Does culture matter to prosocial behavior ? Evidence from a 1 cross-ethnic lab experiment

46 Abstract: Objectives: Recent investigations have uncovered large, consistent 47 deviations from the predictions of Homo economicus that individuals are entirely 48 self-regarding. Our study undertook a cross-cultural study of behavior search for the 49 evidences of other-regarding behaviors and its ethnic difference, and accounted for by 50 anatomy of culture. 51 Method: This study recruited 90 subjects of three ethnic groups from market 52 trade-based (ethnic Han), nomadism-based (ethnic Zang) and agriculture-based 53 (ethnic Bouyei) areas in China and conducted public good provision experiment with 54 stranger-treatment design. 55 Results: Under the assumption of self-regarding preferences, the Nash equilibrium is 56 zero contribution by all in public account using backward induction. However, we 57 found contributions did not reduce to zero over all three sessions. Besides, the 58 differences in contributions between ethnicities strongly depended on the degree of 59 ethnic dominance, and Zang harbored the strongest reciprocal preference generally 60 over all group structures. A particular set of measurable factors was identified as 61 proxies for cultural influences on behavioral differences observed in experiments 62 between ethnicities. The results showed all of the cultural factors accounted for the 63 behavioral differences between the ethnic Han and the other two minor ethnicities. 64 However, behavioral difference between minor ethnicities was attributed to group 65 structure only. 66 Conclusions: (1) People may harbor various forms of prosocial emotions in economic 67 affairs, and especially exhibit stronger at the initial phase rather than what canonical 68 model assumes. (2) Behavioral differences between ethnicities are prominent and can 69 be explained by differences in cultural influence. 70 71

sourced from WVS and regarding family status information, to combine with our own 126 questions (Prosocial Preference Survey, PPS. See Table 7). All the questions were 127 grouped into five independent sections that can provide measurements of dimensions 128 of culture in terms of prosocial preferences 1 : people's attitudes toward participation, 129 outgroup rejection (desirability as neighbors), religion, market interaction, and 130 centralization of power of a family. After the completion of the experimental game, 131 we distributed the questionnaire to collect the survey data from every subject during 132 an interval of 30 to 40 min. Then, we assess whether the differences in various 133 corresponding dimensions of culture exist between ethnicity. This study is intended to 134 account for the ethnic behavioral difference by more enriched anatomy of culture 135 (including both potential cultural dimensions-attitude and objective way of economic 136 life-according to the results of related studies) in order to reveal the pattern of cultural 137 influence on the ethnic behavioral difference more completely. 138 This paper is structured as follows: Section 2 introduces our experimental design. 139 Section 3 first provides the results of investigations into behavioral differences in  144  145  146  147  148  149  150  151  152  153  154  155  156  157  158  159  160  161  162  163  164  165  166  167  168  169 2. Experimental Design 170 The experiment was conducted at the Southwest University for Nationalities, 171 China. A total of 90 subjects were equally divided between three ethnicities. Our 172 sampling has the following considerations. First, we selected subjects from the ethnic 173 Han group, which is the majority, and from two ethnic minority groups, the ethnic 174 Zang, which is the largest minority group, and the ethnic Bouyei, which is a relatively 175 small minority group in China, to make a greater variation in ethnic population scales 2 .  The whole experiment was divided into two phases with 45 participants in each 189 phase. In the first phase, we conducted 3 consecutive treatment sessions, each 190 composed of 10 decision-making rounds. In other words, participants played 30 191 rounds of public good provision games in total. In the first treatment session, all the 192 subjects were randomly divided into 9 groups, and every group had 5 subjects who 193 were aware that they played in a group comprised of different ethnicities (labeled 194 'diverse group'). In the second treatment session, the subjects were randomly divided 195 within a sample of their own ethnicity (labeled 'homogeneous group'). In the third 196 treatment session, we firstly randomly selected 9 subjects equally from three 197 ethnicities and then matched every subject with 4 other subjects different from his/her 198 ethnic affiliation from the rest of the subjects (labeled 'one ethnicity dominant group').

199
To avoid the order effect, in the second phase, we conducted the treatment sessions in chance that subjects will meet in multiple rounds, and the assumption that subjects 205 treat random strangers designs as if they were one-shot experiments is false; thus, a 206 reputation effect may develop. Therefore, the group structures were public 207 information over all sessions, although the subjects were not allowed to obtain the 208 information on recruitment numbers of any ethnicity and types of ethnicities to make 209 them feel there was no chance that they would meet the same person in any other 210 round to rule out reputation effects as well as effects of ethnic stereotype, and thus 211 develop an instinct towards prosocial behavior (Fershtman and Gneezy, 2014).

233
In this section, we provide detailed descriptions and statistical tests of the results. 234 We focused on the behavioral differences from the experiment in Section 3.1. In 235 section 3.2, we assessed to what extent the corresponding indicators from PPS were 236 responsible for the behavioral differences associated with ethnicity observed in the 237 experiment. A description of the variables is presented in Table 1. 238 4 'RMB' is the Chinese currency, and the exchange rate was 6.1 RMB per dollar in November, 2014. Herein, we provide both values as RMB (US$). 5 The final payoff each subject received was determined by the total payoff of the whole game (e.g., Putterman L and Anderson, 2006; Neugebauer, Perote, Schmidt and Malte, 2009) rather than in a random manner (e.g., Anderson, Mellor and Milyo, 2008) because in each round, participants may believe that there is a low possibility (1/30) that this round will be selected and hence do not treat it carefully. Hence, the final payoff is determined by the average payoff of 30 rounds in our experiment.  However, another possible reason to explain the reduction in contribution over time is that subjects are willing to punish free-riders. The less contribution he/she makes to the group account, the less payoff is received by the free-riders from free-riding. This is a potential way to punish free-riders in public good experiments without a punishment mechanism design, and humans reciprocate wrongs by harming the offender, even at a cost to themselves (Fehr and Gä chter, 2000; de Quervain, Fischbacher, Treyer, Schellhammer, Schnyder, Buck, Fehr, 2004). However, we believe subjects' motivation to reduce contributions is out of gradual mistrust in others rather than punishment of free-riders, according to our informal callback survey that inquired about the motivation of subjects to reduce contributions. Han. 346 We divided the panel data by ethnic affiliation and estimated the following  We were able to trace unconditional prosocial preferences by specifying lagged 373

Results from experiment on behavioral differences
AVReal it-1 in regression models. The lagged AVReal it-1 was expected to be 374 uncorrelated with CB because the groups were randomly assigned in each round.
By decomposing culture, The results showed that the cultural factors also

472
(3) Actually, we had not addressed GS (group structure), which we viewed as a 473 very important implicit cultural factor for measuring ethnic identity (ethnic prejudice). 474 We generated various composition of groups in terms of ethnic proportions to structure-bias') may play a more important role in determining behavioral patterns 530 than 'individual-bias'(due to identity or prejudice to individuals out of his/her ethnic 531 affiliation as some studies claim (Becker, 1957(Becker, , 1993Fershtman and Gneezy, 2001)) 532 because we found that subjects of the two minor ethnicities acting as the dominant 533 ethnicity in GS3 contributed much less than they did in GS1. Meanwhile, we also 534 found that the ethnic Han contributed more in GS1 than in GS2, in contrast to the 535 argument that ethnic identity is a means to create boundaries that enable a group to 536 distance themselves from one another (Barth, 1969). Additionally, by using guess 537 values as an independent variable to isolate reciprocity preferences from other        The instructions were read aloud by an experimenter as the students followed 886 along on their computer screens.

887
This is an experiment, funded by a research foundation to study decision making. The instructions are simple. If you follow them carefully, you may earn a considerable amount of money which will be paid to you in cash at the end of experiment and the amount you earn will depend on you and other's decisions. Please make sure you understand the decision process and remember any communication is forbidden.
Group You are about to participate experiment of a group decision-making that consists of three sessions, and every session includes 10 rounds, in other words, you will complete 30 rounds. During each round, you will be placed in a group with other four participants (a group of five). You will not know the identities of the other four members of your group in any given decision round, nor will you be told their identities after the experiment is over. At the beginning of each round, groups will be randomly assigned that you have no chance to meet the same person in any other round, i.e., group composition will be randomly changed from round to round. Moreover, you will not know additional information that we will not provide during the whole process.
Earnings You will receive an initial endowment of 50 token (1 token= 1RMB (0.16US$)) in each round and have to decide on the allocation of your endowment between a private and public good. Each token placed in private one earns one token back while each token placed to public good earn 0.5 times token to each member of group. Your payoff will be determined as: In addition to, you will be asked to guess the mean group contribution after decision-making in allocation in each round. Your payoff from guessing will be determined as follows (in RMB): However, the calculation may be kind of complicated, note that the closer your guess is to the average group contribution, the higher is your payoff. Your total payoff in each round includes the payoff from the group decision as well as from guessing. At the end of the experiment, your earning is the average total payoff in 30 rounds. In each round, you will allowed to have 2 minutes to make decision, and if it is not enough, please let us know and more minutes will be allowed. Scenarios You belong to a different ethnicity. The experiment includes three sessions and each session corresponds to a single scenario. The scenario in the first session is that all of you are randomly divided into several groups and have to be aware that you play in a group probably with participants from different ethnicities. In the second session, you play in a group in which all the other participants belong to the ethnicity of your own. At the beginning of each round in third session, we will randomly select several participants. If you are selected, you will play in a group with other four participants belong to an ethnicity different from yours, and if not, you will play in a group with four other participants, only one of whom belongs to a different ethnicity from yours.