• No results found

Sex differences in trust and trustworthiness : a meta-analysis of the trust game

N/A
N/A
Protected

Academic year: 2021

Share "Sex differences in trust and trustworthiness : a meta-analysis of the trust game"

Copied!
59
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Sex Differences in Trust and Trustworthiness: A Meta-Analysis of the Trust Game Olmo R. van den Akker

(2)

Abstract

Do men and women differ in trusting behavior? This question is directly relevant to social, economic, and political domains yet the answer remains elusive. In this thesis, we present a meta-analytic review of the literature on sex differences in the experimental trust game. The meta-analysis consists of 82 papers encompassing 237 effect sizes and 16,431 participants in 22 countries. We found that men send significantly more than women as first movers, g = 0.22, although a sensitivity analysis sheds some doubt on this result. In addition, we found no sex difference in second mover behavior, g = -0.02. None of the six study characteristics

significantly moderated the two main effects, and neither did a moderator related to social role theory. We can cautiously conclude that a prediction from parental investment theory can help explain sex differences in trust, but we failed to find evidence favoring social role theory as an explanation for sex differences in trustworthiness. When looking at trust games where the sex of the partner is known, we found no evidence that first movers send more in same-sex partnerships than in different-sex partnerships, in contrast to a prediction from social identity theory. When comparing our results to the results of a meta-analysis on sex differences in the gift-exchange game, we noticed a „male multiplier effect‟: men only send more than women when their transfer gets multiplied. Future research needs to substantiate this effect and provide a theoretical

framework to explain it.

(3)

Sex Differences in Trust and Trustworthiness: A Meta-Analysis of the Trust Game Experimental games have been widely used to measure people‟s social preferences. Seven of these games have been especially prevalent in the literature: the prisoner‟s dilemma, the public goods game, the ultimatum game, the dictator game, the trust game, the gift-exchange game, and the third-party punishment game (Camerer & Fehr, 2002). Given the wealth of experimental studies, it is not surprising that many researchers have used meta-analysis to summarize people‟s decisions in these games. However, one salient variable has often been neglected in these analyses: the participants‟ sex. While sex differences have been studied in meta-analyses of the prisoner‟s dilemma and the public goods game (see Balliet, Li, Macfarlan, & Van Vugt, 2011), meta-analysts have not touched upon sex differences in the other games. This study hopes to fill this gap in the literature by performing a meta-analysis of sex differences in the most popular game that measures trust and trustworthiness: the trust game. The goal of this meta-analysis is to identify whether men or women are more trusting and trustworthy.

Understanding sex differences in trust and trustworthiness is important because trusting behavior is relevant in a wide range of situations. For instance, trust fosters cooperation (Balliet & Van Lange, 2013) and facilitates desirable behaviors in organizations (Dirks, & Ferrin, 2001). One specific example of the relevance of trust lies in the context of negotiations. Perceptions of one‟s negotiation partner as trusting or trustworthy can play a major role in determining the outcomes of a negotiation, and negotiators have been found to use all kinds of cues to assess their partner‟s level of trust(worthiness) (Lewicki, & Polin, 2013). If the current meta-analysis shows that men or women are more trusting or trustworthy in the trust game, the partner‟s sex could potentially be used as one of these cues.

(4)

The trust game – originally called investment game – (Berg, Dickhaut, & McCabe, 1995) was developed more than two decades ago and can be described as follows “One player, the first mover, is endowed with a certain amount of money. This first mover has the choice to send a proportion of this money to another player, the second mover. The money he decides to give away is multiplied by a given factor before reaching the other player. Generally this factor is three, but it varies over studies. In the second and final round, the second mover can decide how much of the money he will send back to the first mover. The amount sent by the first mover is seen as a manifestation of trust, whereas the amount returned by the second mover is seen as a manifestation of [trustworthiness].” (Van den Akker, 2016, p. 4).

While several papers have investigated sex differences in the trust game (Croson & Gneezy, 2009; Rau, 2011), few papers included an accompanying rationale to explain these sex differences. In this project, parental investment theory will be used to predict sex differences in trust, while social role theory will be used to predict sex differences in trustworthiness.

Parental investment theory (Trivers, 1972) is based on the idea that the investments of men and women in raising offspring are asymmetric. Women are burdened with a 9-month gestation period and a lactation period after birth that can take years. Men‟s investment, on the other hand, is limited to his sperm contribution, which only requires little time. Because women have to spend a large amount of time raising a single child, they are only able to raise a limited number of children during their lifetime. This means that women must be very selective in choosing a mate as the fitness of the child is influenced greatly by the quality of the father. The higher selectivity of women implies that men have to engage in intrasexual competition to get a mate. In practice, this involves demonstrating the high quality of their genes and/or their capacity to help raise a child as both these traits can improve a child‟s fitness.

(5)

A trait that can be instrumental for males to signal high-quality genes and a high child rearing capacity is risk-taking. Taking risks has been shown to be beneficial in achieving a higher social status because it can lead to the acquisition of resources (e.g. through risky financial investments) or a higher place in the social hierarchy (e.g. through picking fights with higher-status males). This is important because social higher-status is often seen by women as an indicator of a man‟s potential to aid in raising a child (Wilson & Daly, 1985). In addition, men can use risk-taking to signal the high quality of their genes. This is because risk-taking risks is more costly for people with lower fitness than for people with higher fitness, which means that usually only high-fitness individuals take risks (Baker & Maner, 2009). In short, it pays off for males to be more risk-taking and therefore males have evolved a tendency to take more risks. This is evidenced by studies that find men to be more risk taking than women across different ages and cultures (Drevenstedt, Crimmins, Vasunilashorn, & Finch 2008; Barford, Dorling, Smith, & Shaw, 2006; Apicella, Crittenden, & Tobolsky, 2017).

This difference in risk-taking is relevant to the current project because researchers have noted that risk is an essential component of trust as both constructs involve uncertainty (Ben-Ner & Putterman, 2001; Cook & Cooper, 2003). In fact, trust and risk are so related that it is difficult to disentangle these constructs in the trust game; the first transfer inherently involves risk as the first mover does not know whether the second mover will reciprocate (Croson & Gneezy, 2009). Given the risky nature of the first transfer, and given the findings that men are more risk-taking than women on average, we predict that men transfer more money as the first mover in the trust game than women.

To explain sex differences in trustworthiness, we follow Van den Akker (2016) and refer to social role theory: Social role theory states that men and women develop expectations about

(6)

the way they ought to behave based on traditional sex roles, and that men and women behave according to these roles (Eagly, 1987; for an overview see Dulin, 2007). Traditional sex roles prescribe that men are more competitive, independent, task-oriented, and self-confident, while women are more sensitive, nurturing, expressive, and warm. Bakan (1966) labeled these groups of personality characteristics as agentic and communal respectively.

Social role theory predicts that social obligations will have a stronger impact on the behavior of women than on the behavior of men (Buchan, Croson, & Solnick, 2008), which is based on findings that women are more communal and thus more focused on interpersonal relations (Schmitt, Realo, Voracek, & Allik, 2008; Weisberg, DeYoung, & Hirsh, 2011). This suggests that women are less likely than men to violate the trusting relationship by failing to reciprocate. In short, social role theory predicts that women are more trustworthy than men.

Two narrative reviews have been undertaken to summarize the evidence of sex

differences in the trust game (Croson & Gneezy, 2009; Rau, 2011). Both reviews found that men are more trusting and women are more trustworthy, supporting parental investment theory and social role theory as explanations for sex differences. Given this match between theory and empirical findings, we predict that this meta-analysis will show that men transfer more money as the first mover, and women transfer more money as the second mover.

An important study that is relevant to the proposed research is a previous meta-analysis of the trust game (Johnson & Mislin, 2011). The authors of that meta-analysis did not have enough data to study sex differences in trust and reciprocity, which is why we pursue that question in the current project. In their meta-analysis, Johnson and Mislin found that changes in the experimental protocol significantly changed behavior in the trust game. Because it has been suggested that women are more sensitive to the experimental protocols of studies (Chermak &

(7)

Krause, 2002; Croson & Gneezy, 2009), we added six moderators to the analysis regarding the experimental protocol of the trust game. These moderators are: whether participants were paid for their participation, how many times the trust game was played during the experiment, what the multiplication factor was of the first transfer, whether participants played as both the first mover and the second mover, whether the second mover had an initial endowment, and whether the strategy method (Selten, 1967) was used to elicit the decisions of the second movers. As there were no directional hypotheses regarding these moderators were analyzed exploratively.

We also added a moderator to test the prediction of social role theory that women are more trustworthy than men. That prediction can be tested by identifying whether sex differences in trustworthiness are related to Hofstede et al.‟s (2010) masculinity index. The masculinity index is a country-level index that is derived from Hofstede‟s characterization of 76 countries in which he rates these countries on several dimensions. One of those dimensions is masculinity-femininity, where more masculine societies have a preference for competition, assertiveness, and material rewards, while more feminine societies have a preference for cooperation, modesty, and achieving consensus. More relevant to our study, however, is the idea that masculine societies clearly separate social gender roles, while in more feminine societies these social gender roles overlap (Hofstede, 2001). As social role theory relies on clearly distinct social gender roles, we expected the predicted sex difference in trustworthiness to be more prominent in masculine societies than in feminine societies. More concretely, we expected that females are relatively more trustworthy than men in more feminine societies than in more masculine societies.

Finally, during the search for eligible papers, we found that several papers included a modified version of the trust game in which the sex of the partner was known to the players. In order to get a broader view of sex differences in trust and trustworthiness we decided to do

(8)

separate meta-analyses on those papers as well. Social identity theory (Tajfel & Turner, 1979) offers predictions regarding the behavior of both sexes in these modified trust games.

Specifically, social identity theory assumes that people attribute positive qualities to people they have placed in the same category as themselves. Categories can be defined in all kinds of ways, but given that a person‟s sex is such a salient characteristic, people often categorize others by their sex. Regarding trust and trustworthiness, social role theory states that people assess same-sex others to be both more trustworthy and more trusting than opposite-same-sex others, and will therefore send more to same-sex partners in both roles in the trust game. In short, we predicted that both first and second movers in the trust game send more in same-sex partnerships than in mixed-sex partnerships.

In summary, our predictions were: (1) men are more trusting in the trust game than women, (2) women are more trustworthy in the trust game than men, and this effect is moderated by a society‟s masculinity, and (3) first movers and second movers send more in same-sex

partnerships in the trust game than in mixed-sex partnerships. Other moderating effects were looked at exploratively.

Method Search strategy

Our search strategy was influenced by the fact that Johnson and Mislin (2011) already carried out an exhaustive search for trust game papers published in the years up to 2011. Because we judged their search to be comprehensive we decided to use their search results for the papers up to 2010 and carry out our own search for papers published from 2011 onwards. For our own search, we employed four search strategies. First, we used the terms “trust game” and

(9)

“investment game” in searches on the following databases: PsycINFO, EconLit, and the Web of Science Core Collection. SSRN was searched for unpublished papers and OpenGrey was

searched for grey literature. These searches were carried out in April and May 2017. We did not cross-reference the search terms with terms pertaining to a person‟s sex or gender because many studies routinely measure the sex of their participants. Therefore, we listed all studies that met the inclusion criteria, regardless of whether they provided data on the participants‟ sex. Second, we checked the Web of Science database for papers citing the original trust game paper, the paper in which the game was developed (Berg, Dickhaut, & McCabe, 1995). Third, we looked at references in review articles and other relevant articles that we found using the first two search strategies. Examples of such articles are the articles by Croson and Gneezy (2009) and Rau (2011). Fourth, we sent out a call for papers in the Economic Science Association‟s experimental methods discussion group (https://groups.google.com/forum/#!forum/esa-discuss). The call for papers can be found in Appendix A. All search strategies combined yielded a total of 1,199 unique papers in the database. For a visual overview of the search for papers, see Appendix B Inclusion criteria

We used several inclusion criteria to select which of the studies of the 1,199 papers would be included in the analysis. First, because of language barriers, we decided to only include papers in English. Second, only studies with student samples or adult samples were included because there are indications that behavior in trust games is different for children and adults (Sutter & Kocher, 2007). Third, participants had to be from a sample that is not characterized by physical or psychological dysfunctions. Examples of excluded studies were studies that used participants with Parkinson‟s disease (Javor, et al., 2015) and borderline personality disorder (Ebert, et al., 2013). Fourth, we only included studies that involve the trust game as described in the

(10)

introduction. More specifically, that means a game with two players, where the first mover can transfer a certain amount of money to the second mover, the money is multiplied by a given factor, and the second mover can return some of the money to the first mover. Any games that deviate from this design were not included. Examples of reasons for exclusion were trust games that were non-continuous (i.e. games with fewer than eleven choice options as the first or second mover, e.g. Servátka, et al., 2008; Simpson & Eriksson, 2009), trust games in which participants could communication with each other (e.g. Kimbrough & Rubin, 2015), trust games in which players had personal information about the other player other than their sex (e.g. Hargreaves Heap, et al., 2009), and trust games wherein players knew that they were not playing against a real player (e.g. Kirkebøen, et al., 2011). Finally, we only included one-shot trust games or trust games with random partner matching. In games with random partner matching the game is played multiple times, but every time with a different, random partner. Repeated games with the same partner were excluded because reputational concerns may alter the nature of the game. In all, we found 190 papers with one or more studies eligible for inclusion. Of those papers, we were able to retrieve 237 effect sizes from 82 papers. An overview of all the papers included in the meta-analyses can be found in Table 1, Table 2, and Table 3.

(11)

Table 1

Studies Included in the Meta-Analyses on Sex Differences in Trust and Trustworthiness

Paper Condition Pay Both It Sec SM Mtp Masc gt gtw

Ainsworth et al. (2014) Experiment 1 - Non-depletion Yes No 1 NA NA 3 62 0.58 NA Ainsworth et al. (2014) Experiment 2 - Non-depletion

& No future meeting

Yes No 1 NA NA 3 62 0.59 NA

Ainsworth et al. (2014) Experiment 3 - Non-depletion & No information

Yes No 1 NA NA 3 62 -0.07 NA

Atlas & Putterman (2011) Baseline Yes No 1 No No 3 NA 0.12 -0.03

Babin (2016) Baseline Yes No 1 Yes No 3 62 0.29 0.07

Barrera & Simpson (2012) Study 1 - Control Yes Yes 2 No Yes 3 NA 0.40 0.14 Barrera & Simpson (2012) Study 2 - Control Yes Yes 2 No Yes 3 NA 0.95 0.34

Batsaikhan (2016) Trust game Yes Yes 2 No Yes 3 NA 0.56 0.24

Becker et al. (2012) First data set - Multiplier x2 Yes Yes 4 No Yes 2 66 0.15 NA

Bereczkei et al. (2015) Trust game Yes Yes 12 No No 3 88 0.36 -0.20

Böckler, et al. (2016) Trust game Yes No 1 NA NA 3 66 0.75 NA

Boero et al. (2009) Experiment 1 - Baseline Yes Yes 10 Yes No 3 70 0.61 0.07

Bourgeois-Gironde & Corcos (2011)

OSG Yes No 1 No No 3 43 0.34 -0.22

Bravo, et al. (2012) Investment game Yes No 10 Yes No 3 70 -0.03 -0.20

Breuer & Höwe (2014) Trust game Yes No 1 No Yes 3 66 0.49 -0.22

Brülhart & Usunier (2012) Equal endowment Yes No 1 Yes No 3 70 0.32 0.10

Buchan, et al. (2002) Direct treatment - China Yes No 1 Yes No 3 66 0.41 0.48 Buchan, et al. (2002) Direct treatment - Japan Yes No 1 Yes No 3 95 -0.44 -1.13

Buchan, et al. (2002) Direct treatment - USA Yes No 1 Yes No 3 62 0.52 -0.89

Buchan, et al. (2008) Trust game - Both names Yes No 1 Yes No 3 62 0.38 -0.45

Buser (2012) Trust game Yes Yes 2 Yes No 3 14 0.23 -0.13

Butler, et al. (2015) Experiment 1 Yes Yes 12 Yes Yes 3 70 0.25 0.02

Calabuig et al. (2016) NOPUN(10,10) Yes No 1 Yes No 3 42 0.25 0.03

Cameron et al. (2014) Trust game Yes Yes 2 No No 3 61 0.04 0.32

(12)

Paper Condition Pay Both It Sec SM Mtp Masc gt gtw Chaudhuri & Gangadharan

(2007)

Trust game Yes Yes 2 Yes Yes 3 61 -0.52 0.29

Chaudhuri, et al. (2016) Study 1 - Private knowledge Yes No 10 Yes No 3 58 0.14 -0.43 Chaudhuri, et al. (2016) Study 1 - Common knowledge Yes No 10 Yes No 3 58 0.32 0.76 Chaudhuri, et al. (2016) Study 1 - Context neutral Yes No 10 Yes No 3 58 0.01 0.18 Chaudhuri, et al. (2016) Study 2 - Context neutral Yes No 10 Yes No 3 58 0.20 -0.56

Clots-Figueras, et al. (2016) Baseline Yes No 1 Yes No 3 42 0.28 0.17

Courtiol, et al. (2009) Trust game Yes No 1 Yes No 3 43 0.02 0.00

Dean & Ortoleva (2015) Trust game Yes Yes 2 No Yes 3 62 0.03 -0.20

Di Bartolomeo & Papa (2016a) T1 - Control Yes Yes 8 Yes No 3 70 0.39 -0.15 Di Bartolomeo & Papa (2016b) Counterfactual No No 1 Yes No 3 70 0.30 -0.23 Di Bartolomeo & Papa (2016c) Treatment 1 - Direct-response

method

Yes No 1 Yes No 3 70 0.00 -0.63

Di Bartolomeo & Papa (2016c) Treatment C1 Yes No 1 Yes No 3 70 -0.05 0.22

Dilger, Müller, & Müller Trust game Yes No 1 Yes Yes 3 66 0.33 0.32

Dreber et al. (2012) Trust game Yes No 1 No Yes 3 62 0.29 0.23

Evans & Revelle (2008) Study 2 - Send-Only Yes No 1 No No 3 62 -0.11 0.55

Friebel et al. (2017) Stage 1 Yes Yes 2 No Yes 3 66 0.33 -0.18

Galeotti & Zizzo (2014) Baseline Yes Yes 4 No No 3 66 -0.14 0.22

Haesevoets, et al. (2015) Trust game Yes No 1 NA NA 3 54 0.63 NA

Haile, et al. (2008) No information Yes Yes 2 Yes Yes 3 63 -0.26 0.22

Hargreaves-Heap and Zizzo (2009)

Stage 1 Yes Yes 3 No No 3 66 0.32 0.13

Hergueux & Jacquement (2014) Online Yes No 1 Yes Yes 3 43 0.23 -0.38

Hergueux & Jacquement (2014) InLab Yes No 1 Yes Yes 3 43 -0.05 -0.66

Heyes & List (2016) Control No No 1 Yes Yes 3 62 1.61 0.37

Houser, et al. (2010) Trust-1 Yes No 1 Yes No 3 66 0.51 -0.27

Johnsen & Kvaløy (2016) Non-strategic - Part 1 Yes No 4 Yes No 3 8 0.71 0.08

Kanageratnam et al. (2009) One-shot rounds Yes Yes 2 Yes No 3 52 0.44 -0.01

Kausel & Connoly (2014) Study 2 - Neutral player B Yes No 1 No No 3 62 0.59 -1.12 Keck & Karelaia (2012) Experiment 1 - Baseline Yes No 1 Yes No 3 43 -0.35 -0.03

(13)

Paper Condition Pay Both It Sec SM Mtp Masc gt gtw Koranyi & Rothermund (2012) Experiment 1 - Control Yes No 1 NA NA 3 66 0.37 NA Koranyi & Rothermund (2012) Experiment 2 - Control Yes No 1 NA NA 3 66 -0.21 NA

Kvaløy & Luzuriaga (2014) Baseline Yes No 1 Yes No 3 8 1.14 -0.09

Lee & Schwartz (2012) Study 1 - Odourless water Yes No 1 NA NA 4 62 -0.72 NA

Lev-On (2009) No information Yes No 1 NA NA 3 47 -0.30 NA

Luini, et al. (2014) NO-INFO Yes No 20 NA NA 3 70 0.05 NA

Malcman, et al. (2015) Standard mechanism Yes No 1 Yes No 3 47 NA 0.34

Malcman, et al. (2015) Virtual money No No 1 Yes No 3 47 NA 0.30

Markowska-Przybyła & Ramsey (2016)

Trust game Yes No 1 No No 3 64 0.22 0.02

Martinez & Zeelenberg (2015) Experiment 1 - Control No No 1 Yes No 3 31 0.44 0.23 Martinez & Zeelenberg (2015) Experiment 3 - Control Yes No 1 Yes No 3 31 -0.28 -0.55

Migheli (2012) Oslo No No 1 No No 3 8 -0.07 -0.05

Migheli (2012) Leuven No No 1 No No 3 54 0.37 0.05

Migheli (2012) Torino No No 1 No No 3 70 0.30 0.16

Mislin, et al. (2015) Neutral video - Multiplier x2 Yes No 1 Yes No 2 62 1.08 0.52 Mislin, et al. (2015) Neutral video - Multiplier x4 Yes No 1 Yes No 4 62 -0.26 -0.44

Moretto, et al. (2013) Control - Endowment of 12 Yes No 9 NA NA 3 70 1.13 NA

Piff et al. (2010) Study 3 - Trust game No No 1 NA NA 3 62 -0.48 NA

Qin, Shen, & Meng (2011) Community members Yes No 1 Yes Yes 3 66 -0.05 0.67

Riedl & Smeets (2013) Trust game Yes No 1 Yes Yes 3 14 0.03 -0.04

Sapienza, et al. (2013) Trust game Yes Yes 2 No Yes 3 62 -0.20 -0.08

Schniter, et al. (2015) Trust game Yes No 1 No No 3 62 0.18 0.37

Sellaro et al. (2015) Control Yes No 1 NA NA 3 14 1.18 NA

Shen & Qin (2014) Trust game Yes No 1 No Yes 3 66 0.57 0.22

Smith (2011) Treatment 1 Yes Yes 11 No No 3 52 0.52 -0.27

Swope et al. (2008) Trust game Yes No 1 No No 3 62 0.34 -0.25

Takahashi, et al. (2016) Trust game Yes No 1 Yes No 3 95 0.51 -0.32

Tepe (2016) Trust game Yes Yes 2 No No 3 66 0.68 0.03

Tsutsui & Zizzo (2014) Stage 1 Yes Yes 3 No No 3 66 0.10 0.13

Tu & Bulte (2010) Trust game Yes Yes 2 No Yes 3 66 0.21 0.11

(14)

Paper Condition Pay Both It Sec SM Mtp Masc gt gtw Vyrastekova & Onderstal (2005) Trust game Yes Yes 2 No Yes 3 14 0.10 0.07

Wu, et al. (2016) Control Yes Yes 2 No No 3 62 0.00 -0.35

Yamagishi, et al. (2015) Trust game Yes Yes 2 No Yes 3 95 0.11 -0.04

Zak, et al. (2005) Intention Yes No 1 Yes No 3 62 0.39 0.53

Zheng et al. (2016) Study 3 - Low-power transgressor

Yes No 1 NA NA 3 14 0.10 NA

Zheng et al. (2016) Study 3 - High-power

transgressor Yes No 1 NA NA 3 14

-0.20 NA

Zhong et al. (2012) Trust game Yes Yes 2 Yes Yes 3 48 0.14 -0.06

Note: The ‘Condition’ column indicates which of the conditions in the paper was included. The ‘Pub’ column indicates whether participants of the study were paid for their participation. The ’Both’ column indicates whether players in the trust game had to play as both the first mover and the second mover. The ‘It’ column indicates the number of iterations of the trust game. The ‘Sec’ column indicates whether the second mover in the trust game was allocated an initial endowment. The ‘SM’ column indicates whether the second mover decisions were elicited using the strategy method. The ‘Mtp’ column indicates the multiplication factor in the study. The ‘Masc’ column indicates Hofstede’s masculinity level for the country in which the study was carried out. The ‘gt‘ column indicates the effect size of a sex difference with respect to the first mover’s decision. The ‘gtw‘ column indicates the effect size of a sex difference with respect to the first mover’s decision. If the effect sizes are positive, that means that men send more than women.

(15)

Table 2

Papers Included in the Meta-analysis on Sex Differences in Trust with Sex of Partner Known

Paper Condition Masc gmm.mw gwm.ww gmm.wm gmw.ww gmm.ww gmw.wm

Buchan, Croson, & Solnick (2008) Trust game - Both names 62 0.20 -0.06 0.84 0.75 0.81 0.78 Eckel & Petrie (2011) Treatment 2 - Forced picture 62 0.16 0.40 0.10 0.32 0.45 -0.06 Feinberg, Willer, & Keltner (2012) Study 4 - Neutral 62 -0.06 0.24 0.54 0.79 0.71 0.62

Greig & Bohnet (2008) Trust game 41 -0.03 0.11 -0.10 0.03 -0.02 -0.08

Lev-On (2009) Information 47 -0.10 0.10 0.27 0.49 0.36 0.39

Slonim & Guillen (2010) No selection 62 0.49 0.31 0.24 0.15 0.52 -0.21

Sutter & Kocher (2007) Students 79 -0.11 0.01 0.19 0.30 0.20 0.30

Sutter & Kocher (2007) Working professionals 79 -0.27 0.00 -0.23 0.05 -0.25 0.05

Sutter & Kocher (2007) Retired persons 79 NA -0.59 0.52 NA -0.04 NA

Wilson & Eckel (2006) Trust game 62 0.07 -0.52 0.32 -0.30 -0.21 0.27

Note: The ‘Condition’ column indicates which of the conditions in the paper was included. The ‘Masc’ column indicates Hofstede’s masculinity level for the country in which the study was carried out. The other columns indicate effect sizes, where ‘m’ stands for man, and ‘w’ stands for women. When letters are coupled like in ‘mw’ this means that the sex related to the first letter sends toward the sex related to the second letter. In the case of the example a man sends money toward a woman. In addition, when letters are coupled like mm.mw, the effect size relates to a comparison between the coupled letters before and after the dot. For example, the ‘gmm.mw‘ column indicates the effect size where the amount sent by men toward men is compared to the amount sent by men toward women.

(16)

Table 3

Papers Included in the Meta-analysis on Sex Differences in Trust with Sex of Partner Known

Paper Condition Masc gmm.mw gwm.ww gmm.wm gmw.ww gmm.ww gmw.wm

Buchan, et al. (2008) Trust game - Both names 62 -0.18 0.21 -0.02 0.40 0.19 0.16 Eckel & Petrie (2011) Treatment 2 - Forced picture 62 0.41 0.09 0.05 -0.28 0.14 -0.37

Greig & Bohnet (2008) Trust game 41 0.18 0.49 -0.47 -0.31 -0.07 -0.66

Sutter & Kocher (2007) Students 79 0.00 -0.04 0.22 0.24 0.20 0.25

Sutter & Kocher (2007) Working professionals 79 -0.09 -0.11 0.86 0.58 0.78 0.65

Sutter & Kocher (2007) Retired persons 79 NA 0.39 0.42 NA 1.04 NA

Wilson & Eckel (2006) Trust game 62 0.27 0.02 0.22 0.02 0.26 0.00

Note: The ‘Condition’ column indicates which of the conditions in the paper was included. The ‘Masc’ column indicates Hofstede’s masculinity level for the country in which the study was carried out. The other columns indicate effect sizes, where ‘m’ stands for man, and ‘w’ stands for women. When letters are coupled like in ‘mw’ this means that the sex related to the first letter sends toward the sex related to the second letter. In the case of the example a man sends money toward a woman. In addition, when letters are coupled like mm.mw, the effect size relates to a comparison between the coupled letters before and after the dot. For example, the ‘gmm.mw‘ column indicates the effect size where the amount sent by men toward men is compared to the amount sent by men toward women.

(17)

Running head: SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 17 Data collection

Extracting the required information for the meta-analyses proved to be difficult because only five papers included the required data for us to calculate the effect sizes. Two of those could be used to extract effect sizes for trust and trustworthiness, and three of those could be used to extract effect sizes related to games where the sex of the partner was known. The remaining papers in our database only used sex as a control variable or did not mention sex at all, so in those cases we had to contact the authors to request the required information. We first contacted the corresponding authors of each paper, and if we received no response, we sent out a reminder e-mail about three weeks later. If we still did not receive a response after six weeks, we sent out a final data request e-mail to the co-author(s) of the paper with a remark that we already tried to reach the corresponding author. Templates of the different e-mails can be found in Appendix C. Authors could either provide us with the raw data or with the summary statistics we needed to calculate the effect sizes ourselves. From the 185 papers for which we contacted the authors, we received the data 77 times, we did not receive the data 58 times, and we were not able to contact the authors 50 times.

Coding procedure

To measure trust we used the proportion of the first transfer to the initial endowment. This meant that we had to retrieve the following information: the mean first transfer for both sexes, the standard deviations of those means, the number of men and women, and the initial endowment. If we were able to retrieve these values for a particular study we were able to calculate an effect size of sex differences in trust for that study. The calculation of effect sizes is described in the section Statistical analysis.

(18)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 18 Finding a good measure of trustworthiness was more complex because it can be argued that the second transfer by itself is not a good measure of trustworthiness. This is because the concept of trustworthiness is only relevant with regard to a preceding behavior, in this case the first transfer. For that reason, we chose to use the second transfer divided by the multiplied first transfer as the measure of trustworthiness. We calculated this proportion for every individual participant, and then calculated the mean and standard deviation of those proportions, for both sexes. Coupled with the number of men and women, we were then able to calculate the effect sizes of sex differences in trustworthiness.

Regarding the trust games where the gender of the partner was known, we calculated effect sizes for every possible sex pairing.

In addition to coding all effect sizes, we coded for several moderators. Six moderators refer to the protocol of the trust game, masculinity was coded to test a prediction of social role theory, and the other three moderators were coded to carry out sensitivity analyses: whether the study was published in a scientific journal, whether the experiment had potential confounding factors, and whether sex differences were part of the main hypothesis in the paper. Table 4 provides an overview of these moderators.

To assess the validity of our coding efforts we recoded four moderator variables that were already coded by Johnson and Mislin (2011): the multiplication factor, whether the second mover was allocated with an initial endowment, whether the strategy method was used, and whether participants played both as the first and second mover. Codes corresponded 22 out of 22 times for the multiplier moderator, 20 out of 22 times for the second mover endowment moderator, 21 out of 22 times for the strategy method moderator, and 20 out of 22 times for the both roles moderator. In all, we can conclude there is good intercoder reliability.

(19)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 19 Table 4

Moderators Analyzed in the Current Meta-Analysis

Moderator Coded

Payment Coded as 1 when the monetary reward of the participant depended on their decisions in the trust game, coder as 0 otherwise

Both roles Coded as 1 when the participants played both as the first mover and as the second mover in the trust game, coded as 0 otherwise

Second mover endowment Coded as 1 when the second mover was allocated an initial endowment, coded as 0 otherwise

Strategy method Coded as 1 when second mover behavior was elicited using the strategy method, coded as 0 otherwise

Iterations The number of one-shot rounds the trust game was played Multiplication factor The factor that was used to multiply the first transfer in the

trust game

Masculinity The masculinity level of the country the study was carried out in, as measured by Hofstede et al. (2010)

Published Coded as 1 when the study has been published in a scientific journal, coded as 0 otherwise

Potential confounds Coded as 1 when the study included one or more unrelated tasks before the trust game that could have influenced behavior in the trust game

Sex difference hypothesis Coded as 1 when the paper included a directional hypothesis with regard to sex differences in trust or trustworthiness, coded as 0 otherwise

Statistical analysis

To calculate the effect sizes of individual studies, we used the Hedges‟ g effect size measure, which is preferred over Cohen‟s d because the latter is biased for small sample sizes (Hedges & Olkin, 1985). Hedges‟ g is calculated as follows:

𝑔 =(x̅ − x̅ )

s 𝑐(n , n )

where x̅ is the mean for men, x̅ is the mean for women, s is the pooled standard deviation, and 𝑐(n , n ) is a constant that depends on group sizes. Both s and 𝑐(n , n ) are defined below.

(20)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 20

s = √(n − 1)s + (n − 1)s n + n − 2

𝑐(n , n ) ≈ 1 − 3

4(n + n − 2) − 1

where n is the number of males, n is the number of females, s is the variance for males, and s is the variance for females. Finally, we also estimated the variance of the Hedges‟ g effect size measure:

s = n + n

n n +

𝑔

2(n + n − 2)

To combine all individual effect sizes into an overall effect size we planned to use several methods. This is preferred over using only one method because it allows checking the robustness of the results (Steegen, et al., 2016). The meta-analytic methods we planned to use besides a standard random effects analysis were PET-PEESE (Stanley & Doucouliagos, 2014), p-curve (Simonsohn, Nelson, & Simmons, 2014b), and p-uniform (Van Assen, Van Aert, & Wicherts, 2015). The PET-PEESE approach calculates the intercept of the regression line when the effect sizes of individual studies are regressed on the standard errors of those effect sizes. At the intercept the standard error is zero, so this intercept can be seen as an unbiased estimate of the overall effect size. The p-curve and p-uniform approaches estimate the overall effect size by computing the level of right-skewedness of the distribution of significant p-values with the aim to correct for publication bias. The p-curve and p-uniform approaches differ in that they use different methods to do this (Van Aert, Wicherts, & Van Assen, 2016).

However, most of these techniques come with fairly stringent assumptions, the major one being a homogeneous set of studies. Two different comparisons of meta-analytic methods have shown that PET-PEESE, p-curve and p-uniform all lead to a significant bias in estimating the

(21)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 21 overall effect size when study heterogeneity is present (Carter et al., 2017; Stanley, 2017).

Unfortunately, a heterogeneity analysis indicated substantial heterogeneity between the studies in the current meta-analysis. Therefore, based on recommendations from Carter et al. and Stanley we ruled out the use of these methods. A priori we already ruled out the use of the trim-and-fill method (Duval and Tweedie, 2000a; 2000b) because it shows bias both when publication bias is present (Simonsohn, Nelson, & Simmons, 2014a) and when publication bias is not present (Terrin et al, 2003).

The model we did use was a random effects model, a model in which the true effect size is allowed to vary between studies (i.e. there can be heterogeneity between studies). The ultimate goal of a random effects model is to estimate the true mean of the distribution of the different true effect sizes.

The main downside of using a random effects model is that it leads to biased estimates in the presence of publication bias – the tendency to publish significant findings more often than non-significant findings. Publication bias has been prevalent in many meta-analyses (Bakker, Van Dijk, & Wicherts, 2012) and is troublesome because it unjustly inflates the overall effect size. A random effects meta-analysis does not correct for publication bias like the other methods, so should only be used when publication bias is unlikely or absent. Conveniently, that is the case here (see the publication bias analysis in the Results section).

The fact that publication bias is not present makes intuitive sense because only ten of the 82 papers included in the meta-analysis specifically hypothesized sex differences and 32 out of 82 did not even include any data related to the participants‟ sex in the paper. This means that these papers were not selected on the basis of significant sex differences in trust or

(22)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 22 sample that studies with significant sex differences were published at a higher rate than studies with non-significant sex differences (i.e. there is probably no publication bias). In all, due to the large heterogeneity of the studies in the meta-analysis and the fact that publication bias is not likely we used a random-effects model to compute the overall effect sizes.

The random effects meta-analysis was complemented by a moderator analysis in which we regressed the effect size of sex differences in trust and trustworthiness on each moderator variable separately (see Table 6 and Table 7). Because this involves multiple significance tests, we applied the Hochberg procedure to control for false positives. The Benjamini-Hochberg procedure can be used to adjust the original p-values by first ranking the p-values of the original significance tests (with a rank of 1 for the smallest value). Then, the adjusted p-value is the smallest of either (1) the original p-p-value times the number of tests, divided by the rank of that p-value, or (2) the adjusted p-value of the p-value that is ranked one higher. This procedure is preferred over the widely used Bonferroni correction because it is less conservative (Benjamini, & Hochberg, 1995).

To assess heterogeneity we computed both the Q-statistic, which tests the null hypothesis of no heterogeneity, and the 𝐼 -statistic, which measures the extent of the heterogeneity.

Finally, we carried out sensitivity analyses in which we used several criteria to select subsets of the studies in the meta-analysis. We then ran the meta-analyses on the studies in those subsets only. Criteria we used to select subsets of studies were the potential for confounding variables in a study, whether sex differences were part of the main hypotheses in the paper, and the sample size of the study (see the sensitivity analysis in the Results section).

(23)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 23 Results

Overview of the meta-analysis

We conducted several meta-analytic reviews totaling 82 papers encompassing 237 effect sizes and 16,431 participants in 22 countries. The meta-analysis regarding trust involved 74 papers and 90 studies, each with one effect size, while the meta-analysis regarding

trustworthiness encompassed 62 papers and 75 studies, also with one effect size each. The total number of participants included in these two meta-analyses was 14,014. Regarding the trust games where the sex of the partner was known 11 studies within 9 papers yielded 44 effect sizes related to trust, and 7 studies within 5 papers yielded 28 effect sizes related to trustworthiness. The total number of participants in these meta-analyses was 2,417.

Heterogeneity analysis

To assess the diversity of the studies in the meta-analyses on trust and trustworthiness, we performed statistical tests of heterogeneity. The tests for both trust, Q(89) = 224.78, p < .0001, and trustworthiness, Q(81) = 108.08, p = .006, show that the effect sizes in the analysis are heterogeneous. To assess the size of the heterogeneity, we computed the I2 index. The trust studies appear to be substantially heterogeneous with 63.76% of the variance accounted for by between-study variance, while the trustworthiness studies appear to be moderately heterogeneous with 25.70% of the variance accounted for by between-study variance.

Publication bias analysis

To test for the presence of publication bias in the meta-analysis, we produced funnel plots of the data. It is assumed that an asymmetric funnel plot is a sign of publication bias (Egger, Smith, Schneider, & Minder, 1997). In this case the effect sizes for both trust and trustworthiness are distributed evenly on both sides of the mean. This is supported by Egger‟s regression test for

(24)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 24 funnel plot asymmetry, which gives a non-significant intercept for trust studies, z = 1.24, p = .217, and trustworthiness studies, z = -0.34, p = .734. This result can be visually confirmed when looking at the funnel plot of the studies on trust (see Figure 1) and the funnel plot of the studies on trustworthiness (see Figure 2).

The lack of publication bias is also evidenced by the fact that there was no systematic variation between the unpublished studies and the published studies in our meta-analysis. This was estimated by regressing the effect sizes on a dummy variable indicating whether a study was published in a scientific journal or not. The regression showed a non-significant coefficient for both trust, β1 = 0.03, p = .812, and trustworthiness, β1 = 0.08, p = .278, which means that the effect sizes of the published studies do not significantly differ from the effect sizes of the

unpublished studies. This is another sign that it is unlikely that our sample of studies is tainted by publication bias.

Figure 1

(25)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 25 Figure 2

Funnel Plot of the Studies on Sex Differences in Trustworthiness

Main effects analysis

Consistent with the prediction from parental investment theory, the random effects analysis indicated that males were more trusting in the trust game than females, although the effect was small, g = 0.22, 95% CI = [0.15, 0.29], p < .001. On the other hand, contrary to the expectation of social role theory, the analysis on trustworthiness did not show a significant sex difference, g = -0.02, 95% CI = [-0.07, 0.04]. In addition, the masculinity index of the countries in which the studies were carried out did not appear to play a role in sex differences in

trustworthiness either, β1 = -0.0002, p = .871. Table 5 includes a stem-and-leaf plot of the effect sizes for the studies on sex differences in trust and trustworthiness. One can tell from the plot that the peak of effect sizes of the trust studies lies around 0.2, whereas the peak of effect sizes for the trustworthiness studies lies around 0.0.

(26)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 26 When looking at trust games where the sex of the partner is known, we found no

evidence that first movers send more in same-sex partnerships than in different-sex partnerships. We found that men did not send more toward other men than toward women, g = 0.05, 95% CI = [-0.11, 0.20], p = .557, and we found that women did not send more toward other women than toward men, g = -0.10, 95% CI = [-0.23, 0.02], p = .109. Similarly, we did not find that second movers send more in same-sex partnerships: men did not send more toward men than they did toward women, g = 0.13, 95% CI = [-0.11, 0.36], p = .283, and women did not send more toward women than they did toward men, g = -0.18, 95% CI = [-0.39, 0.03], p = .099.

Aside from these main findings we did find some significant effects relating to first mover behavior. We found that men send more toward other men than women send to men, g = 0.32, p = .006, that women send more toward other women than men send to women, g = -0.36, p = .006, that men send more toward other men than women send toward other women, g = 0.36, p = .006, and that men send more toward women than women send toward men, g = 0.25, p = .049. These latter four p-values were adjusted using the Benjamini-Hochberg procedure.

When comparing second mover behavior when the sex of the partner was known we found no significant effects.

(27)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 27 Table 5

Stem-and-leaf Plot of the Effect Sizes for the Studies on Sex Differences in Trust and Trustworthiness

g 0.1 units of the g value for

Trust

0.1 units of the g value for Trustworthiness -1.1 32 -1.0 -0.9 -0.8 9 -0.7 2 -0.6 863 -0.5 2 85 -0.4 84 43 -0.3 50 852 -0.2 866100 7753322000 -0.1 941 85433 -0.0 775530 9886544331 0.0 0123345 0223357778 0.1 000124458 013346788 0.2 01233558899 222233459 0.3 0022233344677899 0224477 0.4 01449 68 0.5 112267899 2235 0.6 138 79 0.7 15 6 0.8 0.9 5 1.0 8 1.1 348 1.2 1.3 1.4 1.5 1.6 1

(28)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 28 Moderator analysis

Because of the possibility of inflated error rates, we decided to adjust all p-values in the moderator analysis using the Benjamini-Hochberg prodedure. We did this separately for the trust studies and the trustworthiness studies.

None of the features of the experimental setting proved to moderate the effect on sex differences in trust: whether the participants got paid for participating, β1 = 0.003, p = 1, the number of iterations of the trust game, β1 = -0.0001, p = 1, the size of the multiplier, β1 = -0.30, p = .765, whether players played as both the first mover and the second mover, β1 = -0.05, p = 1, whether the second mover was endowed with their own endowment, β1 = -0.02, p = 1, and whether the strategy method was used to elicit the behavior of the second mover, β1 = -0.08, p = 1.

The same holds for the effect on sex differences in trustworthiness: whether the participants got paid for participating, β1 = -0.11, p = 1, the number of iterations of the trust game, β1 = -0.002, p = 1, the size of the multiplier, β1 = -0.46, p = .908, whether players played as both the first mover and the second mover, β1 = 0.03, p = 1, whether the second mover was endowed with their own endowment, β1 = -0.05, p = 1, and whether the strategy method was used to elicit the behavior of the second mover, β1 = -0.006, p = 1.

A summary of all the moderator effects on trust can be found in Table 6, while a summary of all the moderator effects on trustworthiness can be found in Table 7.

(29)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 29 Table 6

Summary of the Moderator Effects on Sex Differences in Trust

Moderator Q k g 95% CI Payment 0.001 Yes 83 0.22*** [0.15, 0.29] No 7 0.29 [-0.14, 0.72] Both roles 0.34 Yes 27 0.19** [0.07, 0.31] No 63 0.24*** [0.15, 0.33]

Second mover endowment 0.06

Yes 42 0.22*** [0.10, 0.33] No 32 0.23*** [0.13, 0.33] Strategy method 1.23 Yes 24 0.19** [0.05, 0.33] No 50 0.26*** [0.18, 0.34] Iterations 0.01 90 Multiplier 2.32 90 Masculinity 0.14 86 Note: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 Table 7

Summary of the Moderator Effects on Sex Differences in Trustworthiness

Moderator Q k g 95% CI Payment -0.10 Yes 69 -0.03 [-0.09, 0.03] No 6 0.08 [-0.10, 0.26] Both roles 0.03 Yes 26 -0.01 [-0.08, 0.07] No 49 -0.04 [-0.13, 0.05]

Second mover endowment -0.05

Yes 44 -0.04 [-0.13, 0.04]

No 31 0.01 [-0.07, 0.08]

Strategy method -0.01

Yes 23 -0.01 [-0.12, 0.10]

(30)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 30 Moderator Q k g 95% CI Iterations 0.03 75 Multiplier 2.06 75 Masculinity 0.03 71 Note: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 Sensitivity analyses

To gauge the robustness of the sex difference in trust that we found in the main analysis, we performed several sensitivity analyses. To that end, we used several variables to create subsets of papers (with a higher than average quality) on which we ran the main analysis again. First, we looked at a subset of studies that did not have a main hypothesis regarding to sex. When only those studies were included, the overall effect size remained significant, g = 0.22, 95% CI = [0.15, 0.30]. In line with that finding, we did not find a difference in effect size between studies that had a main hypothesis regarding to sex (k = 7) and studies that did not (k = 83), β1 = -0.04, p = 0.793. Second, we looked at a subset of studies that we judged to have no potential confounds biasing the trust game experiment. When only those studies were included, the overall effect size remained significant, g = 0.26, p < .001. Again, we did not find a difference in effect size

between studies that had potential confounds (k = 39) and studies that did not (k = 51), β1 = -0.08, p = 0.275. Third, we looked at a subset of studies that were published in a scientific journal. When only those studies were included, the overall effect size still remained significant, g = 0.22, p < .001. In this case as well, we did not find a difference in effect size between studies that were published in a scientific journal (k = 80) and studies that were not (k = 10), β1 = 0.03, p = 0.812. An overview of these sensitivity analyses can be found in Table 8.

(31)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 31 Table 8

Summary of the Sensitivity Analyses on Sex Differences in Trust

Moderator Q k g 95% CI

Sex as main hypothesis 0.07

Yes 7 0.18 [-0.12, 0.49] No 83 0.22*** [0.15, 0.30] Potential confounds 1.19 Yes 39 0.18** [0.07, 0.29] No 51 0.26*** [0.17, 0.35] Published 0.06 Yes 80 0.22*** [0.14, 0.31] No 10 0.19* [0.04, 0.35] Note: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001

Finally, we have carried out several sensitivity analyses using subsets of studies with different sample sizes. To this end, we ran several power analyses to find the required effect sizes corresponding to varying a priori estimated effect sizes and a power of 0.8. The first column of Table 9 provides the a priori estimated effect sizes, while the second column provides the

corresponding required sample size per group. We ran several random-effects analyses with only the studies that matched these required sample sizes. For example, the first analysis was run with only studies that had an average sample size per group of at least 394. The third column provides the number of studies that fulfilled this requirement and the remaining columns provide the results from this particular sensitivity analysis. Figure 3 illustrates the information in Table 9 graphically. A discussion of this analysis is provided in the Discussion as it may have

(32)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 32 Table 9

Overview of the Sensitivity Analysis on Sex Differences in Trust Using Sample Size as the Subset Variable

Estimated a

priori ES Required N per group Number of studies left Effect size of Meta-analysis 95% CI

0.2 394 2 0.10 [-0.01, 0.20] 0.25 253 4 0.05 [-0.12, 0.22] 0.3 176 6 0.08 [-0.04, 0.19] 0.35 130 11 0.19** [0.05, 0.33] 0.4 100 16 0.27*** [0.14, 0.39] 0.45 79 22 0.22*** [0.11, 0.32] 0.5 64 28 0.19*** [0.08, 0.29] Note: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 Figure 3

A Graphical Representation of the Sensitivity Analysis on Sex Differences in Trust Using Sample Size as the Subset Variable

(33)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 33 Discussion

In this meta-analysis, we investigated the literature on sex differences in the trust game. In the trust game, the decision of the first mover is said to measure trust, while the decision of the second mover is said to measure trustworthiness. Based on parental investment theory and social role theory, we predicted that men send more than women as first movers, and that women send more than men as second movers. In addition, in line with social identity theory, we

expected both first and second movers to send more to their partners in same-sex partnerships than in mixed-sex partnerships. We ran separate random-effects analyses for all main hypotheses, and we included six moderators to see whether the experimental protocol is relevant for

explaining sex differences in the trust game.

In line with a key prediction from parental investment theory, we found that male first movers send more than female first movers (g = 0.22), although the effect is small. Additionally, inconsistent with a prediction from social role theory, we failed to find sex differences in second mover behavior. Finally, contrary to a prediction from social identity theory, we found no

evidence that first movers and second movers in same-sex partnerships send more than first and second movers in different-sex partnerships. Below we discuss the main results and their implications in more detail. We conclude by comparing our results to a recent meta-analysis on sex differences in the gift-exchange game, because we believe this yields important theoretical insights.

Trust

While the finding that men send more as first movers was expected a priori, a sensitivity analysis raised some concerns about the robustness of this effect. More specifically, we found that the overall effect size is non-significant when only the six largest studies are included. While

(34)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 34 this could be interpreted as a sign of publication bias, we found no evidence for publication bias, which implies we have to look for another explanation. To find another explanation we analyzed the six largest studies to see whether there were any structural differences between those studies and the 84 other studies. For example, it could be the case that the samples of the larger studies are more representative of the wider population than the samples of the smaller studies.

However, this was not the case: while one large study can indeed be described as having a representative sample (Yamagishi et al. 2015), four of the six largest studies had standard student samples (Becker, Deckers, Dohmen, Falk, & Kosse, 2011; Markowska-Przybyla & Ramsey, 2016; Sapienza, Toldra-Simats, & Zingales, 2013; Zhong et al., 2012), and one had a

homogeneous sample of financial investors (Riedl & Smeets, 2014). We also looked for other structural differences, but we were unable to find any. Because the large studies and the smaller studies do not seem to differ aside from their sample size we decided to lump all studies together and stick to the results from the main analysis: men send more money as first movers than women. However, it is important to keep in mind that the effect size of g = 0.22 is small.

To get an intuitive idea of the size of this effect, we created a fictional „average study‟ based on the information in our meta-analysis. In such a study, the group size of men is 66, the group size of women is 68, the standard deviation for men is 0.31, and the standard deviation for women is 0.27. All these numbers are based on the averages in our meta-analysis. Filling out this information in the formula for Hedges‟ g and assuming that women send 50% of their

endowment, we find that an effect size of 0.22 corresponds to men sending 57% of their endowment. This 7%-point difference can indeed be characterized as a small effect.

While the effect is small, the finding that men send more in the trust game is predominant in the existing literature; both the narrative reviews of Croson and Gneezy (2009) and Rau

(35)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 35 (2011) find men to be more trusting. The current meta-analysis corroborates this finding. The theoretical implications will be discussed later in this paper.

Trustworthiness

Regarding sex differences in second mover behavior, the conclusion from our analysis is more clear-cut. With an averaged effect size of g = -0.02 there seems to be no evidence that men and women differ in trustworthiness. An additional prediction of social role theory was not supported either: sex differences were not more pronounced in feminine societies than in

masculine societies. In all, we can conclude that the predictions from social role theory regarding sex differences in second mover behavior in the trust game are not supported. In fact, when looking at the mean amounts sent by men and women in all included studies, there appears to be no sex difference in second mover behavior at all.

This finding stands in contrast with previous narrative reviews of the trust game (Croson & Gneezy, 2008; Rau, 2011), which indicated that women generally send more than men as second movers. This discrepancy may be due to the fact that narrative reviews are more sensitive to bias than meta-analyses (Johnson & Eagly, 2000). While meta-analytic guidelines require researchers to document the procedures used in their meta-analyses in high detail, narrative review usually do not involve such guidelines. This leaves the door open for biases in authors‟ selection of studies and the conclusions they draw from that selection. Because of this larger sensitivity to bias and because our meta-analysis covers a larger number of studies than the narrative reviews we believe the results from this meta-analysis are more reliable. Therefore, we conclude that there is no support for the claim that women are more trustworthy than men.

(36)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 36 Trust and trustworthiness when the sex of the partner is known

To predict behavior of men and women in trust games where the sex of the partner is known we referred to social identity theory (Tajfel & Turner, 1979). Social identity theory predicted that first movers and second movers would send more when they were coupled with someone from the same sex than when they were coupled with someone from the opposite sex. This proved not to be the case. We did not find any significant difference in behavior in same-sex partnerships vs. opposite-sex partnerships. It appears that social identity does not properly

explain the behavior of both sexes in trust games where the sex of the partner is known. However, we do need to keep in mind that the trust games in this meta-analysis did not involve a direct comparison between male and female recipients. Both first and second movers were faced with only one other player and could not allocate their resources between two or more players. To more accurately assess the prediction of social identity theory we could employ a game in which senders in the trust game have to allocate a certain amount of money between a male and a female recipient. In this way, we can more directly measure people‟s preferences for same-sex or mixed-sex partners, and we can more accurately test the prediction from social identity theory.

Moderating Variables of Sex Differences in Trust and Trustworthiness

A priori, we speculated that women would be more sensitive to the experimental protocol of economic games than men. This idea was proposed by several authors (Chermak & Krause, 2002; Croson & Gneezy, 2009) so we decided to test that idea by including six moderators pertaining to the experimental protocol of the trust game: whether participants were paid for their participation, how many times the trust game was played during the experiment, what the

(37)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 37 and the second mover, whether the second mover had an initial endowment, and whether the strategy method was used to elicit the decisions of the second movers. Our analyses indicated that none of these moderators could explain sex differences in the trust game, which leads to the conclusion that there is no evidence that men and women are differentially influenced by the experimental protocol of the trust game.

This conclusion makes sense as there was no a priori reason to expect women to be more sensitive to the experimental protocol. However, some reservations need to be made regarding this conclusion. First, we may have failed to find a moderating effect of the six study protocol variables because the analyses we used lacked statistical power (Hedges & Pigott, 2004; Hempel et al., 2013). Especially moderator analyses are prone to low power as the number of studies in a category of a moderator variable could be relatively low. For example, in our case only 7 studies out of the 90 in the trust analysis did not involve payment of participants. Because of the low number of studies in this „no payment category‟ it could be that our analysis did not have enough power to detect an effect, and therefore caution is warranted when interpreting this null result as well as the other null results we found.

Second, we did not include all possible variations in the experimental protocol as it was impossible to code for some features of the experiment like the duration of the experiment or possible experimenter effects.

Third, our analysis only involves the trust game, so we cannot generalize our conclusion to other economic games. More specifically, it is possible that women are more sensitive to the experimental context in public good games as some findings suggest (Chermak and Krause, 2002;Croson & Gneezy, 2009 ).

(38)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 38 In all, more research is needed to find out if, and under what circumstances, there is a sex difference in the sensitivity to the protocols of experimental games.

A comparison with the gift-exchange game

To make sense of the findings of this meta-analysis we believe it is informative to relate its results to those from a recent meta-analysis by Van den Akker (2016) of sex differences in a related economic game: the bilateral gift-exchange game (Fehr, Kirchler, Weichbold, & Gächter, 1998). The gift-exchange is equivalent to the trust game aside from several important

differences. First, in the gift-exchange game the second mover sometimes has the option to reject the first mover‟s offer. A rejection phase can have important consequences for the first mover‟s behavior as he or she may be scared that the offer will be rejected (which leads to a payoff of zero). Second, the experimental instructions in the gift-exchange game are often framed in a labor context. That is, the first mover is referred to as the firm or the employer, while the second mover is referred to as the worker or the employee. This labor context could also have

implications for the behavior of both players.

While these are important differences between the trust game and the gift-exchange game, they are not fundamental as there are gift-exchange games that are framed neutrally and gift-exchange games in which rejection is not possible. The fundamental difference is that the trust game involves an efficiency gain through the decision of the first mover (as the offer of the first mover is tripled), and the gift-exchange involves an efficiency gain through the decision of the second mover or a combination of the first and second mover‟s decisions. Instructive in our case are the gift-exchange studies in which the efficiency gain comes about through the decision of the second mover only. In a way, these studies can be seen as „reversed‟ trust games – the

(39)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 39 games are equivalent, but in the gift-exchange game the multiplication effect resides in the second instead of the first transfer.

In his meta-analysis of the gift-exchange game, Van den Akker (2016) found no sex difference in first mover behavior, but he did find that men send more than women as the second mover. In a way, these results are „reversed‟ when compared to the results we found in our current trust game meta-analysis. More specifically, this suggests that men and women differ in their behavior only when a multiplication factor is involved. When a multiplication factor is involved, men tend to send more than women, both as a first mover (in the trust game) and as a second mover (in the gift-exchange game). In other words, there seems to be a „male multiplier effect‟. So how can we explain this „male multiplier effect‟? One possibility is to look for an evolutionary explanation.

Throughout our evolutionary past, men and women have encountered different

evolutionary challenges, which may have caused sex differences in all kinds of psychological traits. A familiar example is the sex difference in risk-taking, but it may also be possible that natural selection led to a sex difference regarding the acquisition of resources. In hunter-gatherer societies that have characterized our evolutionary past, women were largely responsible for taking care of the children, while men were largely responsible for acquiring resources (Hooper, Demps, Gurven, Gerkey, & Kaplan, 2015; Kaplan, Hill, Lancaster, & Hurtado, 2001). There are plenty of findings that show that women prefer male partners with a higher potential to attain resources (Buss, 1989; Fales et al., 2016), and this would help these male partners in propagating their genes to the next generation. In this way, males may have evolved a cognitive drive to acquire resources. Obviously, the first transfer in the trust game and the second transfer in the gift-exchange game are excellent opportunities for resource acquisition as in both cases the

(40)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 40 initial number of resources is multiplied. It might be that men, more so than women, are

triggered by the multiplication factor in both games and send more when that multiplier is present.

However, this explanation hinges on the fact that the other player is seen as part of the sender‟s in-group. In both games, the multiplication occurs only when money is transferred to another player. This means that the senders do not directly attain the benefits of the

multiplication. Only if senders would see their partner as an in-group member would this

explanation make sense. If this is not the case, first mover behavior in the trust game and second mover behavior in the gift-exchange should be explained in another way.

Future directions

A first step in making sense of the trust game and gift-exchange game data is to corroborate the male multiplier effect using a large-scale preregistered study. This study would employ two games that are exactly the same except for which transfer gets multiplied. In the „trust game‟ the first transfer would be multiplied, and in the „gift-exchange game‟ the second transfer would be multiplied. All other features of the games would be identical. If we find that men send more than women as first movers in the trust game and as second movers in the gift-exchange game, we can be more confident that a male multiplier effect actually exists. With that in mind, we can start to develop a testable theoretical framework that can explain this effect and can reconcile the findings in the trust game and gift-exchange game. An interesting starting point would be the evolutionary explanation we outlined above, but there may be other explanations, some of which may be rooted in other fields such as social psychology, or sociology. Given the importance of the acquisition of resources in our daily lives, this would be a worthwhile avenue for researchers from many scientific disciplines.

(41)

SEX DIFFERENCES IN TRUST AND TRUSTWORTHINESS 41 References

References marked with an asterisk indicate studies included in the meta-analysis.

*Ainsworth, S. E., Baumeister, R. F., Ariely, D., & Vohs, K. D. (2014). Ego depletion decreases trust in economic decision making. Journal of Experimental Social Psychology, 54, 40-49.

*Atlas, S., & Putterman, L. (2011). Trust among the avatars: A virtual world experiment, with and without textual and visual cues. Southern Economic Journal, 78(1), 63-86.

*Babin, J. J. (2016). A picture is worth a thousand words: Emojis, Computer-mediated communication, and trust (SSRN Working Paper 2883578). Retrieved from https:// papers.ssrn.com/sol3/papers.cfm?abstract_id=2883578.

Bakan, D. (1966). The duality of human existence. Chicago: Rand McNally.

Baker, M. D., & Maner, J. K. (2009). Male risk-taking as a context-sensitive signaling device. Journal of Experimental Social Psychology, 45(5), 1136-1139.

Bakker, M., Van Dijk, A., & Wicherts, J. M. (2012). The rules of the game called psychological science. Perspectives on Psychological Science, 7(6), 543-554.

Balliet, D., Li, N. P., Macfarlan, S. J., & Van Vugt, M. (2011). Sex differences in cooperation: A meta-analytic review of social dilemmas. Psychological Bulletin, 137(6), 881-909. Balliet, D., & Van Lange, P. A. (2013). Trust, punishment, and cooperation across 18 societies A

meta-analysis. Perspectives on Psychological Science, 8(4), 363-379.

Barford, A., Dorling, D., Smith, G. D., & Shaw, M. (2006). Life expectancy: Women now on top everywhere. British Journal of Medicine, 332, 808.

Referenties

GERELATEERDE DOCUMENTEN

Systems designed to record, classify and even change affective state may entail safety, privacy, and security issues.[ 3,8,11 ] As with pass- words and other private information,

Various charge redistribution mecha- nisms can be at play: charge compensation by nonstoichi- ometry, adsorption of foreign atoms or ions, electronic charge redistribution,

Onthouden, begrijpen en toepassen zijn natuurlijk noodzakelijk voor leren, maar wanneer we leerlingen werkelijk willen prikkelen en bij hen kwaliteiten willen ontwikkelen

equal to the sum of the adhesive layer thickness under the waveguides, the height of the waveguides and the height of the slab that is needed (figure 2) for the sample to operate

terestingly, a single transversal ramus (orange curve)) is more efficient at low and medium velocities and single sensilla (violet curve) are always more efficient than a

Moreover, we expected children enrolled in a number line training to make greater gains on a measure of number line acuity and children enrolled in a counting training to make

The study found that through organisational commitment and team commitment, respectively, trust in co-worker has a positive effect on organisational citizenship

pertains to the trust game only (the multiplication factor), four pertain to the gift-exchange game only (whether the experimental instruction was framed neutrally or was framed in