University of Pennsylvania ScholarlyCommons Operations, Information and Decisions Papers Wharton Faculty Research 5-1999 Gender and Culture: International Experimental Evidence From Trust Games Rachel Croson University of Pennsylvania Nancy Buchan Follow this and additional works at: https://repository.upenn.edu/oid_papers Part of the Education Economics Commons, Income Distribution Commons, and the Other Education Commons Recommended Citation Croson, R., & Buchan, N. (1999). Gender and Culture: International Experimental Evidence From Trust Games. American Economic Review, 89 (2), 386-391. http://dx.doi.org/10.1257/aer.89.2.386 This paper is posted at ScholarlyCommons. https://repository.upenn.edu/oid_papers/292 For more information, please contact [email protected]. Gender and Culture: International Experimental Evidence From Trust Games Disciplines Education Economics | Income Distribution | Other Education This journal article is available at ScholarlyCommons: https://repository.upenn.edu/oid_papers/292 Gender and Culture: International Experimental Evidence from Trust Games By RACHEL CROSON AND NANCY BUCHAN* Gender is rarely included as a factor in eco- sending money, proposers are trusting that nomics models. However, recent work in their partners will return some money to them. experimental economics, as well as in psy- In addition, 24 out of 32 of responders who chology and political science, suggests that received money returned some (the average gender is an important determinant of eco- amount returned was $4.66). In returning nomic and strategic behavior. money, responders are reciprocating the pro- We examine gender differences in bargain- poser's actions. ing using the ``trust game'' introduced by In this paper we look for gender differences Joyce Berg et al. (1995).1 In this two-person in this game. We use data previously collected game, the ``proposer'' is given a choice of from four countries (the United States, China, sending some, all, or none of his or her $10 Japan, and Korea) and report gender differ- experimental payment to an anonymous part- ences in proposer (trusting) behavior and re- ner, the ``responder.'' The experimenter tri- sponder (reciprocating) behavior. We ®nd no ples any money sent. The responder then signi®cant effect of gender on amount sent by chooses how much of his or her total wealth proposers (trust behavior). However, we ®nd (his or her $10 experimental payment plus the that women return (reciprocate) signi®cantly tripled money) to return to the proposer. Any more of their wealth than men, both in the money the responder does not return may be United States and internationally. kept (thus the responder is playing a dictator game with his or her endowment plus three I. The Impact of Gender in times the amount the proposer sent ) . The Previous Experiments unique subgame-perfect Nash equilibrium is for the proposer to send no money and for the Catherine Eckel and Philip Grossman responder to return none. (1999) review a budding literature on gender For U.S. subjects, Berg et al. found that 30 in experimental economics. For purposes of of 32 proposers deviated from this economic this paper, we focus on their discussion of dif- equilibrium and sent some money to their part- ferences in bargaining. ners (the average amount sent was $5.16). In In ultimatum games in the laboratory, gen- der has been observed to in¯uence a variety of decisions.2 Eckel and Grossman (1998b) dem- onstrate chivalry (men accept lower offers * Operations and Information Management Depart- from women than from men) and solidarity ment, The Wharton School, University of Pennsylvania, (women accept lower offers from women than Philadelphia, PA 19104-6366, and School of Business, from men). Sara Solnick (1998) ®nds, in con- University of Wisconsin, Madison, WI 53706-0394, re- spectively. Data in this paper were originally collected for trast, that players of both sexes demand more a cross-cultural experimental study, reported in Buchan et al. (1998). The authors thank Eric Johnson for his help and support, as well as Bingfu Chen, Wujin Chu, and Hotatka Katahira, and their research assistants, for their 2 In the ultimatum game, one player (the proposer) help in collecting the data for this research. makes an offer to another (the responder) of how to divide 1 This game is similar to the trust game in David Kreps a ®xed amount of money. The responder can accept or (1990) and the peasant±dictator game in John Van Huyck reject the proposer's offer. If the offer is accepted, the et al. (1995). All have the same prediction that play money is divided as proposed; if the offer is rejected, both should end immediately, even though strict Pareto im- players earn zero. The unique subgame-perfect equilib- provements to payoffs can be found in later stages. For a rium of this game is for the proposer to offer the responder detailed comparison of the games see Berg et al. (1995). « and for the responder to accept. 386 / 3y16 my69 Mp 386 Friday Dec 10 08:21 AM LP±AER my69 VOL. 89 NO. 2 GENDER AND ECONOMIC TRANSACTIONS 387 from women than from men. Both studies icant gender difference in the riskless reci- found that offers were lower to women than procity decisions and no difference in the risky to men, and that offers from women and men trusting decisions. were not signi®cantly different. Other researchers have investigated gender ef- II. Experimental Design and Procedure fects in the dictator game.3 Eckel and Grossman (1998a) found that women give more than men The experimental design used in this study in these games, while Gary Bolton and Elena involves an examination of culture, commu- Katok (1995) found no signi®cant difference. nication, and the social distance between play- James Andreoni and Lise Vesterlund (1998) ers. For purposes of this study, these factors compared gender behavior in dictator games as are treated as nuisance variables, and are con- the monetary value of the tokens being divided trolled for in the analysis. A complete dis- was varied among players. They found that cussion of the experimental design and women gave more overall and were more likely procedures can be found in Buchan et al. to divide tokens evenly despite different mone- (1998); here we present a shortened version. tary values, while men became less generous as Experimental instructions are available from the value of their tokens increased relative to the the authors upon request. value of the responder's tokens. Finally, Eckel Subjects are randomly assigned to the roles and Grossman (1996) examined gender differ- of proposer or responder and are directed to ences in a punishment game, where subjects separate rooms. There they receive instruc- could choose to divide evenly a $10 (or $12) tions for the trust game and are paid their en- pie with someone who had previously been un- dowment in local currency. Proposers are generous with another subject, or an $8 pie with instructed to place any money they wish to someone who had previously been generous. send to their partner in an envelope. Monitors They found that women were at least as likely collect the envelopes and take them to the ex- as men to punish ungenerous counterparts by perimenter, in a different room, who records choosing to divide the $8 pie. the amount of money sent. She then triples the As Eckel and Grossman (1999) have ob- amounts sent by proposers, places the tripled served, the ®ndings regarding gender seem to money into envelopes, and sends these enve- be conditional on the level of risk present in lopes into the respondents' room via another the experiment. In decisions where risk is in- monitor. The respondents receive their enve- volved, such as for the proposer in ultimatum lopes, and decide how much of their own ex- games, there appear to be no systematic dif- perimental fee plus any tripled money received ferences in behavior across genders. However, to return to their partners. Monitors collect the for decisions involving no risk, such as for dic- envelopes from the responders and give them tators or ``punishers,'' women tend to be more to the experimenter. The experimenter records generous and socially oriented in their behav- any amounts returned, places the money back ior. In this paper, we examine behavior in an into the proposers' original envelopes, and experiment involving both risky and riskless sends them back to the proposers' room for decisions. Proposers take a risk by sending distribution. The experiment is then con- money to the responder. Responders face no cluded; subjects turn in a post-experimental risk when deciding how much money, if any, questionnaire and leave with their earnings. to return. Our results are consistent with Eckel This procedure, though elaborate, ensures a and Grossman's distinction. We ®nd a signif- double-blind experiment. Throughout the ex- periment, subjects remain unaware of their partners' identities (and thus their genders), and the experimenter who is recording 3 In the dictator game, one player, the allocator, is given amounts sent and returned is also unaware of a ®xed amount of money to divide between himself and the subjects' identities. The anonymity af- another player, the recipient. The allocator chooses a di- vision, and the money is divided as proposed. This is not forded by this procedure helps to reduce in- a game in the formal sense, but rather an individual de- clinations on part of subjects to ``please the cision problem. experimenter'' (i.e., to behave in what they / 3y16 my69 Mp 387 Friday Dec 10 08:21 AM LP±AER my69 388 AEA PAPERS AND PROCEEDINGS MAY 1999 believe is the manner expected by the experi- TABLE 1ÐGENDER BY ROLE AND COUNTRY menter). Thus a double-blind experiment pro- Number of subjects vides an environment in which self-interested Role and behavior is as uninhibited as possible within a gender China Japan Korea United States Total controlled setting (Elizabeth Hoffman et al., Proposers 1994).