Religion, ethnicity and cooperation: An experimental study

Journal of Economic Psychology 45 (2014) 33–43

Contents lists available at ScienceDirect

Journal of Economic Psychology journal homepage: www.elsevier.com/locate/joep

Religion, ethnicity and cooperation: An experimental study Swee-Hoon Chuah a, Robert Hoffmann a,⇑,1, Bala Ramasamy b,2, Jonathan H.W. Tan c,3 a b c

School of Economics, Finance and Marketing, RMIT University, 445 Swanston Street, Melbourne, Victoria 3000, Australia China Europe International Business School, 699 Hongfeng Road, Pudong, Shanghai 201206, People’s Republic of China Nottingham University Business School, University of Nottingham, Jubilee Campus, Wollaton Road, Nottingham NG8 1BB, UK

a r t i c l e

i n f o

Article history: Received 11 March 2013 Received in revised form 27 January 2014 Accepted 27 July 2014 Available online 3 September 2014 JEL classification: C72 C91 D64 Z12

a b s t r a c t We investigate how cross-cutting ethnic and religious identities as well as the strength of individual religiosity and fundamentalism affect individual cooperation. In a repeated prisoner’s dilemma experiment, information about subjects’ religious and ethnic identities was either revealed or concealed to examine the individual and joint effects of these influences on subject decisions. While subjects’ knowledge of others’ religious and ethnic difference has no net effect on their cooperativeness, the awareness of similarity increases it. Subject religiosity and fundamentalism have no independent effect on cooperation, but they enhance ethnic and religious intergroup effects. Ó 2014 Elsevier B.V. All rights reserved.

Keywords: Religion Intergroup effects Prisoner’s dilemma Experiments

1. Introduction In recent years religion has been identified as a significant macroeconomic influence on performance differentials and trade links between nations (Barro & McCleary, 2003; Guiso, Sapienza, & Zingales, 2003, 2009). A small but growing literature within experimental economics has since examined the effect of religion on individual behaviour to try to account for these findings (for recent overviews, see Hoffmann (2013) and Tan (in press)). A number of studies have examined whether religion enhances cooperation by instilling pro-social values. Orbell, Goldman, Mulford, and Dawes (1992) found that more religious people were no more cooperative in a prisoner’s dilemma. Similarly, Sosis and Ruffle (2003) found no differences in cooperation between religious and secular kibbutz members respectively in a common-pool resource dilemma. In the two studies by Anderson, Mellor, and Milyo (2010) (Anderson & Mellor, 2009; Anderson et al., 2010) the religious affiliations and service attendance of subjects did not significantly explain their public good contributions. Ahmed and Salas (2009) found no differences in public good contributions between religious and non-religious subjects in three countries. ⇑ Corresponding author. Tel.: +61 (0) 3 9925 5447. E-mail addresses: [email protected] (S.-H. Chuah), [email protected] (R. Hoffmann), [email protected] (B. Ramasamy), [email protected] (J.H.W. Tan). 1 Funding through the British Academy’s Committee for South East Asian Studies (grant number RL2102) is gratefully acknowledged. 2 Tel.: +86 (0) 2128 90 5890. 3 Tel.: +44 (0) 115 846 6602. http://dx.doi.org/10.1016/j.joep.2014.07.002 0167-4870/Ó 2014 Elsevier B.V. All rights reserved.

34

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43

These results suggest that the strength of religious values alone does not influence individual cooperation. However, religion may affect behaviour through religious group identity. Religious difference can cause prejudice (Jackson & Hunsberger, 1999) while common religion may engender ingroup favouritism (Irons, 2001). Tan and Vogel (2008) found mutually greater trust among more highly religious Christian subjects. In the study by Chuah, Fahoum, and Hoffmann (2013), Muslim and Hindu communities in Mumbai trusted their religious ingroup members more. Chuah, Hoffmann, Jones, and Williams (2007) found ingroup favouritism in ultimatum games played between British and Malaysian subjects, however their result may be driven by shared ethnic rather than religious group identities. Observing the effect of religion on individual behaviour is therefore hampered by the confounding simultaneous influences of religious values as well as overlapping religious and ethnic group membership. In this paper we report an experiment designed to disentangle these variables by testing for their individual and combined effects. In particular we observe cooperation among a multi-cultural subject pool where religion and ethnicity are cross-cutting social categories that are either are revealed or concealed in different experimental conditions. In addition, we measure the strength of religious values using multi-dimensional scales (religiosity and fundamentalism) shown to be valid across different religions. These measures enable us to examine whether strength of religious sentiment affects behaviour independently or in concert with religious group identity. Our experiment robustly demonstrates the effects of the interplay between religious values and religious and ethnic affiliations on individual cooperation. The rest of the paper is organised as follows. Section 2 reviews the theoretical framework we employ. Section 3 outlines the design of the experiment and our procedure in conducting it. Section 4 reports the results. Section 5 discusses and concludes. 2. Theory The voluminous literature on social categorisation suggests alternative hypotheses for the simultaneous effects of religious and ethnic identity as well as religiosity on the cooperativeness of our subjects. Intergroup effects, i.e. the preferential treatment of similar people relative to those who are different are well understood and documented for a single social category. They are however more complex when multiple, potentially cross-cutting social categories exist as in realistic settings (e.g. Stangor, Lynch, Duan, & Glas, 1992). For example, religious affiliation and ethnicity typically overlap only imperfectly such that individuals may share one but differ in the other group identity. For two dichotomous categories this graduates social distance between interacting individuals in terms of double ingroup (II) or outgroup identities (OO) as well as partial commonality (IO or OI). A number of models exist that suggest how these latter, ambiguous cross-cutting conditions affect intergroup behaviour. Hewstone, Islam, and Judd (1993) classify these models based on their predicted effects for the four conditions which they call contrasts (see Table 1). All models surveyed agree that condition II results in ingroup favouritism (+) while OO engenders discrimination (). However they differ in the predictions for the cross-cutting conditions IO and OI compared with the two extremes. All possible combinations of positive, negative and neutral (0) effects for the two cross-cutting conditions on cooperation have support from particular contrasts proposed by different models. Overall neutral effects for both IO and OI (contrast 1) arise when the individual influences of the two categories are independent and/or additive. For example if a target person belongs to the same religion but a different ethnicity their respective effects cancel out. Positive effects for both IO and OI are predicted by category conjunction models where shared identity in only one category results overall in an ingroup assessment (contrast 2). In this view another person is considered an ingroup member as long as they share either religion or ethnicity. The same prediction is made by category differentiation models which posit that outgroup discrimination can be eliminated when the differentiating social category concerned is crossed with a shared one. Conversely, negative effects for both (contrast 3) result when dissimilarity in one category suffices to trigger outgroup discrimination irrespective of the other. Here an overall outgroup assessment results from either religious or ethnic difference. When only one of the two categories is used and dominates the other this results in a positive effect for one cross-cutting condition and a negative one for the other (contrast 4). For example Hewstone et al. (1993) find category dominance for religion over nationality in their experiment. In hierarchical category models the prior assessment of one category affects that of another which is attended to subsequently. This arises for instance if a person is identified as an ingroup member according to the first category (+) and, due to approbation, receives further scrutiny which reveals outgroup identity according to the other category () resulting in an

Table 1 A priori contrasts for the four intergroup conditions from two dichotomous social categories. Adapted from Hewstone et al. (1993). Condition

II

IO

OI

OO

Cross-cutting categories are

Contrast Contrast Contrast Contrast Contrast Contrast

+ + + + + +

0 +   0 0

0 +  +  +

     

Independent/additive Differentiating/conjunctive (similarity) Conjunctive (dissimilarity) Dominating Hierarchical (approbation) Hierarchical (derogation)

1 2 3 4 5 6

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43

35

overall neutral assessment. Had the other person first been identified as belonging to an outgroup then shared ingroup identity in the second category would have been overlooked, yielding a negative assessment overall. Stangor et al. (1992) found that immediately accessible categories such as ethnicity and gender lead to instant, spontaneous categorisation. For example if two individuals associate based on a immediately apparent shared ethnicity then any religious difference between them would be discovered. If ethnicity is not shared then a commonality in terms of religion might not. The hierarchical model based on approbation thus produces contrast 5. Finally, if outgroup and not ingroup members receive more attention to the second category (due to derogation) we obtain contrast 6. The resulting six contrasts in Table 1 provide alternative hypotheses for the effects of the four possible social identity conditions we study here, i.e. those arising from subjects’ ethnic and religious identities, on their mutual cooperation. In particular, those contrast supported by our results provide potential interpretations and explanations of them. Additional theories suggest the effect of our third variable, the strength of individual religious sentiment. Tajfel and Turner (1979) propose that a potential motivation behind intergroup behaviour is the desire to create a positive group identity as a source of individual self-esteem. Ingroup favouritism and outgroup discrimination sustain a favourable view of one’s group compared to others which elevates ones’s self view. As a result is has been suggested that both behaviour patterns are positively associated with the strength of an individual’s identification with the group (Farnham, Greenwald, & Banaji, 1999; Smurda, Wittig, & Gokalp, 2006). Individuals seek to avoid cognitive dissonance and to balance their self view and view of their group. High group identification individuals in this view have a greater need to for a positive view of the group and therefore for the intergroup behaviour that supports it. In our present context greater individual religiosity is associated with higher identification with the religious ingroup and is therefore expected to enhance intergroup behaviour based on religion. 3. Experiment We designed an experiment to assess the effects of cross-cutting ethnic and religious group membership as well as religiosity on the cooperativeness of subjects. In particular, guided by the theory, the experiment was designed to tease apart the former two effects and to assess the moderating influence of the latter. The experiment addresses the following specific research questions: do religious and ethnic group membership and religious values have independent effects on cooperation? Are there joint effects between them? What is the direction of any effects? We provide answers by observing subject cooperation when religious affiliations and ethnic identities were either revealed or concealed and relate it to their religiosity and religious fundamentalism. The details of the experimental design and procedure are outlined in the following. 3.1. Subjects The experiment was conducted in Malaysia, a developing South-East Asian nation of some 27 million people with a percapita GDP roughly a third of that of the USA4 and equatorial climate. The Malaysian Ringgit (RM) is the national currency and at the time of the experiment traded at an exchange rate of 3.7 for 1 US $. Malaysian society is multi-cultural and consists of four main ethnic groups associated with distinct cultural traditions, languages and religious beliefs (see Table 2). Its origins lie in mass immigration especially from Southern China and India under British colonial rule in response to local labour shortages. The resulting ethnic division of labour broadly into Malay agriculture, Chinese commerce and Indian industrial manual labour was promoted by British colonial authorities and led to lacking integration, increasing economic disparities and open ethnic conflict between the ethnic groups after independence in 1957. Government subsequently adopted an economic programme to minimise ethnic tensions through affirmative action in favour of the Malay majority. As a result, ethnicity permeates Malaysian social, political and economic life to this day (Verkuyten & Khan, 2012). While the Malay majority is politically and culturally dominant, the ethnic Chinese control much of the economy (Gomez, 2003). The Malaysian business community generally and managerial/professional occupations in particular therefore do not reflect the ethnic composition of the overall population but instead are dominated by the ethnically Chinese (Bhopal & Rowley, 2005, p. 563). Most Malaysians, especially urban and educated ones interact across ethnic lines daily however ethnic integration is limited starting with segregation enshrined in the school and university system which continues in the workplace and private sphere (Bhopal & Rowley, 2005; Montesino, 2012). The educated members of all ethnic groups have good spoken and written English due the colonial past which is often used as the language of communication across ethnic groups. We recruited total of 96 undergraduate subjects at the University of Nottingham Malaysia Campus using flyers, posters and class announcements. The university is private and draws students from all ethnic groups of the country but is predominated by ethnically Chinese due to their socio-economic status. Despite the ethnic divisions in Malaysian society discussed earlier, students have the opportunity to mix socially across ethnic groups in a variety of student societies as well as formal classes. English is the only language of instruction. The subjects we recruited were from a wide range of faculties and years of study. The mean age of subjects was 20.6 years and 55.3% were male. In terms of socio-economic status, our subjects self-categorised in the middle of the income and social class strata, and as relatively urban.

4

The 2012 ppp-adjusted figure is 17,143 current international $ for Malaysia and 49,965 for the USA. Source: World Bank Development Indicators.

36

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43 Table 2 Ethnic origins, religious denominations and sentiment (religiosity and fundamentalism) of subjects and Malaysian society. Christianity includes both Catholics and Protestants. Buddhism includes Confucianism, Taoism and other traditional Chinese religions. Source: 2012 Statistical Handbook; Population Distribution and Basic Demographic Characteristic Report 2010, Department of Statistics, Malaysia.

Ethnicity Chinese Indian Malay Other All

Malaysia

Subject pool

%

n

%

FUND

RELI

24.2 7.3 54.7 13.8 100.0

52 13 12 19 96

54.2 13.5 12.5 19.8 100.0

4.51 4.78 7.19 5.41 5.07

3.13 3.73 4.06 3.52 3.40

0.000

0.001

3.96 5.71 4.23 7.14 3.35 3.24 5.07

2.94 3.92 3.63 4.01 2.88 1.96 3.40

0.000

0.000

ANOVA p Religion Buddhism Christianity Hinduism Islam Other None All ANOVA p

21.1 9.2 6.3 61.3 1.4 0.7 100.0

36 21 10 22 1 6 96

37.5 21.9 10.4 22.9 1.0 6.3 100.0

3.2. Task These subjects were observed in a total of eight experimental sessions with twelve subjects in each. Within each session every subject was matched with every other subject in a round-robin tournament to play the ten-round prisoner’s dilemma game used by Andreoni and Miller (1993) using a computer interface. In each round of this game, two players simultaneously chose between cooperate (c) or defect (d) with payoffs shown in Fig. 1. The equilibrium prediction is mutual defection in every round, Pareto dominated by mutual cooperation. As a result each of the twelve subjects in a session played eleven ten-round games in total. In every but one of these games information about the co-players’ age, ethnicity as well as religious affiliation was provided as shown in the screenshot in Fig. 1. Age was used as a distractor to mitigate experimental demand effects. For every subject the sixth game involved no information about the co-player. We use game 6 with co-players of unknown ethnic and religious affiliation as the baseline condition (UU) in our data analysis. Consequently every game took place under one of five experimental conditions in terms of ethnic and religious social distance between the two players (see Table 3). These include the double outgroup (OO) and ingroup (II) conditions as well as the cross-cutting conditions IO and OI where one but no the other identity is shared. Subjects were randomly allocated to sessions to minimize the possibility of subjects playing with those they knew. Each individual subject was admitted to the laboratory at the point of his or her arrival and randomly assigned to a partitioned computer terminal. This feature was designed to avoid subjects’ mutual identification through their ethnicity or religious affiliation suggested by physical cues. After reading written experimental instructions subjects answered a quiz to ensure

Fig. 1. Screenshot of the experimental task. Action A corresponds to cooperation and B to defection.

37

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43 Table 3 Mean cooperation in % over all rounds in a game and in first rounds by experimental condition. Condition

UU OO OI IO II All

Games

48 27 31 104 78 528

Co-player

Cooperation

Ethnicity

Religion

All rounds

Round 1

Unknown Different Different Same Same

Unknown Different Same Different Same

30.8 30.7 40.0 39.5 45.4 35.2

44.8 41.9 51.6 55.6 55.8 47.7

understanding. Subjects with incorrect answers were individually instructed. The task commenced when all subjects were present and had successfully completed the quiz. One of the eleven games was randomly chosen for payment. Subjects were paid RM 0.40 (US $ 0.13) for each payoff point they had accumulated at the end of the ten rounds of that game. Average earnings were RM 20.50 (US $5.55) for about ninety minutes of experimental play. 3.3. Questionnaire In addition to the task we also used a self-administered paper questionnaire to elicit subjects’ demographic characteristics as well as their religiosity and fundamentalism. We also included non-religious questions relating to subjects’ social values such as trust and Machiavellianism (Glaeser, Laibson, Scheinkman, & Soutter, 2000; Gunnthorsdottir & McCabe, 2002), which served as distractors. Subjects collected the questionnaire and returned the completed copies about two weeks prior to the experiment. This time lag was designed to minimise bias across phases of the experiment. Each subject was paid RM 10 (US $2.70) for completing the questionnaire. Religiosity is commonly conceptualised in terms of four dimensions (Glock & Stark, 1965). These comprise (a) theological (conformity of personal beliefs to the tenets of one’s religion), (b) ritual (the degree of participation in religious activities), (c) experiential (the extent of personal religious experiences), and (d) consequential (the impact of religion on everyday activities) religiosity. We used Rohrbaugh and Jessor’s (1975) 8-item, 5-way Likert scale instrument, which has been shown to be reliable across different religions including the ones among our subject pool (Hill & Hood, 1999, p. 307). It provides a measure that permits testing whether cooperation is positively related to the intensity of religious belief and practice. We also measured strength or religious sentiment in terms of fundamentalism, the exclusive commitment to one’s religion and belief in its absolute literal truth and authority (Altemeyer & Hunsberger, 1992; Kirkpatrick, Hood, & Hartz, 1991). Fundamentalist conviction guides adherents’ social interactions and therefore leads to religious intergroup behaviour (Jackson & Esses, 1997). We used Altemeyer and Hunsberger’s (1992) 20-item, 9-way Likert scale instrument which again has cross-religion reliability (Hunsberger, 1996). Half of the items therein concern attitudes towards other religions specifically, well suited for our context of interactions within a multi-religious subject pool. 4. Results We obtained a data set containing subjects’ demographic characteristics, their scores for religiosity and fundamentalism as well as their own and their co-players’ decisions for every round of eleven games under the five experimental conditions

Fig. 2. Mean cooperation in the five conditions by round.

38

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43 Table 4 Multi-level random effects logit regressions for the effects of cultural similarity on cooperation. Cooperation per round as dependent variable in all models except model 4, with cooperation in round 1. Standard errors in parentheses. Model

(1)

(2)

(3)

(4)

Game (g)

0.040⁄⁄⁄ (0.008) 0.185⁄⁄⁄ (0.009) 0.080 (0.089) 0.208⁄⁄⁄ (0.070) 0.648⁄⁄⁄ (0.112) 0.157 (0.142)

0.041⁄⁄⁄ (0.008) 0.194⁄⁄⁄ (0.009) 0.091 (0.096) 0.281⁄⁄⁄ (0.076) 0.691⁄⁄⁄ (0.117) 0.370⁄⁄ (0.152) 0.692⁄⁄⁄ (0.074)

0.040⁄⁄⁄ (0.008) 0.169⁄⁄⁄ (0.029) 0.150 (0.184) 0.266⁄ (0.138) 0.596⁄⁄⁄ (0.227) 0.089 (0.288)

0.014 (0.028)

0.027 (0.325) 0.262 (0.263) 0.808⁄⁄ (0.404) 0.176 (0.517)

0.465 (0.311) 1056

Round (r) info ethsim relsim ethrelsim agesim ethsim  r

Constant

0.514⁄⁄ (0.193)

0.419 (0.226)

0.011 (0.022) 0.013 (0.031) 0.009 (0.036) 0.012 (0.046) 0.600⁄⁄ (0.230)

N

10,560

9240

10,560

info  r relsim  r ethrelsim  r

⁄

Significance level at 0.10. ⁄⁄ Significance level at 0.05. ⁄⁄⁄ Significance level at 0.01.

outlined in Table 3. Cooperation in a particular round is coded as 1 for cooperation and 0 for defection. Table 3 reports average cooperation by experimental condition over all rounds as well as in round one for all games played. Fig. 2 plots subjects’ average cooperation for each of the ten rounds of their games by condition. The general pattern of behaviour we observed, including declining cooperation towards the equilibrium in every condition, is similar to other studies (e.g. Andreoni & Miller, 1993). We analysed how behaviour is affected by our variables of interest, strength of religious values as well as religious and ethnic affiliation. 4.1. Religious and ethnic identity We examine the effects of the experimental conditions on cooperation using multi-level random effects logit regressions. We include subject and session level random effects, with subjects nested in sessions, due to the potential non-independence of observations from repeated play and random rematching.5 To control for the effects of experience, restart and endgame effects our regressions control, where necessary, for the number of the round ðrÞ as well as the game ðgÞ concerned. We use the independent variable info (=0 for unknown and =1 for known co-player ethnicity and religion) to differentiate the baseline condition UU. The dummy variables ethsim and relsim capture same (=1) or different (=0) ethnicity and religion respectively in conditions where co-player identity is known (i.e. =0 if unknown). In addition, we use the interaction variable ethrelsim to test the combined effect of shared ethnic and religious identities beyond their individual influences. For example, for a pair of players of the same known ethnicity and religion, ethrelsim ¼ 1. Table 4 presents the regression results.6 The dependent variable all models with the exception of model 4 is cooperation in a given round. Model 1 shows that cooperation decays over rounds played in a given game, as round is negative and significant, but cooperation does not decrease across games with different co-players as game is positive and significant. Cooperation increases where subjects have the same ethnic or religious identity, as ethsim and relsim are both positive and significant, with the ethnic effect being of smaller magnitude than the religious one.

5 These results are similar to those from using the alternative specification that considers only session-level random effects, which we do not report in this paper. 6 We also conducted a mirror set of models where we drop observations where the co-player’s ethnicity and religion were unknown so as to test the abovementioned effects against the alternative benchmark of cultural dissimilarity in both ethnicity and religion. The results, which we do not report here, are consistent with our interpretations based on the models provided in the main text.

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43

39

Result 1. Relative to the baseline where the ethnic and religious affiliations of co-players are unknown, subjects cooperate more with others who share ethnic or religious identity. Given the use of variables on known shared social identity, namely ethsim; relsim and ethrelsim, the coefficient for info allows us to gauge the effect of known double outgroup identity of a co-player. It is statistically insignificant, providing no evidence for outgroup discrimination in condition OO. Result 2. Relative to the baseline where the ethnic and religious affiliations of co-players are unknown, there is no evidence that subjects cooperate less with others who share neither ethnic or religious identity. The variable ethrelsim allows us to test the effect of double ingroup identity of the co-player II. It is negative, suggesting a limit to how much people discriminate on ethnic or religious grounds. Result 3. There is no evidence that a second shared social identity enhances cooperation beyond the effect of the first. Model 2 is a robustness check of model 1 controlling for age similarity, with agesim ¼ 1 if the co-player is known to be of the same age and =0 if otherwise.7 While we find that cooperation increases with age similarity, our previous findings are robust to the inclusion or exclusion of this control. Taken together, Results 1–3 are in line with contrast 2 in Table 1. Model 3 tests whether the decay of cooperation over rounds (see Fig. 2) is sensitive to ethnic and religious similarity. The model includes the three social similarity variables interacted with round r which measure the marginal effects of similarity on the rate at which cooperation decays. These time interaction terms are statistically insignificant, while ethsim and relsim remain positive and significant. Shared social identity therefore engenders more cooperation starting from round 1. We further test this with model 4, where the dependent variable is cooperation in round 1 of each stage. In support of this interpretation, ethsim and relsim are both positive.8 Models 5 and 6 in Table 5 test if the positive effects of ethnic or religious similarity are indeed due to ingroup favouritism, rather than to the intrinsic cooperativeness of large group members. Model 5 incorporates control variables for ethnic group membership, chinese; indian, and malay (=1 for members and =0 for non-members). Model 6 controls for religious group membership, atheist; buddhist; christian; hindu, and muslim (=1 for members and =0 for non-members). We find that ethsim and relisim stay positive and significant, i.e. our main results are robust to these controls, and there are no significant between-group differences in cooperativeness.9 Models 7 and 8 in Table 5 replace ethsim and relsim with dummy variables for the co-player’s known ethnic group (cpchinese; cpindian; cpmalay, and cpother) in model 7 or religious group (cpatheist; cpbuddhist; cpchristian; cphindu; cpmuslim, and for other religions cpothreli) in model 8. Doing so shows the raw effects of cultural identity of self and other on aggregate. Group biases previously captured by cultural similarity variables are now captured by variables on the cultural identity of self and other. To interpret the results of these two models, we have to consider their sensitivity to the relative frequencies of specific groups in the distribution (e.g. Chinese get more cooperation due to more frequent ingroup interaction) and the extent to which cooperation varies across outgroup co-players (e.g. Chinese cooperating less with Malays and more with other minority ethnicities than with Indians). Model 1 parsimoniously organises these effects. 4.2. Religiosity and fundamentalism We generated the measures RELI for subjects’ general religiosity and FUND for fundamentalism as standardised meancentred averages of the questionnaire responses to the relevant inventory items scaled 1–5, following Marquardt (1980).10 Subjects’ scores (see Table 2) for one of the constructs are highly correlated with the other (Spearman’s q = 0.74, p < 0.0001). ANOVA reveals that both RELI and FUND differ significantly between groups of subjects based on ethnicity as well as on religious affiliation. In general Muslims Malays score higher than other groups on both. For ethnicity, post hoc tests show that these differences are rooted in greater fundamentalism of Malays than all other groups (Scheffe p < 0.005), and greater religiosity of Malays than the Chinese (p = 0.003). Similarly, when subjects are grouped by religious affiliation, Muslims score higher than all other groups on fundamentalism (p < 0.008). Christians score lower on FUND than Muslims but higher than other groups (p < 0.05). Buddhists and Hindus do not differ in fundamentalism (p = 0.983). Turning to religiosity, subjects of no religion unsurprisingly score less than all others (p < 0.002). Buddhists are less religious than the other three religious groups (p < 0.015) which do not significantly differ in religiosity. In view of these individual and group differences, we consider them in the analysis to follow. We test their influence and also the robustness of our earlier findings of ethnic and religious similarity on cooperation controlling for these influences.

7

Tests with agesim drop data from one session with incorrectly recorded co-player ages. In this model, relsim is statistically significant, while ethsim is not. This is possibly attributable to the smaller number of observations for this test where only one round of ten is used, and consistent with the relatively weaker effect ethsim has relative to that of relsim as mentioned earlier. In all the other models, both ethsim and relsim are statistically significant. 9 We are thankful to a reviewer for suggesting these controls, having pointed out that statistically, members of large groups experience more ingroup interaction than members of small groups, and if members of large groups are intrinsically more cooperative than members of small groups then even without ingroup favouritism we would still observe more cooperation in ingroup interactions. 10 Observations of eight (three) subjects were dropped from the regressions because of incomplete religiosity (fundamentalism) questionnaire responses. 8

40

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43 Table 5 Multi-level random effects logit regressions for the effects of ethnic and religious affiliation of self and other on cooperation. Cooperation per round as dependent variable in all models. Standard errors in parentheses. Model

(5)

(6)

(7)

(8)

Game (g)

0.040⁄⁄⁄ (0.008) 0.185⁄⁄⁄ (0.009) 0.082 (0.089) 0.202⁄⁄⁄ (0.070) 0.649⁄⁄⁄ (0.112) 0.155 (0.142) 0.422 (0.429) 0.219 (0.571) 0.795 (0.598)

0.040⁄⁄⁄ (0.008) 0.185⁄⁄⁄ (0.009) 0.081 (0.089) 0.205⁄⁄⁄ (0.070) 0.650⁄⁄⁄ (0.112) 0.158 (0.142)

0.038⁄⁄⁄ (0.008) 0.185⁄⁄⁄ (0.009)

0.038⁄⁄⁄ (0.008) 0.185⁄⁄⁄ (0.009)

Round (r) info ethsim relsim ethrelsim chinese indian malay atheist

0.471 (0.429) 0.093 (0.571) 0.825 (0.600) 1.095 (1.716) 1.575 (1.606) 1.908 (1.622) 1.806 (1.652) 2.195 (1.633)

buddhist christian hindu muslim cpbuddhist cpchristian cphindu cpmuslim cpatheist cpothreli

1.109 (1.732) 1.454 (1.624) 1.894 (1.639) 1.921 (1.665) 2.234 (1.646) 0.465⁄⁄⁄ (0.082) 0.125 (0.089) 0.055 (0.100) 0.010 (0.090) 0.807⁄⁄⁄ (0.125) 1.319⁄⁄⁄ (0.275)

Constant

0.674⁄ (0.401)

1.252 (1.596)

1.264 (1.612)

0.469⁄⁄⁄ (0.088) 0.026 (0.107) 0.311⁄⁄⁄ (0.116) 0.346⁄⁄⁄ (0.099) 0.670⁄ (0.397)

N

10,560

10,560

10,560

10,560

cpchinese cpindian cpmalay cpother

⁄⁄

Significance level at 0.05. Significance level at 0.10. ⁄⁄⁄ Significance level at 0.01. ⁄

We performed regression analysis to test the effects of these religious values on cooperation individually and in combination with religious affiliation as suggested by theory. The results are reported in Table 6. Model 9 extends model 1 by adding RELI, while model 10 does so with FUND. The results of these models indicate that religiosity and fundamentalism are not significantly related to cooperation rates. These effects are robust to the addition of variables interacting info; ethsim; relsim and ethrelsim by RELI (model 11 and 13) or with FUND (model 12 and 14).11 Models 13 and 14 control for age similarity as a robustness test.

11 We conducted alternative tests on game 6 data (condition UU) only which generate null results. Similarly alternatively constructed religiosity variables were never significant explanators of cooperation in their own right.

41

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43

Table 6 Multi-level random effects logit regressions for the effects of cultural similarity, religiosity and fundamentalism on cooperation. Cooperation per round as dependent variable in all models. Standard errors in parentheses. Model

(9)

(10)

(11)

(12)

(13)

(14)

(15)

(16)

game

0.044⁄⁄⁄ (0.008) 0.181⁄⁄⁄ (0.009) 0.247 (0.207) 0.024 (0.091) 0.227⁄⁄⁄ (0.072) 0.673⁄⁄⁄ (0.115) 0.190 (0.145)

0.038⁄⁄⁄ (0.008) 0.180⁄⁄⁄ (0.009)

0.044⁄⁄⁄ (0.008) 0.181⁄⁄⁄ (0.009) 0.197 (0.230) 0.022 (0.091) 0.263⁄⁄⁄ (0.074) 0.613⁄⁄⁄ (0.120) 0.173 (0.151)

0.038⁄⁄⁄ (0.008) 0.181⁄⁄⁄ (0.009)

0.046⁄⁄⁄ (0.008) 0.190⁄⁄⁄ (0.010) 0.079 (0.243) 0.171⁄ (0.099) 0.364⁄⁄⁄ (0.080) 0.663⁄⁄⁄ (0.124) 0.418⁄⁄⁄ (0.160)

0.039⁄⁄⁄ (0.008) 0.189⁄⁄⁄ (0.009)

0.038⁄⁄⁄ (0.008) 0.180⁄⁄⁄ (0.009)

0.101 (0.097) 0.316⁄⁄⁄ (0.078) 0.561⁄⁄⁄ (0.125) 1.221⁄⁄ (0.480) 0.052 (0.112)

0.075 (0.090) 0.220⁄⁄⁄ (0.071) 0.611⁄⁄⁄ (0.117) 0.141 (0.148) 0.027 (0.103)

0.042⁄⁄⁄ (0.008) 0.178⁄⁄⁄ (0.009) 0.128 (0.312) 0.023 (0.091) 0.284⁄⁄⁄ (0.073) 0.572⁄⁄⁄ (0.122) 1.512⁄⁄⁄ (0.463) 0.036 (0.141)

0.137⁄⁄ (0.056) 0.062 (0.045) 0.195⁄⁄⁄ (0.066) 0.308⁄⁄⁄ (0.088) 0.661⁄⁄⁄ (0.074) 0.409 (0.292)

0.069 (0.051) 0.033 (0.038) 0.022 (0.040)

0.115⁄⁄ (0.054) 0.151⁄⁄⁄ (0.046) 0.237⁄⁄⁄ (0.064) 0.335⁄⁄⁄ (0.085)

0.507⁄⁄ (0.253)

0.407 (0.248)

8910

10,230

9460

round (r) RELI info ethsim relsim ethrelsim FUND

0.073 (0.090) 0.220⁄⁄⁄ (0.071) 0.616⁄⁄⁄ (0.114) 0.150 (0.144) 0.073 (0.092)

info  RELI

0.075 (0.090) 0.230⁄⁄⁄ (0.072) 0.524⁄⁄⁄ (0.121) 1.157⁄⁄ (0.454) 0.025 (0.103)

0.248⁄⁄ (0.123) 0.274⁄⁄⁄ (0.093) 0.428⁄⁄⁄ (0.155) 0.635⁄⁄⁄ (0.203)

0.159 (0.117) 0.229⁄⁄ (0.090) 0.361⁄⁄ (0.149) 0.538⁄⁄⁄ (0.194)

ethsim  RELI relsim  RELI ethrelsim  RELI

0.087⁄ (0.052) 0.095⁄⁄ (0.043) 0.166⁄⁄⁄ (0.062) 0.248⁄⁄⁄ (0.082)

info  FUND ethsim  FUND relsim  FUND ethrelsim  FUND

constant

0.415 (0.258)

0.507⁄⁄ (0.254)

0.411 (0.257)

0.503⁄⁄ (0.251)

0.681⁄⁄⁄ (0.075) 0.299 (0.307)

N

9680

10,230

9680

10,230

8360

agesim

⁄ ⁄⁄

Significance level at 0.10. Significance level at 0.05. Significance level at 0.01.

⁄⁄⁄

Result 4. Religiosity and fundamentalism have no independent effect on cooperation. These interaction variables examine whether the strength of religious values moderates intergroup biases as discussed in Section 2. All ethsim or relsim by RELI or FUND interactions are positive and significant, showing that increased cooperation with fellow ethnic or religious affiliates is further amplified by religiosity and fundamentalism. Consistent with the limit effect reported in model 1, these amplification effects are also limited, judging from the negative and significant ethrelsim by RELI or FUND terms. Put differently, the amplification effects of religious values on ethnicity and religious similarity matter, but only to a certain extent, and adding one to another has a non-increasing effect. To check this, we removed the ethrelsim by FUND term in model 15. Both ethsim by FUND and relsim by FUND are positive but no longer significant (the same result holds if we substitute FUND with RELI), further corroborating our interpretation of a limit effect. Result 5. Religious values amplify ingroup favouritism toward others of shared ethnic or religious identity. The info  FUND and info  RELI terms of models 11–14 test whether outgroup discrimination is enhanced by strength of religiosity or fundamentalism. Model 11 and 13 show that the effect of general religiosity on cooperation with double outgroup co-players is negative, but this effect is statistically significant only when we control for agesim. More convincingly, info  FUND is significant in both models 12 and 14, confirming that the negative effect of cultural dissimilarity increases in fundamentalism. This implies that fundamentalist values in particular rather than religiosity in general increases outgroup prejudice. As a final test of robustness, in Model 16 we included both religiosity and fundamentalism to control for their respective effects as proposed by Kirkpatrick et al. (1991). The significance of the interaction effects are further enhanced.

42

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43

Result 6. Fundamentalism amplifies outgroup discrimination toward others of different ethnic or religious identity.

The positive effects of ethsim and relsim stay robust throughout these models with different controls for individual differences in the strength of religiosity and fundamentalism, with statistical significance of p < 0:01 in all models. 5. Discussion and conclusion In contrast to much of previous experimental work our study provides robust evidence for the effects of both religious affiliation and religious values on cooperative behaviour controlling for ethnic group belonging. Biases in cooperation due to religious identity were observed to be at least as strong, if not stronger, than those from ethnic identity. This is surprising to the extent that the latter is often accompanied by visible cues which are thought to have primacy for decision makers when multiple group categories are available (Stangor et al., 1992). This attests to the strength of religious groups in shaping cooperation. Our results for the four social distance conditions in Table 1 are in line with the predictions of contrast 2. Here a single shared social category suffices to generate ingroup favouritism despite cross-cutting additional categories. There are two separate explanations for this effect between which our results cannot discriminate. One is that shared social identity can be used to eliminate outgroup prejudice created by alternative social categories. The other is that one shared social category suffices to create an ingroup identification. Further work in this direction may identify promising policy avenues to reduce social tensions in multicultural societies such as Malaysia. Both ethnic and religious effects constitute ingroup cooperation rather than increased intergroup conflict. Our finding supports the view, for religious groups at least, that group identification is not necessarily associated with outgroup rejection. For instance, Brewer (1999, p. 442) has argued that ‘‘societies characterised by multiple cross-cutting group divisions are more likely to provide a context in which ingroup attachments and loyalties are not necessarily associated with outgroup antagonism.’’ Strength of religious sentiment (i.e. religiosity and fundamentalism) has, as in previous experimental studies of cooperation, no independent effect in simultaneous-move social dilemmas. Our results suggest that its effective role is in the promotion of cooperation in ethnic or religion-based networks similar to many others studied in the social sciences (e.g. kinship groups, guanxi networks and clubs). It would be interesting to investigate finer grained cultural effects, for example differences at the level of specific ethnic groups and interactions between them. Further, the underlying cultural explanations such as differences in ethnic values or religious doctrine are a promising area of further investigation. We leave these issues for future research. In the present paper, we focused on systematic tendencies at the aggregate level controlling for the potential biases of group and individual level differences. Our finding that ethnic and religious similarity engender cooperation was strong throughout the analysis. Both religious and ethnic group biases were amplified by the degree to which subjects identified religiously and, in particular, how fundamentalistic they were. Evidence exists from other studies that religious priming induces the willingness to help needy ingroup members (Saroglou, Corneille, & Van Cappellen, 2005). Fundamentalism is, however, also connected to the type of prejudice shown towards outgroup members or those who challenge religious doctrine (Hall, Matz, & Wood, 2010). We found not only that fundamentalism and religiosity are predictive of higher ingroup favouritism but also that fundamentalism is a good predictor of higher outgroup prejudice. Religious attitudes that position others in relation to the self may result in decreased cooperation at the dyadic level. While outgroup biases are not prominent at the aggregate level, the prejudicial nature of fundamentalism is unmasked when we controlled for general religiosity. The finding that religiosity and fundamentalism amplify group biases is consistent with certain interpretations of social identity theory, i.e. that intergroup behaviour is positively related to the strength of an individual’s identification with the group (Farnham et al., 1999; Smurda et al., 2006). Similar biases have been found in non-economic survey data with respects to self-reported prejudiced attitudes to strangers (Gorsuch, 1974) and the willingness to help (Jackson & Esses, 1997). References Ahmed, A. M., & Salas, O. (2009). Is the hand of God involved in human cooperation? International Journal of Social Economics, 36(1/2), 70–80. Altemeyer, B., & Hunsberger, B. (1992). Authoritarianism, religious fundamentalism, quest, and prejudice. International Journal for the Psychology of Religion, 2(2), 113–133. Anderson, L., Mellor, J., & Milyo, J. (2010). Did the devil make them do it? The effects of religion in public goods and trust games. Kyklos, 63, 163–175. Anderson, L. R., & Mellor, J. M. (2009). Religion and cooperation in a public goods experiment. Economics Letters, 105(1), 58–60. Andreoni, J., & Miller, J. H. (1993). Rational cooperation in the finitely repeated prisoners’ dilemma. Economic Journal, 103, 570–585. Barro, R. J., & McCleary, R. M. (2003). Religion and economic growth across countries. American Sociological Review, 68(5), 760–781. Bhopal, M., & Rowley, C. (2005). Ethnicity as a management issue: Examples from Malaysia. Asia Pacific Business Review, 11(4), 553–574. Brewer, M. B. (1999). The psychology of prejudice: Ingroup love or outgroup hate? Journal of Social Issues, 55(3), 429–444. Chuah, S. H., Fahoum, R., & Hoffmann, R. (2013). Fractionalization and trust in India: A field-experiment. Economics Letters, 119(2), 191–194. Chuah, S. H., Hoffmann, R., Jones, M., & Williams, G. (2007). Do cultures clash? Evidence from cross-national ultimatum game experiments. Journal of Economic Behavior and Organization, 64(1), 35–48. Farnham, S. D., Greenwald, A. G., & Banaji, M. R. (1999). Implicit self-esteem. In D. Abrams & M. A. Hogg (Eds.), Social identity and social cognition (pp. 230–248). Malden, UK: Blackwell. Glaeser, E. L., Laibson, D. I., Scheinkman, J. A., & Soutter, C. L. (2000). Measuring trust. Quarterly Journal of Economics, 115(3), 811–846. Glock, C. Y., & Stark, R. (1965). Religion and society in tension. Chicago: Rand McNally.

S.-H. Chuah et al. / Journal of Economic Psychology 45 (2014) 33–43

43

Gomez, T. (Ed.). (2003). Chinese business in South-East Asia: Contesting cultural explanations, researching entrepreneurship. London, UK: Routledge. Gorsuch, D. A. (1974). Christian faith and ethnic prejudice: A review and interpretation of research. Journal for the Scientific Study of Religion, 13, 281–307. Guiso, L., Sapienza, P., & Zingales, L. (2003). People’s opium? Religion and economic attitudes. Journal of Monetary Economics, 50(1), 225–282. Guiso, L., Sapienza, P., & Zingales, L. (2009). Cultural biases in economic exchange. Quarterly Journal of Economics, 124(3), 1095–1131. Gunnthorsdottir, A., & McCabe, K. (2002). Using the Machiavellianism instrument to predict trustworthiness in a bargaining game. Journal of Economic Psychology, 23(1), 49–66. Hall, D. L., Matz, D. C., & Wood, W. (2010). Why don’t we practice what we preach? A meta-analytic review of religious racism. Personality and Social Psychology Review, 14(1), 126–139. Hewstone, M., Islam, M. R., & Judd, C. M. (1993). Models of crossed categorization and intergroup relations. Journal of Personality and Social Psychology, 64(5), 779–793. Hill, P. C., & Hood, R. W. (Eds.). (1999). Measures of religiosity. Birmingham, AL: Religious Education Press. Hoffmann, R. (2013). The experimental economics of religion. Journal of Economic Surveys, 27(5), 813–845. Hunsberger, B. (1996). Religious fundamentalism, right-wing authoritarianism, and hostility toward homosexuals in non-christian religious groups. International Journal for the Psychology of Religion, 6(1), 39–49. Irons, W. (2001). Religion as a hard-to-fake sign of commitment. In R. M. Nesse (Ed.), Evolution and the capacity for commitment (pp. 292–309). Russel Sage Foundation. Jackson, L. M., & Esses, V. M. (1997). Of scripture and ascription: The relation between religious fundamentalism and intergroup helping. Personality and Social Psychology Bulletin, 23(8), 893–906. Jackson, L. M., & Hunsberger, B. (1999). An intergroup perspective on religion and prejudice. The International Journal for the Psychology of Religion, 38(4), 509–523. Kirkpatrick, L. A., Hood, R. W., & Hartz, G. (1991). Fundamentalist religion conceptualized in terms of Rokeach’s theory of the open and closed mind: New perspectives on some old ideas. Research in the Social Scientific Study of Religion, 3, 157–179. Marquardt, D. W. (1980). You should standardize the predictor variables in your regression models. Journal of the American Statistical Association, 75, 87–91. Montesino, M. U. (2012). Cross-cultural conflict and affirmative action: Inter- and intra-ethnic dilemmas of Malaysia’s heterogeneous workplace. International Journal of Cross Cultural Management, 12(1), 115–132. Orbell, J., Goldman, M., Mulford, M., & Dawes, R. (1992). Religion, context, and constraint toward strangers. Rationality and Society, 4(3), 291–307. Rohrbaugh, J., & Jessor, R. (1975). Religiosity in youth: A personal control against deviant behavior. Journal of Personality, 43(1), 136–155. Saroglou, V., Corneille, O., & Van Cappellen, P. (2005). Speak, Lord, your servant is listening: Religious priming activates submissive thoughts and behaviors. International Journal or the Psychology of Religion, 19, 143–154. Smurda, J. D., Wittig, M. A., & Gokalp, G. (2006). Effects of threat to a valued social identity on implicit self-esteem and discrimination. Group Processes & Intergroup Relations, 9(2), 181–197. Sosis, R., & Ruffle, B. J. (2003). Religious ritual and cooperation: Testing for a relationship on Israeli religious and secular kibbutzim. Current Anthropology, 44(5), 713–722. Stangor, C., Lynch, L., Duan, C. M., & Glas, B. (1992). Categorization of individuals on the basis of multiple social features. Journal of Personality and Social Psychology, 62(2), 207–218. Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel (Eds.). The social psychology of intergroup relations (Vol. 33, pp. 33–47). Monterey, CA: Brooks/Cole. Tan, J. H. W. (in press). Behavioral economics of religion. In P. Oslington (Ed.), Oxford handbook of christianity and economics. Oxford, UK: Oxford University Press. Tan, J. H. W., & Vogel, C. (2008). Religion and trust: An experimental study. Journal of Economic Psychology, 29(6), 832–848. Verkuyten, M., & Khan, A. (2012). Interethnic relations in Malaysia: Group identifications, indispensability and inclusive nationhood. Asian Journal of Social Psychology, 15(2), 132–139.