Gender differences in the
gift-exchange game: A meta-analysis
Master Thesis by Olmo van den Akker (5843049)
Master of Behavioral Economics and Game Theory
University of Amsterdam
22 November 2016
Supervisor: Matthijs van Veelen
Number of words: 8.335
2
Table of Contents
Abstract………...
3
Introduction………...
4
Previous literature……….……….
8
Added value……….
10
Methods………..
12
Search strategy...……….
12
Inclusion criteria………..
12
Summary table of all included studies……….………
13
Coding procedure……….………
15
Statistical methods………..
15
Results………
18
Heterogeneity analysis………..
18
Publication bias analysis………..
18
Main effect analysis……….
18
Funnel plot of studies on trust………
19
Funnel plot of studies on reciprocity……….. 19
Moderator analysis……….
20
Forest plot of studies on trust.………..
21
Forest plot of studies on reciprocity………..
22
Summary table of moderator analysis of studies on trust………..
23
Summary table of moderator analysis of studies on reciprocity….…… 23
Conclusion and discussion………
24
References……… 27
In-paper references……….. 27
Meta-analysis references………..………... 30
Appendix A – Summed payoff tables for Fooken (2013) and Owens (2011)………. 33
Appendix B – Excluded studies……… 34
3 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, I present a meta-analytic review of the literature on gender differences in the experimental gift-exchange game – an economic game that is frequently used to measure trust and reciprocity of trust. The meta-analysis consists of 32 studies and 1.308 participants. I found no difference between the two genders with regard to trust (Hedges’ g = 0.054) and I found that men reciprocate trusting behavior more than women (Hedges’ g = 0.256). I also considered several moderators, but none of these had an influence on gender differences in trust and reciprocity.
4 Introduction
Trust has been the topic of many studies in the past decades. It is now widely recognized that trust is a defining element of interpersonal relationships (Rempel, Holmes, & Zanna, 1985; Ferris et al., 2009) and that trust is an important factor in increasing cooperation (La Porta et al., 1996; Balliet & Van Lange, 2013) and economic growth (Beugelsdijk, De Groot, & Van Schaik, 2004). In addition, the existing literature seems to indicate positive associations between trust and personal characteristics such as intelligence (Sturgis, Read, and Allum, 2010) and religiosity (Glaeser, Laibson, Scheinkman, & Soutter, 1999; Guiso, Sapienza, & Zingales, 2003). Following up on these individual differences in trust, more and more studies have pursued the topic of gender differences in trust. Many of these studies also investigate the tendency for men and women to reciprocate trusting behavior. This paper will present a meta-analytic review of the studies on trust and reciprocity.
Trust and related concepts are generally measured using either field studies or experiments. The former often focus on surveys in which participants respond to Likert questions of the following form: “To what extent do you agree that you see yourself as someone who is generally trusting?” (e.g. Herd, Carr, & Roan, 2014). Since trust is generally seen as a desirable attribute (Rotter, 1967) this type of phrasing can be problematic due to the possibility of social desirability bias. Therefore, studies often include trust scales where people’s beliefs about the trustworthiness of others are taken as a measure of their level of trust (e.g. Yamagishi & Yamagishi, 1994; Reeskens & Hooghe, 2008). However, these adjusted trust scales still show signs of social desirability bias (Naef & Schupp, 2009) and also fail to show a consistent correlation with behavioral measures of trust (Glaeser et al., 1999). For these reasons, this meta-analysis will focus solely on studies that use experimental games to measure trust. These experimental games also have the added value of measuring people’s willingness to reciprocate trusting behavior.
There are numerous experimental games that are used to measure trust and reciprocity. The most frequently used is quite fittingly named the trust game and is developed by Berg, Dickhaut, and McCabe (1995). The trust game consists of two players and can be described as follows. One player, the first mover, is endowed with a certain amount of monetary units. This first mover has the choice to send a proportion of this monetary endowment to another player, the second mover. The monetary units he decides to give away are multiplied by a given factor before reaching the other player. Generally this factor is 3, but it varies over studies. In the second and final round, the second mover can decide how much of the monetary units 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 reciprocity.
5
A game that is fundamentally similar to the trust game is the bilateral gift-exchange game1
(Fehr, Kirchler, Weichbold, & Gächter, 1998). In the gift-exchange game the first player (the principal,
p) can again allocate a certain amount of monetary units (the wage, 𝑤𝑤𝑝𝑝 – typically an integer between 20 and 120) to the second player (the agent, a). In many (but not all) studies, the agent can either accept or reject this wage offer. In case of a rejection, both players get a payoff of zero. In case of acceptation, the agent has to decide on an effort level, 𝑒𝑒𝑎𝑎. This effort level is costly to himself, but beneficial to the principal. Typically, the payoff for the principal is ∏𝑝𝑝= �120 − 𝑤𝑤𝑝𝑝�𝑒𝑒𝑎𝑎, and the payoff for the agent is ∏𝑎𝑎= 𝑤𝑤𝑝𝑝− 𝑐𝑐(𝑒𝑒𝑎𝑎), where the cost, c is related to the effort level according to Table 1. However, it must be said that the payoff functions for both the principal and the agent vary markedly over studies. Common for all studies, though, is that the wage is seen as a measure of trust while the effort level is seen as a measure of reciprocity.
Table 1
The typical relationship between the agent’s effort and cost in the gift-exchange game Effort level 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Cost of effort 0 1 2 4 6 8 10 12 15 18
The trust game and the gift-exchange game are the most widely used games to measure trust and reciprocity. There are other economic games that measure trust and/or reciprocity, but they have not been used frequently enough to warrant a meta-analysis. Examples of such games are the real-effort dictator game (Heinz, Juranek, & Rau, 2012) and the moonlighting game (Abbink, Irlenbusch, & Renner, 2000). In addition, some games that have been linked to trust are not included because they are more likely to measure other personal characteristics. For example, cooperation in a public good game is more likely to be a signal of irrationality or altruism than it is of trust (Camerer & Fehr, 2002).
It is important to note that this thesis will only focus on gender differences in the gift-exchange game. The reason for this is that an affiliated research group at the Free University in Amsterdam is doing a meta-analysis on gender differences in the trust game. Eventually, these analyses will be synthesized and the resulting paper will be sent out for review at a high-impact scientific journal. Because of reasons of convenience - and because most research on these games is
1 The bilateral gift-exchange game differs from the original gift-exchange game developed by Fehr, Kirchsteiger,
and Riedl (1993) because it involves only two players. In the original game, more than two players participate in a double auction market. Such a market environment adds a competitive element to the game, which makes it more difficult to purely measure trust and reciprocity. That is why in this meta-analysis only bilateral gift-exchange games are used.
6
equally valid for both - I will use the term ‘TG games’ (Trust and Gift-exchange games) when I refer to both of the games.
As this thesis focuses on the gift-exchange game it may be instructive to take a closer look at the game and see what predictions game theory makes about the behavior of the principal and the agent. When assuming that the players in the game are both selfish and rational we would predict the principal to offer the lowest possible wage and the agent to put in the lowest amount of effort. The agent puts in minimum effort because higher effort comes with higher cost (and thus a lower payoff). The principal knows this, so he will provide the lowest wage possible to maximize his payoff. The game theoretical outcome is nicely illustrated in Table 2, where the upper box provides the payoffs for the principal for all possible effort levels and some wage levels (the outcomes of his best responses are highlighted in green), the middle box provides the payoffs for the agent (outcomes of best responses highlighted in green) and the bottom box provides the summed payoffs of both players (most efficient outcomes per wage level highlighted in green). The table clearly indicates that the best response of the principal is to always choose the minimum wage and the best response of the agent is to always choose the minimum effort level. The summed payoff of these choices is highlighted in red. Obviously, this outcome is far from efficient. In fact, it is the lowest possible outcome.
Important to note at this point is that, because the payoff functions of the principal and the agent vary over studies, the payoffs and the summed payoffs do not always correspond to those in Table 2. More specifically, the payoffs can be categorized on the basis of which player is responsible for achieving an efficient outcome. In the case of the payoffs in Table 2 it immediately becomes clear that the principal should always choose the maximum wage of 120 to achieve efficiency. However, there is no dominant choice for an efficiency-maximizing agent. For all wage levels up to and including 91, it is most efficient for the agent to put in maximum effort. However, for higher wage levels it becomes increasingly efficient to provide lower effort levels. From a wage level of 109 onwards it is actually most efficient for the agent to provide the minimum effort level. The reason for this is that the benefit of the agent’s effort to the principal is outweighed by the higher cost of that effort to the agent. In the end, the most efficient outcome comes about when the principal offers the maximum wage and the agent provides the minimum effort.
In this situation, the efficiency in the gift-exchange game depends on a combination of the choices of the principal and the agent, which makes it difficult to see who is primarily responsible for attaining an efficient outcome. However, there are two other situations where it is more clear who determines efficiency. A first example comes from Fooken (2013). In this study, the payoff functions are such that the summed payoffs differ only with relation to the effort level. That is, for each effort level, there is a summed payoff that is constant over all wage levels. An example of the summed
7
payoffs for the Fooken (2013) study can be found in Appendix B. From the payoff table in Appendix B it becomes clear that only the worker is responsible for achieving efficiency in this situation. A second example comes from Owens (2011). In this study, the payoff functions are such that - to achieve efficiency - the principal should choose the maximum wage, while the agent should choose the maximum effort level. This holds irrespective of the other player’s choice. Therefore, in this situation it is both the principal and the agent who (through dominant choices) are responsible for an efficient outcome. In short, it is either the agent, both the agent and the worker, or a combination of their choices that determines efficiency in gift-exchange games. I included the way efficiency is determined as one of the methodological moderators in the meta-analysis (see Table 3).
Because of the differences in payoff functions, the gift-exchange game differs from the trust game, where it is obvious that the added value comes about through the decision of the first mover (as his or her allocation will be tripled by the experimenter). This difference is essential because it can have implications for the behavior of the players in both games. At a later point I will discuss these implications for gender differences in both games.
But first it is important to make sure that TG games actually measure trust and reciprocity as opposed to some other variables. This is essential, because it is possible that TG games do not measure trust in total isolation. For example, it can be argued that giving money to the agent is a manifestation of risk-seeking behavior more so than it is a manifestation of trust. This does seem to make sense, since the principal puts himself in a position of risk when he decides to hand over some of the money. The debate whether TG games measure trust or measure risk preferences has been going on for quite a few years. Several studies have tried to resolve this debate by disentangling trust and risk aversion in trust games. In general, as pointed out by Friebel, Lalanne, Richter, Schwardmann, & Seabright (2013) “the majority of papers have found no effect of risk aversion on
trust decisions.” For this reason, it is concluded that TG games are no consistent measures of risk
preferences, so the premise that TG games measure trust still holds.
One other potential confounding variable has been the focus of discussion: social preferences. Social preferences can be defined as “a concern for the payoffs allocated to other
relevant reference agents in addition to the concern for one’s own payoff” (Carpenter, 2008). It can
be argued that money sent by the principal is indicative of social preferences and not an indication of trust. For instance, people may send money because of an aversion to monetary inequalities or a desire to make others happy. Following up on this idea, Brülhart and Usunier (2012) investigated whether the trust game is likely to measure trust or to measure altruistic tendencies by giving principals information about the neediness of second movers. They found support for the idea that the trust game measures trust, and no support for the idea that the trust game measures altruism. In
8
all, it can be confidently stated that TG games measure trust more so than other personal characteristics.
Table 2
Some of the outcomes of the gift-exchange game
e=0.1 e=0.2 e=0.3 e=0.4 e=0.5 e=0.6 e=0.7 e=0.8 e=0.9 e=1.0
w=20 10 20 30 40 50 60 70 80 90 100 w=40 8 16 24 32 40 48 56 64 72 80 w=60 6 12 18 24 30 36 42 48 54 60 w=80 4 8 12 16 20 24 28 32 36 40 w=100 2 4 6 8 10 12 14 16 18 20 w=120 0 0 0 0 0 0 0 0 0 0 w=20 20 19 18 16 14 12 10 8 5 2 w=40 40 39 38 36 34 32 30 28 25 22 w=60 60 59 58 56 54 52 50 48 45 42 w=80 80 79 78 76 74 72 70 68 65 62 w=100 100 99 98 96 94 92 90 88 85 82 w=120 120 119 118 116 114 112 110 108 105 102 w=20 30 39 48 56 64 72 80 88 95 102 w=40 48 55 62 68 74 80 86 92 97 102 w=60 66 71 76 80 84 88 92 96 99 102 w=80 84 87 90 92 94 96 98 100 101 102 w=100 102 103 104 104 104 104 104 104 103 102 w=120 120 119 118 116 114 112 110 108 105 102 Previous literature
As of yet, no meta-analysis regarding gender differences in trust and reciprocity has been carried out, making this paper the first of its kind. Johnson and Mislin (2011) did do a meta-analysis of trust games in which they looked at the factors influencing people’s trusting behavior. However, only a limited number of studies in their meta-analysis contained information about the gender of the participants. This made it impossible for the authors to systematically study gender differences.
While no meta-analysis has been carried out, there are two systematic narrative reviews on the topic. Croson and Gneezy (2009) reviewed general preference differences between men and women and included eleven experimental studies that specifically pertain to trust and reciprocity. Overall, they found that men trust more in trust games and related games, while women reciprocate more. Rau (2011) arrives at the same conclusion by using data from trust games, gift-exchange games, and real effort tasks. Given these findings, it is likely that this meta-analysis will find two positive effects: men offer higher wages, and women choose higher effort levels.
9
One interesting thing I noticed while scouring the literature was that many authors do not provide an explanation for the gender differences in trust and reciprocity2. That is, no models or
theories are proposed that may account for the fact that men trust more and women reciprocate more. The only authors that do provide a theoretical conjecture are Buchan, Croson, and Solnick (2008) who propose social role theory as the driving force behind gender differences in both trust and reciprocity.
Social role theory is one of the most popular theories to account for gender differences in social behaviors (Eagly, 1987, for an overview see Dulin, 2007). The theory states that men and women develop expectations about the way they ought to behave based on traditional gender roles3, and that men and women behave accordingly. Traditional gender roles prescribe that men are
more competitive, independent, task-oriented, and self-confident, while women are more sensitive, nurturant, expressive, and warm. Bakan (1966) labeled these groups of personality characteristics as agentic and communal respectively, and these labels have been widely used by researchers since. Supporting social role theory, studies have shown that men are indeed perceived as more agentic, and women are indeed perceived as more communal (Eagly & Steffen, 1984; Spence & Buckner, 2000; Madera, Hebl, & Martin, 2009). Moreover, it has been shown that men and women accommodate their behavior to match existing gender roles. That is, women actually behave more communally, and men actually behave more agentic (Eagly & Johnson, 1990; Feingold, 1994; Echabe & Castro, 1999; Taylor, Klein, Lewis, Gruenewald, Guring, & Updegraff, 2000). Social role theory has been used to account for gender differences in all kinds of social behaviors, but only Buchan et al. (2008) tried to apply the theory to gender differences in trust and reciprocity.
In their paper, Buchan et al. (2008) use the widespread notion that the positive expectation of the behavior of another is a crucial determinant of trust (Rousseau, Sitkin, Burt, & Camerer, 1998). This notion has been supported by the finding that expectations account for a very large part of the variance in trusting behavior (Ashraf, Bohnet, & Piankov, 2006; Sapienza, Simats, & Zingales, 2007; Garbarino, & Slonim, 2009). In light of the gift-exchange game, Buchan et al. hypothesize that the principal’s expectations of the agent’s reciprocity will have a stronger impact on men than on women. This makes sense because men, who are prescribed by their social roles to be more competitive and goal-oriented, are more likely to value the final outcome more (Eagly & Johnson, 1990). However, Buchan et al. go on to say that this implies that men will be more trusting than women. This is not a logical assertion. The fact that men’s behavior is more strongly affected by expectations does not necessarily mean that those expectations are always positive. And it is
2 This is likely the case because many studies only included gender in their studies to avoid omitted variable
bias. Their main hypotheses were not related to gender so that explains why they did not discuss this issue.
3 Of course, these traditional gender roles did not come into being out of nothing, but the origin of these
10
precisely those positive expectations that are assumed to determine trust. To my knowledge, neither social role theory nor any other theory makes a prediction about the degree of positive expectations in men and women. There are, however, some empirical findings related to this matter. While some studies find that men have more positive expectations in the trust game (Wang & Yamagishi, 2005; Schwieren & Sutter, 2007) others studies do not (Chaudhuri, & Gangadharan, 2003; Sapienza et al., 2007; Buchan, Croson, & Solnick, 2008). In all, social role theory does not predict any gender difference in trusting behavior and the evidence on positive expectations is mixed. However, as the narrative reviews by Croson and Gneezy (2009) and Rau (2011) did find men to be more trusting than women we still expect to find that in this meta-analysis as well.
Regarding reciprocity, Buchan et al. (2008) also refer to social role theory and in this case their conjecture is more plausible. They hypothesize that social obligations will have a stronger impact on the behavior of females than on the behavior of males. This is based on the idea that women are more communal and thus more focused on interpersonal relations and being perceived as caring and warm (Schmitt, Realo, Voracek, & Allik, 2008). Previous research also indicates that women are more agreeable, meaning they prefer cohesion over conflict (Weisberg, DeYoung, and Hirsh, 2011). These findings suggest that women are not likely to violate the gift-exchange relationship by failing to reciprocate. On the other hand, men, who are more goal-oriented and competitive, are less likely to value the gift-exchange relationship and focus more on the outcome of the exchange (Eagly & Johnson, 1990). This leaves more room to not reciprocate. In all, the theoretical prediction matches nicely with the findings of the narrative reviews that women reciprocate more. Therefore, in this meta-analysis we expect to find that women provide higher effort levels than men.
Added value
The current study is the first meta-analysis of gender differences in trust and reciprocity, so in that sense this study adds to the existing literature. A meta-analysis is preferred over narrative reviews because, as noted by Johnson and Eagly (2000), narrative reviews come with several limitations. The overarching limitation of narrative reviews is the large degree of subjectivity that inherently comes with them. This subjectivity mainly comes about through authors’ selection of studies and their conclusions based on these studies. For example, authors often have certain beliefs about their topic of interest, and their study selection and their evaluation of the evidence may be biased by these beliefs. Specifically, authors may have a tendency to select studies or overweigh studies that support their positions. This leads to confirmation bias (Littell, 2008). Another source of subjectivity is the so-called availability bias, the phenomenon that people “judge frequency or probability by the ease with
11
of narrative reviews, it could be the case that salient article titles are weighed more than non-salient article titles in estimating the magnitude of the relationship between two variables. This is exactly what Bushman and Wells (2001) found. In addition, they found that people who were schooled in meta-analytic techniques did not show this bias, indicating that meta-analysts are less prone to this kind of subjectivity. This bigger tendency for objectivity by meta-analysts makes sense since it is common practice in meta-analyses to include a formal protocol in which the precise selection criteria and the precise weighing criteria of studies are stated. This also holds for this meta-analysis.
Another disadvantage of narrative reviews versus meta-analyses is their focus on the p-values of individual studies (i.e. statistical significance). This focus makes it impossible to come to an overarching quantitative measure since there is no way to synthesize p-values from different studies. In meta-analyses, individual effect sizes are used, which can be combined to create an overall effect size. This has a clear advantage because a quantifiable measure is more objective than a subjective judgment.
In addition to this added value, this study adds to the narrative reviews by including a larger amount of individual studies. Recent years have seen several new studies on the topic of gender differences in trust and reciprocity, all of which also need to be taken into account to come to a trustworthy conclusion. Moreover, the narrative reviews only included studies that intentionally set out to study gender differences but many articles in which trust was studied in a different context also provide information about gender. These studies are also included in this meta-analysis to be sure that as much information is used as possible.
Based on the arguments above, this meta-analysis on gender differences in trust and reciprocity is assumed to be more objective and more informative than the existing narrative reviews on the subject.
12 Methods Search strategy
To find eligible studies for this meta-analysis I employed four search strategies. First, I used the term “gift exchange game” in a Google Scholar search and in a search of the Web of Science Social Sciences Citation Index. I only looked at papers published in 1998 or later because the bilateral gift-exchange game was introduced in that year. Second, I looked at all papers citing the original 1998 reference. Third, I looked at references in review articles and other relevant articles that I found using the first two search strategies. Fourth, I sent out a request for studies in the Economic Science Association’s experimental methods discussion group (http://groups.google.com/group/esadiscuss). The total number of studies that I found using these four search strategies was 1.405. After scanning the abstracts and the introduction of those papers I found that 129 of those papers actually employed a gift-exchange game.
Inclusion criteria
I used several inclusion criteria to select which of the 129 leftover studies would be included in the analysis. First, because of language barriers, I decided to only include English papers. Second, all participants had to be 18 years or older because there are indications that gift-exchange behavior is different for children and adults (Owens, 2011). Third, only studies that employed a one-shot bilateral gift-exchange game were included. All studies that used a different design were excluded to avoid possible confounding factors. I excluded studies employing a repeated game, studies where one principal was coupled with multiple agents, studies where the agent decided before the principal, studies without a control group, studies using a market design, and studies that otherwise had a different setup. Of course, I also excluded studies that did not measure participant’s gender or where the participants were solely male or female. In total, 83 studies were excluded. An overview of the excluded studies, including the reasons for exclusion, can be found in Appendix B. After excluding the 83 studies I was left with 46 suitable studies. Unfortunately, none of the papers included the required data on gender differences, so I had to contact all authors by-email for a data request. I received the data for 32 studies, which corresponds to a response rate of 69.6%. These 32 studies provided 31 effect sizes for wage decisions4, and 32 effect sizes for effort decisions. An overview of
all included studies can be found in Table 3.
4 In the Gose (2013) paper only 1 female participant made a wage decision, which made it impossible to
13 Table 3
Summary of included studies
Paper Study / condition Pub Frame Des Rej SM Eff N1 (m) N2(m) g (W) g (E/W)
Aquino, Gazzale, & Jacobson (2015) Baseline NP Neutral No No No Agent 36 (19) 36 (19) 0.006 0.002 Benjamin, Choi, & Fisher (2016) Control P Labor No No Yes Standard 226 (101) 226 (101) -0.110 -0.018
Bergstresser (2009) Out-group NP Neutral No No No Agent 60 (33) 60 (33) 0.020 0.075
Charness, et al. (2011) Control strangers P Labor Yes No No Both 24 (12) 24 (10) 0.100 0.042 Chaudhuri, Cruickshank, & Sbai (2015) Endowment P Labor Yes Yes No Agent 37 (17) 37 (23) 0.090 0.332 Chaudhuri, Cruickshank, & Sbai (2015) No endowment P Labor Yes Yes No Agent 38 (19) 38 (21) 0.145 0.168 Dariel & Nikiforakis (2014) Gift-exchange game P Labor No No No Agent 24 (10) 24 (14) 0.129 0.283 Dariel & Riedl (2013) Strategy method NP Labor No No Yes Agent 20 (11) 20 (12) -0.291 -0.021
Dariel & Riedl (2013) Actual game NP Labor No No No Agent 20 (11) 20 (8) 0.104 1.159
Eriksson & Villeval (2008) High skill P Neutral Yes Yes No Both 12 (8) 12 (2) 0.861 -0.249 Eriksson & Villeval (2008) Low skill P Neutral Yes Yes No Both 11 (8) 11 (5) -0.147 0.339
Fooken (2013) Australian NP Labor No Yes No Agent 40 (19) 39 (21) -0.095 -0.171
Fooken (2013) Intercultural NP Labor No Yes No Agent 36 (19) 36 (19) -0.033 -0.098
Franke, Gurtoviy, & Mertins (2014) Control P Labor Yes Yes No Standard 47 (15) 47 (26) -0.566 1.528
Gose (2013) 165 – different partners P Labor No Yes No Agent 12 (11) 12 (6) NA -0.173
He, Fooken, & Dulleck (2015) Gift-exchange game P Labor No Yes No Agent 108 (68) 108 (62) 0.557 0.770 Kocher & Sutter (2007) Individual P Neutral No No No Mixed 28 (16) 28 (18) -0.010 0.978
14
Lönnqvist, et al. (2013) Beer-Sheva P Labor No No Yes Agent 36 (18) 36 (18) 0.091 -0.067
Lönnqvist, et al. (2013) Bonn P Labor No No Yes Agent 35 (18) 36 (18) -0.417 0.439
Lönnqvist, et al. (2013) Jerusalem P Labor No No Yes Agent 36 (18) 36 (17) -0.028 0.219
Lönnqvist, et al. (2013) Köln P Labor No No Yes Agent 36 (18) 36 (18) -0.621 0.364
Lönnqvist, et al. (2013) Rammalah P Labor No No Yes Agent 35 (17) 36 (18) -0.118 0.130
Maximiano, Sloof, & Sonnemans (2007) 1-1 P Labor No No No Agent 17 (12) 20 (16) 0.167 0.192 Mertins & Warning (2014) Control NP Labor No No Yes Standard 55 (22) 55 (22) 0.150 0.102
Owens (2011) Adults P Neutral No No No Both 16 (3) 17 (3) -0.293 0.586
Owens (2011) Undergraduates P Neutral No No No Both 27 (17) 27 (16) -0.101 0.406
Owens (2012) Full responsibility P Labor No No No Both 76 (44) 76 (46) 0.376 -0.049
Owens & Kagel (2010) MWtoNO 1-5 P Labor No No No Both 30 (13) 30 (14) 0.578 0.084
Owens & Kagel (2010) NOtoMW 1-5 P Labor No No No Both 28 (18) 28 (17) 0.265 -0.142
Petit (2009) Control NP Neutral No No No Standard 37 (24) 35 (16) 0.538 0.680
Yamamori & Iwata (2015) Gift-exchange game NP Labor No Yes Yes Standard 26 (16) 26 (16) -0.297 -0.290 The study column indicates which of the studies in the paper was included. The ‘Pub’ column indicates whether the study was published in a peer-reviewed journal (‘P’) or not (‘NP’). The column ‘Frame’ indicates whether the experimental instruction of the gift-exchange game was framed neutrally or was framed in a labor context. ‘Des’ states whether the principals in the study had to fill out a desired effort level or not. ‘Rej’ states whether the agents were able to reject the principal’s wage offer or not. ‘SM’ states whether the effort decisions of the agents were elicited using the strategy method or not. ‘Eff’ states whether the efficiency was determined by only the agent, both the agent and the principal, or by a combination of choices (‘standard’). N1 (m) indicates the number of participants (males) that had to make a wage decision. N2 (m) indicates the number of participants (males) that had to make an effort decision. The ‘g(W)’ column provides the effect size for the wage decisions. The ‘g(E/W)’ column provides the effect size for the effort/wage proportions.
15
Coding procedures
Of all eligible studies, I extracted the mean wage and the mean effort/wage proportion, as well as the standard deviations of those means for both genders. I also extracted the number of males and females making the wage decision(s) and effort decision(s). I chose to use the mean effort/wage proportion because it could be the case that, on average, men or women were allocated a higher initial wage. If that was the case, a higher effort level by men would not necessarily indicate a higher level of reciprocity. For example, if men received a mean wage of 100 and women received a mean wage of 200, an effort level of 90 would be high for men but low for women. Using the effort/wage proportion solves this problem.
I included the following five moderators in the analysis: 1) whether the experimental instructions were framed in a labor context or in a neutral context, 2) whether principals could suggest an effort level before the agent’s effort decision, 3) whether the wage decision could be rejected by the agent, 4) whether the strategy method was implemented to measure effort decisions, and 5) whether the efficiency was determined by only the agent, by both the agent and the principal, or by a combination of their choices. Additionally, I looked at the publication status of all studies (published in a peer-reviewed journal or not) to identify and possibly control for publication bias.
Statistical methods
To quantify gender differences in trust and reciprocity I used the Hedges’ g effect size measure. I chose Hedges’ g over Cohen’s d because the latter is biased for small sample sizes (Lakens, 2013). Hedges’ g is calculated as follows.
𝑔𝑔 =(𝑥𝑥̅1− 𝑥𝑥̅2) 𝑠𝑠∗
where 𝑥𝑥̅1 is the mean for men, 𝑥𝑥̅2 is the mean for women, and 𝑠𝑠∗ is the pooled standard deviation that is calculated as:
𝑠𝑠∗= �(𝑛𝑛1− 1)𝑠𝑠12+ (𝑛𝑛2− 1)𝑠𝑠22 𝑛𝑛1+ 𝑛𝑛2− 2
where 𝑛𝑛1 is the number of males, 𝑛𝑛2 is the number of females, 𝑠𝑠12 is the variance for males, and 𝑠𝑠22 is the variance for females.
Even though the experimental designs of the studies included in the meta-analysis are very similar, I decided to use a random effects model to calculate the overall effect size. I did this to account for possible variation in effect sizes (i.e. study heterogeneity). This variation could have come about through different samples (and thus possibly different gender roles) or through slight
16
differences in design (for example, some studies used the strategy method to elicit effort and others did not). Study heterogeneity was measured through Cochran’s Q-statistic:
𝑄𝑄 = � 𝑤𝑤
𝑖𝑖(𝑇𝑇
𝑖𝑖− 𝑇𝑇�)
𝑘𝑘𝑖𝑖=1
²
where 𝑇𝑇𝑖𝑖 is the effect sizes of study i, 𝑇𝑇� is the mean effect size, 𝑤𝑤𝑖𝑖 is the weighting factor for study i, and k is the number of studies. Q is referred to a 𝜒𝜒2 distribution with k-1 degrees of freedom. A significant Q statistic indicates heterogeneity of the studies in the analysis, while a non-significant Q statistic indicates homogeneity. For a more formal explanation of fixed effect models, random effects models, and Cochran’s Q see Appendix B.
In meta-analyses it has become standard practice to see whether the analysis is tainted by publication bias – the tendency to publish significant findings more often than non-significant findings. Publication bias has been prevalent in many meta-analytic reviews (Bakker, Van Dijk, & Wicherts, 2012) and is troublesome because it can unjustly inflate the overall effect size. To identify publication bias I created a funnel plot. A funnel plot is a scatterplot that plots the studies of a meta-analysis against some measure of study size or precision (see Figure 1 for two examples). I used the standard errors of the studies as a measure of precision because of statistical reasons that are outside the scope of this thesis (but see Sterne & Egger, 2001). Because the standard error and the sample size are inversely related it is difficult to properly compare funnel plots that use the standard error and funnel plots that use the sample size. Therefore, the standard error is generally plotted on a reversed scale with higher standard errors closer to the origin. This means that more precise (or larger) studies can be found in the upper part of the funnel plot, while less precise (or smaller) studies can be found in the lower part of the funnel plot. In general, two diagonal lines are included in a funnel plot. These lines indicate the 95% confidence intervals around the mean effect size. Of course, confidence intervals are wider near the bottom because the standard errors are higher there. This implies that more precise studies are expected to be closer to the mean effect size, whereas less precise studies are expected to sometimes deviate significantly from the mean effect size.
So how can we identify publication bias from a funnel plot? The key here lies in the smaller (less precise) studies. It is assumed that large and precise studies are published regardless of whether the study effects are positive, negative or non-existent. In contrast, smaller studies that do not find an effect are often rejected by scientific journals or even remain in a researcher’s file drawer because the result is deemed uninteresting. So, small studies without an effect (or with a negative effect) are often not published. Visually, this leads to a ‘gap’ in the funnel plot (see the right panel of Figure 1). In such a situation, the effect sizes are not scatted randomly around the mean effect size as would be the case if all studies were published (see the left panel of Figure 1). Instead, the scatterplot becomes
17
asymmetrical due to this publication bias gap5. It is exactly this asymmetry that is used to identify
publication bias. Of course, visually estimating funnel plot asymmetry is informative, but there are ways to quantify the (a)symmetry of a funnel plot. I decided to do that through computing Kendall’s tau, a commonly used method to measure funnel plot asymmetry. Kendall’s tau calculates the rank correlation between the standard error and the effect size. A lack of correlation indicates symmetry, while a positive or negative correlation indicates asymmetry. I calculated Kendall’s tau for both the wage and the effort decisions.
Figure 1
A funnel plot without (left) and with (right) publication bias (adapted from Scherer, 2012)
5 This is a situation of one-sided publication bias, but there are also situations of two-sided publication bias. In
such situations, there are studies in the bottom left and bottom right corner of the funnel plot, but no studies in the bottom center (as that is the location that indicates no effects were found).
18 Results Heterogeneity analysis
To assess the diversity of the studies in the meta-analysis, I performed a statistical test of heterogeneity for both the wage decisions and the effort decisions. For the wage decisions, the test showed the studies in the analysis to be homogeneous, Q(30) = 30.263, p = 0.452. For the effort decisions, the test showed that the studies in the meta-analysis have systematic differences, Q(31) = 49.496, p = .019. The later finding indicates that the choice of a random effects model was justified.
Publication bias analysis
To test for the presence of publication bias in the meta-analysis, I produced funnel plots of the data. It is assumed that asymmetric funnel plots are a sign of publication bias (Egger, Smith, Schneider, & Minder, 1997). Fortunately, in this case the data points are distributed evenly on both sides of the mean, both for the wage decision, Kendall’s tau = -0.082, p = 0.518, as well as for the effort decision,
Kendall’s tau = 0.188, p = .132. This can be visually confirmed when looking at the funnel plot of the
studies on trust (see Figure 2) and the funnel plot of the studies on reciprocity (see Figure 3).
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 calculating Cochran’s Q, where
𝑔𝑔
𝑖𝑖 is now the effect size of a subgroup as opposed to the effect size of an individual study. As only two groups are compared, the degrees of freedom of the Q-statistic now becomes k-1 = 1. For the wage studies, Q(1) = 0.060, p = 0.807. For the effort studies, Q(1) = 1.589, p = 0.207. This non-significance makes sense from an intuitive point of view, because most of the studies included in the meta-analysis did not hypothesize gender differences and did not even include them in the paper. This means that a bias towards publishing a significant gender difference in trust or reciprocity was unlikely. For more information on identifying publication bias in meta-analyses see Scherer (2012).Main effect analysis
The random effects analysis indicated no significant difference between males and females with regard to trusting behavior, Hedges’ g = 0.054, 95% CI = [-0.056, 0.164]. Figure 4 includes a forest plot of the studies on trust, in which you can see that the individual effect sizes are scattered around the 0-axis (indicating the absence of an effect)
On the other hand, the analysis regarding the effort decision did show a significant gender difference, Hedges’ g = 0.256, 95% CI = [0.108, 0.403]. This difference means that males reciprocated more than females. Figure 5 includes a forest plot of the studies on reciprocity, in which you can see
19
that individual effect sizes of most studies lie to the right of the 0-axis (indicating that in most studies men reciprocated more than women).
Figure 2
Funnel plot of the studies on trust
Figure 3
Funnel plot of the studies on reciprocity
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 0,0 0,2 0,4 0,6 0,8 St an dar d E rro r Hedges's g
Funnel Plot of Standard Error by Hedges's g
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 0,0 0,2 0,4 0,6 0,8 St an dar d E rro r Hedges's g
20
Moderator analysis
Because of the possibility of inflated error rates, I decided to test all 10 moderator effects using a Bonferroni correction (Armstrong, 2014). This leads to a confidence level of 𝛼𝛼 = 0.0510 = 0.005. The effects of the moderators were tested using Cochran’s Q, where
𝑔𝑔
𝑖𝑖 is now again the effect size of a subgroup as opposed to the effect size of an individual study.Trust
With regard to trust, I found no significant differences in effect size when comparing studies with a labor frame and studies with a neutral frame, Q(1) = 0.124, p = 0.724, when comparing studies where the principal had to provide a desired effort level versus studies where the principal did not have to,
Q(1) = 0.165, p = 0.684, when comparing studies where the agent could reject the principal’s wage
offer versus studies where the agent could not, Q(1) = 0.019, p = 0.889, when comparing studies where the agent’s decisions were elicited using the strategy method versus where the agent’s decision were elicited behaviorally, Q(1) = 5.233, p = 0.022, and when comparing studies where the efficiency was determined only by the agent, by the principal and the agent, or by a combination of their choices, Q(2) = 3.054, p = 0.217. A summary of the moderator analysis of gender differences in trust can be found in Table 4.
Reciprocity
With regard to reciprocity, I found no significant differences in effect size when comparing studies with a labor frame and studies with a neutral frame, Q(1) = 0.383, p = 0.536, when comparing studies where the principal had to provide a desired effort level versus studies where the principal did not have to, Q(1) = 0.595, p = 0.441, when comparing studies where the agent could reject the principal’s wage offer versus studies where the agent could not, Q(1) = 0.233, p = 0.629, when comparing studies where the agent’s decisions were elicited using the strategy method versus where the agent’s decision were elicited behaviorally, Q(1) = 2.298, p = 0.130, and when comparing studies where the efficiency was determined only by the agent, by the principal and the agent, or by a combination of their choices, Q(2) = 2.357, p = 0.308. A summary of the moderator analysis of gender differences in reciprocity can be found in Table 5.
21 Figure 4
22 Figure 5
23 Table 4
Summary of the moderator analysis on gender differences in trust
Variable and class Q k n g 95% CI
Frame 0.124
Neutral 8 227 0.099 [-0.165, 0.363]
Labor 23 1.075 0.046 [-0.089, 0.181]
Desired effort level? 0.165
No 25 1.133 0.065 [-0.053, 0.182] Yes 6 169 -0.009 [-0.341, 0.324] Possibility to reject? 0.019 No 22 947 0.034 [-0.093, 0.162] Yes 9 355 0.056 [-0.229, 0.342] Strategy method? 5.233 No 21 752 0.164 [0.019, 0.308] Yes 10 550 -0.093 [-0.260, 0.073] Efficiency through? 3.054 Agent 17 659 0.061 [-0.090, 0.213] Both 8 224 0.265 [-0.001, 0.532] Standard 6 419 -0.059 [-0.314, 0.195] * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001 Table 5
Summary of the moderator analysis on gender differences in reciprocity
Variable and class Q k n g 95% CI
Frame 0.383
Neutral 8 226 0.332 [0.065, 0.599]
Labor 24 1.082 0.231 [0.057,0.406]
Desired effort level? 0.595
No 26 1.139 0.211 [0.070, 0.353] Yes 6 169 0.431 [-0.109, 0.971] Possibility to reject? 0.233 No 22 366 0.175 [0.046, 0.303] Yes 10 942 0.274 [-0.109, 0.658] Strategy method? 2.298 No 22 765 0.314 [0.109, 0.518] Yes 10 543 0.109 [-0.059, 0.277] Efficiency through? 2.357 Agent 18 666 0.266 [0.102, 0.429] Both 8 225 0.072 [-0.194, 0.338] Standard 6 417 0.472 [-0.052, 0.996] * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001
24
Conclusion and discussion
In this meta-analysis, I investigated the literature on gender differences in the gift-exchange game. I found no gender differences in trusting behavior, but I did find gender differences in reciprocity – men reciprocated trust more than women. Both findings are unexpected because the previous reviews on gender differences and reciprocity found men to be more trusting and women to be more reciprocal. However, in hindsight, the finding that neither men nor women are more trusting is not that surprising as there was no theoretical justification for a gender difference in trust. The discrepancy between my findings and the findings of Croson and Gneezy (2008) and Rau (2011) could have been caused by the fact that narrative reviews are inherently biased. The current meta-analysis is more objective and takes into account much more information than the previous reviews. Therefore, the finding that men and women do not differ in trusting behavior is assumed to be valid.
The situation is more complex for the finding that men reciprocate more than women. This result is the polar opposite of the prediction, which was theoretically justified and supported by previous findings. We can again assume that the narrative reviews were biased and the finding from this meta-analysis is more reliable, but that still does not explain why men, and not women, reciprocate more in the gift-exchange game. Some insight may be gained when looking at the moderators in this meta-analysis. It could very well be that one of the moderators accounted for the fact that males reciprocate more than females. For example, it could be that labor-framed experimental instructions caused different behaviors in men and women than neutrally framed instructions. This is plausible as it has been shown that women are more sensitive to the specific contexts of experimental games (Eckel, & Grossman, 2001; Solnick, 2001; Cox & Deck, 2006). However, this was not the case; a moderator analysis indicated that the overall effect size did not change for labor-framed studies and neutrally framed studies. Other potential moderators of gender differences in reciprocity could have come about through the fact that the gift-exchange games were designed slightly differently from study to study. However, no significant difference in effect sizes was found when looking at studies where principals could or could not suggest an effort level, when looking at studies where the wage decision could or could not be rejected, when looking at studies where the strategy method was or was not implemented, and when looking at studies in which the efficiency was determined by the agent, both the principal and the agent, or a combination of their choices. In all, the moderators in this meta-analysis could not explain the gender difference in reciprocity.
A different, highly speculative, explanation is that the threshold may be lower for men to feel connected to someone else. Maybe men are able to establish relationships with only minimal contact, while women need more interpersonal communication and emotional bonding. This idea makes sense in the light of previous research that has shown that women value talking, emotional
25
sharing, and discussing personal problems in friendships, while men show an emphasis on sharing activities and doing things (Williams, 2015). If men indeed have a lower relationship threshold, it could be that men established a connection with the other player in the gift-exchange game merely because they were playing the same game, while for women this was not enough to establish a connection. The idea that people form a connection in a short-term and anonymous game may be implausible, but such connections have been documented previously in the minimal group paradigm (Tajfel, 1970). Unfortunately, gender differences have not been investigated in that paradigm. With respect to the current analysis, this speculative explanation cannot be quantitatively assessed because of a lack of data. Other explanations face the same problem so the reason why men reciprocate more in the gift-exchange game remains unclear.
This puzzle is exacerbated by the preliminary results of the meta-analysis on gender differences in the trust game at the Free University. Their tentative results indicate that women, not men, reciprocate trust more. This means that there may be gender differences in the measures of reciprocity. One explanation for this may be that the trust game is often carried out using psychology students and the gift-exchange game is often carried out using economics students. Indeed, Van Lange, Schippers and Balliet (2011) found personality differences between psychology students and economy students (psychologists were found to be more prosocial while economists were found to be more individualistic). It could be that these or other personality traits interact with gender to explain the difference between the gift-exchange game outcomes and the trust game outcomes.
Alternatively, the design differences between the gift-exchange game and the trust game may be relevant. There are three design differences that are worth noting. 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 principal’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 principal is referred to as the firm or the employer, while the agent is referred to as the worker or the employee. This labor context could also have implications for the behavior of both players.
However, there are gift-exchange games that are framed neutrally and gift-exchange games in which rejection is not possible so these differences are not fundamental. The fundamental difference is that in the trust game, the added value of the exchange comes from the first transaction (i.e. the first mover or the principal) because the offer of the first mover is tripled by the experimenter before the money arrives at the second mover. In the gift-exchange game, in contrast, the added value can come about through the decision of the agent, the decision of both the principal and the agent, and a combination of their choices. This essential difference may have accounted for the gender differences in both games.
26
However, before investigating this issue more closely, it is necessary to wait for the final results of the trust game meta-analysis. But whatever the final results may be, comparing the gift-exchange game to the trust game will certainly be interesting. To be continued.
27
In-paper references
Abbink, K., Irlenbusch, B., & Renner, E. (2000). The moonlighting game: An experimental study on reciprocity and retribution. Journal of Economic Behavior & Organization, 42(2), 265-277. Armstrong, R. A. (2014). When to use the Bonferroni correction. Ophthalmic and Physiological
Optics, 34(5), 502-508.
Ashraf, N., Bohnet, I., & Piankov, N. (2006). Decomposing trust and trustworthiness. Experimental
Economics, 9(3), 193-208.
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., & Van Lange, P. A. (2013). Trust, conflict, and cooperation: A meta-analysis. Psychological
Bulletin, 139(5), 1090-1112.
Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity, and social history. Games and Economic
Behavior, 10(1), 122-142.
Beugelsdijk, S., De Groot, H. L., & Van Schaik, A. B. (2004). Trust and economic growth: A robustness analysis. Oxford Economic Papers, 56(1), 118-134.
Brülhart, M., & Usunier, J. C. (2012). Does the trust game measure trust? Economics Letters, 115(1), 20-23.
Buchan, N. R., Croson, T. A., & Solnick, S. (2008). Trust and gender: An examination of behavior and beliefs in the investment game. Journal of Economic Behavior & Organization, 68, 466-476. Bushman, B. J., & Wells, G. L. (2001). Narrative impressions of literature: The availability bias and the
corrective properties of meta-analytic approaches. Personality and Social Psychology
Bulletin, 27(9), 1123-1130.
Carpenter, J. (2008). Social preferences. In S. N. Durlauf & L. E. Blume (Eds.), The new Palgrave
dictionary of economics (2nd ed.).
Camerer, C. F., & Fehr, E. (2002). Measuring social norms and preferences using experimental games: A guide for social scientists (Working Paper No. 97). Retrieved from Caltech website:
http://authors.library.caltech.edu/27315.
Chaudhuri, A., & Gangadharan, L. (2003). Gender differences in trust and reciprocity (Working Paper 248). Retrieved from Research Space of University of Auckland website:
https://researchspace.auckland.ac.nz/handle/2292/136.
Cox, J. C., & Deck, C. A. (2006). When are women more generous than men? Economic Inquiry, 44(4), 587-598.
Croson, R., & Gneezy, U. (2009). Gender differences in preferences. Journal of Economic Literature,
28
DerSimonian, R., & Laird, N. (1986). Meta-analysis in clinical trials. Controlled Clinical Trials, 7(3), 177-188.
Dulin, A. M. (2007). A lesson on social role theory: An example of human behavior in the social environment theory. Advances in Social Work, 8(1), 104-112.
Eagly, A.H. (1987). Sex differences in social behavior: A social-role interpretation. Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Eagly, A. H., & Johnson, B. T. (1990). Gender and leadership style: A meta-analysis. Psychological
Bulletin, 108(2), 233-256.
Echabe, A. E., & Castro, J. L. G. (1999). The impact of context on gender social identities. European
Journal of Social Psychology, 29(2-3), 287-304.
Eckel, C. C., & Grossman, P. J. (2001). Chivalry and solidarity in ultimatum games. Economic
Inquiry, 39(2), 171-188.
Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in meta-analysis detected by a simple, graphical test. Bmj, 315(7109), 629-634.
Fehr, E., Kirchler, E., Weichbold, A., & Gächter, S. (1998). When social norms overpower competition: Gift exchange in experimental labor markets. Journal of Labor economics, 16(2), 324-351.
Feingold, A. (1994). Gender differences in personality: A meta-analysis. Psychological Bulletin, 116(3), 429-456.
Ferris, G. R., Liden, R. C., Munyon, T. P., Summers, J. K., Basik, K. J., & Buckley, M. R. (2009). Relationships at work: Toward a multidimensional conceptualization of dyadic work relationships. Journal of Management, 35(6), 1379-1403.
Friebel, G., Lalanne, M., Richter, B., Schwardmann, P., & Seabright, P. (2013). Selectivity and
opportunism: Two dimensions of gender differences in trust games and network formation. In B. Richter (Ed.), Essays in behavioral economics: Evidence on self-selection into jobs, social
networks, and leniency (Doctoral dissertation). Retrieved from
http://publikationen.ub.uni-frankfurt.de/frontdoor/index/index/docId/31704
Garbarino, E., & Slonim, R. (2009). The robustness of trust and reciprocity across a heterogeneous US population. Journal of Economic Behavior & Organization, 69(3), 226-240.
Glaeser, E. L., Laibson, D., Scheinkman, J. A., & Soutter, C. L. (1999). What is social capital? The determinants of trust and trustworthiness (Working Paper No. 7216). Retrieved from National Bureau of Economic Research website: http://www.nber.org/papers/w7216. Guiso, L., Sapienza, P., & Zingales, L. (2003). People's opium? Religion and economic
29
Heinz, M., Juranek, S., & Rau, H. A. (2012). Do women behave more reciprocally than men? Gender differences in real effort dictator games. Journal of Economic Behavior & Organization, 83(1), 105-110.
Herd, P., Carr, D., & Roan, C. (2014). Cohort profile: Wisconsing Longitudinal Study (WLS).
International Journal of Epidemiology, 43(1), 34-41.
Johnson, B. T., & Eagly, A. H. (2000). Quantitative synthesis of social psychological research. In H. T. Reis & C. M. Judd (Eds.), Handbook of Research Methods in Social and Personality Psychology (pp. 496-528). London: Cambridge University Press.
Johnson, N. D., Mislin, A. A. (2011). Trust games: A meta-analysis. Journal of Economic Psychology,
32, 865-889.
La Porta, R., Lopez-de-Silane, F., Shleifer, A., & Vishny, R. W. (1996). Trust in large
organizations (Working Paper No. 5864). Retrieved from National Bureau of Economic Research website: http://www.nber.org/papers/w5864.
Littell, J. H. (2008). Evidence-based or biased? The quality of published reviews of evidence-based practices. Children and Youth Services Review,30(11), 1299-1317.
Marín-Martínez, F., & Sánchez-Meca, J. (2010). Weighting by inverse variance or by sample size in random-effects meta-analysis. Educational and Psychological Measurement, 70(1), 56-73. Naef, M., & Schupp, J. (2009). Measuring trust: Experiments and surveys in contrast and combination
(IZA Discussion Paper No. 4087). Retrieved from EconStor website:
http://www.econstor.eu/dspace/handle/10419/35525.
Rau, H. A. (2011). Trust and trustworthiness: A survey of gender differences. In S. McGeown (Ed.),
Psychology of Gender Differences. New York: Nova Science Publishers.
Reeskens, T., & Hooghe, M. (2008). Cross-cultural measurement equivalence of generalized trust: Evidence from the European Social Survey (2002 and 2004). Social Indicators Research, 85(3), 515-532.
Rempel, J. K., Holmes, J. G., & Zanna, M. P. (1985). Trust in close relationships. Journal of Personality
and Social Psychology, 49(1), 95.
Rousseau, D. M., Sitkin, S. B., Burt, R. S., & Camerer, C. (1998). Not so different after all: A cross-discipline view of trust. Academy of Management Review, 23(3), 393-404.
Rotter, J. (1967). A new scale for the measurement of interpersonal trust. Journal of Personality,
35(4), 651-665.
Sapienza, P., Toldra-Simats, A., & Zingales, L. (2007). Understanding trust. The Economic Journal, 123(573), 1313-1332.
Scherer, S. (2012). Bias in Psychology: Bring in all significant results [Blog post]. Retrieved from
30
Schmitt, D. P., Realo, A., Voracek, M., & Allik, J. (2008). Why can't a man be more like a woman? Sex differences in Big Five personality traits across 55 cultures. Journal of Personality and Social
Psychology, 94(1), 168.
Solnick, S. J. (2001). Gender differences in the ultimatum game. Economic Inquiry, 39(2), 189. Stanley, T. D. (2005). Beyond publication bias. Journal of Economic Surveys, 19(3), 309-345.
Sterne, J. A., & Egger, M. (2001). Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. Journal of clinical epidemiology, 54(10), 1046-1055.
Sturgis, P., Read, S., & Allum, N. (2010). Does intelligence foster generalized trust? An empirical test using the UK birth cohort studies. Intelligence, 38(1), 45-54.
Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American, 223, 96-102. Taylor, S. E., Klein, L. C., Lewis, B. P., Gruenewald, T. L., Gurung, R. A., & Updegraff, J. A. (2000).
Biobehavioral responses to stress in females: Tend-and-befriend, not fight-or-flight. Psychological Review, 107(3), 411.
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive psychology, 5(2), 207-232.
Wang, F., & Yamagishi, T. (2005). Group-based trust and genfder differences in China. Asian Journal
of Social Psychology, 8(2), 199-210.
Weisberg, Y. J., DeYoung, C. G., & Hirsh, J. B. (2011). Gender differences in personality across the ten aspects of the Big Five. Frontiers in Psychology, 2, 178.
Williams, G. (2015). Men and friendship: An exploration of male perceptions of same-sex friendships. (Unpublished doctoral dissertation). University of New Orleans.
Yamagishi, T. & Yamagishi, M. (1994). Trust and commitment in the United States and Japan.
Motivation and Emotion, 18, 129-166.
Meta-analysis references
Aquino, P. Gazzale, R. & Jacobson, S. (2015). When do punishment institutions work?
(Working Paper No. 15). Retrieved from RePEc website:
http://econpapers.repec.org/paper/wilwileco/2015-15.htm
Benjamin, D. J., Choi, J. J., & Fisher, G. (2016). Religious identity and economic
behavior. Review of Economics and Statistics.
Bergstresser, K. D. (2009). Effects of in-group bias in a gift-exchange transaction: A theory of
employee ownership and evidence from a laboratory experiment. (Unpublished
doctoral dissertation). University of Maryland.
31
Charness, G., Cobo-Reyes, R., Jiménez, N., Lacomba, J. A., & Lagos, F. (2012). The hidden
advantage of delegation: Pareto improvements in a gift exchange game. The
American Economic Review, 102(5), 2358-2379.
Chaudhuri, A., Cruickshank, A., & Sbai, E. (2015). Gender differences in personnel
management: Some experimental evidence. Journal of Behavioral and Experimental
Economics, 58, 20-32.
Colasante, A., & Russo, A. (2014). Reciprocity in the labour market: Experimental
evidence (Working Paper No. 404). Retrieved from the RePEc website:
https://ideas.repec.org/p/anc/wpaper/404.html
.
Dariel, A., & Nikiforakis, N. (2014). Cooperators and reciprocators: A within-subject analysis
of pro-social behavior. Economics Letters, 122(2), 163-166.
Dariel, A., & Riedl, A. M. (2013). Reciprocal Preferences and the Unraveling of Gift-exchange.
(Working Paper). Maastricht University.
Eriksson, T., & Villeval, M. C. (2008). Performance-pay, sorting and social motivation. Journal
of Economic Behavior & Organization, 68(2), 412-421.
Franke, J., Gurtoviy, R., & Mertins, V. (2016). Workers’ participation in wage setting: A
gift-exchange experiment. Journal of Economic Psychology, 56, 151-162.
Fooken, J. (2013). The importance of individual characteristics in experimental economic
research. (Unpublished doctoral dissertation). Queensland University of Technology
Gose, K. (2013). Fair wages survive multiple sources of income inequality.Economics
Letters, 121(3), 473-477.
He, Y., Fooken, J., & Dulleck, U. (2015). Facing a changing labour force in China:
Determinants of trust and reciprocity in an experimental labour market. Economics
Bulletin, 35(3), 1525-1530.
Kocher, M. G., & Sutter, M. (2007). Individual versus group behavior and the role of the
decision making procedure in gift-exchange experiments. Empirica, 34(1), 63-88.
Lönnqvist, J. E., Verkasalo, M., Wichardt, P. C., & Walkowitz, G. (2013). Personal values and
prosocial behaviour in strategic interactions: Distinguishing value-expressive from
value-ambivalent behaviours. European Journal of Social Psychology, 43(6), 554-569.
Maximiano, S., Sloof, R., & Sonnemans, J. (2007). Gift exchange in a multi-worker firm. The
32
Owens, M. F. (2011). Do other-regarding preferences change with age? Evidence from a gift
exchange experiment. The Journal of Socio-Economics,40(6), 868-878.
Owens, M. F. (2012). Responsibility-alleviation and other-regarding preferences with peer
workers in labor markets: an experimental investigation. Journal of Labor
Research, 33(3), 353-369.
Owens, M. F., & Kagel, J. H. (2010). Minimum wage restrictions and employee effort in
incomplete labor markets: An experimental investigation. Journal of Economic
Behavior & Organization, 73(3), 317-326.
Petit, E. (2009). Does indignation lead to generosity? An experimental investigation (Working
Paper No. 10). University of Bordeaux.
33
Appendix A – Summed payoff tables for Fooken (2013) and Owens (2011)
Summed payoff table for Fooken (2013)
e=1 e=2 e=3 e=4 e=5 e=6 e=7 e=8 e=9 e=10
w=5 110 126 142 158 174 190 206 222 238 254 w=10 110 126 142 158 174 190 206 222 238 254 w=15 110 126 142 158 174 190 206 222 238 254 w=20 110 126 142 158 174 190 206 222 238 254 w=25 110 126 142 158 174 190 206 222 238 254 w=30 110 126 142 158 174 190 206 222 238 254 w=35 110 126 142 158 174 190 206 222 238 254 w=40 110 126 142 158 174 190 206 222 238 254 w=45 110 126 142 158 174 190 206 222 238 254 w=50 110 126 142 158 174 190 206 222 238 254 w=55 110 126 142 158 174 190 206 222 238 254 w=60 110 126 142 158 174 190 206 222 238 254 w=65 110 126 142 158 174 190 206 222 238 254 w=70 110 126 142 158 174 190 206 222 238 254 w=75 110 126 142 158 174 190 206 222 238 254 w=80 110 126 142 158 174 190 206 222 238 254 w=85 110 126 142 158 174 190 206 222 238 254 w=90 110 126 142 158 174 190 206 222 238 254 w=95 110 126 142 158 174 190 206 222 238 254 w=100 110 126 142 158 174 190 206 222 238 254
Summed payoff table for Owens (2011)
e=1 e=2 e=3 e=4 e=5 e=6 e=7 e=8 e=9 e=10
w=1 28 32 36 40 44 48 52 56 60 64 w=2 32 36 40 44 48 52 56 60 64 68 w=3 36 40 44 48 52 56 60 64 68 72 w=4 40 44 48 52 56 60 64 68 72 76 w=5 44 48 52 56 60 64 68 72 76 80 w=6 48 52 56 60 64 68 72 76 80 84 w=7 52 56 60 64 68 72 76 80 84 88 w=8 56 60 64 68 72 76 80 84 88 92 w=9 60 64 68 72 76 80 84 88 92 96 w=10 64 68 72 76 80 84 88 92 96 100
