12
ANALYSIS Upstreamdownstream transactions and watershed externalities: Experimental evidence from Kenya B. Kelsey Jack Harvard University, 79 JFK Street, Cambridge, Massachusetts 02138, United States ARTICLE DATA ABSTRACT Article history: Received 12 May 2008 Received in revised form 20 October 2008 Accepted 3 December 2008 Available online 15 January 2009 Where environmental policies or projects seek behavioral change, understanding underlying norms and preferences is essential to securing environmental outcomes. This study models a payment for environmental services intervention in an experimental field laboratory in Nyanza Province, Kenya. Upstream and downstream individuals are paired in a standard investment game, in which the upstream mover's investment represents land use decisions and the downstream mover responds with a choice of compensation payment. The experimental intervention introduces an enforcement treatment on the downstream movers' compensation decisions for a single round. Underlying social preferences and identity appear to shape individual transactions between upstream and downstream individuals. Upstream first movers are sensitive to the removal of the enforcement on their downstream partners in the second round, and make decisions consistent with crowding out of social preferences. The results suggest that environmental interventions may affect resource decisions for individuals who are not themselves direct targets of enforcement. © 2008 Elsevier B.V. All rights reserved. Keywords: Payments for environmental services Trust game Investment game Cooperation Crowding out Social preferences 1. Introduction Individual land use decisions are a common source of negative environmental externalities, affecting quantity and quality of downstream water supplies, reducing biodiversity and emit- ting carbon into the atmosphere. A Coasian solution to land use externalities has gained recent popularity in conservation circles and proposes to align private and social costs through conditional payments from those affected by the externality to the landholder (Engel et al., 2008; Jack et al., 2008). Many payments for environmental services (PES) programs are led by non-governmental organizations in developing countries. These actors often have little authority to enforce land use outcomes and operate on project length budgets, which may result in weak or short-run PES interventions. Whether, in the long run, such interventions improve outcomes depends both on the efficacy of the intervention itself and on its lasting impact on behavior. This study develops and implements an artefactual field experiment designed to explore individual behavior in a PES-like interaction. Data collection took place in a rural, developing country setting affected by water-related environmental extern- alities, and experimental transactions were between upstream producers of the externalities and the affected downstream ECOLOGICAL ECONOMICS 68 (2009) 1813 1824 I wish to thank Iris Bohnet, Brent Swallow, William C. Clark, Juan-Camilo Cardenas, Luz-Angela Rodriguez, Fiona Greig, two anonymous reviewers, seminar participants at Harvard University and the enumerators and field team during the data collection process. Financial support was provided by the Harvard University Center for the Environment, the Georgio Ruffolo Fellowship in Sustainability Science, and the Vicki Norberg-Bohm Fellowship. Additional field support was provided by the Water and Food Challenge Program of the CGIAR through the SCALES project. E-mail address: [email protected]. 0921-8009/$ see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolecon.2008.12.002 available at www.sciencedirect.com www.elsevier.com/locate/ecolecon

Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ l oca te /eco l econ

ANALYSIS

Upstream–downstream transactions and watershedexternalities: Experimental evidence from Kenya☆

B. Kelsey JackHarvard University, 79 JFK Street, Cambridge, Massachusetts 02138, United States

A R T I C L E D A T A

☆ I wish to thank Iris Bohnet, Brent Swaanonymous reviewers, seminar participantsFinancial support was provided by the HarvaScience, and the Vicki Norberg-Bohm FellowsCGIAR through the SCALES project.

E-mail address: [email protected]

0921-8009/$ – see front matter © 2008 Elsevidoi:10.1016/j.ecolecon.2008.12.002

A B S T R A C T

Article history:Received 12 May 2008Received in revised form20 October 2008Accepted 3 December 2008Available online 15 January 2009

Where environmental policies or projects seek behavioral change, understandingunderlying norms and preferences is essential to securing environmental outcomes. Thisstudy models a payment for environmental services intervention in an experimental fieldlaboratory in Nyanza Province, Kenya. Upstream and downstream individuals are paired ina standard investment game, in which the upstream mover's investment represents landuse decisions and the downstream mover responds with a choice of compensationpayment. The experimental intervention introduces an enforcement treatment on thedownstream movers' compensation decisions for a single round. Underlying socialpreferences and identity appear to shape individual transactions between upstream anddownstream individuals. Upstream first movers are sensitive to the removal of theenforcement on their downstream partners in the second round, and make decisionsconsistent with crowding out of social preferences. The results suggest that environmentalinterventions may affect resource decisions for individuals who are not themselves directtargets of enforcement.

© 2008 Elsevier B.V. All rights reserved.

Keywords:Payments for environmental servicesTrust gameInvestment gameCooperationCrowding outSocial preferences

1. Introduction

Individual land use decisions are a common source of negativeenvironmental externalities, affecting quantity and quality ofdownstream water supplies, reducing biodiversity and emit-ting carbon into the atmosphere. A Coasian solution to landuse externalities has gained recent popularity in conservationcircles and proposes to align private and social costs throughconditional payments from those affected by the externality tothe landholder (Engel et al., 2008; Jack et al., 2008). Manypayments for environmental services (PES) programs are ledby non-governmental organizations in developing countries.

llow, William C. Clark,at Harvard University andrd University Center forhip. Additional field supp

rd.edu.

er B.V. All rights reserved

These actors often have little authority to enforce land useoutcomes and operate on project length budgets, which mayresult in weak or short-run PES interventions. Whether, in thelong run, such interventions improve outcomes depends bothon the efficacy of the intervention itself and on its lastingimpact on behavior.

This study develops and implements an artefactual fieldexperimentdesignedtoexplore individualbehavior inaPES-likeinteraction. Data collection took place in a rural, developingcountry settingaffectedbywater-relatedenvironmental extern-alities, and experimental transactions were between upstreamproducers of the externalities and the affected downstream

Juan-Camilo Cardenas, Luz-Angela Rodriguez, Fiona Greig, twothe enumerators and field team during the data collection process.the Environment, the Georgio Ruffolo Fellowship in Sustainabilityort was provided by the Water and Food Challenge Program of the

.

Page 2: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

1814 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

communities. Individual land use decisions and subsequentcompensation for resulting environmental services, particularlyin the case ofwater quality- or quantity- related projects, can bemodeled by a standard investment game (e.g., Berg et al., 1995).In this sequential two-person game, both first and secondmovers begin with an initial endowment. The first moverchooses a transfer amount (X), which is tripled on its way tothe second mover, who then decides how much of the initialendowment plus the tripled transfer to return to the first mover(Y). Interpreted as a PES interaction, the upstream landholderacts as first mover and chooses a level of investment inenvironmental service generating activities. Investments areprivately costly and produce positive environmental external-ities. The services are received by downstream individuals, whoare then in a position to choose how much to compensate thelandholder forhisorher investment.Higher levelsof investmentby theupstream landholder are likely to yield efficiencygains, ascaptured by themultiplication of his or her investment decisionin the game set up, insofar as the value of water improvementsdownstream exceed the private cost of investment.1

Existing PES programs encourage investment and compen-sation through a variety of interventions, including reductionsin transaction costs, provision of positive incentives forupstream land use outcomes, and increases in the costs ofunder-compensation by downstream communities (seerecent journal issues devoted to the topic: Bulte et al., 2008;Daily and Matson, 2008; Engel and Palmer, 2008). An interven-tion is successful if it increases environmental serviceprovision, modeled in the experiment as first mover invest-ment. This study exploits a standard experimental gamedesign to investigate the importance of social identity andexternally imposed enforcement in a PES-like transaction.Assigning experimental roles to reflect the actual relation-ships between players within a watershed increases externalvalidity of the findings by measuring a stylized version of thereal-world interactions of interest. Section 2 summarizesrelated literature. Section 3 describes the study setting,experimental design and implementation. Empirical resultsare presented in Section 4. Section 5 concludes with adiscussion of the findings and suggestions for future research.

2 In their typology of field experiments, Harrison and Listdifferentiate artefactual and framed field experiments based on

2. Related literature

2.1. Experimental economics in the field

Experimental economics increasingly uses field laboratoriesto understand decision-making in real world contexts. Whilesocial preference measures, including those elicited throughthe investment game, may suffer in terms of generalizabilityof the estimated parameters due to the heightened scrutiny ofthe experimental environment (Levitt and List, 2007), studies

1 Assigning downstream individuals to the role of first mover inthe investment game may apply well to certain real-worldpayment for environmental services schemes. However, theresulting interaction creates a less general match between mostwatershed externalities and the standard investment gamedesign, in which the efficiency generating transfer is made bythe first mover.

of populations of interest potentially generate more accuratefindings than studies of a convenience sample of students.The importance of the field setting increases when theexperimental decision captures information or experiencespecific to the context (Harrison and List, 2004).2 Severalstudies of natural resource users highlight the relationshipbetween experimental setting and real world behavior.Carpenter and Seki (2005) show that Japanese fishermenwho face a social dilemma situation in their daily lives similarto that modeled in the experiment exhibit different prefer-ences than student players, and that the fishermen's prefer-ences are more generalizable to non-experimentalproductivity data. Bouma et al. (2008) also link experimentalmeasures of social preference to non-experimental data. Theyfind that decisions in an investment game are poor predictorsof soil and water conservation decisions at the individuallevel, but that the samemeasures do explain actual conserva-tion decisions at the village level. Cardenas et al. (2000) useframed experiments with rural communities in Colombia tostudy environmental policy impacts, specifically the crowdingout implications of external restrictions on firewood collec-tion. Vollan (2008) also studies collective action dilemmas andcrowding out from regulation in the context of overgrazing inNamibia and South Africa. Outside of the natural resourcedomain, other field experiments have showna strong relation-ship between subject decisions and local economic conditions(Henrich et al., 2004), social norms (Greig and Bohnet, 2008) andproductivity in theworkplace (Barr and Serneels, 2004), or haveoffered alternative interpretations of subject decisions (Karlan,2005).

2.2. The investment game

In the two player investment game's sequential decisions, selfinterested, payoff maximizing second movers will not returnanything (Y=0) to the first mover and, knowing this, the firstmovers will not invest (X=0). However, empirical evidencesuggests that both players transfer at levels above the Nashequilibrium of (X,Y)= (0,0). Efforts to interpret or untangle themotivation behind the observed decisions suggest that bothfirst and second movers are sensitive to conditional (expecta-tion-based) and unconditional (preference-based) factors(Cox, 2004). For first movers, positive investment levels maybe attributed to trust, with decisions conditional on expecta-tions of reciprocity, to social preferences, including altruismand inequality aversion, or to risk preferences, due to theuncertain returns associated with the investment decision.For second movers, positive return transfers may be due toreciprocity or trustworthiness,which is conditional on positive

the latter's provision of field context in “either the commodity,task, or information set that the subjects can use” (2004: 1014).The task presented here captures field context by embeddingexisting participant relationships in the experimental design.Though references to water quality and quantity were provided inthe experimental instructions, the decisions pertained to abstractexperimental tokens. Together these features place the currentstudy somewhere between an artefactual and a framed fieldexperiment.

Page 3: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

1815E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

investment by the first mover, or social preferences, whichmay be independent of first-mover decisions (Ashraf et al.,2006). More complex game designs are required to untanglethese different motivators of positive transfers (e.g., Cox, 2004;Ashraf et al., 2006; Bohnet et al., 2008). Such a design goesbeyond the current study's aim to capture upstream invest-ment decisions in context, inclusive of all motivators, thoughpossible interpretations of observed behavior are discussed inrelation to the existing literature.

2.3. Social identity and enforcement

Experimental studies of the effects of enforcement on experi-mental decisions investigate both sanctions chosen by theparticipants and sanctions imposed by the experimenter. Inan investment game, Fehr and Rockenbach (2003) show thatsanctions determined to be “fair” by the second mover aremore likely to increase compensatory transfers than aresanctions that are perceived to be selfish or greedy. Thisfinding indicates that perceptions of the norms introduced bythe intervention may determine its success. In a study using adifferent game design, Falk and Kosfeld (2006) show that morecontrolling behavior by the first mover signals distrust to thesecond mover and results in lowered levels of reciprocatingbehavior. These and other results suggest that externalenforcement or control may actually impede efficient transac-tion by crowding out social preferences (e.g., Gneezy andRustichini, 2000; Frey and Jegen, 2001; Bohnet et al., 2002).Crowding out of cooperative behavior in the context of naturalresource management is directly investigated in the work byCardenas et al. (2000) and Vollan (2008) mentioned earlier.

Several studies examine the relationship between socialidentity and enforcement, though they focus primarily onvoluntary enforcement of social norms. Bernhard et al. (2006)find that individuals in New Guinea are more likely to punishdeviations from cooperative behavior if the deviation isdirected at a member of the decision maker's tribe. Similarresults are found for Swiss army platoons (Goette et al., 2006).Investment games designed to study the role of social identityeither explicitly select for participant variation based onexogenous factors or induce variation through the experi-mental design. Barr (2003) takes the former approach and findsa divergence in investment game decisions for traditional andre-settled villages in Zimbabwe. Using experimentally inducedvariation in both direct and indirect reputation, Bohnet andHuck (2004) examine the lasting impacts of temporary trustpromoting interventions in a binary choice investment game.Generally, greater social distance is associated with lowerlevels of trust or investment by firstmovers and lower levels ofcompensation or reciprocity by second movers (Bohnet, 2008).

3 The inclusion of a large number of villages in the samplereduced the risk of information spillovers across days. Daydummy variables do not significantly predict transfer amountsfurther indicating that communication did not affect the experi-mental results.

3. Experimental context and design

This study matched the roles in the standard investmentgame to participant location in a watershed in NyanzaProvince in the western part of Kenya, where upstream landuse decisions impact downstream individuals through qualityand quantity of water flows. In the experiment described here,members of upstream communities acted as the first movers

and were paired with members of downstream communities.An investment game captures many of the key features of theinteraction of interest. No existing institutions govern naturalresource externalities between the communities, and as aresult, downstream individuals (second movers) have littledirect economic incentive to compensate upstream commu-nities for investments that increase the provision of positiveenvironmental externalities or mitigate negative externalities.Upstream individuals (first movers), in turn, have littleeconomic incentive to invest in resource decisions that benefittheir downstream counterparts. The investment game tem-porarily removes other barriers to Pareto improving transac-tion and explores the underlying social identity andpreferences that could affect investment and compensationdecisions between the communities.

3.1. Setting

The experiment took place at one upstream and one down-stream site, and drew on a random sample of participantsinvited several days in advance from village household listscovering 19 upstream and 10 downstream villages. The experi-ment was conducted over three days in the two sites, which arelocated about 5 miles from each other, a distance that takesmore than one hour to cover by road.3 In addition to theexperimental data, survey data were collected from theparticipants, revealing differences in both demographic mea-sures and beliefs between the upstream and downstreamcommunities (Table 1).

Downstream villages are ethnically homogenous, whileupstream villages are made up of five different ethnicitieswith the majority constituting just over half of the sample.Communities in both sites are agrarian, though upstreamparticipants are slightly more educated (n.s.) with largerfamilies than their downstream partners. Upstream partici-pants perceive significantly higher levels of cooperation intheir communities than do downstream participants, indi-cated by the estimated number of neighbors who participatein a public project, and both communities believe thatdownstream individuals are less trustworthy than upstreamindividuals. The perception that upstream communities livebetter than downstream communities is common in bothsites. This geographic distribution contrasts with a commonassumption in conservation circles that upstream commu-nities are marginalized (e.g., Noordwijk et al., 2004). Agricul-tural productivity is higher in the uplands, which are sparedthe externalities from erosion, including flash floods andpollution, which are experienced downstream (Swallow et al.,2005). Decisions in the investment gamemay be influenced bythe differences between sites insofar as these differencesaffect social identity and individual preferences that relate tointeractions across the watershed.

Based on the survey data, communities in both locationsare aware of the land use externalities that contaminate

,

Page 4: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

5 Learning between rounds may explain some within subjectvariation between rounds, however, there is no reason to believethat learning would be different for the treatment and controlgroups. A strangers design (e.g., Buchan et al. 2002) with manyrepetitions would also isolate the effects of enforcement removalfrom reputational factors and reduce early round variationcreated as participants learn how to play the game.

Table 1 – Means or response percentages from individual survey of upstream and downstream populations with t-test forequal group meansa

Upstream Downstream t-statistic

Ethnicity Kalenjin 56% Luo 100%Kipsigis 33%Kisii 6%Gusii 3%Bantu 1%

Number of villages sampled 19 10Age 40.5 (1.56) 38.0 (1.90) −1.0076Greater than primary education 26.6% (0.06) 18.8% (0.05) −1.0521Family size 6.8 (2.74) 5.7 (2.35) −2.3533⁎Female participants 36% (0.06) 53% (0.06) 1.9706Downstream people should compensate for water 92% (0.03) 59% (0.06) −4.5999⁎⁎⁎Your water use decreases availability downstream 50% ((0.06) 30% (0.06) −2.3804⁎Think downstream people are more trustworthy than upstream people 17% (0.05) 25% (0.05) 1.0797Think upstream people have a better lifestyle than downstream people 61% (0.06) 78% (0.05) 2.1333⁎Number of neighbors who cooperate in a public project, out of 10 7.54 (0.23) 6.76 (0.29) −2.1013⁎Upstream and downstream communities can reach agreement on water issues 92% (0.03) 73% (0.06) −2.8797⁎

T-test for equality of group means: ⁎⁎⁎Significant at pb0.001, ⁎⁎ significant at pb0.01, ⁎significant at pb0.05.a Standard deviations are in parentheses.

1816 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

downstream water sources, with 92% of the sample respond-ing that upstream people are responsible for contamination ofwater sources. Though inter-community negotiation has notoccurred, it is not clear whether this failure is due totransaction costs, a lack of trust, or a failure of institutionalsupport for environmental service payments. Around three-quarters of downstream participants and over 90% ofupstream participants responded positively to a surveyquestion asking whether upstream and downstream groupscan reach an agreement on water issues.

3.2. Experimental design

The standard investment game design was modified in threeways to investigate upstream–downstream cooperation in aPES-type interaction. First, the treatment group was subjectedto a weak enforcement intervention to encourage down-stream compensation for upstream investments. Down-stream second movers who returned an amount less thantheir partners' investment (YbX) were subject to enforcement,which subtracted the amount of under-compensation (f=X–Y)from the second mover's payoffs. Players were aware that thesubtracted amount was not returned to the first mover. Thisenforcement was referred to as “fine” in the experimentalinstructions. To reflect the enforcement difficulties associatedwith PES programs in developing countries, the enforcementwas implemented with a probability p=0.5, determined by theflip of a coin. As shown in Table 2, the enforcement treatmentdoes not change Nash equilibrium behavior for the para-meters used in this experiment.4 Furthermore, by restrictingthe effects of punishment to the secondmover, any differencein first mover investment between the treatment and controlis mediated by a shift in expectation or preferences. Partici-pants were assigned to either the treatment or control group.

4 The expected cost of the fine is 0.5(X−Y) while the cost ofcompliance is X−Y for all XNY.

Second, the game was repeated for two independentrounds with cumulative payoffs for the participants. For thetreatment group, enforcement was imposed during the firstround and removed in the second round to allow betweensubject analyses of the effects of enforcement removal. Thechange in rules between rounds for the enforcement treat-ment was public information, and players did not learn thefirst round outcomes before making second round decisions.The second round of the enforcement treatment was thusidentical to the control, allowing between subject analysis ofthe effects of enforcement removal. A repeated one-shot gameisolates the effects of enforcement removal from decisionsbased on reputation or the history of play.5 Environmentalexternalities generated through upstream land use decisionsare aggregated in the watershed, such that individual reputa-tion is unlikely play a significant role in partners' transferdecisions.6 In addition, temporal delays between investmentupstream and the accrual of benefits downstream slow thereputation building process. These delays may lead to situa-tions where compensation decisions are required before theoutcomes of upstream investment are observed.

Third, specific reference to the location of the players in thewatershed and examples referring to upstream players'impacts on downstreamwater quality and quantity generateda frame within which participants could base their decisions.Further details of the instructions provided to participants areexplained in the description of implementation.

6 Though individual decisions are aggregated via the environ-mental service, the PES approach encourages individual levelinvestments upstream and individual level payments down-stream. As a result, a game between individuals reasonablymimics actual transactions.

Page 5: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

Table 2 – Experimental design and payoffs

Design Payoffs

Round 1 Round 2 First mover Second mover

Control (N=64) No enforcement No enforcement 8−X 8+3X−YTreatment (N=64) Enforcement (p=0.5) No enforcement 8−X Y≥X: 8+3X−Y

YbX : p : 8 + 3X − Y − X − Yð Þ1 − p : 8 + 3X − Y

1817E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

Initial endowments were set at 8 tokens and the firstmover's investment decision was tripled before it reached thesecond mover. First mover decisions were constrained toincrements of two and second mover responses were con-strained to increments of four for every possible amount sentby the first mover (Table 3).

3.3. Implementation

Experimental sessions were run with groups of eight partici-pants in the upstream and downstream sites simultaneously,with two morning and two afternoon session for each of onefull day and two half days to generate a total sample size of128. Upstream sessions were conducted in an elementaryschool and downstream sessions in a community center, withidentical display and informational materials to assist withinstructions in each site (see Appendix). The experimentalinstructions were reverse translated into two languages:Kalenjin for the upstream sessions and Luo for the down-stream sessions. Experimenter effects cannot be ruled out,though they are controlled for to the extent possible in theanalyses. Participants were given a show up fee of Ksh 200(USD 3.30). The games were played with tokens, each of whichwas worth Ksh 15, and participants were told the currencyequivalent during the instructions. Average earnings in thegame were Ksh 440 (USD 7.30), or around four times theaverage daily wage.When participants were invited to join thestudy, they were told about the show up fee and the chance toearn additional money by participating.7

Participants were welcomed as a group of sixteen and givenconsent forms, then divided into two separate rooms wheretheywere read the instructions, providedexamples, andaskedaseries of questions to test for their understanding. In each of thesessions, participants were told the general location of theirpartner, with specific reference to the location in the watershed(upstream or downstream).8 It was explained that the experi-menters would communicate by mobile phone to determineoutcomes.

7 The experiments described here are a subset of a large numberof experimental games implemented by ICRAF — The WorldAgroforestry Centre in collaboration with Professor Juan-CamiloCardenas (Universidad de los Andes) under the SCALES projectduring June 2007. Higher payoffs were used in one of thetreatments in this study to make first round earnings consistentwith another set of treatments, not described here. Averagetransfers are not affected by this higher token value.8 In another set of treatments, examples of the game decisions

made explicit reference to upstream impacts on downstreamwater quality. This framing had no impact on either upstream ordownstream transfers.

Following the instructions, participantswere called out of theroom one by one, where the possible moves and associatedoutcomes were once again explained by the enumerator beforeparticipants were asked to make a decision. Though theexperiments were not double-blind, player decisions wereanonymous and based on assigned identification numbers. Amonitor remained in the room while decisions were elicited toensure thatparticipantsdidnotdiscussdecisionsas theywaited.

The strategy method was used to elicit conditionalresponses by second movers to all possible decisions by thefirst mover. The strategy method provides data on the fullresponse function for second movers and allowed for all datacollection to occur before players learned their partners'transfer decisions. Strategy method data was collected byasking for a second player return decision for each possibleamount transferred by the first mover. Expectations of partnertransfers were also measured at the time of first and secondround decisions, using an incentive compatible elicitationmethod. Rewards for accuracy of expectations were calculatedby the formula: 10/ (|E(received)−received|+1). Participantswere told that their payments would be increased by a smallamount, the closer their expectation to the real amount.Following the first round, all participants were told that theywould play for one more round with the same partner. At thestart of the game, participants were not given specificinformation about how many times the game would beplayed, and none asked.9 When the rules of the game changedbetween rounds (the enforcement treatment), new instruc-tions and payoff tables were provided (see Appendix forsamples). Participants were told that the rounds wereindependent and that they would be paid the sum of theoutcomes of the rounds at the end of the game. Theenforcement outcome for the treatment sessions was deter-mined publicly by a coin flip at end of the second round, toavoid any effects from the random enforcement outcome onsecond round decisions. Following completion of all experi-mental decisions, a short survey was administered to eachparticipant. During this time, the experimenters communi-cated by mobile phone and calculated payoffs. Total payoffswere delivered privately in sealed envelopes.

4. Empirical results

From an environmental services perspective, an effectiveintervention increases levels of investment by upstream com-munities, represented by higher levels ofX, by changing costs ofinvestment (the value of X), expectations of returns (E(Y)) or

9 Some loss of control may result from this design feature iplayers had heterogeneous expectations of length of play.

f

Page 6: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

Table 3 – Possible decisions by first and second movers ineach rounda

First mover'sendowment=8

x3 Second mover'sendowment=8

Send (X): 0 →0 Return (Y): 0, 4, 82 →6 0, 4, 8, 124 →12 0, 4, 8, 12, 16, 206 →18 0, 4, 8, 12, 16, 20, 248 →24 0, 4, 8, 12, 16, 20, 24, 28, 32

a The decision table is from J. Cardenas, following other experi-ments in this project.

Table 4 –Mean transfer amounts and t-tests for equalmeans

Round 1 Round 2 Within subject(round 1=round 2) a

First mover (X)Control 3.375 (2.28) 4.875 (2.80) −3.5496⁎⁎⁎Treatment 3.44 (2.21) 2.81 (1.88) 2.2504⁎Between subject(control=treatment)b

−0.1575 4.8918⁎⁎⁎

Second mover (Y /X)Control 1.54 (1.00) 1.35 (1.14) 0.9300Treatment 2.08 (1.56) 2.41 (1.53) −1.2249Between subject(control=treatment)b

−1.9847⁎ −3.6864⁎⁎⁎

1818 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

preferences. If, on the other hand, investment is decreasedeither during the intervention or following its removal, then thespecific intervention, in this case weak and temporary enforce-ment of downstream decisions, may have detrimental environ-ment and livelihood consequences.10 Both first and secondmover transfer decisions are analyzed. Simple means compar-isons between rounds and between treatment groups arepresented together for both players, then first and secondmover decisions are examined separately in regression modelsthat introduce additional explanatory and control variables. Toavoid attribution to a specific behavioralmotivation, firstmovertransfers are referred to as investments and second movertransfers referred to as compensation when discussing results.

Within subject comparison of mean transfer decisionsbetween rounds, shown in Table 4, reveals that meaninvestments by first movers (X) significantly increased for thecontrol group, and significantly decreased for the enforcementtreatment group. Between treatment and control subjects, firstround investments (X1) were statistically indistinguishablewhile second round investments (X2) were significantly lowerin the treatment group than in the control group. First moversin the treatment group do not respond to the enforcementintervention itself (first round), but are sensitive to the removalof the enforcement after the first round (second round). Meanexpectations (E(Y)) of return transfers were statistically thesame both within and between subjects, which suggests thatthe effects of the treatment functioned through preferences.This interpretation is explored further under a regressionmodel in the next sub-section. Overall, the fraction of theendowment invested by first movers is within the rangeobserved by other investment experiments in developingcountries (summarized in Cardenas and Carpenter, 2008).Firstmover expectationswere veryhigh,withexpected returnson investment always greater than one, which contrasts withsurvey responses that suggest a general perception thatdownstream individuals are less trustworthy than upstreamindividuals (see Table 1).

For downstream second movers, the average amountreturned as a fraction of quantity invested (Y/X), does notsignificantly differ between rounds for either the treatment orcontrol groups. In both rounds of the treatment, downstreamindividuals returned significantly more than those in the control

10 Other types of enforcement, including penalties for low-levelinvestment by first movers or strictly enforced penalties for eitherplayer may, of course, generate different effects on transferdecisions.

group. The variance in fraction returned was also significantlyhigher in the treatment group for both rounds, possibly becauseof heterogeneity in individual response to the treatment.

Mean transfers are further illustrated by the bar graphs inFig. 1. The pattern of actual decisions (left) follows expecta-tions (right) for both first (top) and second (bottom) movers.

4.1. Upstream investment decisions

A regression framework further examines the between subjectdifferences in the treatment and control groups for observableindividual- and session-specific covariates, including expecta-tions. For the firstmover, the amount sent,Xi,t, is individual andround specific. Since the primary difference in behavior acrosstreatments manifests as the difference in behavior acrossrounds, all round dependent variables are differenced betweenrounds in the OLS models unless otherwise noted. In the OLSregression model: Xi,2−Xi,1=α+E(Yi,2−Yi,1) ⁎β+T ⁎γ+controls, theamount sent is regressed onto expectations of the amountreturned, a treatment dummy, and individual controls, whichincludes transfer levels from the first round. The effect ofenforcement on expectation is explored through a similarmodel: E(Yi,2−Yi,1)=α+(Xi,2−Xi,1) ⁎β+T ⁎γ+controls.

For the first mover, the difference in the amount sentbetween rounds is significantly decreased by enforcement andexpectations significantly predict decisions (Table 5: columns 1,2 and 3). For the most part the regression analysis follows themeans comparison in the sample. The complete models(Table 5: columns 2 and 3) indicate a possible experimentereffect (Facilitator) related to variation created by the differentfacilitators used in the experiment. A survey measure ofperceptions of relative lifestyle quality for upstream and down-stream communities (Better lifestyle upstream) is included inthe complete regression models and significantly correlateswith first mover transfer decisions. Upstream individuals whobelieve that they generally enjoy a better lifestyle thanindividuals downstream increase contributions betweenrounds more than do individuals who responded otherwise.

Standard deviations are in parentheses. t-statistics are reported formeans comparison tests. ⁎⁎⁎Significant at pb0.001, ⁎⁎ significant atpb0.01, ⁎significant at pb0.05.a Mean comparison test for paired data (t-test).b Group mean comparison test (t-test).

Page 7: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

Fig. 1 –Decisions and expectations for first and second movers, by round and treatment.

1819E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

The increase in investments between the first and secondround in the control condition is difficult to compare with otherexperimental results since first round outcomes were notknown prior to second round decisions. Increased secondround transfers could be explained by the larger endowmentin the second round if individuals treat earnings or expectedearnings as cumulative. Evaluating transfer decisions as afraction of first movers' cumulative endowment in the secondround, including expected returns from the first round, showsfraction sent decreasing in the second round for both thetreatment and control groups.11 However, in the regressionmodel, expected returns from the first round (Expected returns(round 1)) do not significantly explain the change in transferamounts between rounds (Table 5: column 3). Inclusion of aninteraction term (Interaction) for expected returns and enforce-ment also indicates that changes in expectation relate directlyto transfer decisions, rather than indirectly through thetreatment.12

These results suggest that the effects of enforcement oninvestment decisions are not driven by changes in expecta-tions alone. Instead, the removal of enforcement in the second

11 The mean fraction of endowment sent in round 1 isapproximately 42.5 in both the control and treatment groups. Inthe second round, the fraction of expected endowment trans-ferred is 25.8% in the control group and to 14.1% in the treatmentgroup.12 As a robustness check, including the two expectation variablesin the model presented in column 3 (Table 5) one at a time doesnot change the significance of any model parameters.

round seems to crowd out preference-based transfers byupstream first movers, resulting in a decline in investmentfollowing removal of the enforcement, as compared with thecontrol group.While the gamedesigndoesnot allow for testingof the motivation behind the relative decrease in investmentdue to enforcement, the significance of the perception thatupstream individuals are better suggests that larger increasesin transfers between rounds may be driven by considerationsof social preferences such as equity or altruism.

4.2. Downstream compensation decisions

Analyzing only actual amounts returned by second moverscould lead to incorrect attribution of a treatment effect,depending on the shape of the individual responsefunctions.13 Exploiting the strategy method data provides agraph of average secondmover “response functions”with andwithout enforcement.

The top set of lines show the average total amount returned,as a fraction of the amount sent, while the bottom set of linesshow the difference in amount returned between the first andsecond rounds, also as a fraction of amount sent. Secondmovers show decreasing returns to first mover investment,which has been found by experimenters in other developing

13 For example, in the context of a common pool resource gameVelez et al. (2005) describe different contribution functions resultingfrom differences in participant motivation.

,

Page 8: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

15 This variable is insignificant in predicting first mover beha-vior, or expectations for either player.

Table 5 – OLS regression model for first mover decisions,differenced across roundsa

Sent Expectation

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

Enforcement −2.262⁎⁎⁎ −2.611⁎⁎⁎ −2.592⁎⁎⁎ −0.786(0.50) (0.46) (0.47) (0.93)

Expectation −0.223⁎⁎ −0.267⁎⁎⁎ −0.286⁎⁎⁎(0.10) (0.07) (0.08)

Interactionb −0.216(0.24)

Expected returns(round 1)

−0.295 −0.862⁎⁎⁎(0.26) (0.15)

Sent (round 1) −0.664⁎⁎⁎ −0.648⁎⁎⁎ −0.664⁎⁎⁎(0.13) (0.12) (0.16)

Better lifestyleupstreamc

0.926⁎ 0.986⁎ 1.005(0.51) (0.56) (0.84)

Female −0.084 (0.56) −0.565(0.50) 0.013 (0.64)

Education 0.192 (0.52) −0.351(0.30) 0.206 (0.26)

Age 0.013 (0.32) 0.014(0.02) 0.016 (0.02)

Facilitator −0.503⁎ (0.02) −0.402(0.26) −0.570⁎⁎ (0.26)

Constant 4.253⁎⁎⁎ 3.935⁎⁎⁎ 4.569⁎⁎⁎ 2.873⁎⁎⁎(0.61) (1.21) (1.29) (1.04)

a OLS model used in all regressions with N=64 (upstream).Amounts sent and expectations are all differenced between rounds.Standard errors are in parentheses. ⁎⁎⁎ significant at pb0.01, ⁎⁎significant at pb0.05, ⁎significant at pb0.10.b Interaction term for the difference in expectation and theenforcement treatment.c Dummy variable for survey response to question: “Who has abetter lifestyle, upstream or downstream people?”

1820 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

country contexts (e.g., Greig and Bohnet, 2008). In this case, themean fraction returned is greater than one for every quantitysent by the first mover and for both rounds, such that firstmovers are, on average, always made better off by theirinvestments. Examining the full strategy of responses for thesecond movers, shown in Fig. 2, indicates that differences inamounts returned across experimental conditions are drivenlargely by the differences in amount received.14

For the second mover, the amount returned is conditionalon amount sent, j, and is analyzed as a panel of responses foreach possible decision by the first mover, also differencedacross rounds for all time-dependent variables. The panelregression model ∑j

(Yi,j,2−Yi,j,1) /Xj=α+T ⁎γ+controls regresses the

set of conditional responses onto expectations, a treatmentdummy variable (Enforcement) and controls that includeslevels from the first round. The GLSmodel (Table 6: columns 2and 3) is compared with an OLS regression for actual amountsreturned (Table 6: column 1).

For the second mover the strategy method of elicitationproves important in separating behavior by the second moverfrom first mover responses to the treatment. When the

14 None of the average strategy responses differ significantlybetween the treatment and control groups in two sample t-testsof group means.

data are analyzed only for second mover transfer outcomes(Table 6: column 1), enforcement appears to have a positiveimpact on the difference between first and second rounddecisions. However, when the full panel of responses isanalyzed, this effect goes away. Though none of the demo-graphic variables are important predictors of the panel ofstrategy responses, a surveymeasure of norms (Norm response)among second movers does significantly explain changes incompensation between rounds (Table 6: column 3). The ques-tion asked respondents whether they thought downstreamindividuals should compensate upstream communities forthe quantity and quality of water they provide. Those whoresponded “yes” increased the amount returned betweenrounds significantly more than those who answered “no”.The overall average amount returned is also positively andsignificantly explained by this survey response at the pb0.02level.15

Data for the second movers show no effect of theenforcement treatment on transfer decisions and clearlyillustrate the importance of the strategy method for rigorousinvestigation of individual response to changes in institutions.Since the design of the enforcement treatment applieddirectly to second movers only, the weakness of the enforce-ment may have been more salient to them than to theirupstream partners. Baseline return amounts were high, suchthat only 14% of second movers in the control group chose toreturn amounts that would have been eligible for enforce-ment. The similarity in response functions between thetreatment and control group may have been due to theleniency of the enforcement treatment or to the high baselinelevels of compliance. While second movers do show somepreference based variation in transfer decisions, as capturedby the surveymeasure, specific motivations behind individualtransfers are not differentiated by the game design and theenforcement treatment appears not to have impacted thisrelationship. An enforcement mechanism with higher enfor-cement probabilities or with stricter requirements on secondmovers may have generated a stronger decision response.

4.3. Distributional considerations

While efficiency is determined entirely by the decision of thefirst mover, the second mover determines the distributionaloutcomes of the game by deciding how to divide the totalavailable quantity following the first mover's investmentdecision.16 These measures are sensitive to the parametersselected by the experimenter, which are also likely to affectparticipant decisions (Roth, 1995). Analysis of first moverdecisions in Section 4.1 suggests that distributional preferencesmay be important determinants of investment decisions, asmeasured by individual survey responses.

16 Note that the payoff outcomes only incorporate the realizedtransfers by second movers, and should not be interpreted assignaling downstream distributional preferences. As discussionin Section 4.2, effects of the treatment on second mover decisions(and thus any effect on preferences) should be analyzed as aresponse function using the strategy method data.

Page 9: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

Table 6 – Regression results for second mover decisions(Y /X) a

Actual fractionreturnedb

Strategyresponses c

(1) (2) (3)

Fig. 2 –Second mover response functions.

1821E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

Distributional impacts from the enforcement treatmentcan be evaluated in the context of the game by comparingpayoff outcomes for each player. The accrual of efficiencygains to the secondmover in the control condition result fromincrease in first mover investment, and can be clearly seen inthe higher payoffs that second movers receive in the controlgroup in that round (Fig. 3). Overall, aggregate payoffs in thewatershed are higher in the control condition, though more“fair” under the enforcement treatment. Thus, enforcementleads to amore equitable distribution of payoffs not because itencourages fairness norms in the second mover, but becauseit discourages the unevenly distributed gains from higherfirst mover investment in the second round. However, sincefirst movers increase investment decisions more, indepen-dent of expectations of compensation, when they feel thatupstream lifestyles are better than downstream lifestyles, thisdistributional outcome may be consistent with first moverpreferences.

Enforcement 1.006⁎⁎ 0.039 0.026(0.47) (0.18) (0.18)

Levels (round 1) −0.740⁎⁎⁎ −0.638⁎⁎⁎ −0.576⁎⁎⁎(0.16) (0.07) (0.08)

Normresponsed

0.682 0.414⁎⁎(0.42) (0.18)

Female 1.064⁎⁎⁎ −0.035(0.38) (0.18)

Education 1.013⁎⁎ 0.069(0.41) (0.18)

Age 0.027 −0.008(0.02) (0.01)

Facilitator 0.088 0.096(0.23) (0.09)

Constant −2.608⁎ 0.878⁎⁎⁎ 0.361(1.43) (0.15) (0.61)

N 47 256 248

a Second mover transfer decisions as a fraction of first movertransfer,differencedacrossrounds. ⁎⁎⁎Significantatpb0.01, ⁎⁎significantat pb0.05, ⁎significant at pb0.10.b OLS regression model.c Random effects GLS regression model with standard errors areclustered at the individual level.d Dummy variable for affirmative response to survey question:Should downstream individuals compensate upstreamcommunities for the quantity and quality of water they provide?

5. Conclusion

The design of the experiment described here measuresupstream–downstream transactions, modeled on the increas-ing number of payment for environmental services projects indeveloping countries. Implementation of the experiment inrural Kenya highlights the role of social identity betweencommunities in a watershed for increasing the positiveinvestment externalities generated by the first movers,located upstream. Transfers under a weak, temporary enfor-cement treatment imposed on the second mover are com-pared with a control group, based on a standard investmentgame design.

The results suggest that temporary enforcement of com-pensatory transfers by downstream individuals may actuallybe detrimental to the environment. This effect is not driven byresponses of the directly enforced individuals under theintervention, as suggested by other research on the crowdingout impacts of environmental regulation (Cardenas et al.,2000; Vollan, 2008). Instead, removal of the enforcementmechanism in the second round of the game is associatedwith lower levels of investment by first movers in the

treatment group. This response is not fully explained bychanges in expectations of compensation, indicating thatsome portion of relatively lower levels of investment may bedue to crowding out of social preferences that occur as anindirect impact of the enforcement mechanism. Better under-standing the indirect effects of PES interventions and othertypes of environmental regulation, both on participants notdirectly targeted by enforcement and on environmentaldecisions following termination of an intervention, will helppredict the full ramifications of policies and projects.

Theexperimental findingsalso indicate that social identity isan important component of individual transfers. Survey mea-sures of upstream and downstream attitudes explain a sig-nificant portion of both investment and returndecisions. Like inFehr and Rockenbach (2003), downstream second moverswhose norms are enforced by the intervention are moreresponsive to the treatment. For the existing beliefs andpreferences of the communities studied, results suggest thatmitigating transaction costs between upstream landholdersand downstream parties may be the most effective course ofaction, particularly since enforcement comes at an additionalcost to the implementing agency. On the other hand, thedecreasing response function observed in the downstreamcommunities provides little incentive for upstream individualsto increase land use investments, and may lead cause invest-ments to collapse to someminimal positive level over time.

The current design says nothing about decisions underrepeated interaction and offers no conclusions about the

Page 10: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

Fig. 3 –Payoff outcomes for each player.

1822 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

potential efficacy of an enforcementmechanism that changesdominant strategy response by second movers. While takingexperimental economics to the field may enhance general-izability of the findings to the population of interest, extra-polation requires caution, particularly to developing countryPES interventions. Tracing the flow of environmental servicesacross physical space is scientifically difficult, with high levelsof uncertainty and temporal delays between upstream invest-ment and downstream receipt of services. Though thesubjects in this study demonstrated an awareness of thewatershed externalities caused by upstream land use, manyrural communities are poorly informed about the relationshipbetween land use decisions and downstream water flows, orlack information on how upstream investments mightenhance downstream service provision. Finally, the value ofenvironmental services and the costs of land use investmentsare difficult to quantify (Shogren, 2005).

Variations on the game that incorporate some of thesecomplexities, such as the uncertainty in the service flows fromland use investments or the collective action aspects bothservice provision and compensation would highlight aspectsof the PES environment that were neglected in this study. Evenmore importantly, studies on the long run efficiency impactsof interventions aimed at enhancing investment and encoura-ging compensatory payments for improved land use practicesshould be undertaken in the context of ongoing PES programsin developing countries. Controlled studies of the crowdingeffect of incentive payments outside of the laboratory willprovide important insights for the design and implementationof future environmental interventions.

Appendix A. Examples and payoff tables shown toparticipants

1. Examples shown to upstream participants in round one ofthe enforcement treatment during instructions. Comparableexamples, with roles reversed, were shown to downstreamparticipants.

Example 1: 0 returned is as much as 0 sent so no coin flip.

Example 2: 5 returned is more than 4 sent so no coin flip

Example 3: 6 returned is less than 8 sent, so coin flip

Page 11: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

1823E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

2. Sample payoff table showing upstream (first mover)decisions: one table for every possible transfer decisions wasdisplayed (two units invested, as seen by first movers, is shownhere).

3. Sample payoff table showing downstream decisions inround one of the enforcement treatment: one table for everypossible transfer decision was displayed (response optionsassociated with 6 units invested, as seen by first movers isshown here).

R E F E R E N C E S

Ashraf, N., Bohnet, I., et al., 2006. Decomposing trust andtrustworthiness. Experimental Economics 9 (3), 193–208.

Barr, A., 2003. Trust and expected trustworthiness: experimentalevidence from Zimbabwean villages⁎. The Economic Journal113 (489), 614–630.

Barr, A., Serneels, P., 2004.Wages andReciprocity in theWorkplace.Centre for the Study of African Economies Series WPS/2004-18.

Berg, J., Dickhaut, J., et al., 1995. Trust, reciprocity, and socialhistory. Games and Economic Behavior 10 (1), 122–142.

Bernhard, H., Fehr, E., et al., 2006. Group affiliation and altruisticnorm enforcement. American Economic Review 96 (2), 217–221.

Bohnet, I., 2008. Trust in experiments. In: Durlauf, S.N., Blume, L.E.(Eds.), NewPalgraveDictionary of Economics. PalgraveMacmillan.

Bohnet, I., Huck, S., 2004. Repetition and reputation: Implicationsfor trust and trustworthiness when institutions change.American Economic Review 94 (2), 362–366.

Bohnet, I., Frey, B.S., et al., 2002. More order with less law: oncontract enforcement, trust, and crowding. American PoliticalScience Review 95 (01), 131–144.

Bohnet, I., Greig, F., et al., 2008. Betrayal aversion: evidence fromBrazil, China, Oman, Switzerland, Turkey, and the UnitedStates. American Economic Review 98 (1), 294–310.

Bouma, J., Bulte, E., et al., 2008. Trust and cooperation: socialcapital and community resource management. Journal ofEnvironmental Economics and Management 56 (2), 155–166.

Buchan, N., Croson, R., et al., 2002. Swift neighbors and persistentstrangers: a cross-cultural investigation of trust and reciprocityin social exchange. American Journal of Sociology 108 (1),168–206.

Bulte, E.H., Lipper, L., et al., 2008. Payments for ecosystemservicesand poverty reduction: concepts, issues, and empiricalperspectives. Environment and Development Economics 13(03), 245–254.

Cardenas, J.C., Carpenter, J.P., 2008. Behavioural developmenteconomics: Lessons from field labs in the developing world.Journal of Development Studies 44 (3), 311–338.

Cardenas, J.C., Stranlund, J., et al., 2000. Local environmentalcontrol and institutional crowding-out. World Development 28(10), 1719–1733.

Carpenter, J.P. and Seki, E., (2005). Do Social Preferences IncreaseProductivity? Field Experimental Evidence from Fishermen inToyama Bay. IZA Discussion Paper Series, No. 1697, IZADiscussion Paper Series: No. 1697.

Cox, J.C., 2004. How to identify trust and reciprocity. Games andEconomic Behavior 46 (2), 260–281.

Daily, G.C., Matson, P.A., 2008. Ecosystem services: from theory toimplementation. Proceedings of the National Academy ofSciences 105 (28), 9455–9456.

Engel, S., Palmer, C., 2008. Payments for environmental services asan alternative to logging under weak property rights: The caseof Indonesia. Ecological Economics 65 (4), 799–809.

Engel, S., Pagiola, S., et al., 2008. Designing payments forenvironmental services in theory and practice: An overview ofthe issues. Ecological Economics 65 (4), 663–674.

Falk, A., Kosfeld, M., 2006. The hidden costs of control. AmericanEconomic Review 96 (5), 1611–1630.

Fehr, E., Rockenbach, B., 2003. Detrimental effects of sanctions onhuman altruism. Nature 422, 137–140.

Frey, B.S., Jegen, R., 2001. Motivation crowding theory. Journal ofEconomic Surveys 15 (5), 589–611.

Gneezy, U., Rustichini, A., 2000. Pay enough or don't pay at all. TheQuarterly Journal of Economics 115 (3), 791–810.

Goette, L., Huffman, D., et al., 2006. The impact of groupmembership on cooperation and norm enforcement: Evidenceusing random assignment to real social groups. The AmericanEconomic Review 96, 212–216.

Page 12: Upstream–downstream transactions and watershed ...sites.tufts.edu/kjack/files/2011/08/Jack_Ecol-Econ.pdf · downstream movers' compensation decisions for a single round. Underlying

1824 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 9 ) 1 8 1 3 – 1 8 2 4

Greig, F., Bohnet, I., 2008. Is there reciprocity in a reciprocalexchange economy? Evidence of gendered norms from a slumin Nairobi, Kenya. Economic Inquiry 46 (1), 77–83.

Harrison, G.W., List, J., 2004. Field experiments. Journal ofEconomic Literature 42, 1009–1055.

Henrich, J., Boyd, R., et al., 2004. Foundations of Human Sociality:Economic Experiments and Ethnographic Evidence fromFifteen Small-Scale Societies. Oxford University Press, Oxford.

Jack, B.K., Kousky, C., et al., 2008. Designing payments forecosystem services: Lessons from previous experience withincentive-based mechanisms. Proceedings of the NationalAcademy of Sciences 105 (28), 9465–9470.

Karlan, D., 2005. Using experimental economics to measure socialcapital and predict financial decisions. American EconomicReview 95 (5), 1688–1699.

Levitt, S.D., List, J.A., 2007. What do laboratory experimentsmeasuring social preferences reveal about the real world?Journal of Economic Perspectives 21 (2), 153–174.

Noordwijk, M.v., Chandler, F.J., et al., 2004. An introduction to theconceptual basis of RUPES: Rewarding upland poor for the

environmental services they provide. World AgroforestryCentre, Bogor, Indonesia.

Roth, A.E., 1995. Individual bargaining. In: Kagel, J.H., Roth, A.E.(Eds.), Handbook of Experimental Economics. PrincetonUniversity Press, Princeton, N.J.

Shogren, J.F., 2005. Experimental methods and valuation. In:Mäler, K., Vincent, J. (Eds.), Handbook of EnvironmentalEconomics. North-Holland, Amsterdam.

Swallow, B., Onyango, L., et al., 2005. Dynamics of poverty,livelihoods and property rights in the Lower Nyando basin ofKenya. International workshop on 'African Water laws: Plurallegislative Frameworks for Rural Water Management in Africa',Gauteng, South Africa.

Velez, M., Stranlund, J., et al., 2005. What Motivates Common PoolResourceUsers? Experimental Evidence fromtheField.Universityof Massachusetts Department of Resource Economics WorkingPaper Series No. 2005-4.

Vollan, B., 2008. Socio-ecological explanations for crowding-outeffects from economic field experiments in southern Africa.Ecological Economics.