2004).

14

2000). Compared to non-whites, whites are more critical of maternal employment and considerit more harmful for young children (Dugger 1988). Ethnicity is therefore controlled for, not onlybecause it can affect the child’s beliefs about gender roles, but because it may also determine who thefamily breadwinner is through its effect on parents’ norms (since children have the same ethnicityas their parents). Beyond ethnicity, I also control for whether the child was born in the UK, thechild’s religion and the importance of religion to her life.

Furthermore, child characteristics related to cognitive abilities may have a confounding effect.Research relates higher intelligence to more socially liberal views (Deary et al. 2008); evidence alsoshows that low birth weight is associated with development delays and thereby lower intellectualcapacity (Ramey et al. 1999). Since the data do not contain objective measures of innate ability,the best way to control for this possible confounder is to indirectly account for it by controlling forbirth weight. Further proxies are also available in the data, such as whether the child has specialeducation needs and disabilities. All of these variables are included to control for cognitive abilitiesas accurately as possible.

Last, as I am interested in examining how successfully parents socialise their children, I accountfor an important determinant of success: parental effort in socialisation. Since no perfect measureexists, I use proxies capturing the quality and quantity of contact time between parents and children,which has been emphasised in the literature as an important factor in gender stereotyping (Asgariet al. 2010; Bem 1985; Carrell et al. 2010; Cooksey and Fondell 1996). For a detailed description ofthese proxies, see Table B.1 in appendix B.

6 Results

6.1 Main Results

Table 1 presents the main results. The first column shows the unconditional difference in the meanprobability of expressing traditional norms between children of modern and traditional families. Ifind evidence of intergenerational persistence of gender norms, with children in modern families being3.2% points less likely than children in traditional families to express traditional norms. Column (2)shows that this persistence holds regardless of the child’s sex. It also reveals that females overall are10.9% points less likely than boys to express traditional views, with this difference being statisticallysignificant at the 1% level.

To investigate between-sex heterogeneity, column (3) introduces an interaction term between thetype of family and the child’s sex. The coefficient of this variable is not only highly statisticallysignificant, but, importantly, it is positive, which has interesting implications for between-groupcomparisons. First, this result tells us that there is heterogeneity in the probability of developingtraditional norms between girls and boys living in modern families. Girls are 6.1% points morelikely than boys to state that women should not work full-time. In other words, vertical socialisationthrough the family is more successful for boys than girls. More intriguing however, is the comparisonbetween girls living in modern families and girls living in traditional families. Girls in modernfamilies are actually 1.9% points more likely to express traditional views, compared to girls intraditional families. Hence, girls react to the family norm when it is in opposition to the social norm,and are more likely to adopt the latter instead, rendering vertical socialisation totally unsuccessful.

The immediate concern is that these results may be driven by omitted variable bias. A wide rangeof controls is next introduced, with results shown in column (4). The full estimates of all covariates’coefficients are shown in Table B.2 of appendix B. Despite the vast number of control variables,the previous findings cannot be explained away. If anything, the between-group comparisons arenow even larger and more statistically significant. Conditional on the covariates, the results show

15

that girls living in modern families are 6.5% points more likely than boys living in modern families,and 4.4% points more likely than girls living in traditional families, to develop traditional views.A notable change is in the comparison between boys living in modern families and boys living intraditional families. In contrast to the results for girls, living in modern families is associated withmore modern gender norms among boys. The introduction of the covariates, however, reduces thebetween-group likelihood difference from 4.2% points (column 3) to 2.1% points (column 4).

Which background characteristics are associated with developing more traditional views? Start-ing with family structure and parental characteristics, I find households with more dependent chil-dren to be associated with children expressing more traditional beliefs, although the marital statusof parents appears to be insignificant. While the age of each parent is also statistically significant,its economic significance is too negligible to have any meaningful interpretation, due to the tinycoefficient sizes.

An important characteristic that proxies for culture is religion, which I also find to be a highlystatistically significant (at the 1% level) determinant of children’s beliefs. Moreover, the likelihoodof expressing traditional norms is higher for children whose parents state that religion is importantto their way of life. This result is in line with my expectations and previous findings (Guiso et al.2003), as prescriptions on appropriate behaviours and social roles characterise all religions.

The results showing a relation between parental socio-economic characteristics and norms alsoconfirm previous research (Dee 2004; Kanazawa 2010; Stankov 2009). Children of parents withlower education, lower socio-economic class, and lower income are more likely to develop traditionalnorms. Living in more deprived areas, however, has the opposite effect. Based on the IDACI andIMD data, higher local deprivation is related with a lower likelihood of holding traditional views10.This finding, therefore, highlights that parental education, socio-economic class, and income do notrelate one-to-one to the poverty level of the area in which a family lives. This may be becauseparental characteristics have a vertical socialisation effect, but geographical characteristics do not,as they are not family-specific. Moreover, other families’ education and social class may not beeasily observable by children, whereas local poverty and deprivation are. Children may therefore beusing their own neighbourhood deprivation level as a reference point from which to expand theiraspirations window.

With regard to child attributes, country of birth and ethnicity seem to matter. The likelihoodof developing traditional norms is lower for UK-born children, while it is the highest for whites,consistent with earlier research findings (Dugger 1988). The only exception is the black Africancategory, while the most conservative ethnic group appears to be the Bangladeshi. Consistent withthe effect of parental religiosity, children who state that religion is important to their way of life aremore likely to develop traditional norms. Furthermore, higher birth weight is associated with moremodern views, giving support to the findings of Deary et al. (2009) and Ramey et al. (1999) on thepositive relation among birth weight, cognitive ability, and socially liberal views. This finding is alsosupported by the positive relation between having special education needs and expressing traditionalviews. However, this result may also be driven by the fact that the majority of SEN teachers andcaregivers are female (Department for Education 2014), which may affect an SEN child’s view onappropriate gender roles.

Last, I consider the proxies for parental-socialisation effort. Conditional on all the other covari-ates, parental effort does not seem to affect the likelihood of expressing traditional views.

10Note that the opposite signs of these variables’ coefficients are not contradictory: a higher IDACI rank indicatesa poorer area, while a higher IMD rank indicates a richer area.

16

6.2 Robustness checks

In this section I show that the results are robust to other possible sources of confoundedness. Table2 presents the results11. I first address a possible concern regarding the data used to define a modernfamily. As explained previously, these data are taken from just one wave (the first), and therefore,may be affected by transitory income, hence not reflecting the true earnings trajectory of the family.To account for this, the specification in column (1) of Table 2 introduces controls for whether theemployment status of the mother and father has remained constant over all the years for whichemployment data exist. This spans five years before the survey and up to wave 4. The results arevery stable, showing that the previous estimates were not affected by this potentially confoundingfactor. Albeit very small, the changes are a 0.1% point decrease in the coefficient of the modernfamily dummy and a 0.1% point increase in that of the interaction term. This result implies a0.1% point increase in the difference in the likelihood of expressing traditional norms between boysin modern and boys in traditional families, and between girls in modern and girls in traditionalfamilies. These changes are intuitive; the more permanent the division of income in the family is,the more permanent the family type (modern versus traditional) is and the stronger is its effect onthe norm development of children through vertical socialisation.

Next, what if children’s gender norms are actually affected by the children’s observation ofparental differences in either education or job status, which can be correlated with who the bread-winner is, driving the results I am finding? To rule this out, I control for both of these factors. Ascolumns (2) and (3) show, neither the relative education nor the job status of the mother matter.The mother’s role as breadwinner is still what drives the results.

As a further check, I re-estimate the regression after excluding the (154) cases in my sample inwhich both parents earn zero income. These would be classified as families in which the motheris not the breadwinner; the problem is that in these families, there is neither a male nor a femalebreadwinner. This makes gender socialisation in such families opaque as there is no breadwinnerfor the child to observe. Excluding these cases is my preferred specification because it makes theattempt to identify the effect of having a female breadwinner more robust. Column (4) shows thatthe estimated between-group differences are even larger. This result is intuitive since these are thecases in which socialisation takes a clear direction.

Finally, what if the results are driven by the fact that in some families, one of the parents doesnot work? This would imply that what matters for socialisation may not be that the mother earnsmore than the father (and vice versa), but that (s)he is simply the only parent working. To rule thisout, I run the same regression on the sub-sample of children whose parents both work. As column(5) shows, results are again not affected. If anything, the results are even stronger, (in terms of sizeand statistical significance) when only dual-earner families are considered.

Finally, I check that my results are not affected by the imposition of a linear probability model. Ire-estimate all regressions using a probit and logit specification and find nearly identical coefficientsacross all three approaches12.

11For the results including all the coefficient estimates, see Table B.3 in appendix B.12The probit and logit results are available upon request.

17

Table 1: Main Results (LPM). Dependent var: Traditional Norm

(1) (2) (3) (4)Modern Family -0.032* -0.035* -0.042*** -0.021***

(0.017) (0.016) (0.011) (0.005)Female -0.109*** -0.138*** -0.126***

(0.024) (0.013) (0.010)Modern Family⇥Female 0.061** 0.065***

(0.023) (0.014)Constant 0.349*** 0.406*** 0.400*** 0.472*

(0.003) (0.012) (0.005) (0.229)Family Characteristics ⇥ ⇥ ⇥ XChild Characteristics ⇥ ⇥ ⇥ XParental-Socialisation Effort ⇥ ⇥ ⇥ XRegion Fixed Effects ⇥ ⇥ ⇥ XArea Type Fixed Effects ⇥ ⇥ ⇥ XObservations 1,640 1,640 1,640 1,640R

2 0.000 0.014 0.020 0.116Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

18

Table 2: Robustness Checks (LPM). Dependent var: Traditional Norm

(1) (2) (3) (4) (5)Modern Family -0.021*** -0.020*** -0.020*** -0.025*** -0.086***

(0.005) (0.005) (0.005) (0.007) (0.024)Female -0.126*** -0.126*** -0.126*** -0.141*** -0.135***

(0.010) (0.010) (0.010) (0.009) (0.016)Modern Family⇥Female 0.066*** 0.065*** 0.066*** 0.082*** 0.176***

(0.016) (0.015) (0.016) (0.020) (0.042)Father Employment: Stable 0.005 0.005 0.005 0.014 0.016

(0.019) (0.019) (0.019) (0.023) (0.023)Mother Employment: Stable 0.012 0.012 0.011 0.007 0.004

(0.031) (0.031) (0.031) (0.029) (0.020)Mother More Educated -0.006

(0.007)Mother Higher Job Status -0.006

(0.011)Constant 0.459 0.452 0.460 -0.160 -0.573*

(0.262) (0.269) (0.262) (0.199) (0.292)Family Characteristics X X X X XChild Characteristics X X X X XParental-Socialisation Effort X X X X XRegion Fixed Effects X X X X XArea Type Fixed Effects X X X X XObservations 1,640 1,640 1,640 1,486 1,125R

2 0.116 0.116 0.116 0.122 0.141Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

6.3 Other Outcomes Related to Gender Norms

I now consider, as an extension, two further outcomes associated with gender norms. They arederived from the following survey questions: “Do you agree that having a job that pays well isimportant?” (asked in wave 1) and “Would you like to study for a science degree at university?”(asked in wave 3). Both are coded as binary variables (Yes versus No). The first question is relatedto gender norms because it captures how children envision their financial independence in adulthood.For example, if children have traditional norms, I expect to find boys more likely to agree that highearnings are important, due to their future role as breadwinners. Similarly, girls with traditionalnorms, expecting their spouses to be the breadwinner, should place less importance on high futureearnings.

The second question is related to the well-known educational gender gap in science (and STEMsubjects more generally) (OECD 2012). In fact, a significant proportion of the gender pay gapamong university graduates can be attributed to gender gaps in entry into science degree programmes(Brown and Cororan 1997; Hunt et al. 2012; Weinberger 1999). The sciences have diachronicallybeen considered as ‘masculine’ disciplines and have led to stereotypes about the appropriate degreesand, thus, professions for each gender. It therefore becomes important to examine whether familysocialisation exacerbates this phenomenon by also propagating traditional gender norms in this

19

dimension. To examine the effect of family socialisation on these outcomes, I estimate the followingmodel:

Pr(Other Outcome)i

= �0 + �1ModernFamilyi

+ �2Femalei

+ �3ModernFamilyi

⇥ Femalei

+ Z0i

⇠ + vi

(6)

The results, shown in Table 313, are in line with my previous findings. I find that amongchildren in modern families, girls are 6.4% points less likely than boys to state that high futurewages are important. They are also 19.6% less likely to want to study science at university. Themost interesting comparison, however, is between girls in modern and traditional families. I againfind that girls in modern families are more likely to adopt traditional norms, in opposition to theirfamily’s norm but in line with society’s. Specifically, they are 4% points less likely to believe thathigh wages are important, and 9% points less likely to want to study science, compared to girls intraditional families.

Table 3: Other Outcomes. Dependent Var: Traditional Norm

(1) (2)Believe High Wages Important Want to Study Science

Modern Family 0.024* 0.106**(0.012) (0.031)

Female -0.046* -0.056***(0.022) (0.007)

Modern Family⇥Female -0.064*** -0.196***(0.011) (0.053)

Constant 0.487*** 0.186(0.129) (0.207)

Family Characteristics X XChild Characteristics X XParental-Socialisation Effort X XRegion Fixed Effects X XArea Type Fixed Effects X XObservations 1,486 1,486R2/ Pseudo-R2 0.125 0.229Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

6.4 Are Female-Breadwinner Families not Actually Transmitting Modern Views?

This study’s results suggest that girls growing up in modern families develop views opposite towhat we would expect. Instead of adopting their parents’ modern values, the girls are becomingmore traditional. One potential explanation is that parental actions do not coincide with beliefs,and so the families I am categorising as modern may not actually be transmitting modern views.Consider, for instance, families that hold traditional gender views but are forced by their economiccircumstances to have a female breadwinner. Not only will their actions not match their beliefs, buttheir dissatisfaction with this situation may make it even clearer to children that having mothers

13Full regression coefficient estimates are shown in Table B.4 in appendix B.

20

who are breadwinners is not the accepted norm. If this is true, it may explain why girls in modernfamilies develop more traditional views.

While this explanation is possible, a plethora of evidence suggests that beliefs and actions areindeed aligned. First, Figures 2 and 3 imply this alignment: there is strong cross-country correlationbetween gender views and female labour supply. Fortin (2015), who focuses on the US, reachesthe same conclusion. She shows that changes in gender-role attitudes are the strongest predictorof changes in female labour-force participation (particularly the recent ‘opting-out’ phenomenon).Further evidence comes from Johnston et al. (2014), who use data from the 1970 British CohortStudy. They find a strong correlation between the mother’s gender views and her hours worked.Interestingly, they also find children’s beliefs to be strongly related to maternal labour supply, overand above the mother’s gender views. This evidence emphasises that views and actions are aligned,but also that parents’ actions do matter for how children are socialised, irrespective of the parents’beliefs.

Nevertheless, I also examine my data for the possibility that female breadwinners are not actuallytransmitting modern views. Because the survey does not ask parents about their own gender views,I approach this differently. I look at two situations in which families may be ‘forced’ to have a femalebreadwinner: (1) the father cannot work due to a disability, and (2) the family is very poor. Todiscover whether my results are driven by constrained families, I split my sample into groups thatdiffer by how constrained they are based on these two factors, and repeat my analysis by sub-sampleto see if constrained families are driving my results.

Table 4 shows the results. Each column corresponds to a different sub-sample. The table showsthat constrained families do not drive the results. If anything, the opposite holds. Columns (1) and(2), for instance, show that results are driven purely by families where the father is not disabled.Columns (3) and (4) split the sample into those above and below the median family income. Again,I find that the results are driven by families that are not financially constrained. In the last column,I restrict the sample to families in which both the father is not disabled, and total family income isabove the median level. Again, the main finding holds. These results show that my findings overallcannot be explained by female breadwinner families being ‘forced’ into modern-family status.

An alternative explanation could be that female breadwinners do second shifts. Mothers who earnmore may try to ‘compensate’ for their violation of traditional norms by taking on a more traditionalgender role in the household through increased domestic work. I therefore examine whether secondshifts are more prevalent among modern families. Because the data on weekly hours of domesticwork are not continuous but banded in duration categories, I take two different approaches: (a) Icreate a dummy for whether the mother provides any positive amounts of domestic work, and (b) Icreate a continuous variable by assigning to each mother the mid-point of the category she is in. Ithen use these measures as the outcomes in specification (6). Table 5 shows the results. Column (1)corresponds to case (a) and column (2) to case (b). There is no statistically significant differencein either the prevalence or duration of second shifts between modern and traditional families. Thisfinding is consistent with other UK evidence. Female relative earnings and domestic work are notpositively related in either the 2000 Time Use Survey (Washbrook 2007) or the British HouseholdPanel Survey (Kan 2008).

21

Table 4: Dependent var: Traditional Norm

(1) (2) (3) (4) (5)Father not Father Family Income Family Income Family Income Above MedianDisabled Disabled Above Median Below Median and Father not Disabled

Modern Family -0.051*** 0.125 -0.085*** 0.036 -0.074***(0.011) (0.081) (0.015) (0.032) (0.013)

Female -0.147*** -0.167 -0.148*** -0.137** -0.157***(0.010) (0.102) (0.028) (0.054) (0.039)

Modern Family⇥Female 0.121*** -0.242* 0.149*** 0.039 0.138***(0.019) (0.102) (0.013) (0.071) (0.011)

Constant -0.603** 3.482* 1.130*** -0.935*** 0.169(0.202) (1.409) (0.126) (0.175) (0.166)

Family Characteristics X X X X XChild Characteristics X X X X XParental-Socialisation Effort X X X X XRegion Fixed Effects X X X X XArea Type Fixed Effects X X X X XObservations 1,353 132 820 665 767R2 0.127 0.685 0.186 0.191 0.189Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

Table 5: Mothers’ Second Shift

(1) (2)Mother Does Second Shift (LPM) Duration of Mother’s Second Shift

Modern Family 0.023 -0.361(0.022) (0.558)

Female -0.003 0.048(0.012) (0.239)

Modern Family⇥Female -0.013 0.054(0.015) (0.324)

Constant -0.375*** 3.739(0.091) (3.765)

Family Characteristics X XChild Characteristics X XParental-Socialisation Effort X XRegion Fixed Effects X XArea Type Fixed Effects X XObservations 1,485 1,485R

2 0.090 0.079Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

22

6.5 How do These Results Square With the Literature on Maternal LabourSupply?

The literature has focused on the relation between mothers’ hours worked and either daughters’gender attitudes or daughters’ own labour supply (Del Boca et al. 2000; Fernández et al. 2004;Morrill and Morrill 2013; Olivetti et al. 2016). These studies find a positive relation, suggesting atransmission of modern gender norms. Here, I show that looking at hours worked, without takinginto account which parent works or earns more, can be misleading. Table 6 shows the results when Ialso consider maternal labour supply in my analysis. In column (1), I include mother’s hours workedas an additional regressor. In column (2), I use a dummy for whether the mother works more hoursthan the father. The results in column (1) are consistent with the literature. I find that girls are lesslikely to develop traditional gender attitudes, the longer their mothers work. What is striking is howthe results change in column (2). When the mother works more hours than the father, the resultsare diametrically opposite. Girls are now more likely to develop traditional views, both comparedto sons of mothers who work longer than their fathers, and compared to daughters of mothers whowork less than their fathers. This finding highlights that what matters is not just how long motherswork, but how many hours they work (and how much they earn) relative to their husbands.

Table 6: Incorporating Mother’s Hours Worked. Dependent Var: Traditional Norm

(1) (2)Modern Family -0.041** -0.017*

(0.016) (0.007)Female -0.111*** -0.166***

(0.020) (0.012)Modern Family⇥Female 0.107*** 0.062**

(0.018) (0.018)Mother’s Hours Worked 0.001

(0.001)Mother’s Hours Worked⇥Female -0.002**

(0.001)Mother Works More Hours -0.090***

(0.022)Mother Works More Hours⇥Female 0.159**

(0.046)Constant -0.576* -0.112

(0.297) (0.191)Family Characteristics X XChild Characteristics X XParental-Socialisation Effort X XRegion Fixed Effects X XArea Type Fixed Effects X XObservations 1,486 1,486R2/Pseudo-R2 0.149 0.126Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

23

7 Examining the Mechanism: A Behavioural Explanation

The results reveal that family socialisation is not successful for girls in modern families. Not onlyare they more likely to develop traditional views compared to boys in modern families, but they arealso more likely to do so relative to girls in traditional families. These findings naturally lead to thefollowing question: why are girls brought up in modern families more likely to reject their familynorm and to develop traditional norms instead?

7.1 Preference for Conformity to the Family’s Norm

One key parameter in the model presented in Section 3 is the strength of the preference for conformityto the family. The model predicted that, when the family norm is more liberal than the social norm,a stronger preference for conformity to the family implies the development of relatively more modernnorms. In the same way, a weaker preference for conformity to the family then implies less modernnorms, i.e. more traditional norms, which is what I observe for girls. Could a weaker preferencefor conformity to the family’s norm therefore explain my results? To test this prediction, I run thefollowing regression:

Pr(Conformity to Family)i

= �0+ �1ModernFamilyi

+ �2Femalei

+ �3ModernFamilyi

⇥Femalei

+Z0i

�+⌘i

(7)

Two measures of conformity to the family are explored. The first comes from the first wave, at apoint in time when children have already chosen which subjects to study at the General Certificateof Secondary Education (GCSE) level (the last stage of compulsory education in the UK). This isan important decision as it affects what one can then study at the General Certificate of Education(GCE) Advanced Level, which directly affects university entrance. The particular measure I willexploit is agreement with the statement “I chose what to study at GCSE level based on what myparents wanted.” The second is a measure of whether children argue often with their parents, andis available from the second wave. All conformity indicators are constructed as binary (Yes versusNo) variables.

Table 7 shows the main results14. The findings support the model predictions, confirming thatgirls in modern families are less conformist to their family. Compared to boys in modern families,girls in modern families are 15.5% points less likely to have had their GCSE subjects chosen bytheir parents and 3.7% points more likely to argue often with their parents. Compared to girls intraditional families, girls in modern families are less conformist regarding whether parents chose theGCSE subjects and on the frequent-arguing measure by 7.3% and 6.4% points, respectively15.

14Full regression coefficient estimates are shown in Table B.8 in appendix B.15While in the case of the second measure (table 7) the coefficient of the Modern Family dummy is not individually

statistically significant, I reject the null hypothesis H0 : �1 + �3 = 0 at the 7% level (p-value 0.0658).

24

Table 7: Preferences for Conformity

(1) (2)Parents chose GCSE Subjects Argue Often With Parents

Modern Family 0.082** 0.027(0.028) (0.031)

Female 0.002 0.020(0.021) (0.015)

Modern Family⇥Female -0.155** 0.037***(0.047) (0.008)

Constant 0.753*** 1.643***(0.154) (0.173)

Family Characteristics X XChild Characteristics X XParental-Socialisation Effort X XRegion Fixed Effects X XArea Type Fixed Effects X XObservations 1,486 1,486R2/Pseudo-R2 0.096 0.085Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

7.2 But Why are Girls in Modern Families Less Conformist?

If girls in modern families are more traditional because they are less conformist to their family, whyare they less conformist? Is there an underlying cause making them less conformist? To answerthis, I draw from research findings in social psychology. Beginning with the seminal work andexperiments of Asch (1951; 1952; etc.) the literature has established that people have a strongpreference to conform to the view of the majority. With regard to between-gender differences inconformity preferences, it has also been shown that girls are much more susceptible to the majority’sviews than boys are (Eagly and Carli 1981; Santee and Jackson 1982). I combine these findings withthe fact that girls in modern families are growing up in a norm-minority family, in the sense thatonly 13.2% of families with dependent children in the UK have female breadwinners16. I use thesefacts to formulate the following hypothesis: living in a modern family causes girls to reject theirfamily’s norm-minority status, because it violates the oppositional social norm. Hence, growing upin a modern family makes girls less conformist to their family, which in turn makes them adopt moretraditional gender views. I next test this hypothesis.

7.3 Identification Using Regression Discontinuity

The negative relationship between growing up in and conforming to a modern family, as shown tohold for girls in Section 6.4, can be interpreted as causal only if we are willing to assume selection onobservables. Given the vast range of control variables, this is not such a heroic assumption. A morecompelling argument, however, can be made by exploiting quasi-experimental variation in modern-family status and applying a regression discontinuity design. This is possible because growing up

16This statistic is based on the Next Steps Survey. Using the UK Family Resources Survey, I find a very similarestimate (14.7%).

25

in a modern family is a deterministic function of the relative family income earned by the mother,which thereby creates a discontinuity in the probability of treatment.

Within the RD design language, the treatment is living in a modern family, and the assignmentvariable is the mother’s share (s

i

) of family income, defined as

s

i

=Mother

0s Income

i

Mother

0s Income

i

+ Father

0s Income

i

(8)

I will exploit the discontinuous jump in treatment at the 0.5 threshold of the assignment variable:

Pr(Treatment)i

=

(1 if s

i

> 0.5

0 if si

0.5(9)

I will estimate different versions of the following model:

Pr(Conformity to Family)i

= ⌧0 + ⌧1Treatment

i

+ f(si

,�) + Treatment

i

⇤ f(si

, ) +$

i

(10)

f(si

, ; ) is a polynomial function with parameter vector � that controls for the assignment variableand that controls for the interaction between the assignment variable and treatment status. ⌧1is the causal effect of living in a modern family on the probability of conforming. For robustness,some specifications will also include the vector of controls Z

i

used in the previous analysis.

7.4 The Identifying Assumption

Identification of ⌧1 requires local random assignment of the assignment variable. This means thatfamilies cannot precisely choose where they are around the 0.5 threshold of the mother’s incomeshare (Lee and Lemieux 2010). In a recent paper using US data, Bertrand et al. (2015) (henceforthBKP) show that wives’ income shares do indeed exhibit a discontinuity around 0.5. While the BKPfinding would invalidate my empirical strategy, there is no reason why it should necessarily hold inother countries. Eriksson and Stenberg (2015), for instance, repeat the BKP exercise for Swedenand find no evidence of sorting at the 0.5 threshold.

In this section I show evidence against sorting and in support of the identifying assumption. Ibegin by examining the probability density function of the mother’s income share to graphically testfor potential manipulation at the cutoff. If parents can precisely choose the mother’s income share,I should observe sorting around the cutoff. Figure 5a plots the number of families in each bin of s

i

and shows that the assignment variable varies smoothly across the 0.5 threshold. I also perform aMcCrary test and fail to reject the null of no discontinuity at the cutoff. Hence, both the graphicaland statistical evidence supports the identifying assumption.

However, one concern may be that my sample size is too small to reveal any sorting. To overcomethis concern, I test this assumption by using another UK dataset, the Family Resources Survey(FRS). The FRS is the primary source of income data in the UK. It includes cross-sections of morethan 20,000 households per year and contains detailed information on respondents’ incomes. Due toits high quality and sample size, it serves as the primary source of information for the Departmentfor Work and Pensions (DWP) to guide UK welfare policy (DWP 2015).

Following BKP, I use the FRS to calculate individual total income as the sum of wage and self-employment income, and estimate the wife’s income share. To make the FRS sample comparable tothe families in my Next Steps data, I restrict my analysis to dual-parent, heterosexual families, inwhich there is at least one dependent child and at least one parent working, and consider the years1997-2004.

26

Figure 5b plots the distribution of the wife’s income share from the FRS. Note that here Ihave followed BKP and did not differentiate between wage and self-employment income. Similarto BKP’s findings, I find significant bunching at the 0.5 threshold, implying that the assignmentvariable is manipulated. However, a major concern with the BKP method is that no distinction ismade between wage and self-employed income; these are aggregated into a single measure of income.Of course, self-employed individuals have much more control over their exact incomes, but thosewith only wage income may only be able to sort at shares of 0 (or 1) by having only the father(mother) work (this is also evident in Figure 5a with sorting at 0). It is therefore unclear whetherthe BKP results are driven by the self-employed.

To clarify this, I next split the sample into two groups: families with at least one self-employedspouse, and families with no self-employed spouses. Figures 5c and 5d show the correspondingdensities. I now find a large discrepancy that speaks against BKP’s results. While there is verysignificant bunching among the self-employed at the 0.5 cutoff (Figure 5c), there is none amongwage-earning families (Figure 5d). The McCrary tests again confirm the visual evidence. Hence,BKP’s findings are not universal; in the UK, at least, they only hold for the self-employed.

Eriksson and Stenberg (2015) also highlight the importance of self-employment earning in manip-ulating the wife’s share. When they examine the characteristics of couples around the 0.5 threshold,they find that the fraction of couples with at least one self-employed spouse is 70.6% at the 0.5 share,compared to just 13% in the surrounding bins. Moreover, this finding is also in line with the recentbunching literature on behavioural responses to taxation17. While economic theory predicts thatworkers should bunch at convex kinks of the income tax schedule, the evidence shows that bunchingis mainly driven by the self-employed. This is because wage earners face significant labour-marketfrictions that constrain them from freely choosing their income levels (Chetty et al. 2011). To ex-pect, then, that couples can precisely choose where to locate around the 0.5 threshold would assumenot only that workers can choose their wage earnings precisely, but that they do so also in responseto their spouse’s wages.

The evidence presented in this section supports the RD identifying assumption regarding theassignment variable, at least among wage earners. While the self-employed are not a large groupin the Next Steps data (they are approximately 10%), I will present results with and without theirinclusion in the sample for robustness. Before proceeding with estimation, I also test whetherbaseline covariates exhibit discontinuities across the threshold. If treatment is randomised, theseshould be locally balanced on each side of the 0.5 cutoff. Given the very large number of covariatesI consider, some discontinuities may be statistically significant by random chance (Lee and Lemieux2010). Hence, as long as the proportion of statistically significant covariates is substantially small(5-10% of covariates only), this will not pose a threat to identification (Jacob and Zhu 2012). I testthis by estimating (10) for each covariate separately, substituting each element of Z

i

for Y

i

and notconditioning on other covariates. I also do this for various sample sizes, where I, in turn, trim theoutermost 1%, 5%, 10%, 20%, 40%, 60% and 80% of the sample for robustness. Table 8 shows theresults, confirming that baseline covariates are balanced overall. I find only a very small proportionof covariates to exhibit any discontinuity that is statistically significant, especially at the 5% or 1%levels. Importantly, this finding holds as I symmetrically exclude more and more observations awayfrom the cutoff, and thus provides further support to my identifying assumption.

17For a review, see Kleven (2015).

27

Figure 5: The Distribution of the Mother’s Income Share

(a) Next Steps SurveyMcCrary Test:Discontinuity est.= .089(.3)

050

100

150

Num

ber

of O

bse

rvatio

ns

per

Bin

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Mother’s Income Share

(b) Family Resources Survey - Earnings IncludeWage and Self-Employment Income

McCrary Test:Discontinuity est.= .232***(.052)

0500

1000

1500

Num

ber

of O

bse

rvatio

ns

per

Bin

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Mother’s Income Share

(c) Family Resources Survey - Only Families withSelf-Employment Income

McCrary Test:Discontinuity est.= .799***(.108)

050

100

150

200

250

Num

ber

of O

bse

rvatio

ns

per

Bin

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Mother’s Income Share

(d) Family Resources Survey - Families with OnlyWage Income

McCrary Test:Discontinuity est.= .007(.059)

0500

1000

1500

Num

ber

of O

bse

rvatio

ns

per

Bin

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Mother’s Income Share

Notes: These figures plot the distribution of the assignment variable, the mother’s income share, for differentsamples. The bin size in each figure is 0.02. Figure (a) corresponds to the Next Steps Survey, and Figures (b)- (d) to the Family Resources Survey. While figure (b) includes all families, figure (c) restricts the sample toonly families where at least one spouse has income from self-employment, and figure (d) restricts the sampleto families with only wage income. Each figure reports the result of the McCrary test, which tests the null ofno discontinuity at the 0.5 threshold (vertical line) of the assignment variable. The null of no discontinuitycannot be rejected at the 10% level in cases (a) and (d), but is rejected at the 1% level in cases (b) and (c).Results show that sorting at the 0.5 threshold is driven by the self-employed.

Table 8: Testing Discontinuities in Covariates

% of outer sample trimmed 0% 1% 5% 10% 20% 40% 60% 80%Proportion of statistically significant covariates

10% significance level 0 5 4 3 6 4 6 45% significance level 6 4 5 4 4 2 0 21% significance level 0 0 0 0 0 0 0 0

Notes: This table shows the % of covariates that exhibit a statistically significant discontinuity (at the 10%,5% and 1% level) at the 0.5 cutoff, from RD regressions where each covariate is treated as the dependentvariable. Results are shown for the full sample as well as smaller subsamples, where observations are droppedsymmetrically (trimmed) around the 0.5 cutoff in the assignment variable.

28

7.5 RD Results

I first examine whether a discontinuity at the cutoff can be identified visually. Figure 6 shows therelationship between the share of mother’s income and the two conformity measures. In both cases,there is a discontinuity in the probability of each statement being true around the cutoff. Comparedto those just below the cutoff, girls whose mothers have a share of family income just above 0.5 aremuch less likely to have chosen their GCSE subjects based on what their parents wanted, and aremuch more likely to argue often with them. I next estimate the size of these discontinuities. Theresults are presented in Table 9. Panel A shows estimates of the effect on the probability that agirl studied what her parents wanted at GCSE, and panel B shows estimates of the effect on theprobability that a girl argues often with her parents. Panel A1 (B1) estimates the effect using thewhole sample. In light of the differences in sorting between wage earners and self-employed in theFRS, panel A2 (B2) restricts the sample to only families in which neither parent has earnings fromself-employment.

The results show that these discontinuities are statistically significant and robust to the f(si

, ; )specification, the inclusion of controls and the restriction of the sample to wage earners. Althoughthe results do not differ substantially, my preferred specification is the one estimated on only wage-earner parents. The results support my hypothesis of a causal effect of living in a modern family.The modern family treatment causes girls to have a lower preference for conformity to the family.This, in turn, makes them adopt more traditional views regarding gender roles.

29

Figure 6: Regression Discontinuity Graphs: Conformity

0.2

.4.6

.81

0 .2 .4 .6 .8 1Mother’s Income Share

Parents Chose GCSE Subjects

0.2

.4.6

.81

0 .2 .4 .6 .8 1Mother’s Income Share

Argue Often with Parents

Notes: The figure shows for females, the probability that children agree with each measure of conformityabove and below the cutoff. The first is whether children chose what GCSE subjects to study based on whattheir parents wanted, and the second is whether children argue with their parents often. Observations aredivided into bins of width 0.05. Individuals to the right of the cutoff receive the treatment of living in amodern family, while those to the left do not.

Table 9: RD Results. Dependent Var: Conformity Measures

Panel A: Parents Chose GCSE Subjects

Panel A1: Wage Earners and Self-Employed Parents(1) (2) (3) (4) (5) (6)

Modern Family -0.313*** -0.356*** -0.334*** -0.299*** -0.328*** -0.378***(0.0296) (0.0507) (0.0755) (0.0482) (0.0928) (0.0973)

Controls ⇥ ⇥ ⇥ X X X1st-order polynomial X ⇥ ⇥ X ⇥ ⇥2nd-order polynomial ⇥ X ⇥ ⇥ X ⇥3rd-order polynomial ⇥ ⇥ X ⇥ ⇥ XObservations 779 779 779 779 779 779R-squared 0.023 0.024 0.025 0.164 0.165 0.167

Panel A2: Only Wage-Earner Parents(7) (8) (9) (10) (11) (12)

Modern Family -0.325*** -0.410*** -0.396*** -0.333*** -0.413*** -0.476***(0.0363) (0.0453) (0.0812) (0.0459) (0.0841) (0.0911)

Controls ⇥ ⇥ ⇥ X X X1st-order polynomial X ⇥ ⇥ X ⇥ ⇥2nd-order polynomial ⇥ X ⇥ ⇥ X ⇥3rd-order polynomial ⇥ ⇥ X ⇥ ⇥ XObservations 682 682 682 681 681 681R-squared 0.023 0.025 0.027 0.187 0.189 0.191

Panel B: Argue Often With Parents

Panel B1: Wage Earners and Self-Employed Parents(13) (14) (15) (16) (17) (18)

Modern Family 0.105** 0.247*** 0.153** 0.154*** 0.226*** 0.156*(0.0405) (0.0205) (0.0514) (0.0284) (0.0304) (0.0777)

Controls ⇥ ⇥ ⇥ X X X1st-order polynomial X ⇥ ⇥ X ⇥ ⇥2nd-order polynomial ⇥ X ⇥ ⇥ X ⇥3rd-order polynomial ⇥ ⇥ X ⇥ ⇥ XObservations 779 779 779 779 779 779R-squared 0.003 0.010 0.011 0.170 0.171 0.173

Panel B2: Only Wage-Earner Parents(19) (20) (21) (22) (23) (24)

Modern Family 0.105* 0.300*** 0.217*** 0.158** 0.237*** 0.175*(0.0561) (0.0195) (0.0434) (0.0515) (0.0491) (0.0838)

Controls ⇥ ⇥ ⇥ X X X1st-order polynomial X ⇥ ⇥ X ⇥ ⇥2nd-order polynomial ⇥ X ⇥ ⇥ X ⇥3rd-order polynomial ⇥ ⇥ X ⇥ ⇥ XObservations 661 661 661 660 660 660R-squared 0.005 0.015 0.016 0.212 0.214 0.214Robust standard errors (clustered by bin) in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

8 Conclusion

In this paper I examined whether gender norms are passed on from parents to children when thenorms of the family oppose those of the society in which the family is embedded. While boys raisedin modern families (i.e. in which the mother is the breadwinner) are less likely to develop traditionalnorms, girls raised in modern families are actually more likely to do so, in opposition to their family’snorm but in line with society’s. I argued, based on the predictions of a gender-identity model, thatthese results can be explained by a weaker preference among girls for conforming to the family.Drawing on research from social psychology and using a regression discontinuity design, I showedthat this weaker preference is, in fact, driven by a causal effect of living in a modern family.

My results question the literature on the relation between mothers’ and daughters’ labour supplyor gender attitudes (Del Boca et al. 2000; Fernández et al. 2004; Morrill and Morrill 2013; Olivettiet al. 2016). Previous studies’ finding of a positive relation is interpreted as transmission of moderngender norms. I replicate their finding when only considering mothers’ hours worked, but show thatresults change drastically when I consider whether the mother works or earns more than the fatherdoes. I therefore show that a positive relation between mothers’ and daughters’ labour supply is notsufficient for the transmission of modern views. What matters is not just how long mothers work,but how many hours they work (and how much they earn) relative to their husbands.

My results reveal that horizontal socialisation is very important for the development of girls’gender norms. In fact, it is so strong that it leads to ‘reactionary’ behaviour by girls when theirfamilies violate the traditional social norm. If we are interested in reducing gender inequalitiesby promoting more modern gender norms, we must therefore focus on changing the social norm.Teenage girls are exposed to social norms through the mass media and through schools. Policy musttherefore address how gender norms are portrayed in the media, but more important, it must focuson the role of schools and teachers in transmitting social norms to children.

Moreover, a critical assessment of the current family policy in England seems pertinent. At thetime of writing, the UK’s Statutory Paternity Pay entitles fathers to just two weeks of paid paternityleave, payable at the minimum of 90% of previous weakly earnings and £139.58. While this leavecan be extended up to 26 weeks, it is conditional on the mother returning to work before the endof her maternity leave period, payable again at just £139.58 per week. As a result, less than 1% offathers extend their leave beyond two weeks (TUC 2015). The current policy has failed to providefathers with incentives to provide child care, especially in cases where their foregone earnings arehigher than their wives’. Unless paternity leave becomes much more generous, the social norm willcontinue to define child care as an exclusively maternal responsibility.

32

References

[1] Akerlof, G.A. and R.E. Kranton (2000) “Economics and Identity,” Quarterly Journal of Eco-nomics, vol. 115(3), pp. 715-753.

[2] Akerlof, G.A. and R.E. Kranton (2002) “Identity and Schooling: Some Lessons for the Eco-nomics of Education,” Journal of Economic Literature, vol. 40(3), pp. 1167-1201.

[3] Akerlof, G.A. and R.E. Kranton (2010) Identity Economics. New Jersey: Princeton UniversityPress.

[4] Alesina, A. and P. Giuliano (2010) “The Power of the Family,” Journal of Economic Growth,vol. 15, pp. 93-125.

[5] Alesina, A., P. Giuliano and N. Nunn (2013) “On the Origins of Gender Roles: Women and thePlough,” Quarterly Journal of Economics, vol. 128(2), pp. 469-530.

[6] Asch, S.E. (1951) “Effects of Group Pressure on Modification and Distortion of Judgments,” InGuetzkow, H., eds., Groups, Leadership and Men. Pittsburgh: Carnegie Press.

[7] Asch, S.E. (1952) Social Psychology. New Jersey: Prentice Hall.

[8] Asgari, S. N. Dasgupta and N.G. Cote (2010) “When Does Contact with successful IngroupMembers Change Self-Stereotypes?” Social Psychology, vol. 41(3), pp. 203-211.

[9] Becker, G. (1964) Human Capital. Columbia University Press: New York.

[10] Bem, S L. (1985) “Androgyny and Gender Schema Theory: A conceptual and Empirical Inte-gration,” Psychology and Gender, vol. 32, pp. 179-226.

[11] Bénabou, R. and J. Tirole (2007) “Identity, Dignity and Taboos: Beliefs as Assets,” CEPRDiscussion Paper 6123.

[12] Bertrand, M., E. Kamenica and J. Pan (2015) “Gender Identity and Relative Income WithinHouseholds,” Quarterly Journal of Economics, forthcoming.

[13] Bisin, A., E. Patacchini, T. Verdier and Y. Zenou (2011) “Formation and Persistence of Oppo-sitional Identities,” European Economic Review, vol. 55, pp. 1046-1071.

[14] Blau, F.D. and L.M. Kahn (2000) “Gender Differences in Pay,” Journal of Economic Perspec-tives, vol. 14(4), pp. 75-99.

[15] Blunch, N.H. and M.B. Das (2014) “Changing Norms about Gender Inequality in Education:Evidence from Bangladesh,” IZA Discussion Paper No. 8365.

[16] Bonke, J. (2008) “Income Distribution and Financial Satisfaction Between Spouses in Europe,”The Journal of Socio-Economics, vol. 37, pp. 2291-2303.

[17] Bonke, J. and M. Browning (2009) “The Distribution of Well-Being and Income Within theHousehold,” Review of the Economics of the Household, vol. 7, pp. 31-42.

[18] Brown, C. and M. Cororan (1997) “Sex-Based Differences in School Content and the Male-Female Wage Gap,” Journal of Labor Economics, vol. 15(3), pp. 431-465.

33

[19] Butikofer, A. (2013) “Revisiting ‘Mothers and Sons’ Preference Formation and the Female LaborForce in Switzerland,” Labour Economics, vol. 20, pp. 82-91.

[20] Carrell, S.E., M.E. Page and J.E. West (2010) “Sex and Science: How Professor Gender Per-petuates the Gender Gap,” Quarterly Journal of Economics, vol. 125(3), pp. 1101-1144.

[21] Chetty, R., J. Friedman, T. Olsen and L. Pistaferri (2011) “Adjustment Costs, Firm Responsesand Micro vs. Macro Labor Supply Elasticities: Evidence from Danish Tax Records,” QuarterlyJournal of Economics, vol. 126(2), pp. 749-804.

[22] Cooksey, E.C. and M.M. Fondell (1996) “Spending Time With his Kids: Effects of FamilyStructure on Fathers’ and Children’s Lives,” Journal of Marriage and the Family, vol. 58, pp.693-707.

[23] Cooray, A. and N. Potrafke (2010) “Gender Inequality in Education: Political Institutions orCulture and Religion?” Konstanz Working Paper 2010-01.

[24] Costa-Font, J. and F.A. Cowell (2015) “Social Identity and Redistributive Preferences,” Journalof Economic Surveys, vol. 29, pp. 357-374.

[25] Deary, I.J., G.D. Batty and C.R. Gale (2008) “Bright Children Become Enligthened Adults,”Perspectives on Psychological Science, vol. 19(1), pp. 1-6.

[26] Dee, T.S. (2004) “Are There Civic Returns to Education?” Journal of Public Economics, vol.88, pp. 1697- 1720.

[27] Del Boca, D., M. Locatelli and S. Pasqua (2000) “Employment Decisions of Married Women:Evidence and Explanations,” Labour, vol. 14(1), pp. 35-52.

[28] Department for Education (2014) “School Workforce in England: November 2013,” StatisticalFirst Release 11/2004.

[29] Department for Work and Pensions (2015) “Family Resources Survey – United Kingdom2013/14,” National Statistics Release 06/2015.

[30] Dollar, D. and R. Gatti (1999) “Gender Inequality, Income, and Growth: Are Good TimesGood For Women,” World Bank Working Paper Series No. 1.

[31] Dugger, K. (1988) “Social Location and Gender-Role Attitudes: a Comparison of Black andWhite Women,” Gender and Sociology, vol. 2(4), pp. 425-448.

[32] Eagly, A.H. and L.L. Carli (1981) “Sex of Researchers and Sex-Typed Communications as De-terminants of Sex Differences in Influenceability: A Meta-Analysis of Social Influence Studies,”Psychological Bulletin, vol. 90(1), pp. 1-20.

[33] Eckstein, Z. and O. Lifshitz (2011) “Dynamic Female Labor Supply,” Econometrica, vol. 79(6),pp. 1675-1726.

[34] Eriksson, K.H and A. Stenberg (2015) “Gender Identity and Relative Income Within House-holds: Evidence from Sweden,” IZA Discussion Paper No. 9533.

[35] Epstein, M. and M.L. Ward (2011) “Exploring Parent-Adolescent Communication About Gen-der. Results from Adolescent and Emerging Adult Samples,” Sex Roles, vol. 65, pp. 108-118.

34

[36] FAO (2009) “Women and Rural Employment: Fighting Poverty by Redefining Gender Roles”Economic and Social Perspectives, Policy Brief 5.

[37] Farre, L. and F. Vella (2013) “The Intergenerational Transmission of Gender Role Attitudesand its Implications for Female Labor Force Participation,” Economica, vol. 80, pp. 219-247.

[38] Fernández, R. and A. Fogli (2009) “Culture: an Empirical Investigation of Beliefs, Work andFertility,” American Economic Journal: Macroeconomics, vol. 1(1), pp. 146-177.

[39] Fernández, R., A. Fogli and C. Olivetti (2004) “Mothers and Sons: Preference Formation andFemale Labor Force Dynamics,” Quarterly Journal of Economics, vol. 119(4), pp. 1249-1299.

[40] Fortin, N.M. (2015) “Gender Role Attitudes and Women’s Labor Market Participation: Opting-Out, AIDS, and the Persistent Appeal of Housewifery,” Annals of Economics and Statistics, vol.117-118, pp. 379-401.

[41] Georgiadis, A. and A. Manning (2013) “One Nation Under a Groove? Understanding NationalIdentity,” Journal of Economic Behavior and Organization, vol. 93, pp. 166-185.

[42] Goldin, C. (1995) “The U-Shaped Female Labor Force Function in Economic Development andEconomic History,” NBER Working Paper No. 4707.

[43] Goldin, C. and L.F. Katz (2002) “The Power of the Pill: Oral Contraceptives and Women’sCareer and Marriage Decisions,” Journal of Political Economy, vol. 110(4), pp. 730-770.

[44] Gonzalez de San Roman, A. and S. de la Rica Goiricelaya (2012) “Gender Gaps in PISA TestScores: The Impact of Social Norms and the Mother’s Transmission of Role Attitudes,” IZADiscussion Paper No. 6338.

[45] Guiso, L., P. Sapienza and L. Zingales (2003) “People’s Opium? Religion and Economic Atti-tudes,” Journal of Monetary Economics, vol. 50, pp. 225-282.

[46] Heineck, G. (2004) “Does Religion Influence the Labor Supply of Married Women in Germany?”Journal of Socio-Economics, vol. 33, pp. 307-328.

[47] Hognas, R.S. and M.J. Carlson (2012) “Like Parent, Like Child? The Intergenerational Trans-mission of Nonmarital Childbearing,” Social Science Research, vol. 41(6), pp. 1480-1494.

[48] Hunt, J., J.P. Garant, H. Herman and D.J. Munroe (2012) “Why Don’t Women Patent?” NBERWorking Paper No. 17888.

[49] IPPR (2015) “Who’s Breadwinning in Europe? A Comparative Analysis of Maternal Bread-winning in Great Britain and Germany,” IPPR Report.

[50] Jacob, R., P. Zhu, M.A. Somers and H. Bloom (2012) “A Practical Guide to Regression Dis-continuity,” MDRC.

[51] Johnston, D.W., S. Schurer and M.A. Schields (2014) “Maternal Gender Role Attitudes, HumanCapital Investment, and Labour Supply of Sons and Daughters,” Oxford Economics Papers, vol.66, pp. 631-659.

[52] Kan, M.Y. (2008) “Does Gender Trump Money? Housework Hours of Husbands and Wives inBritain,” Work, Employment and Society, vol. 22(1), pp. 45-66.

35

[53] Kanazawa, S. (2010) “Why Liberals and Atheists are More Intelligent,” Social Psychology Quar-terly, vol. 73, pp. 33–57.

[54] Kane, E.W. (2000) “Racial and Ethnic Variations in Gender-Related Attitudes,” Annual Reviewof Sociology, vol. 26, pp. 419-439.

[55] Keshet, J.K. (2001) “Uncounted Stepfamilies: Remarried Fathers and their Wives,” In Cath,S.H. and Shopper, M., eds., Stepparenting: Creating and Recreating Families in America Today.New Jersey: The Analytic Press.

[56] Klasen, S. (1999) “Does Gender Inequality Reduce Growth and Development? Evidence fromCross-Country Regressions,”. World Bank Working Paper Series No. 7.

[57] Kleven, H. (2015) “Bunching,” Draft Prepared for the Annual Review of Economics.

[58] Klor, E.F. and M. Shayo (2010) “Social Identity and Preferences Over Redistribution,” Journalof Public Economics, vol. 92, pp. 944-961.

[59] Knowles, S., P. Lorgelly and D. Owen (2002) “Are Educational Gender Gaps a Brake on Eco-nomic Development? Some Cross-Country Empirical Evidence,” Oxford Economic Papers, vol.54, pp. 118–149.

[60] Lee, D.S. and T. Lemieux (2010) “Regression Discontinuity Designs in Economics,” Journal ofEconomic Literature, vol. 48, pp. 281-355.

[61] Lehrer, E.L. (1995) “The Effects of Religion on the Labor Supply of Married Women,” SocialScience Research, vol. 24, pp. 281-301.

[62] Lorber, J. (1994) Paradoxes of Gender. New Haven: Yale University Press.

[63] Lytton, H. and D.M. Romney (1991) “Parents’ Differential Socialization of Boys and Girls: AMeta-Analysis,” Psychological Bulletin, vol. 109(2), pp. 267-296.

[64] Mammen, K. and C. Paxson (2000) “Women’s Work and Economic Development,” Journal ofEconomic Perspectives, vol. 14(4), pp. 141-164.

[65] Maneschiold, P. and B. Haraldsson (2007) “Religious Norms and Labour Supply of MarriedWomen in Sweden,” Finnish Economic Papers, vol. 20, pp. 41-56.

[66] McCrary, J. (2008) “Manipulation of the Running Variable in the Regression DiscontinuityDesign: A Density Test,” Journal of Econometrics, vol. 142, pp. 698–714.

[67] Mincer, J. and S. Polachek (1974) “Family Investments in Human Capital: Earnings of Women.”Journal of Political Economy, vol. 82(2), pp. S76-S108.

[68] Morrill, M.S. and T. Morrill (2013) “Intergenerational Links in Female Labor Force Participa-tion,” Labour Economics, vol. 20, pp. 38-47.

[69] Neighbourhood Renewal Unit (2004) The English Indices of Deprivation 2004. The Office ofthe Deputy Prime Minister: London.

[70] OECD (2004) “Female Labour Force Participation: Past Trends and Main Determinants inOECD Countries,” OECD Working Papers Series No. 376.

[71] OECD (2012) Closing the Gender Gap: Act Now. OECD Publishing.

36

[72] Olivetti, C., E. Patacchini and Y. Zenou (2016) “Mothers, Peers and Gender Identity,” WorkingPaper.

[73] Parker, K. and W. Wang (2013) “Modern Parenthood -Roles of Moms and Dads Converge asthey Balance Work and Family,” PEW Research Center Report.

[74] Polachek, S.W. (2008) “Earnings Over the Life Cycle: The Mincer Earnings Function and ItsApplications,” Foundations and Trends in Microeconomics, vol. 4(3), pp. 165-272.

[75] Ramey, C.T., F.A. Campbell and S.L. Ramey (1999) “Early Intervention: Successful Pathwaysto Improving Intellectual Development,” Developmental Neuropsychology, vol. 16, pp. 385-392.

[76] Santee, R.T. and S.E. Jackson (1982) “Identity Implications of Conformity: Sex Differences inNormative and Attributional Judgments,” Social Psychology Quarterly, vol. 45(2), pp. 121-125.

[77] Stankov, L. (2009) “Conservatism and Cognitive Ability,” Intelligence, vol. 37, pp. 294–304.

[78] Tajfel, H. (1978) Differentiation Between Social Groups: Studies in the Psychology of IntergroupRelations. Academic Press.

[79] TUC (2015) “Statistical Briefing on Fathers’ Leave and Pay” Trade Union Congress Report.

[80] UNDP (2014) Human Development Report 2014. Sustaining Human Progress: Reducing Vul-nerabilities and Building Resilience. UNDP: New York.

[81] Vella, F. (1994) “Gender Roles and Human Capital Investment: The Relationship BetweenTraditional Attitudes and Female Labour Market Performance,” Economica, vol. 61, pp. 191-211.

[82] Washbrook, E. (2007) “Explaining the Gender Division of Labour: The Role of the GenderWage Gap,” CMPO Working Paper Series No. 07/174.

[83] Weinberger, C.J. (1999) “Mathematical College Majors and the Gender Gap in Wages,” Indus-trial Relations, vol. 38(4). pp. 407-413.

[84] World Bank (2006) World Development Report 2006: Equity and Development. The WorldBank: Washington.

37

Appendix A. Other comparative statics from the gender-identity modelSection 3 presented a simple model of gender identity. It was shown that children choose their

optimal belief about gender roles using the following rule:

x

⇤ = argmax

x

U = x(xF,

x

S

, Z) =c

F

(xF

, x

S

, Z)xF

+ c

S

(xF

, x

S

, Z)xS

c

F

(xF

, x

S

, Z) + c

S

(xF

, x

S

, Z)(11)

The main text focused on a particular comparative static ( @x⇤

@c

F

) and went to lengths to show thatit holds in the data. In this appendix I briefly discuss another comparative static and show that italso confirmed by the empirical results. Here I look at @x

⇤

@x

F

, i.e. how living in a more traditionalfamily changes the child’s optimal belief. Recall that higher values of x⇤, x

F

and x

S

indicate moretraditional values, and that in the empirical specification, x

S

> x

F

. The prediction is as follows18:

@x

⇤

@x

F

=c

2F

(xF

, x

S

, Z) + c

F

(xF

, x

S

, Z)cS

(xF

, x

S

, Z) + (xS

� x

F

)⇣c

F

(xF

, x

S

, Z) @c

S

@x

F

� c

S

(xF

, x

S

, Z) @cF@x

F

⌘

(cF

(xF

, x

S

, Z) + c

S

(xF

, x

S

, Z))2

(12)

The prediction is ambiguous, while empirically I found @x

⇤

@x

F

> 0 for boys and @x

⇤

@x

F

< 0 for girls.The prediction depends on the signs of both @c

S

@x

F

and @c

F

@x

F

. While there is no proxy in the datato estimate @c

S

@x

F

, the evidence from the social psychology literature (Asch 1951; 1952; etc.) implies@c

S

@x

F

� 0. The reasoning is as follows. The literature has established that individuals have strongpreferences for conforming to society. Furthermore, the social norm regarding gender roles in theUK is very traditional. Hence, the more traditional a family’s gender norm is, the more the familynorm aligns with the social norm that individuals have a strong preference to conform to. Hence,the strength of their preference for conforming to the social norm must be non-decreasing in x

F

.The prediction therefore ultimately depends on the sign of @c

F

@x

F

.To show how my empirical results are consistent with the model, first consider boys. A sufficient

condition for @x

⇤

@x

F

> 0 is @c

F

@x

F

< 0, i.e. that boys in modern families have stronger preferencesfor conforming to their family than boys in traditional families. The results of Table 7 confirmthis. The coefficient on Modern Family, which compares boys in modern to boys in traditionalfamilies, is positive and statistically significant, confirming that boys in modern families have strongerpreferences for conformity.

Now consider the prediction for girls. A necessary (though not sufficient condition) for @x

⇤

@x

F

< 0

is @c

F

@x

F

> 0, i.e. that girls in modern families have weaker preferences for conforming to their familythan girls in traditional families. This was the focus of all of Section 7, which confirmed this weakerpreference among girls in modern families.

18Here I have for simplicity assumed that the family norm does not depend on the social norm. Introducing thisdependence does not change the prediction.

38

Appendix B. Summary Statistics and Full Coefficient Estimates of Results Shown in

Main Text

This apendix includes the table of summary statistics, and the full regression coefficient estimates of all the tables shown in themain text. In the case of a binary dependent variable, the estimates presented here are based on linear probability models.

Table B.1: Summary Statistics

Variables Description Mean St. Dev.

Main Variables of Interest

Modern Family 1 if mother earns more than father 0.132 0.338

Traditional Norm 1 if child believes women with young children should never work full-time 0.327 0.469

High Wage Important 1 if child believes earning a high wage is important 0.651 0.477

Study Science 1 if child wants to pursue a science degree at university level 0.416 0.493

Family Chose GCSEs 1 if the child’s family chose what she will study at GCSE level 0.375 0.484

Argue Often with Parents 1 if child argues with parents at least once a week 0.596 0.491

Family Characteristics

Parents not Married 1 if child’s parents are not married 0.055 0.229

Parents Natural 1 if both are child’s biological parents 0.899 0.301

Household Size No. of people in child’s household 4.711 1.249

Dependents Number of dependent children in household 2.326 1.100

Father’s Age Male parent’s age in years 43.176 5.916

Mother’s Age Female parent’s age in years 44.480 6.374

Religion Important to Parents 1 if any parent states that religion is important to family’s way of life 0.607 0.489

Father’s Religion: Christian 1 if father is Christian 0.792 0.406

Father’s Religion: Muslim 1 if father is Muslim 0.132 0.338

Father’s Religion: Minority 1 if father is Sikh, Hindu, Buddhist, Jewish, Atheist 0.070 0.255

Mother’s Religion: Christian 1 if mother is Christian 0.693 0.461

Mother’s Religion: Muslim 1 if mother is Muslim 0.133 0.340

Mother’s Religion: Minority 1 if mother is Sikh, Hindu, Buddhist, Jewish, Atheist 0.168 0.374

Family Education: Degree 1 if highest education between parents is university degree level or higher 0.229 0.420

Family Education: Below Degree 1 if highest education between parents is higher education below degree

level

0.187 0.390

Family Education: GCE A Level 1 if highest education between parents is GCE A Level or equivalent 0.195 0.396

Family Education: GCSE 1 if highest education between parents is GCSE level grades A-C, or

equivalent

0.232 0.422

Family Education: Level 1 1 if highest education between parents is at level 1 or below 0.038 0.191

Family Education: Other 1 if highest education between parents is informal/undefined (below

degree)

0.010 0.098

Family Education: None 1 if neither parent has any education qualifications 0.110 0.313

Father’s Social Class: Higher Mang. 1 if father’s occupation is ‘Higher Managerial or Professional’ 0.073 0.260

Father’s Social Class: Lower Mang. 1 if father’s occupation is ‘Lower Managerial or Professional’ 0.276 0.447

Father’s Social Class: Intermediate 1 if father’s occupation is ‘Intermediate’ 0.165 0.371

Father’s Social Class: Small Employer 1 if father’s occupation is ‘Small Employer/Own Accounts Worker’ 0.040 0.195

Father’s Social Class: Lower Supervisor 1 if father’s occupation is ‘Lower Supervisory/Technical’ 0.084 0.278

Father’s Social Class: Semi-Routine 1 if father’s occupation is ‘Semi-Routine’ 0.196 0.397

Father’s Social Class: Routine 1 if father’s occupation is ‘Routine’ 0.100 0.300

Father’s Social Class: Long-Term u/e 1 if father is long-term unemployed/never worked 0.066 0.249

Mother’s Social Class: Higher Mang. 1 if mother’s occupation is ‘Higher Managerial or Professional’ 0.161 0.368

39

Mother’s Social Class: Lower Mang. 1 if mother’s occupation is ‘Lower Managerial or Professional’ 0.235 0.424

Mother’s Social Class: Intermediate 1 if mother’s occupation is ‘Intermediate’ 0.071 0.256

Mother’s Social Class: Small Employer 1 if mother’s occupation is ‘Small Employer/Own Accounts Worker’ 0.080 0.271

Mother’s Social Class: Lower Supervisor 1 if mother’s occupation is ‘Lower Supervisory/Technical’ 0.136 0.343

Mother’s Social Class: Semi-Routine 1 if mother’s occupation is ‘Semi-Routine’ 0.100 0.300

Mother’s Social Class: Routine 1 if mother’s occupation is ‘Routine’ 0.121 0.326

Mother’s Social Class: Long-Term u/e 1 if mother is long-term unemployed/never worked 0.097 0.296

Family Income Total annual family income (sum of father’s and mother’s income) 31522.6 33649.02

IDACI score Income Deprivation Affecting Children Index, range: 0-1 0.189 0.170

IMD Rank Index of Multiple Deprivation rank, range: 1-32,482 21.579 16.588

Mother More Educated 1 if mother has higher level of education than father 0.278 0.448

Mother Higher Job Status 1 if mother has higher job status level than father (defined by occupation

type)

0.494 0.500

Father Employment Stable 1 if father did not change employment 5 years before wave 1 up to wave 4 0.987 0.112

Mother Employment Stable 1 if mother did not change employment 5 years before wave 1 up to wave 4 0.991 0.095

Mother’s Hours Worked mother’s hours worked per week 26.765 10.713

Mother Works More Hours 1 if mother works more hours per week than father 0.208 0.406

Mother Does Domestic Chores 1 if mother does positive amounts of domestic chores 0.159 0.366

Mother’s Hours of Domestic Chores mother’s hours of domestic chores per week 3.931 14.576

Child Characteristics

Child Sex 1 if female 0.520 0.500

Child Ethnicity: White 1 if child’s (and hence parents’) ethnicity is white 0.750 0.433

Child Ethnicity: Mixed 1 if child’s (and hence parents’) ethnicity is mixed 0.022 0.147

Child Ethnicity: Indian 1 if child’s (and hence parents’) ethnicity is indian 0.073 0.260

Child Ethnicity: Pakistani 1 if child’s (and hence parents’) ethnicity is pakistani 0.068 0.251

Child Ethnicity: Bangladeshi 1 if child’s (and hence parents’) ethnicity is bangladeshi 0.040 0.197

Child Ethnicity: Black Caribbean 1 if child’s (and hence parents’) ethnicity is black caribbean 0.010 0.098

Child Ethnicity: Black African 1 if child’s (and hence parents’) ethnicity is black african 0.014 0.118

Child Ethnicity: Other 1 if child’s (and hence parents’) ethnicity is none of above 0.023 0.150

Religion Important to Child 1 if child states that religion is important to her way of life 0.212 0.409

Child Religion: Christian 1 if child is Christian 0.745 0.436

Child Religion: Muslim 1 if child is Muslim 0.133 0.340

Child Religion: Minority 1 if child is Sikh, Hindu, Buddhist, Jewish, Atheist 0.118 0.322

Birth Weight Child’s birth weight in kgs 3.299 0.595

SEN 1 if child has Special Education Needs 0.137 0.343

Disability 1 if child has a known disability 0.029 0.167

UK born 1 if child was born in the UK 0.956 0.206

Parental Socialisation Effort

Family Evenings: Often 1 if parents spend evenings with child, at least 3 times per month 0.943 0.232

Family Evenings: Sometimes 1 if parents spend evenings with child, once per month or less 0.028 0.165

Family Evenings: Never 1 if parents never spend evenings with child 0.029 0.169

Curfew: Always 1 if parents always set curfew for child 0.965 0.185

Curfew: Rarely 1 if parents rarely (or never) set curfew for child 0.057 0.231

Family Activities: Often 1 if go out together as family, at least 3 times per month 0.896 0.306

Family Activities: Sometimes 1 if go out together as family, once per month or less 0.083 0.277

Family Activities: Never 1 if never go out together as family 0.021 0.142

Talk about School: Often 1 if parents frequently talk to child about school day 0.473 0.499

40

Talk about School: Rarely 1 if parents rarely (or never) talk to child about school day 0.527 0.499

Geographic Characteristics

Region: North East 1 if child lives in North East 0.062 0.240

Region: North West 1 if child lives in North West 0.151 0.358

Region: Yorkshire and The Humber 1 if child lives in Yorkshire and The Humber 0.108 0.311

Region: East Midlands 1 if child lives in East Midlands 0.092 0.289

Region: West Midlands 1 if child lives in West Midlands 0.140 0.347

Region: East of England 1 if child lives in East of England 0.099 0.298

Region: London 1 if child lives in London 0.116 0.320

Region: South East 1 if child lives in South East 0.147 0.355

Region: South West 1 if child lives in South West 0.085 0.279

Area: Sparce Urban 1 if child lives in sparce urban area 0.001 0.025

Area: Sparce Town and Fringe 1 if child lives in sparce town and fringe area 0.006 0.078

Area: Sparce Village 1 if child lives in sparce village 0.005 0.070

Area: Sparce Hamlet and Isolated Dwelling 1 if child lives in sparce hamlet and isolated dwelling 0.005 0.074

Area: Urban 1 if child lives in urban area 0.800 0.400

Area: Town and Fringe 1 if child lives in town and fringe area 0.083 0.277

Area: Village 1 if child lives in village 0.077 0.266

Area: Hamlet and Isolated Dwelling 1 if child lives in hamlet and isolated dwelling 0.023 0.150

41

Table B.2: Main Results (LPM). Dependent Var: Traditional Norm

(1) (2) (3) (4)

Modern Family -0.032* -0.035* -0.042*** -0.021***

(0.017) (0.016) (0.011) (0.005)

Female -0.109*** -0.138*** -0.126***

(0.024) (0.013) (0.010)

Modern Family⇥Female 0.061** 0.065***

(0.023) (0.014)

Parents both Natural 0.050

(0.036)

Parents not Married 0.033

(0.018)

Household Size -0.003

(0.005)

Dependents 0.023***

(0.006)

Father Age 0.004***

(0.001)

Mother Age -0.003*

(0.001)

Religion Important to Parents 0.035***

(0.009)

Father Religion: Muslim -0.111***

(0.026)

Father Religion: Minority 0.397***

(0.054)

Mother Religion: Muslim 0.116***

(0.032)

Mother Religion: Minority -0.142***

(0.006)

Family Education: Below Degree -0.007

(0.015)

Family Education: GCE A Level 0.002

(0.015)

Family Education: GCSE 0.030

(0.030)

Family Education: Level 1 0.089*

(0.044)

Family Education: Other 0.142***

(0.013)

Family Education: None 0.013

(0.032)

Father Social Class: Lower Mang. 0.026

(0.029)

Father Social Class: Intermediate 0.039*

(0.020)

Father Social Class: Small Employer 0.108***

42

Table B.2: Main Results (LPM). Dependent Var: Traditional Norm

(1) (2) (3) (4)

(0.027)

Father Social Class: Lower Supervisor 0.052

(0.035)

Father Social Class: Semi-Routine 0.126***

(0.011)

Father Social Class: Routine 0.148***

(0.030)

Father Social Class: Long-Term u/e 0.210***

(0.028)

Mother Social Class: Lower Mang. 0.022

(0.013)

Mother Social Class: Intermediate 0.056***

(0.011)

Mother Social Class: Small Employer 0.090**

(0.027)

Mother Social Class: Lower Supervisor 0.033*

(0.015)

Mother Social Class: Semi-Routine 0.063**

(0.023)

Mother Social Class: Routine 0.061**

(0.019)

Mother Social Class: Long-Term u/e 0.099***

(0.028)

Family Income -0.000***

(0.000)

IDACI score -0.399**

(0.120)

IMD rank 0.004**

(0.001)

Child Born in UK -0.090***

(0.025)

Child Ethnicity: Mixed -0.084**

(0.030)

Child Ethnicity: Indian -0.279***

(0.045)

Child Ethnicity: Pakistani -0.174***

(0.041)

Child Ethnicity: Bangladeshi -0.282***

(0.032)

Child Ethnicity: Black Caribbean -0.173***

(0.006)

Child Ethnicity: Black African 0.044*

(0.020)

Child Ethnicity: Other -0.013

(0.044)

43

Table B.2: Main Results (LPM). Dependent Var: Traditional Norm

(1) (2) (3) (4)

Religion Important to Child 0.072***

(0.014)

Child Religion: Muslim 0.074

(0.054)

Child Religion: Minority -0.035

(0.058)

Child Birth Weight -0.020*

(0.010)

Child has SEN statement 0.067***

(0.018)

Child has Disability 0.049

(0.032)

Family Evenings: Often -0.018

(0.058)

Family Evenings: Sometimes -0.119

(0.077)

Curfew: Always 0.097

(0.053)

Family Activities: Often 0.013

(0.073)

Family Activities: Sometimes 0.041

(0.069)

Talk About School: Often -0.010

(0.012)

Constant 0.349*** 0.406*** 0.400*** 0.472*

(0.003) (0.012) (0.005) (0.229)

Region Fixed Effects ⇥ ⇥ ⇥ XArea Type Fixed Effects ⇥ ⇥ ⇥ XObservations 1,640 1,640 1,640 1,640

R

2 0.000 0.014 0.020 0.116

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, male, at leastone parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian, mother’s religion:Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher managerial/professional,child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SEN status, childdoes not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

44

Table B.3: Main Robustness Checks (LPM). Dependent Var: Traditional Norm

(1) (2) (3) (4) (5)

Modern Family -0.021*** -0.020*** -0.020*** -0.025*** -0.086***

(0.005) (0.005) (0.005) (0.007) (0.024)

Female -0.126*** -0.126*** -0.126*** -0.141*** -0.135***

(0.010) (0.010) (0.010) (0.009) (0.016)

Modern Family⇥Female 0.066*** 0.065*** 0.066*** 0.082*** 0.176***

(0.016) (0.015) (0.016) (0.020) (0.042)

Father Employment: Stable 0.005 0.005 0.005 0.014 0.016

(0.019) (0.019) (0.019) (0.023) (0.023)

Mother Employment: Stable 0.012 0.012 0.011 0.007 0.004

(0.031) (0.031) (0.031) (0.029) (0.020)

Mother More Educated -0.006

(0.007)

Mother Higher Job Status -0.006

(0.011)

Parents both Natural 0.050 0.050 0.049 0.031 0.036

(0.037) (0.037) (0.038) (0.039) (0.052)

Parents not Married 0.033 0.033 0.033 0.032 0.079**

(0.018) (0.018) (0.018) (0.023) (0.033)

Household Size -0.003 -0.003 -0.003 0.002 -0.016

(0.005) (0.005) (0.005) (0.006) (0.012)

Dependents 0.023*** 0.023*** 0.023*** 0.024*** 0.035***

(0.006) (0.006) (0.006) (0.006) (0.007)

Father Age 0.004*** 0.004*** 0.004*** 0.003*** 0.006***

(0.001) (0.001) (0.001) (0.001) (0.001)

Mother Age -0.003 -0.003 -0.003 -0.002 -0.000

(0.002) (0.002) (0.002) (0.002) (0.002)

Religion Important to Parents 0.034** 0.035*** 0.034** 0.049*** 0.047**

(0.010) (0.010) (0.010) (0.009) (0.015)

Father Religion: Muslim -0.112*** -0.113*** -0.111*** -0.126*** -0.493***

(0.025) (0.025) (0.027) (0.025) (0.092)

Father Religion: Minority 0.398*** 0.398*** 0.399*** 0.382*** 0.377***

(0.053) (0.053) (0.054) (0.047) (0.025)

Mother Religion: Muslim 0.116** 0.117** 0.115** 0.142*** 0.355***

(0.037) (0.037) (0.037) (0.038) (0.070)

Mother Religion: Minority -0.143*** -0.143*** -0.143*** -0.147*** -0.130***

(0.006) (0.006) (0.006) (0.003) (0.017)

Family Education: Below Degree -0.006 -0.006 -0.007 -0.008 -0.028

(0.015) (0.015) (0.015) (0.016) (0.028)

Family Education: GCE A Level 0.002 0.002 0.002 0.005 0.021

(0.015) (0.015) (0.015) (0.016) (0.018)

Family Education: GCSE 0.030 0.030 0.030 0.028 0.048**

(0.031) (0.031) (0.031) (0.024) (0.018)

Family Education: Level 1 0.088 0.088* 0.087 0.054 0.083

(0.047) (0.046) (0.048) (0.047) (0.058)

Family Education: Other 0.140*** 0.141*** 0.138*** 0.196*** 0.484***

45

Table B.3: Main Robustness Checks (LPM). Dependent Var: Traditional Norm

(1) (2) (3) (4) (5)

(0.017) (0.018) (0.018) (0.019) (0.087)

Family Education: None 0.012 0.011 0.012 -0.003 0.045**

(0.033) (0.034) (0.033) (0.024) (0.017)

Father Social Class: Lower Mang. 0.025 0.026 0.025 0.018 -0.055***

(0.029) (0.029) (0.029) (0.022) (0.015)

Father Social Class: Intermediate 0.039* 0.039* 0.038* 0.037 -0.044*

(0.020) (0.020) (0.020) (0.021) (0.022)

Father Social Class: Small Employer 0.108*** 0.108*** 0.107*** 0.104*** 0.041

(0.026) (0.026) (0.027) (0.028) (0.046)

Father Social Class: Lower Supervisor 0.052 0.052 0.050 0.054 -0.051

(0.035) (0.035) (0.034) (0.030) (0.046)

Father Social Class: Semi-Routine 0.128*** 0.128*** 0.125*** 0.136*** 0.090**

(0.012) (0.012) (0.010) (0.012) (0.026)

Father Social Class: Routine 0.148*** 0.148*** 0.145*** 0.156*** 0.120*

(0.031) (0.031) (0.030) (0.025) (0.058)

Father Social Class: Long-Term u/e 0.209*** 0.209*** 0.207*** 0.131***

(0.027) (0.027) (0.026) (0.015)

Mother Social Class: Lower Mang. 0.023 0.024 0.025 0.024 0.029**

(0.014) (0.014) (0.017) (0.013) (0.010)

Mother Social Class: Intermediate 0.056*** 0.057*** 0.059*** 0.053** 0.098***

(0.011) (0.011) (0.014) (0.017) (0.020)

Mother Social Class: Small Employer 0.092*** 0.093*** 0.094*** 0.118*** 0.113**

(0.025) (0.025) (0.022) (0.024) (0.039)

Mother Social Class: Lower Supervisor 0.034* 0.035* 0.037 0.017 0.012

(0.017) (0.017) (0.022) (0.017) (0.012)

Mother Social Class: Semi-Routine 0.064** 0.066** 0.068*** 0.068** 0.065**

(0.021) (0.021) (0.019) (0.028) (0.027)

Mother Social Class: Routine 0.061** 0.063** 0.066*** 0.072** 0.117**

(0.019) (0.019) (0.019) (0.026) (0.042)

Mother Social Class: Long-Term u/e 0.099*** 0.100*** 0.102*** 0.017

(0.027) (0.026) (0.023) (0.045)

Family Income -0.001*** -0.001*** -0.001*** -0.001* -0.000*

(0.000) (0.000) (0.000) (0.000) (0.000)

IDACI score -0.402*** -0.404*** -0.403*** -0.693*** -0.800***

(0.112) (0.111) (0.112) (0.091) (0.142)

IMD rank 0.004*** 0.004*** 0.004*** 0.007*** 0.008***

(0.001) (0.001) (0.001) (0.001) (0.001)

Child Born in UK -0.092*** -0.092*** -0.093*** -0.071** -0.061

(0.020) (0.020) (0.020) (0.027) (0.045)

Child Ethnicity: Mixed -0.082** -0.082** -0.081* -0.054*** -0.062***

(0.034) (0.034) (0.035) (0.006) (0.007)

Child Ethnicity: Indian -0.279*** -0.280*** -0.281*** -0.279*** -0.235***

(0.045) (0.044) (0.043) (0.038) (0.051)

Child Ethnicity: Pakistani -0.173*** -0.174*** -0.175*** -0.134*** 0.349***

(0.043) (0.042) (0.040) (0.031) (0.092)

46

Table B.3: Main Robustness Checks (LPM). Dependent Var: Traditional Norm

(1) (2) (3) (4) (5)

Child Ethnicity: Bangladeshi -0.283*** -0.283*** -0.285*** -0.234*** -0.178***

(0.031) (0.031) (0.028) (0.017) (0.013)

Child Ethnicity: Black Caribbean -0.175*** -0.175*** -0.176*** -0.192*** -0.172***

(0.012) (0.011) (0.013) (0.013) (0.015)

Child Ethnicity: Black African 0.048 0.046 0.046 0.094** 0.072***

(0.028) (0.027) (0.025) (0.030) (0.008)

Child Ethnicity: Other -0.013 -0.014 -0.015 -0.024 -0.080*

(0.045) (0.044) (0.043) (0.034) (0.038)

Religion Important to Child 0.072*** 0.072*** 0.072*** 0.078*** 0.039

(0.013) (0.013) (0.013) (0.013) (0.028)

Child Religion: Muslim 0.074 0.073 0.074 0.036

(0.058) (0.058) (0.059) (0.053)

Child Religion: Minority -0.036 -0.036 -0.036 -0.026 -0.053**

(0.060) (0.060) (0.060) (0.052) (0.017)

Child Birth Weight -0.020* -0.020* -0.020* -0.007 -0.014

(0.010) (0.010) (0.010) (0.010) (0.010)

Child has SEN statement 0.068** 0.068** 0.068** 0.086*** 0.074***

(0.020) (0.020) (0.020) (0.019) (0.020)

Child has Disability 0.051 0.052 0.052 0.008 0.043

(0.030) (0.030) (0.030) (0.040) (0.044)

Family Evenings: Often -0.019 -0.018 -0.019 -0.020 -0.020

(0.059) (0.060) (0.060) (0.057) (0.073)

Family Evenings: Sometimes -0.120 -0.120 -0.120 -0.137 -0.112

(0.078) (0.078) (0.078) (0.085) (0.090)

Curfew: Always 0.097 0.097 0.097 0.106 0.116

(0.054) (0.054) (0.053) (0.057) (0.104)

Family Activities: Often 0.013 0.013 0.012 0.010 0.111

(0.072) (0.072) (0.070) (0.076) (0.100)

Family Activities: Sometimes 0.040 0.040 0.039 0.053 0.120

(0.067) (0.067) (0.066) (0.079) (0.093)

Talk About School: Often -0.010 -0.010 -0.010 -0.010 -0.005

(0.012) (0.012) (0.012) (0.012) (0.019)

Constant 0.459 0.452 0.460 -0.160 -0.573*

(0.262) (0.269) (0.262) (0.199) (0.292)

Region Fixed Effects X X X X XArea Type Fixed Effects X X X X XObservations 1,640 1,640 1,640 1,486 1,125

R

2 0.116 0.116 0.116 0.122 0.141

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, male, father’semployment unstable, mother’s employment unstable, father has same or higher education, father has same or higher job status, at least one parentis not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian, mother’s religion: Christian,family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher managerial/professional, child not bornin UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SEN status, child does not havedisabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

47

Table B.4: Other Outcomes. Dependent Var: Traditional Norm

(1) (2)

Believe High Wage Important Want to Study Science

Modern Family 0.024* 0.106**

(0.012) (0.031)

Female -0.046* -0.056***

(0.022) (0.007)

Modern Family⇥Female -0.064*** -0.196***

(0.011) (0.053)

Father Employment: Stable -0.005 -0.099***

(0.010) (0.008)

Mother Employment: Stable -0.019 -0.062

(0.012) (0.050)

Parents both Natural 0.049 -0.035

(0.040) (0.062)

Parents not Married 0.083** -0.116

(0.029) (0.105)

Household Size -0.009 0.033***

(0.011) (0.007)

Dependents -0.003 -0.048***

(0.006) (0.005)

Father Age -0.001 -0.006**

(0.001) (0.002)

Mother Age -0.001 0.001

(0.002) (0.001)

Religion Important to Parents -0.009 0.022

(0.024) (0.020)

Father Religion: Muslim 0.276*** -0.863***

(0.019) (0.087)

Father Religion: Minority -0.024 -0.314**

(0.063) (0.118)

Mother Religion: Muslim -0.007 0.361***

(0.040) (0.094)

Mother Religion: Minority -0.036 0.047

(0.036) (0.066)

Family Education: Below Degree -0.106*** -0.052

(0.024) (0.069)

Family Education: GCE A Level 0.002 -0.111**

(0.022) (0.045)

Family Education: GCSE -0.001 -0.147**

(0.009) (0.048)

Family Education: Level 1 0.047* -0.287***

(0.022) (0.078)

Family Education: Other 0.056 -0.143

(0.083) (0.085)

Family Education: None -0.064*** -0.027

(0.009) (0.065)

48

Table B.4: Other Outcomes. Dependent Var: Traditional Norm

(1) (2)

Believe High Wage Important Want to Study Science

Father Social Class: Lower Mang. -0.059 -0.027

(0.047) (0.040)

Father Social Class: Intermediate 0.029 -0.100

(0.026) (0.120)

Father Social Class: Small Employer 0.061 -0.027

(0.047) (0.089)

Father Social Class: Lower Supervisor -0.067 0.027

(0.062) (0.029)

Father Social Class: Semi-Routine 0.025 -0.029

(0.028) (0.091)

Father Social Class: Routine 0.030 0.065

(0.025) (0.105)

Father Social Class: Long-Term u/e -0.056* -0.004

(0.028) (0.066)

Mother Social Class: Lower Mang. 0.018 -0.118**

(0.012) (0.038)

Mother Social Class: Intermediate 0.129*** -0.228***

(0.017) (0.041)

Mother Social Class: Small Employer 0.120 -0.062**

(0.093) (0.026)

Mother Social Class: Lower Supervisor 0.069* -0.192***

(0.032) (0.035)

Mother Social Class: Semi-Routine 0.056** -0.244**

(0.018) (0.082)

Mother Social Class: Routine 0.098*** -0.178***

(0.023) (0.030)

Mother Social Class: Long-Term u/e 0.081*** -0.375***

(0.018) (0.024)

Family Income -0.000 -0.002

(0.000) (0.001)

IDACI score 0.127* -0.140

(0.056) (0.216)

IMD rank 0.000 0.004

(0.000) (0.003)

Child born in UK 0.021 0.013

(0.027) (0.097)

Child Ethnicity: Mixed 0.136*** -0.028*

(0.004) (0.013)

Child Ethnicity: Indian 0.225*** 0.449***

(0.044) (0.122)

Child Ethnicity: Pakistani 0.117*** 0.332**

(0.026) (0.113)

Child Ethnicity: Bangladeshi 0.227*** 0.491***

(0.019) (0.116)

49

Table B.4: Other Outcomes. Dependent Var: Traditional Norm

(1) (2)

Believe High Wage Important Want to Study Science

Child Ethnicity: Black Caribbean 0.211*** -0.028

(0.014) (0.048)

Child Ethnicity: Black African 0.280*** 0.328**

(0.031) (0.127)

Child Ethnicity: Other 0.060*** 0.210

(0.016) (0.223)

Religion Important to Child -0.036*** -0.069

(0.005) (0.040)

Child Religion: Muslim -0.236*** 0.447***

(0.034) (0.124)

Child Religion: Minority 0.092*** -0.106*

(0.019) (0.054)

Child Birth Weight 0.028* -0.057*

(0.014) (0.024)

Child has SEN statement -0.016 -0.018

(0.022) (0.048)

Child has Disability -0.086*** -0.178**

(0.022) (0.060)

Family Activities: Often -0.143*** 0.070

(0.012) (0.085)

Family Activities: Sometimes -0.051 -0.189

(0.027) (0.164)

Curfew: Always 0.026 0.158***

(0.039) (0.043)

Family Evenings: Often -0.176*** 0.066

(0.013) (0.086)

Family Evenings: Sometimes -0.158*** 0.043

(0.016) (0.028)

Talk About School: Often 0.011 -0.021

(0.022) (0.029)

Constant 0.487*** 0.186

(0.129) (0.207)

Region Fixed Effects X XArea Type Fixed Effects X XObservations 1,486 1,486

Pseudo-R2 0.125 0.042

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, male, at leastone parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian, mother’s religion:Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher managerial/professional,child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SEN status, childdoes not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

50

Table B.5: Sub-Group Analysis (LPM). Dependent var: Traditional Norm

(1) (2) (3) (4) (5)

Father not Father Family Income Family Income Family Income Above Median

Disabled Disabled Above Median Below Median and Father not Disabled

Modern Family -0.051*** 0.125 -0.085*** 0.036 -0.074***

(0.011) (0.081) (0.015) (0.032) (0.013)

Female -0.147*** -0.167 -0.148*** -0.137** -0.157***

(0.010) (0.102) (0.028) (0.054) (0.039)

Modern Family⇥Female 0.121*** -0.242* 0.149*** 0.039 0.138***

(0.019) (0.102) (0.013) (0.071) (0.011)

Father Employment: Stable 0.022 -0.220*** 0.044* 0.018 0.053*

(0.029) (0.053) (0.019) (0.036) (0.025)

Mother Employment: Stable 0.005 0.015 -0.005 -0.002 -0.007

(0.038) (0.081) (0.028) (0.051) (0.037)

Parents both Natural 0.014 0.177 0.066 0.029 0.054

(0.036) (0.179) (0.068) (0.043) (0.077)

Parents not Married 0.020 0.300 0.099*** 0.043 0.100***

(0.021) (0.235) (0.021) (0.058) (0.023)

Household Size 0.008 -0.213* -0.028*** 0.033** -0.018*

(0.006) (0.096) (0.008) (0.011) (0.008)

Dependents 0.020** 0.159*** 0.048*** -0.001 0.045***

(0.007) (0.026) (0.011) (0.014) (0.008)

Father Age 0.003** 0.006* 0.003*** 0.001 0.004**

(0.001) (0.003) (0.001) (0.002) (0.001)

Mother Age -0.002 -0.001 0.001 -0.004 0.000

(0.003) (0.004) (0.003) (0.003) (0.004)

Religion Important to Parents 0.059*** -0.144* 0.059* 0.026 0.064**

(0.010) (0.070) (0.027) (0.037) (0.026)

Father Religion: Muslim -0.302*** 0.691** -0.515*** -0.001 -0.547***

(0.033) (0.211) (0.010) (0.050) (0.050)

Father Religion: Minority 0.398*** 1.497*** 0.503*** 0.259*** 0.515***

(0.052) (0.111) (0.065) (0.067) (0.065)

Mother Religion: Muslim 0.411*** -0.537** -0.044 0.005 0.165

(0.039) (0.153) (0.276) (0.077) (0.606)

Mother Religion: Minority -0.134*** -0.582*** -0.116*** -0.149*** -0.110***

(0.006) (0.061) (0.008) (0.020) (0.013)

Family Education: Below Degree -0.015 -0.097 -0.011 0.048 -0.026

(0.021) (0.125) (0.030) (0.036) (0.028)

Family Education: GCE A Level 0.008 0.262 0.012 0.037 0.010

(0.017) (0.145) (0.023) (0.027) (0.020)

Family Education: GCSE 0.015 0.130* -0.007 0.103** -0.023

(0.016) (0.057) (0.031) (0.042) (0.026)

Family Education: Level 1 0.020 0.374** -0.015 0.109** -0.101

(0.034) (0.095) (0.235) (0.037) (0.155)

Family Education: Other 0.173 0.273 -0.389 0.293*** -0.318

(0.099) (0.209) (0.682) (0.074) (0.732)

Family Education: None -0.027 0.238** -0.000 0.119*** -0.073*

51

Table B.5: Sub-Group Analysis (LPM). Dependent var: Traditional Norm

(1) (2) (3) (4) (5)

Father not Father Family Income Family Income Family Income Above Median

Disabled Disabled Above Median Below Median and Father not Disabled

(0.020) (0.090) (0.080) (0.029) (0.038)

Father Social Class: Lower Mang. 0.018 -0.175 -0.008 0.071* -0.009

(0.024) (0.101) (0.020) (0.034) (0.022)

Father Social Class: Intermediate 0.054* -0.517** 0.004 0.075 0.023

(0.028) (0.129) (0.027) (0.047) (0.033)

Father Social Class: Small Employer 0.168*** -0.717** 0.055* 0.118* 0.111***

(0.025) (0.201) (0.029) (0.053) (0.018)

Father Social Class: Lower Supervisor 0.053 -0.468** -0.086 0.086* -0.084

(0.043) (0.139) (0.054) (0.038) (0.069)

Father Social Class: Semi-Routine 0.148*** -0.225** 0.109*** 0.128** 0.124***

(0.023) (0.078) (0.011) (0.042) (0.016)

Father Social Class: Routine 0.204*** -0.481** 0.360*** 0.089** 0.411***

(0.030) (0.176) (0.082) (0.032) (0.061)

Father Social Class: Long-Term u/e 0.170*** -0.467** 0.331*** 0.016 0.350***

(0.011) (0.139) (0.082) (0.025) (0.056)

Mother Social Class: Lower Mang. 0.015 0.231* 0.031*** 0.043 0.017***

(0.008) (0.109) (0.007) (0.057) (0.004)

Mother Social Class: Intermediate 0.043** 0.041 0.075** 0.064 0.069**

(0.014) (0.177) (0.023) (0.093) (0.021)

Mother Social Class: Small Employer 0.087** 0.164 0.078* 0.154*** 0.057

(0.033) (0.219) (0.037) (0.022) (0.041)

Mother Social Class: Lower Supervisor 0.007 0.243*** 0.004 0.043 -0.001

(0.016) (0.038) (0.017) (0.052) (0.027)

Mother Social Class: Semi-Routine 0.051 0.081 0.056 0.076 0.048

(0.031) (0.092) (0.036) (0.054) (0.044)

Mother Social Class: Routine 0.073** 0.128 0.049** 0.106* 0.017

(0.024) (0.107) (0.020) (0.050) (0.010)

Mother Social Class: Long-Term u/e 0.013 0.569** 0.645*** -0.061 0.634***

(0.055) (0.217) (0.075) (0.034) (0.086)

Family Income -0.000 0.001 -0.001*** -0.002 -0.001***

(0.000) (0.002) (0.000) (0.002) (0.000)

IDACI score -0.641*** -2.583*** -0.831*** -0.520*** -0.825***

(0.095) (0.491) (0.157) (0.078) (0.196)

IMD rank 0.006*** 0.031*** 0.010*** 0.004*** 0.010***

(0.001) (0.005) (0.001) (0.001) (0.002)

Child Born in UK -0.075*** 0.031 -0.072* -0.106** -0.085***

(0.016) (0.340) (0.038) (0.036) (0.022)

Child Ethnicity: Mixed -0.085*** 1.057*** -0.206*** 0.039 -0.198***

(0.009) (0.092) (0.011) (0.020) (0.015)

Child Ethnicity: Indian -0.289*** -0.698*** -0.324*** -0.286*** -0.340***

(0.041) (0.131) (0.035) (0.018) (0.057)

Child Ethnicity: Pakistani -0.169*** 0.091 0.164 -0.128*** 0.103

(0.036) (0.286) (0.096) (0.024) (0.135)

52

Table B.5: Sub-Group Analysis (LPM). Dependent var: Traditional Norm

(1) (2) (3) (4) (5)

Father not Father Family Income Family Income Family Income Above Median

Disabled Disabled Above Median Below Median and Father not Disabled

Child Ethnicity: Bangladeshi -0.257*** -0.653* -0.481*** -0.208*** -0.458***

(0.013) (0.269) (0.032) (0.029) (0.045)

Child Ethnicity: Black Caribbean -0.165*** -1.005*** -0.206*** -0.169*** -0.166***

(0.016) (0.138) (0.019) (0.033) (0.021)

Child Ethnicity: Black African 0.073 0.589*** 0.056** 0.049** 0.050

(0.039) (0.125) (0.019) (0.019) (0.040)

Child Ethnicity: Other -0.055 -0.185 -0.108*** 0.098* -0.118***

(0.033) (0.141) (0.013) (0.045) (0.021)

Religion Important to Child 0.086*** 0.245*** 0.008 0.173*** 0.014

(0.017) (0.052) (0.013) (0.028) (0.011)

Child Religion: Muslim -0.048 -0.178 0.697* -0.002 0.524

(0.059) (0.282) (0.317) (0.040) (0.663)

Child Religion: Minority -0.045 -0.164** -0.103 0.058 -0.102

(0.066) (0.046) (0.071) (0.050) (0.082)

Child Birth Weight 0.000 -0.133** -0.003 -0.021 0.003

(0.012) (0.045) (0.009) (0.020) (0.009)

Child has SEN statement 0.089*** 0.203 0.051 0.083** 0.066

(0.015) (0.213) (0.030) (0.026) (0.037)

Child has Disability 0.019 -0.535* 0.055 0.019 0.086**

(0.040) (0.237) (0.030) (0.068) (0.025)

Family Evenings: Often 0.069 -0.517*** -0.065 0.029 -0.010

(0.056) (0.045) (0.039) (0.064) (0.059)

Family Evenings: Sometimes -0.056 -0.708*** -0.070 -0.203** -0.012

(0.083) (0.137) (0.060) (0.065) (0.084)

Curfew: Always 0.089 0.009 0.070* 0.220* 0.080*

(0.061) (0.192) (0.034) (0.107) (0.041)

Family Activities: Often 0.025 0.313*** -0.177 0.159*** -0.184**

(0.056) (0.077) (0.107) (0.035) (0.071)

Family Activities: Sometimes 0.081 0.296** -0.201* 0.229*** -0.164*

(0.062) (0.110) (0.087) (0.033) (0.074)

Talk About School: Often 0.003 -0.213*** -0.026 -0.008 -0.019

(0.015) (0.029) (0.019) (0.020) (0.015)

Constant -0.603** 3.482* 1.130*** -0.935*** 0.169

(0.202) (1.409) (0.126) (0.175) (0.166)

Region Fixed Effects X X X X XArea Type Fixed Effects X X X X XObservations 1,353 132 820 665 767

R

2 0.127 0.685 0.186 0.191 0.189

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, male, at leastone parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian, mother’s religion:Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher managerial/professional,child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SEN status, childdoes not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

53

Table B.6: Mothers’ Second Shift

(1) (2)

Mother Does Second Shift Duration of Mother’s Second Shift

Modern Family 0.023 -0.361

(0.022) (0.558)

Female -0.003 0.048

(0.012) (0.239)

Modern Family⇥Female -0.013 0.054

(0.015) (0.324)

Father Employment: Stable 0.009 -0.940**

(0.013) (0.381)

Mother Employment: Stable 0.025 0.269***

(0.013) (0.066)

Parents both Natural -0.022 -0.376

(0.034) (0.480)

Parents not Married 0.093*** 2.598***

(0.018) (0.506)

Household Size -0.010 -0.278**

(0.007) (0.111)

Dependents 0.026*** 0.116

(0.006) (0.105)

Father Age -0.001 -0.054*

(0.001) (0.023)

Mother Age 0.007** 0.007

(0.002) (0.076)

Religion Important to Parents 0.020 0.517*

(0.011) (0.231)

Father Religion: Muslim 0.062* -0.639

(0.027) (1.311)

Father Religion: Minority 0.071 -2.616*

(0.069) (1.244)

Mother Religion: Muslim -0.045 -4.397***

(0.027) (0.150)

Mother Religion: Minority 0.017 -1.304**

(0.018) (0.411)

Family Education: Below Degree 0.027** -0.155

(0.009) (0.703)

Family Education: GCE A Level 0.001 -0.416

(0.015) (0.286)

Family Education: GCSE 0.026 -0.579

(0.017) (0.627)

Family Education: Level 1 0.133*** 6.596***

(0.027) (0.632)

Family Education: Other -0.065 -1.647

(0.048) (1.190)

Family Education: None -0.048*** -2.264**

(0.013) (0.697)

54

Table B.6: Mothers’ Second Shift

(1) (2)

Mother Does Second Shift Duration of Mother’s Second Shift

Father Social Class: Lower Mang. 0.072*** 1.102***

(0.020) (0.245)

Father Social Class: Intermediate 0.081*** 1.443**

(0.019) (0.454)

Father Social Class: Small Employer 0.050 1.850**

(0.029) (0.547)

Father Social Class: Lower Supervisor 0.049*** 0.944**

(0.009) (0.328)

Father Social Class: Semi-Routine 0.038** 1.682***

(0.015) (0.230)

Father Social Class: Routine 0.041 1.392

(0.031) (0.810)

Father Social Class: Long-Term u/e -0.097*** -0.540

(0.022) (0.394)

Mother Social Class: Lower Mang. 0.004 -0.511

(0.005) (0.791)

Mother Social Class: Intermediate 0.022 -0.863**

(0.023) (0.308)

Mother Social Class: Small Employer 0.035 0.622

(0.036) (1.279)

Mother Social Class: Lower Supervisor -0.006 -1.181

(0.035) (1.067)

Mother Social Class: Semi-Routine -0.017 -1.644**

(0.013) (0.613)

Mother Social Class: Routine -0.023** -0.607

(0.008) (0.515)

Mother Social Class: Long-Term u/e -0.156*** -4.408***

(0.012) (0.574)

Family Income -0.001** -0.022***

(0.000) (0.003)

IDACI score 0.052 0.491

(0.055) (1.721)

IMD rank 0.001 0.014

(0.001) (0.021)

Child Born in UK -0.018 0.406

(0.024) (0.946)

Child Ethnicity: Mixed -0.109*** -1.776***

(0.010) (0.359)

Child Ethnicity: Indian -0.112*** 2.330**

(0.024) (0.772)

Child Ethnicity: Pakistani -0.062** 3.709***

(0.021) (0.822)

Child Ethnicity: Bangladeshi -0.067** 2.274*

(0.025) (1.200)

55

Table B.6: Mothers’ Second Shift

(1) (2)

Mother Does Second Shift Duration of Mother’s Second Shift

Child Ethnicity: Black Caribbean -0.085*** -2.387***

(0.012) (0.426)

Child Ethnicity: Black African -0.111*** -1.668**

(0.015) (0.487)

Child Ethnicity: Other -0.005 7.666***

(0.032) (1.684)

Religion Important to Child 0.015 1.397

(0.019) (1.163)

Child Religion: Muslim 0.046 1.348

(0.030) (1.013)

Child Religion: Minority 0.012 1.248

(0.057) (0.931)

Child Birth Weight -0.013* -0.572***

(0.006) (0.056)

Child has SEN statement 0.007 1.463**

(0.034) (0.598)

Child has Disability 0.074** 7.855***

(0.029) (0.741)

Family Evenings: Often 0.164*** 4.089***

(0.005) (0.475)

Family Evenings: Sometimes 0.175*** 4.436***

(0.018) (0.237)

Curfew: Always -0.036 -2.764**

(0.041) (0.990)

Family Activities: Often -0.082** 1.121*

(0.033) (0.541)

Family Activities: Sometimes -0.025** 1.668**

(0.009) (0.620)

Talk About School: Often 0.003 0.298

(0.005) (0.176)

Constant -0.375*** 3.739

(0.091) (3.765)

Region Fixed Effects X XArea Type Fixed Effects X XObservations 1,486 1,486

R

2/Pseudo-R2 0.090 0.105

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, male, at leastone parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian, mother’s religion:Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher managerial/professional,child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SEN status, childdoes not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

Table B.7: Incorporating Mother’s Hours Worked. Dependent Var: Traditional Norm

(1) (2)

Modern Family -0.041** -0.017*

56

Table B.7: Incorporating Mother’s Hours Worked. Dependent Var: Traditional Norm

(1) (2)

(0.016) (0.007)

Female -0.111*** -0.166***

(0.020) (0.012)

Modern Family⇥Female 0.107*** 0.062**

(0.018) (0.018)

Mother’s Hours Worked 0.001

(0.001)

Mother’s Hours Worked⇥Female -0.002**

(0.001)

Mother Works More Hours -0.090***

(0.022)

Mother Works More Hours⇥Female 0.159**

(0.046)

Father Employment: Stable 0.008 0.013

(0.015) (0.022)

Mother Employment: Stable -0.004 0.009

(0.030) (0.029)

Parents both Natural 0.037 0.034

(0.028) (0.039)

Parents not Married 0.049* 0.029

(0.023) (0.024)

Household Size 0.015** 0.002

(0.006) (0.007)

Dependents 0.029** 0.026***

(0.011) (0.006)

Father Age 0.007*** 0.003***

(0.001) (0.001)

Mother Age -0.002 -0.002

(0.002) (0.002)

Religion Important to Parents 0.040** 0.052***

(0.015) (0.009)

Father Religion: Muslim -0.034 -0.146***

(0.036) (0.020)

Father Religion: Minority 0.415*** 0.373***

(0.041) (0.045)

Mother Religion: Muslim -0.093 0.150***

(0.052) (0.040)

Mother Religion: Minority -0.143*** -0.141***

(0.019) (0.004)

Family Education: Below Degree -0.018 -0.008

(0.017) (0.015)

Family Education: GCE A Level 0.014 0.007

(0.009) (0.015)

Family Education: GCSE 0.039** 0.029

(0.015) (0.022)

57

Table B.7: Incorporating Mother’s Hours Worked. Dependent Var: Traditional Norm

(1) (2)

Family Education: Level 1 0.074 0.055

(0.041) (0.044)

Family Education: Other 0.303** 0.190***

(0.103) (0.017)

Family Education: None 0.010 -0.013

(0.013) (0.023)

Father Social Class: Lower Mang. 0.016 0.023

(0.012) (0.023)

Father Social Class: Intermediate 0.024 0.036

(0.021) (0.022)

Father Social Class: Small Employer -0.111 0.109***

(0.154) (0.026)

Father Social Class: Lower Supervisor 0.037 0.051

(0.026) (0.030)

Father Social Class: Semi-Routine 0.123*** 0.136***

(0.024) (0.014)

Father Social Class: Routine 0.151*** 0.159***

(0.024) (0.027)

Father Social Class: Long-Term u/e 0.146*** 0.139***

(0.031) (0.012)

Mother Social Class: Lower Mang. 0.036 0.026*

(0.021) (0.013)

Mother Social Class: Intermediate 0.055* 0.054**

(0.024) (0.016)

Mother Social Class: Small Employer 0.148*** 0.118***

(0.036) (0.026)

Mother Social Class: Lower Supervisor 0.012 0.020

(0.028) (0.017)

Mother Social Class: Semi-Routine 0.065 0.069*

(0.041) (0.030)

Mother Social Class: Routine 0.059 0.075**

(0.038) (0.026)

Mother Social Class: Long-Term u/e 0.088*** 0.028

(0.015) (0.047)

Family Income -0.001*** -0.000*

(0.000) (0.000)

IDACI score -0.731*** -0.684***

(0.142) (0.099)

IMD rank 0.008*** 0.007***

(0.002) (0.001)

Child Born in UK -0.054 -0.076**

(0.040) (0.027)

Child Ethnicity: Mixed -0.028** -0.059***

(0.009) (0.005)

Child Ethnicity: Indian -0.320*** -0.279***

58

Table B.7: Incorporating Mother’s Hours Worked. Dependent Var: Traditional Norm

(1) (2)

(0.023) (0.038)

Child Ethnicity: Pakistani 0.047 -0.131***

(0.063) (0.032)

Child Ethnicity: Bangladeshi -0.368*** -0.244***

(0.007) (0.017)

Child Ethnicity: Black Caribbean -0.186*** -0.206***

(0.011) (0.016)

Child Ethnicity: Black African 0.082*** 0.114**

(0.010) (0.035)

Child Ethnicity: Other -0.074*** -0.032

(0.021) (0.034)

Religion Important to Child 0.072*** 0.077***

(0.010) (0.013)

Child Religion: Muslim 0.186*** 0.049

(0.050) (0.052)

Child Religion: Minority -0.032 -0.022

(0.046) (0.051)

Child Birth Weight -0.019* -0.010

(0.009) (0.012)

Child has SEN statement 0.072*** 0.082***

(0.019) (0.019)

Child has Disability 0.009 0.002

(0.040) (0.040)

Family Evenings: Often -0.050 -0.020

(0.059) (0.057)

Family Evenings: Sometimes -0.168 -0.133

(0.094) (0.085)

Curfew: Always 0.153*** 0.105

(0.034) (0.058)

Family Activities: Often 0.040 0.015

(0.098) (0.076)

Family Activities: Sometimes 0.028 0.053

(0.104) (0.081)

Talk About School: Often -0.008 -0.009

(0.010) (0.012)

Constant -0.576* -0.112

(0.297) (0.191)

Region Fixed Effects X XArea Type Fixed Effects X XObservations 1,320 1,485

R

2 0.149 0.126

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, mother worksless hours than father, male, at least one parent is not child’s biological parent, parents are married, religion not important to parents, father’sreligion: Christian, mother’s religion: Christian, family education: degree, father’s social class: higher managerial/professional, mother’s socialclass: higher managerial/professional, child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian,child does not have SEN status, child does not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school:rarely.

59

Table B.8: Preferences for Conformity

(1) (4)

Parents Chose GCSEs Argue Often with Parents

Modern Family 0.082** 0.027

(0.028) (0.031)

Female 0.002 0.020

(0.021) (0.015)

Modern Family⇥Female -0.155** 0.037***

(0.047) (0.008)

Father Employment: Stable -0.025 -0.001

(0.023) (0.020)

Mother Employment: Stable -0.008 -0.034

(0.013) (0.021)

Parents both Natural 0.051 0.044**

(0.034) (0.018)

Parents not Married 0.003 0.003

(0.043) (0.045)

Household Size 0.016** -0.036***

(0.005) (0.009)

Dependents -0.014 0.048***

(0.014) (0.007)

Father Age -0.004*** -0.003

(0.001) (0.002)

Mother Age -0.000 -0.006*

(0.003) (0.003)

Religion Important to Parents 0.008 0.019

(0.036) (0.020)

Father Religion: Muslim -0.028 0.208***

(0.022) (0.055)

Father Religion: Minority -0.165*** 0.103**

(0.009) (0.039)

Mother Religion: Muslim 0.137*** 0.213***

(0.032) (0.045)

Mother Religion: Minority -0.116*** -0.035

(0.015) (0.052)

Family Education: Below Degree 0.007 -0.035

(0.013) (0.040)

Family Education: GCE A Level -0.048*** 0.017

(0.009) (0.016)

Family Education: GCSE -0.057 -0.064

(0.036) (0.044)

Family Education: Level 1 -0.095* -0.013

(0.040) (0.029)

Family Education: Other -0.012 0.112

(0.068) (0.077)

Family Education: None -0.041 -0.083*

(0.026) (0.039)

60

Table B.8: Preferences for Conformity

(1) (4)

Parents Chose GCSEs Argue Often with Parents

Father Social Class: Lower Mang. -0.018 0.074

(0.022) (0.058)

Father Social Class: Intermediate -0.002 0.076*

(0.019) (0.038)

Father Social Class: Small Employer 0.010 -0.039

(0.028) (0.038)

Father Social Class: Lower Supervisor 0.032 0.071

(0.035) (0.040)

Father Social Class: Semi-Routine -0.086*** 0.100**

(0.016) (0.034)

Father Social Class: Routine -0.114*** 0.084**

(0.027) (0.025)

Father Social Class: Long-Term u/e -0.032 -0.029

(0.026) (0.020)

Mother Social Class: Lower Mang. -0.014 -0.074*

(0.019) (0.038)

Mother Social Class: Intermediate -0.032 0.027

(0.063) (0.044)

Mother Social Class: Small Employer -0.044** -0.024

(0.013) (0.080)

Mother Social Class: Lower Supervisor -0.001 -0.034

(0.024) (0.029)

Mother Social Class: Semi-Routine 0.008 -0.030

(0.046) (0.047)

Mother Social Class: Routine 0.007 0.049*

(0.020) (0.023)

Mother Social Class: Long-Term u/e 0.001 -0.132*

(0.027) (0.056)

Family Income 0.000 -0.001*

(0.000) (0.000)

IDACI score -0.231** 0.200

(0.082) (0.138)

IMD rank 0.002 -0.003**

(0.001) (0.001)

Child Born in UK -0.059*** -0.038

(0.010) (0.029)

Child Ethnicity: Mixed 0.149*** 0.065***

(0.003) (0.009)

Child Ethnicity: Indian 0.381*** -0.148***

(0.054) (0.021)

Child Ethnicity: Pakistani 0.053 -0.275***

(0.035) (0.022)

Child Ethnicity: Bangladeshi 0.229*** -0.580***

(0.033) (0.020)

61

Table B.8: Preferences for Conformity

(1) (4)

Parents Chose GCSEs Argue Often with Parents

Child Ethnicity: Black Caribbean 0.070** -0.066***

(0.021) (0.018)

Child Ethnicity: Black African 0.157*** 0.091***

(0.026) (0.020)

Child Ethnicity: Other 0.152*** -0.099***

(0.032) (0.027)

Religion Important to Child 0.017 -0.084***

(0.012) (0.017)

Child Religion: Muslim -0.271*** -0.214*

(0.061) (0.100)

Child Religion: Minority -0.027 -0.036

(0.024) (0.025)

Child Birth Weight -0.060*** -0.015

(0.014) (0.012)

Child has SEN statement 0.019 0.002

(0.010) (0.032)

Child has Disability -0.016 -0.019

(0.070) (0.079)

Family Evenings: Often -0.039 -0.099***

(0.034) (0.019)

Family Evenings: Sometimes 0.072 -0.117*

(0.087) (0.057)

Curfew: Always -0.046 0.049

(0.026) (0.073)

Family Activities: Often -0.060 0.111**

(0.052) (0.033)

Family Activities: Sometimes -0.097* 0.085

(0.047) (0.062)

Talk About School: Often 0.043*** -0.024

(0.008) (0.012)

Constant 0.753*** 1.643***

(0.154) (0.173)

Region Fixed Effects X XArea Type Fixed Effects X XObservations 1,486 1,486

R

2/Pseudo-R2 0.096 0.085

Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditional family, male, at leastone parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian, mother’s religion:Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher managerial/professional,child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SEN status, childdoes not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

62

Table B.9: RD Results on Full Sample. Dependent Var: Parents Chose GCSE Subjects

(1) (2) (3) (4) (5) (6)

Modern Family Treatment Effect -0.313*** -0.356*** -0.334*** -0.299*** -0.328*** -0.378***

(0.030) (0.051) (0.075) (0.048) (0.093) (0.097)

Mother’s Share 0.202*** 0.437 1.057 0.164** -0.151 1.347

(0.040) (0.544) (0.836) (0.060) (0.889) (0.978)

Modern Family⇥Mother’s Share 0.665*** 0.778 -1.449 0.638*** 1.936 -0.041

(0.075) (1.039) (2.453) (0.178) (1.238) (2.758)

Mother’s Share2 0.422 3.317 -0.589 6.567

(0.973) (4.021) (1.582) (3.812)

Modern Family⇥Mother’s Share2 -1.070 4.855 -1.281 -5.742

(1.880) (13.479) (2.292) (15.627)

Mother’s Share3 3.583 8.990*

(5.275) (4.848)

Modern Family⇥Mother’s Share3 -15.240 -12.662

(18.229) (20.890)

Father’s Employment: Stable -0.003 -0.006 -0.009

(0.045) (0.043) (0.042)

Mother’s Employment: Stable -0.001 -0.001 -0.001

(0.037) (0.037) (0.037)

Parents both Natural 0.032 0.030 0.028

(0.092) (0.093) (0.095)

Parents not Married -0.016 -0.019 -0.010

(0.106) (0.110) (0.108)

Household Size 0.046 0.043 0.041

(0.025) (0.024) (0.025)

Dependents -0.055*** -0.052*** -0.049***

(0.015) (0.013) (0.013)

Father Age -0.007 -0.007 -0.007

(0.006) (0.006) (0.006)

Mother Age -0.002 -0.002 -0.002

(0.004) (0.004) (0.004)

Religion Important to Parents 0.037 0.037 0.038

(0.038) (0.038) (0.039)

Father Religion: Muslim 0.346* 0.345* 0.339

(0.186) (0.186) (0.189)

Father Religion: Minority -0.219 -0.213 -0.220

(0.212) (0.205) (0.198)

Mother Religion: Muslim 0.010 0.029 0.026

(0.343) (0.347) (0.353)

Mother Religion: Minority -0.140** -0.141** -0.142**

(0.055) (0.055) (0.056)

Family Education: Below Degree -0.034 -0.039 -0.039

(0.055) (0.055) (0.055)

Family Education: GCE A Level -0.105* -0.102* -0.101

(0.056) (0.055) (0.056)

Family Education: GCSE -0.128 -0.130 -0.128

63

Table B.9: RD Results on Full Sample. Dependent Var: Parents Chose GCSE Subjects

(1) (2) (3) (4) (5) (6)

(0.087) (0.087) (0.089)

Family Education: Level 1 -0.087 -0.088 -0.078

(0.136) (0.133) (0.136)

Family Education: Other 0.057 0.053 0.053

(0.143) (0.143) (0.144)

Family Education: None -0.124 -0.122 -0.125

(0.143) (0.141) (0.138)

Father Social Class: Lower Mang. -0.029 -0.025 -0.022

(0.061) (0.061) (0.062)

Father Social Class: Intermediate 0.055 0.054 0.052

(0.060) (0.055) (0.057)

Father Social Class: Small Employer 0.008 0.018 0.008

(0.085) (0.088) (0.089)

Father Social Class: Lower Supervisor 0.055 0.058 0.058

(0.077) (0.077) (0.077)

Father Social Class: Semi-Routine -0.006 -0.005 -0.008

(0.044) (0.043) (0.045)

Father Social Class: Routine -0.102* -0.094* -0.097*

(0.047) (0.045) (0.050)

Father Social Class: Long-Term u/e 0.164* 0.186* 0.200*

(0.089) (0.087) (0.099)

Mother Social Class: Lower Mang. 0.002 0.003 0.007

(0.055) (0.060) (0.059)

Mother Social Class: Intermediate 0.009 0.008 0.008

(0.070) (0.073) (0.073)

Mother Social Class: Small Employer -0.051 -0.057 -0.061

(0.110) (0.111) (0.110)

Mother Social Class: Lower Supervisor 0.035 0.032 0.032

(0.040) (0.043) (0.042)

Mother Social Class: Semi-Routine 0.044 0.043 0.047

(0.077) (0.084) (0.084)

Mother Social Class: Routine -0.071 -0.075 -0.071

(0.068) (0.069) (0.071)

Mother Social Class: Long-Term u/e 0.045 0.059 0.074

(0.092) (0.096) (0.098)

Family Income -0.000 -0.000 -0.000

(0.000) (0.000) (0.000)

IDACI score -0.369 -0.380 -0.362

(0.302) (0.298) (0.294)

IMD rank 0.003 0.003 0.003

(0.003) (0.003) (0.003)

Child born in UK -0.185* -0.185* -0.183

(0.097) (0.097) (0.102)

Child Ethnicity: Mixed 0.150 0.161 0.164

(0.098) (0.103) (0.104)

64

Table B.9: RD Results on Full Sample. Dependent Var: Parents Chose GCSE Subjects

(1) (2) (3) (4) (5) (6)

Child Ethnicity: Indian 0.483** 0.490** 0.497**

(0.176) (0.159) (0.153)

Child Ethnicity: Pakistani 0.090 0.087 0.098

(0.070) (0.068) (0.073)

Child Ethnicity: Bangladeshi 0.164 0.165 0.167

(0.194) (0.199) (0.197)

Child Ethnicity: Black Caribbean 0.035 0.033 0.019

(0.129) (0.136) (0.152)

Child Ethnicity: Black African 0.108 0.122 0.115

(0.215) (0.222) (0.224)

Child Ethnicity: Other 0.269* 0.274* 0.277*

(0.123) (0.130) (0.132)

Religion Important to Child 0.090 0.087 0.090

(0.055) (0.050) (0.050)

Child Religion: Muslim -0.610*** -0.621*** -0.618***

(0.169) (0.167) (0.168)

Child Religion: Minority -0.049 -0.053 -0.051

(0.088) (0.089) (0.090)

Child Birth Weight -0.060** -0.061** -0.061**

(0.020) (0.020) (0.020)

Child has SEN statement 0.085 0.081 0.079

(0.056) (0.055) (0.055)

Child has Disability 0.093 0.094 0.088

(0.084) (0.085) (0.085)

Family Evenings: Often -0.130 -0.133 -0.124

(0.100) (0.098) (0.100)

Family Evenings: Sometimes -0.065 -0.071 -0.056

(0.061) (0.059) (0.065)

Curfew: Always -0.153 -0.149 -0.146

(0.114) (0.113) (0.113)

Family Activities: Often -0.109 -0.109 -0.116

(0.171) (0.172) (0.169)

Family Activities: Sometimes -0.145 -0.142 -0.149

(0.160) (0.163) (0.162)

Talk About School: Often 0.021 0.019 0.016

(0.040) (0.039) (0.040)

Constant 0.424*** 0.446*** 0.474*** 1.212 1.218 1.300*

(0.026) (0.034) (0.033) (0.698) (0.683) (0.694)

Region Fixed Effects ⇥ ⇥ ⇥ X X XArea Type Fixed Effects ⇥ ⇥ ⇥ X X XObservations 779 779 779 779 779 779

R

2 0.023 0.024 0.025 0.164 0.165 0.167

Notes: Robust standard errors (clustered by bin) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditionalfamily, male, at least one parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian,mother’s religion: Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher manage-rial/professional, child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SENstatus, child does not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

65

Table B.10: RD Results on Sample of Wage Earner Parents. Dependent Var: Parents Chose GCSE Subjects

(1) (2) (3) (4) (5) (6)

Modern Family Treatment Effect -0.325*** -0.410*** -0.396*** -0.333*** -0.413*** -0.476***

(0.036) (0.045) (0.081) (0.046) (0.084) (0.091)

Mother’s Share 0.219*** 0.775 2.075** 0.157 0.111 1.747*

(0.063) (0.545) (0.731) (0.100) (0.775) (0.859)

Modern Family⇥Mother’s Share 0.605*** 0.628 -3.065 0.720*** 2.293** 0.427

(0.087) (0.895) (2.810) (0.180) (0.771) (2.575)

Mother’s Share2 0.996 7.074* -0.097 7.764*

(0.966) (3.558) (1.387) (3.772)

Modern Family⇥Mother’s Share2 -2.068 4.944 -2.765 -9.230

(1.652) (14.060) (1.905) (13.375)

Mother’s Share3 7.525 9.925*

(4.673) (5.050)

Modern Family⇥Mother’s Share3 -24.783 -11.923

(17.763) (19.889)

Father’s Employment: Stable 0.001 -0.001 -0.001

(0.045) (0.044) (0.044)

Mother’s Employment: Stable 0.009 0.009 0.010

(0.032) (0.032) (0.031)

Parents both Natural 0.036 0.037 0.034

(0.084) (0.083) (0.082)

Parents not Married 0.011 0.010 0.020

(0.084) (0.085) (0.084)

Household Size 0.053 0.051 0.047

(0.036) (0.035) (0.036)

Dependents -0.054** -0.051** -0.046**

(0.021) (0.019) (0.018)

Father Age -0.005 -0.005 -0.005

(0.006) (0.006) (0.006)

Mother Age -0.005 -0.005 -0.005

(0.003) (0.003) (0.003)

Religion Important to Parents 0.048 0.050 0.052

(0.055) (0.054) (0.056)

Father Religion: Muslim 0.301** 0.292* 0.289*

(0.130) (0.130) (0.131)

Father Religion: Minority -0.260 -0.246 -0.256

(0.211) (0.203) (0.199)

Mother Religion: Muslim 0.085 0.105 0.104

(0.273) (0.283) (0.293)

Mother Religion: Minority -0.118* -0.118* -0.119*

(0.063) (0.062) (0.063)

Family Education: Below Degree -0.033 -0.042 -0.043

(0.051) (0.054) (0.055)

Family Education: GCE A Level -0.121* -0.118* -0.117*

(0.055) (0.056) (0.057)

Family Education: GCSE -0.138 -0.139 -0.135

66

Table B.10: RD Results on Sample of Wage Earner Parents. Dependent Var: Parents Chose GCSE Subjects

(1) (2) (3) (4) (5) (6)

(0.094) (0.096) (0.097)

Family Education: Level 1 -0.058 -0.060 -0.050

(0.119) (0.119) (0.121)

Family Education: Other 0.067 0.051 0.052

(0.145) (0.141) (0.138)

Family Education: None -0.134 -0.136 -0.138

(0.161) (0.158) (0.158)

Father Social Class: Lower Mang. -0.021 -0.011 -0.004

(0.068) (0.069) (0.069)

Father Social Class: Intermediate 0.067 0.075 0.077

(0.068) (0.062) (0.062)

Father Social Class: Small Employer 0.028 0.046 0.070

(0.189) (0.199) (0.200)

Father Social Class: Lower Supervisor 0.079 0.085 0.089

(0.081) (0.083) (0.083)

Father Social Class: Semi-Routine -0.015 -0.005 -0.004

(0.053) (0.050) (0.049)

Father Social Class: Routine -0.080 -0.066 -0.064

(0.047) (0.047) (0.048)

Father Social Class: Long-Term u/e 0.182*** 0.204*** 0.223***

(0.051) (0.047) (0.064)

Mother Social Class: Lower Mang. -0.008 -0.008 -0.003

(0.057) (0.060) (0.059)

Mother Social Class: Intermediate 0.007 0.007 0.006

(0.073) (0.077) (0.075)

Mother Social Class: Small Employer -0.633** -0.611* -0.561*

(0.261) (0.300) (0.302)

Mother Social Class: Lower Supervisor 0.037 0.031 0.030

(0.049) (0.052) (0.049)

Mother Social Class: Semi-Routine 0.041 0.036 0.039

(0.088) (0.094) (0.093)

Mother Social Class: Routine -0.073 -0.080 -0.076

(0.075) (0.076) (0.080)

Mother Social Class: Long-Term u/e 0.000 0.013 0.029

(0.088) (0.093) (0.100)

Family Income -0.001* -0.001* -0.001*

(0.000) (0.000) (0.000)

IDACI score -0.542 -0.557* -0.535

(0.304) (0.303) (0.307)

IMD rank 0.004 0.004 0.004

(0.003) (0.004) (0.004)

Child born in UK -0.184 -0.182 -0.180

(0.129) (0.129) (0.139)

Child Ethnicity: Mixed 0.199* 0.209** 0.213**

(0.091) (0.092) (0.088)

67

Table B.10: RD Results on Sample of Wage Earner Parents. Dependent Var: Parents Chose GCSE Subjects

(1) (2) (3) (4) (5) (6)

Child Ethnicity: Indian 0.450** 0.447** 0.455**

(0.177) (0.160) (0.158)

Child Ethnicity: Pakistani 0.086 0.076 0.091

(0.091) (0.083) (0.084)

Child Ethnicity: Bangladeshi 0.128 0.123 0.126

(0.204) (0.204) (0.203)

Child Ethnicity: Black Caribbean 0.080 0.070 0.055

(0.143) (0.156) (0.169)

Child Ethnicity: Black African -0.016 0.002 -0.005

(0.217) (0.222) (0.226)

Child Ethnicity: Other 0.279* 0.275* 0.279*

(0.142) (0.150) (0.152)

Religion Important to Child 0.115* 0.114* 0.116*

(0.062) (0.058) (0.061)

Child Religion: Muslim -0.626*** -0.633*** -0.633***

(0.185) (0.189) (0.193)

Child Religion: Minority -0.024 -0.029 -0.027

(0.097) (0.099) (0.100)

Child Birth Weight -0.057** -0.059** -0.060**

(0.021) (0.021) (0.021)

Child has SEN statement 0.049 0.041 0.038

(0.059) (0.059) (0.058)

Child has Disability 0.140* 0.138* 0.131

(0.073) (0.074) (0.073)

Family Evenings: Often -0.174 -0.175 -0.162

(0.106) (0.105) (0.108)

Family Evenings: Sometimes -0.056 -0.057 -0.038

(0.102) (0.102) (0.111)

Curfew: Always -0.112 -0.111 -0.112

(0.106) (0.106) (0.106)

Family Activities: Often -0.112 -0.113 -0.120

(0.204) (0.200) (0.200)

Family Activities: Sometimes -0.148 -0.142 -0.148

(0.199) (0.198) (0.200)

Talk About School: Often 0.052 0.050 0.047

(0.036) (0.035) (0.036)

Constant 0.430*** 0.483*** 0.541*** 0.915 0.909 0.966

(0.032) (0.036) (0.030) (0.659) (0.609) (0.642)

Region Fixed Effects ⇥ ⇥ ⇥ X X XArea Type Fixed Effects ⇥ ⇥ ⇥ X X XObservations 682 682 682 681 681 681

R

2 0.023 0.025 0.027 0.187 0.189 0.191

Notes: Robust standard errors (clustered by bin) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditionalfamily, male, at least one parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian,mother’s religion: Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher manage-rial/professional, child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SENstatus, child does not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

68

Table B.11: RD Results on Full Sample. Dependent Var: Argue Often with Parents

(1) (2) (3) (4) (5) (6)

Modern Family Treatment Effect 0.105** 0.247*** 0.153** 0.154*** 0.226*** 0.156*

(0.040) (0.020) (0.051) (0.028) (0.030) (0.078)

Mother’s Share -0.056 -1.102*** -0.057 -0.068 -0.683** 0.552

(0.127) (0.154) (0.457) (0.109) (0.238) (0.503)

Modern Family⇥Mother’s Share -0.282* 0.043 0.527 -0.164 0.130 -0.566

(0.133) (0.290) (1.921) (0.147) (0.529) (2.307)

Mother’s Share2 -1.886*** 2.996 -1.124** 4.768*

(0.273) (2.459) (0.409) (2.373)

Modern Family⇥Mother’s Share2 3.222*** -10.145 1.690 -7.176

(0.526) (11.128) (1.122) (15.343)

Mother’s Share3 6.051* 7.398**

(3.228) (3.203)

Modern Family⇥Mother’s Share3 5.222 -3.499

(15.019) (17.103)

Father’s Employment: Stable 0.003 -0.000 -0.002

(0.019) (0.019) (0.019)

Mother’s Employment: Stable -0.024 -0.021 -0.021

(0.039) (0.039) (0.039)

Parents both Natural 0.020 0.013 0.011

(0.081) (0.082) (0.081)

Parents not Married -0.006 -0.013 -0.006

(0.107) (0.105) (0.108)

Household Size -0.047 -0.048* -0.050*

(0.027) (0.026) (0.027)

Dependents 0.028 0.031 0.033

(0.033) (0.033) (0.035)

Father Age -0.003 -0.004 -0.004

(0.003) (0.003) (0.003)

Mother Age -0.011*** -0.010*** -0.010***

(0.003) (0.003) (0.003)

Religion Important to Parents 0.076** 0.073** 0.074**

(0.028) (0.029) (0.030)

Father Religion: Muslim 0.364 0.380 0.371

(0.314) (0.283) (0.283)

Father Religion: Minority 0.086 0.079 0.072

(0.156) (0.156) (0.151)

Mother Religion: Muslim 0.182 0.184 0.180

(0.403) (0.393) (0.388)

Mother Religion: Minority -0.034 -0.034 -0.036

(0.065) (0.065) (0.067)

Family Education: Below Degree -0.044 -0.045 -0.045

(0.056) (0.055) (0.056)

Family Education: GCE A Level 0.078 0.078 0.078

(0.059) (0.059) (0.060)

Family Education: GCSE -0.059 -0.061 -0.060

69

Table B.11: RD Results on Full Sample. Dependent Var: Argue Often with Parents

(1) (2) (3) (4) (5) (6)

(0.081) (0.080) (0.081)

Family Education: Level 1 0.080 0.083 0.089

(0.087) (0.087) (0.088)

Family Education: Other 0.228 0.246 0.246

(0.231) (0.239) (0.240)

Family Education: None -0.126 -0.120 -0.122

(0.115) (0.114) (0.115)

Father Social Class: Lower Mang. 0.154 0.154 0.157*

(0.084) (0.085) (0.083)

Father Social Class: Intermediate 0.171* 0.161 0.161

(0.092) (0.094) (0.091)

Father Social Class: Small Employer 0.129 0.135 0.129

(0.127) (0.127) (0.125)

Father Social Class: Lower Supervisor 0.068 0.065 0.067

(0.107) (0.109) (0.106)

Father Social Class: Semi-Routine 0.226** 0.217** 0.217**

(0.088) (0.090) (0.084)

Father Social Class: Routine 0.218* 0.215* 0.213*

(0.111) (0.111) (0.108)

Father Social Class: Long-Term u/e 0.080 0.094 0.106

(0.110) (0.097) (0.085)

Mother Social Class: Lower Mang. -0.047 -0.041 -0.038

(0.072) (0.071) (0.071)

Mother Social Class: Intermediate 0.051 0.054 0.054

(0.082) (0.081) (0.080)

Mother Social Class: Small Employer 0.081 0.087 0.084

(0.080) (0.080) (0.080)

Mother Social Class: Lower Supervisor -0.037 -0.032 -0.032

(0.100) (0.100) (0.101)

Mother Social Class: Semi-Routine -0.038 -0.032 -0.028

(0.073) (0.072) (0.071)

Mother Social Class: Routine 0.024 0.029 0.036

(0.104) (0.103) (0.107)

Mother Social Class: Long-Term u/e 0.084 0.096 0.106

(0.150) (0.143) (0.137)

Family Income -0.000 -0.000 -0.000

(0.000) (0.000) (0.000)

IDACI score 0.282 0.284 0.301

(0.288) (0.298) (0.291)

IMD rank -0.003 -0.003 -0.003

(0.003) (0.003) (0.003)

Child born in UK -0.099 -0.103 -0.103

(0.098) (0.099) (0.096)

Child Ethnicity: Mixed 0.015 0.024 0.029

(0.106) (0.112) (0.113)

70

Table B.11: RD Results on Full Sample. Dependent Var: Argue Often with Parents

(1) (2) (3) (4) (5) (6)

Child Ethnicity: Indian -0.318* -0.303* -0.295*

(0.143) (0.146) (0.141)

Child Ethnicity: Pakistani -0.460*** -0.454*** -0.439***

(0.107) (0.112) (0.094)

Child Ethnicity: Bangladeshi -0.736*** -0.728*** -0.724***

(0.139) (0.140) (0.139)

Child Ethnicity: Black Caribbean -0.231 -0.226 -0.231

(0.170) (0.168) (0.169)

Child Ethnicity: Black African 0.196 0.202 0.195

(0.288) (0.297) (0.296)

Child Ethnicity: Other -0.197 -0.183 -0.178

(0.152) (0.155) (0.150)

Religion Important to Child -0.106 -0.110 -0.110

(0.082) (0.079) (0.077)

Child Religion: Muslim -0.371* -0.384* -0.378*

(0.172) (0.176) (0.176)

Child Religion: Minority 0.058 0.058 0.059

(0.091) (0.092) (0.093)

Child Birth Weight -0.004 -0.003 -0.004

(0.046) (0.047) (0.047)

Child has SEN statement -0.059 -0.060 -0.060

(0.075) (0.077) (0.078)

Child has Disability -0.070 -0.067 -0.069

(0.046) (0.046) (0.047)

Family Evenings: Often -0.083 -0.088 -0.082

(0.156) (0.158) (0.158)

Family Evenings: Sometimes -0.011 -0.019 -0.007

(0.151) (0.158) (0.158)

Curfew: Always 0.207** 0.204* 0.204*

(0.088) (0.090) (0.093)

Family Activities: Often -0.040 -0.038 -0.041

(0.162) (0.159) (0.164)

Family Activities: Sometimes -0.061 -0.061 -0.064

(0.134) (0.131) (0.135)

Talk About School: Often -0.094*** -0.093*** -0.096***

(0.026) (0.027) (0.027)

Constant 0.598*** 0.499*** 0.545*** 0.890** 0.866* 0.936*

(0.040) (0.013) (0.017) (0.393) (0.414) (0.443)

Region Fixed Effects ⇥ ⇥ ⇥ X X XArea Type Fixed Effects ⇥ ⇥ ⇥ X X XObservations 758 758 758 757 757 757

R

2 0.003 0.010 0.011 0.170 0.171 0.173

Notes: Robust standard errors (clustered by bin) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditionalfamily, male, at least one parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian,mother’s religion: Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher manage-rial/professional, child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SENstatus, child does not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

71

Table B.12: RD Results on Sample of Wage-Earner Parents. Dependent Var: Argue Often with Parents

(1) (2) (3) (4) (5) (6)

Modern Family Treatment Effect 0.105* 0.300*** 0.217*** 0.158** 0.237*** 0.175*

(0.056) (0.020) (0.043) (0.052) (0.049) (0.084)

Mother’s Share 0.008 -1.237*** -0.656 -0.099 -0.681* -0.058

(0.162) (0.153) (0.467) (0.119) (0.348) (0.733)

Modern Family⇥Mother’s Share -0.445** -0.582* 0.718 -0.185 -0.087 0.430

(0.175) (0.287) (1.553) (0.129) (0.897) (1.654)

Mother’s Share2 -2.246*** 0.467 -1.063* 1.922

(0.248) (2.602) (0.561) (4.631)

Modern Family⇥Mother’s Share2 4.797*** -8.356 1.925 -7.453

(0.513) (9.083) (1.303) (15.551)

Mother’s Share3 3.364 3.763

(3.450) (6.614)

Modern Family⇥Mother’s Share3 10.503 4.755

(12.350) (10.740)

Father’s Employment: Stable -0.014 -0.016 -0.017

(0.031) (0.030) (0.029)

Mother’s Employment: Stable -0.035 -0.034 -0.033

(0.045) (0.045) (0.045)

Parents both Natural -0.009 -0.015 -0.016

(0.066) (0.067) (0.068)

Parents not Married -0.029 -0.035 -0.031

(0.112) (0.110) (0.115)

Household Size -0.041 -0.042 -0.043

(0.036) (0.035) (0.037)

Dependents 0.006 0.009 0.010

(0.044) (0.045) (0.047)

Father Age -0.001 -0.001 -0.001

(0.004) (0.004) (0.004)

Mother Age -0.011*** -0.011*** -0.011***

(0.002) (0.003) (0.003)

Religion Important to Parents 0.088** 0.085** 0.086**

(0.032) (0.032) (0.033)

Father Religion: Muslim 0.466 0.479 0.471

(0.368) (0.346) (0.357)

Father Religion: Minority 0.149 0.138 0.133

(0.139) (0.139) (0.139)

Mother Religion: Muslim 0.144 0.144 0.142

(0.448) (0.443) (0.441)

Mother Religion: Minority -0.051 -0.051 -0.053

(0.057) (0.057) (0.059)

Family Education: Below Degree -0.037 -0.038 -0.037

(0.055) (0.051) (0.051)

Family Education: GCE A Level 0.098* 0.097* 0.096*

(0.050) (0.049) (0.050)

Family Education: GCSE -0.077 -0.079 -0.078

72

Table B.12: RD Results on Sample of Wage-Earner Parents. Dependent Var: Argue Often with Parents

(1) (2) (3) (4) (5) (6)

(0.083) (0.081) (0.083)

Family Education: Level 1 0.078 0.077 0.081

(0.143) (0.140) (0.145)

Family Education: Other 0.170 0.188 0.188

(0.247) (0.259) (0.261)

Family Education: None -0.168 -0.163 -0.163

(0.103) (0.103) (0.104)

Father Social Class: Lower Mang. 0.198* 0.198 0.202*

(0.106) (0.109) (0.110)

Father Social Class: Intermediate 0.227** 0.218** 0.221**

(0.089) (0.091) (0.087)

Father Social Class: Small Employer -0.412** -0.387** -0.375**

(0.132) (0.124) (0.124)

Father Social Class: Lower Supervisor 0.120 0.117 0.121

(0.138) (0.141) (0.139)

Father Social Class: Semi-Routine 0.294** 0.286** 0.289**

(0.101) (0.102) (0.098)

Father Social Class: Routine 0.250* 0.248* 0.250*

(0.131) (0.131) (0.128)

Father Social Class: Long-Term u/e 0.102 0.115 0.124

(0.168) (0.156) (0.145)

Mother Social Class: Lower Mang. -0.020 -0.014 -0.013

(0.066) (0.064) (0.062)

Mother Social Class: Intermediate 0.057 0.061 0.061

(0.083) (0.081) (0.080)

Mother Social Class: Small Employer 0.171 0.211* 0.229*

(0.111) (0.111) (0.124)

Mother Social Class: Lower Supervisor -0.023 -0.017 -0.017

(0.110) (0.109) (0.110)

Mother Social Class: Semi-Routine -0.022 -0.015 -0.014

(0.091) (0.088) (0.086)

Mother Social Class: Routine 0.059 0.066 0.072

(0.102) (0.101) (0.102)

Mother Social Class: Long-Term u/e 0.094 0.104 0.108

(0.153) (0.140) (0.131)

Family Income -0.001* -0.001 -0.001

(0.000) (0.000) (0.000)

IDACI score 0.112 0.119 0.132

(0.282) (0.287) (0.299)

IMD rank -0.002 -0.002 -0.002

(0.002) (0.002) (0.002)

Child born in UK -0.091 -0.095 -0.095

(0.115) (0.119) (0.117)

Child Ethnicity: Mixed 0.009 0.016 0.021

(0.099) (0.105) (0.109)

73

Table B.12: RD Results on Sample of Wage-Earner Parents. Dependent Var: Argue Often with Parents

(1) (2) (3) (4) (5) (6)

Child Ethnicity: Indian -0.281 -0.266 -0.261

(0.166) (0.171) (0.172)

Child Ethnicity: Pakistani -0.475*** -0.463*** -0.450***

(0.115) (0.111) (0.106)

Child Ethnicity: Bangladeshi -0.662*** -0.654*** -0.651***

(0.134) (0.134) (0.135)

Child Ethnicity: Black Caribbean -0.308* -0.301* -0.297*

(0.162) (0.155) (0.150)

Child Ethnicity: Black African 0.246 0.250 0.244

(0.345) (0.360) (0.357)

Child Ethnicity: Other -0.183 -0.168 -0.163

(0.143) (0.147) (0.148)

Religion Important to Child -0.128* -0.130** -0.132**

(0.058) (0.057) (0.054)

Child Religion: Muslim -0.452*** -0.460*** -0.454***

(0.119) (0.117) (0.118)

Child Religion: Minority 0.020 0.022 0.023

(0.066) (0.067) (0.067)

Child Birth Weight 0.036 0.038 0.037

(0.052) (0.052) (0.053)

Child has SEN statement -0.020 -0.018 -0.018

(0.085) (0.088) (0.089)

Child has Disability -0.070 -0.067 -0.067

(0.051) (0.049) (0.049)

Family Evenings: Often -0.117 -0.120 -0.116

(0.156) (0.160) (0.159)

Family Evenings: Sometimes 0.027 0.021 0.028

(0.195) (0.203) (0.196)

Curfew: Always 0.301*** 0.306*** 0.304***

(0.087) (0.083) (0.086)

Family Activities: Often -0.130 -0.127 -0.128

(0.174) (0.171) (0.175)

Family Activities: Sometimes -0.170 -0.169 -0.169

(0.185) (0.182) (0.184)

Talk About School: Often -0.098*** -0.097*** -0.097***

(0.026) (0.027) (0.026)

Constant 0.611*** 0.493*** 0.519*** 1.429** 1.368** 1.399**

(0.053) (0.015) (0.016) (0.458) (0.481) (0.524)

Region Fixed Effects ⇥ ⇥ ⇥ X X XArea Type Fixed Effects ⇥ ⇥ ⇥ X X XObservations 661 661 661 660 660 660

R

2 0.005 0.015 0.016 0.212 0.214 0.214

Notes: Robust standard errors (clustered by bin) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Base dummy variables are traditionalfamily, male, at least one parent is not child’s biological parent, parents are married, religion not important to parents, father’s religion: Christian,mother’s religion: Christian, family education: degree, father’s social class: higher managerial/professional, mother’s social class: higher manage-rial/professional, child not born in UK, child’s ethnicity: white, religion not important to child, child’s religion: Christian, child does not have SENstatus, child does not have disabilities, family evenings: never, curfew: rarely, family activities: never, talk about school: rarely.

74