Can gender diversity prevent risky choice shifts? The effect of gender composition on group decisions under risk

Lima de Miranda, Katharina; Detlefsen, Lena; Schmidt, Ulrich Article — Published Version Can gender diversity prevent risky choice shifts? The effect of gender composition on group decisions under risk Experimental Economics Provided in Cooperation with: Kiel Institute for the World Economy – Leibniz Center for Research on Global Economic Challenges Suggested Citation: Lima de Miranda, Katharina; Detlefsen, Lena; Schmidt, Ulrich (2025) : Can gender diversity prevent risky choice shifts? The effect of gender composition on group decisions under risk, Experimental Economics, ISSN 1573-6938, Cambridge University Press, Cambridge, Iss. FirstView, pp. 1-22, https://doi.org/10.1017/eec.2024.4 This Version is available at: https://hdl.handle.net/10419/330837 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0 Experimental Economics (2025), 1–22 doi:10.1017/eec.2024.4 ORIGINAL PAPER Can gender diversity prevent risky choice shifts? The effect of gender composition on group decisions under risk Katharina Lima de Miranda1, Lena Detlefsen1and Ulrich Schmidt1,2,3 1Kiel Institute for the World Economy, University of Kiel, Kiellinie, Kiel, Germany 2Department of Economics, University of Kiel, Kiel, Germany 3Deparment of Economics and Econometrics, University of Johannesburg, Johannesburg, South Africa Corresponding author: Lena Detlefsen; Email: [email protected] (Received 2 October 2024; accepted 2 October 2024) Abstract Our study contributes to the literature on choice shifts in group decision-making by analyzing how the level of risk-taking within a group is influenced by its gender composition. In particular, we investigate experimentally whether group composition affects how preferences ‘shift’ when comparing individual and group choices. Consistent with hypotheses derived from previous literature, we show that male-dominated groups shift toward riskier decisions in a way that is not explained by any simple preference aggregation mechanism. We discuss potential channels for the observed pattern of choice shifts. Keywords: Experiment; gender; group decisions; risk-taking; risky shift JEL Codes: D71; D81; D91; J16 1. Introduction Many important economic and politically relevant decisions are made by groups, such as boards of directors, supervisory boards, parliaments, and other political bodies, as well as smaller-scale working groups and teams. Increasing gender diversity in these decision-making bodies, which are often male-dominated, has been a key political goal. To achieve this, measures such as gender quotas and mentorship programs (Flory et al., 2021; Kofoed, 2019) have been implemented with varying success. The focus on gender diversity has spurred a growing body of research examining its impact on decision-making. Research in policymaking yields mixed results, ranging from no impact of quotas (Bagues & Campa, 2021; Ferreira & Gyourko, 2014) to improved qualifications of politicians (Baltrunaite et al., 2014) and the adoption of more egalitarian policies (Ranehill & Weber, 2022). Studies of corporate boards similarly provide mixed evidence regarding the impact of quotas on firm performance and valuation (Ahern & Dittmar, 2012; Eckbo et al., 2016; Johansen & Sandnes, 2008; Matsa & Miller, 2013; Nygaard, 2011). However, research on gender-diverse management teams finds some evidence for better performance of more diverse teams (Adams & Ragunathan, 2015; Apesteguia et al., 2012; Bansak et al., 2011; Cueva & Rustichini, 2015; Hoogendoorn et al., 2013). This paper focuses on risk-taking, a crucial aspect of many economic and political decisions. Risk-taking is, for instance, important for firm performance (e.g., Gilley et al., 2002; Walls & Dyer, 1996), and there are compelling reasons to believe that gender composition significantly affects © The Author(s), 2025. Published by Cambridge University Press on behalf of Economic Science Association. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 2 Katharina Lima de Miranda et al. group decisions under risk. Literature indicates that women and men differ in economic preferences (see Croson & Gneezy, 2009, for an overview). There exists abundant evidence that women are more risk-averse than men in financial risk-taking and many other domains (Byrnes et al., 1999; Croson & Gneezy, 2009), though the extent and economic relevance of this gender difference remain debated (Filippin & Crosetto, 2016). However, group decisions do not always reflect the individual preferences of group members. A crucial yet underexplored question is how men and women change their decisions in a group setting and how gender diversity influences these shifts. This paper investigates how the level of risktaking within a group is influenced by its gender composition and how this composition affects the shift from individual to group decisions. Our study contributes to the literature on choice shifts in group decision-making by examining how choice shifts are affected by gender composition. Understanding the impact of gender composition on group decision-making under risk has important implications for policymakers and organizations aiming to increase gender diversity and improve decision-making outcomes. The difference between individual and group risk-taking has been extensively debated in social psychology, dating back to Stoner’s (1961) seminal work, and has recently gained attention in the economics literature. A common pattern is the polarization of group decisions, where groups, on average, tend to take more risks than individuals, a phenomenon known as the “risky shift.” Early evidence on risky shifts was based on studies employing choice dilemma questionnaires. Recently, several studies analyzed group decisions under risk with monetarily incentivized experiments, producing mixed results. Some studies find evidence of risky shifts (Nieboer, 2015; Sutter, 2009), while others report cautious shifts, where groups take less risk than individuals (Baker et al., 2008; Masclet et al., 2009; Shupp & Williams, 2008), or observe no systematic differences (Harrison et al., 2013). Several reasons for these choice shifts have been proposed, including the conformity hypothesis (Cialdini & Goldstein, 2004; Jagau & Offerman, 2018), the diffusion of responsibility (Eliaz et al., 2006; Wallach et al., 1962,1964), and the risk-as-value hypothesis (Bauer & Turner, 1974; Vidmar, 1970).1 Evidence that gender plays a significant role in group decisions under risk is provided by Nieboer (2015) and Bogan et al. (2013). Using an investment task and an investment portfolio management decision task, respectively, they find that gender is the only individual characteristic that significantly affects risk-taking in group settings. Moreover, they observe that the presence of men increases risktaking, although the effect is not linear in the case of Bogan et al. (2013). We add to this literature by using a simple lottery choice experiment and a one-shot decision task, avoiding potential priming effects related to investment tasks that can impact risk attitudes (Eckel & Grossman, 2008), as well as learning effects or changes in group decision-making over multiple rounds. In contrast to previous studies that rely on between-subject comparisons, we use a within-subject design to directly compare individual and group decision-making under risk. Despite the importance of understanding choice shifts in group decision-making, direct evidence of gender’s role remains limited. Our study directly addresses this gap. Daly and Wilson (2001) offer indirect evidence by comparing individual risky decisions made in private with those made in public, where subjects had to announce their individual choice in front of a group of peers. They found that men took significantly more risk in public than in private, whereas women showed no such difference. This indirect evidence suggests that male group members may take more risks and appear more risk tolerant in group settings, thereby enhancing the risky shift with a higher proportion of males in the group. Further indirect evidence is provided by Ertac and Gurdal (2012), who find that men are generally more willing to lead groups. Those men willing to lead also take more risks on the group’s behalf than those not willing to lead. For women, they find no differences. This evidence is underpinned by 1Further details on these theories are provided in Section 3. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 3 literature focusing on leadership. Recent studies confirm that women are less willing to lead, particularly when the leader’s gender is revealed to group members, with female leaders being most willing to lead in all-female groups (e.g., Grossman et al., 2015; Li et al., 2020). This result also holds in a simple risky environment where decisions involve how much of a fixed endowment to invest in a risky asset. When placed in groups of five, women are more likely to refuse to decide on their group’s behalf (Ertac & Gurdal, 2012). Recent discussions suggest that these differences arise because, on average, men are more selfconfident, more likely to run for elections, have more influence, and are ranked higher than equally performing women (e.g., Born et al., 2022; Chen & Houser, 2019; Kanthak & Woon, 2015). In this paper, we analyze the impact of gender composition on group decisions under risk and choice shifts by using a simple lottery choice experiment with monetary incentives. Groups of three with varying gender compositions were formed, and their group choices were compared to the individual choices of the group members. We varied whether the group or the individual decision was made first, and all group decisions were made face-to-face. Our results reveal several important findings. First, our results show a clear and significant impact of gender composition on group decision-making. Female-dominated groups take significantly less risk than male-dominated groups. Second, and crucially, we find that gender composition plays an important role in choice shifts. While female-dominated groups show no significant shifts, maledominated groups exhibit substantial risky shifts, taking on average more risk than the average and median preferences of group members would imply. Finally, we discuss potential channels for these observed choice shifts, highlighting the influence of male presence in driving the shift toward riskier decisions. The paper is structured as follows: Section 2 outlines the experimental design, Section 3 formalizes our gender-specific hypothesis more precisely, and Section 4 presents the results. The conclusion follows in Section 5. 2. Experimental design 2.1. Participants and procedure The experiment, involving 492 participants, was conducted in the student canteen of the University of Kiel, Germany.2The gender ratio was balanced (51.53% women), and the average age was 23.24 (SD 4.95). Participants were recruited among the canteen’s customers to take part in an economic experiment, which was conducted in designated quiet areas of the canteen. Potential participants were told they would receive a €2 participation fee and could gain additional money by playing a lottery. To discreetly control the group’s gender composition, recruiters used a systematic approach. Participants were approached individually and invited to participate without revealing the genderbased aspect of the grouping. As individuals agreed to participate, their gender was recorded, and recruiters tracked the required gender balance for each group composition (FFF, FFM, FMM, MMM).3To ensure that group members were unfamiliar with each other, participants were asked if they knew anyone present before groups were formed. The recruiters then formed groups of three people, ensuring a mix of gender compositions according to the experimental design. This process allowed the researchers to control the gender composition of each group discreetly without making it apparent to the participants that gender was 2One subject in treatment GF did not return after the individual phase, so most of the analysis is done for 491 subjects only. 3In the post-experiment survey in treatment IF, we asked participants about the criteria they thought were used to determine the groups after the treatment. The majority, ca. 75%, answered that it was by chance. Only 6.5% thought of demographic characteristics. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 4 Katharina Lima de Miranda et al. Table 1 Gender composition of groups Nb of groups Nb of participants Gender composition Treatment: GF Treatment: IF Total Women Men Total FFF 22 19 41 123 0 123 MMM 21 16 37 0 110 110 FMM 21 21 42 42 84 126 FFM 21 23 44 88 44 132 Overall 85 79 164 253 238 491 Notes: The left side of the table displays the number of participating groups in the two treatments sorted by their gender composition. The right side of the table shows the number of women and men who participated in the experiment, sorted by the gender composition of their group. a factor in their grouping.4All other group characteristics were randomly allocated.5The sample size was determined based on previous studies using the incentivized Eckel and Grossman (2002) task. A meta-analysis by Filippin and Crosetto (2016) reports that the average gender effect size for this task is equal to Cohen’s d =.55 on the individual level.6Overall, data from 41 purely female groups (further mentioned as FFF), 44 groups with two women and one man (FFM), 42 groups with 1 woman and 2 men (FMM), and 37 purely male groups (MMM) were collected; an overview is given in Table 1. A critical feature of experimental designs that analyze group polarization is the order in which individual and group choices are elicited. Most studies elicited individual preferences first, although some have reversed the order, each approach has its distinct advantages and drawbacks. When participants make multiple evaluations, earlier decisions can serve as an anchor, biasing subsequent choices (Ariely et al., 2003). As a result, anchoring tends to favor eliciting individual preferences first. Recent research, however, suggests that individual decisions are also heavily influenced by the social context, particularly gender composition. For instance, Castillo et al. (2015) let subjects make individual risktaking decisions while sitting in a room with other people. Although the decisions are private and not revealed to other subjects, the gender composition in the room systematically impacts individual preferences. Consequently, in the case of group decisions, the gender composition of the group might also influence individual risk preferences. However, this effect is at least reduced if subjects anchor on the individual decision they made before in private, that is, without any influence of the social context. Given the significant role of social context in decision-making, we implemented two experimental treatments to vary the order of the choices: one with the group decision first (GF) and the other with the individual decision first (IF). Consistent with previous literature on choice shifts, participants in the IF treatment were unaware of their group’s composition when making individual choices. We tested whether the order of decisions systematically affected the group and individual preferences and found no significant impacts on either group or individual decisions (see Section 4). This supports previous work by Harrison et al. (2013), which found that the order of decisions does not influence individual risk preferences. For both treatments, participants were approached as they were about to leave the canteen, minimizing the chance they could share information with others who had yet to participate. While we 4In the IF treatment, we asked participants after the treatment about the criteria they thought were used to determine the groups. The majority, ca. 75%, answered that it was by chance – only 6.5% thought of demographic characteristics. 5Standardized differences of covariates between treatments, gender domination in the group, and gender, as well as means and standard deviations of covariates by group types, can be found in Tables A1 and A2 in Appendix A. 6Assuming that the median voter is more likely to be female in female-dominated and male in male-dominated groups, we collected data from 85 female-dominated and 79 male-dominated groups to detect a medium-sized gender effect in group choices (with β = 0.90, α = 0.05). https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 5 cannot entirely exclude this possibility of information exchange, the risk is inherent in consecutive sessions of lab experiments as well. Conducting the experiment in a natural canteen setting, rather than in a lab, allowed for a more realistic group context. Although the risk of information transmission was potentially greater in this environment, we think it did not bias our results. First, subjects were unaware that gender composition was central to the study and could, therefore, not transmit this information. Second, even if some information about the experiment was shared, this would have no systematic impact on the differences we observe related to group composition. In the GF treatment, groups were formed at the beginning of the experiment, and it was explained to the participants that they had to make a risky decision as a group first, then had to fill out a questionnaire on their own, and in the end, had to reunite in their initial group to receive their payment. After the groups were formed, a card displaying six lottery options was provided to them (see Fig. 1). Participants were told that the group had to choose exactly one of these lotteries by consensus. There were no time constraints for discussion and reaching a consensus and no fallback option (no group took longer than five minutes to reach a consensus). When participants within a group agreed on a lottery, they stated their choice to the experimenter and were handed the questionnaires, which had to be filled out in private. The questionnaire included an individual risk preference task (which equaled the group task with all amounts divided by three), basic demographic questions (gender, age, highest educational degree7), happiness (self-reported happiness on a five-point Likert scale) and their level of satisfaction with the group choice on a five-point Likert scale. After completing the questionnaires, participants reunited in their groups for the payoff. In the IF treatment, participants first had to decide on the individual risk preferences task in private, without being aware of the group composition they were allocated to in the sequel. We had four quiet places (each for one subject only) for these private decisions in different corners of the canteen, so subjects could not see that other subjects were also making a private decision simultaneously. After the group decided, subjects had to fill out the individual questionnaire again in private and finally reunited for the payoff. The payoff mechanism was the same in both treatments and took place at the very end of the experiment. To determine the payoff, a coin was flipped twice. The first coin flip indicated whether the group or the individual lottery choice would be relevant for payment. The second coin flip determined the outcome – high or low payoff – according to the group or individual choice. 2.2. Methods Risk preferences. A well-established task developed by Eckel and Grossman (2002) was used to elicit risk preferences. The task is distributed along a one-dimensional spectrum. The groups had to choose exactly one out of six lotteries depicted in Fig. 1 by consensus. The lotteries were represented with coins that had two colored sides indicating the size of a gain (in Euro) – orange (high gain) and pink (low gain). For all six lotteries, the chances to win the high or low gain were equal (50% probability). The lotteries increased in risk and expected value starting from lottery 1 with a sure gain of €12 (or €4 for each group member) to lottery 5 with an expected value of €15 (€8 or €2 for each group member). Lottery 6 had the same expected value as lottery 5 but a higher risk (€9 or €1 for each group member) and allowed the detection of risk-loving attitudes. The lottery number chosen by the group will be referred to as group choice (GC) in the sequel. Generally, the higher the GC, the lower the given group’s degree of risk aversion. To control for individual risk preferences, the questionnaire included the same lottery task but with individual gains (i.e., group amount divided by three). Responses to this task will be termed individual choice (IC). Again, a higher number of IC indicates lower risk aversion. The group shift is given by GC – IC where GC – IC >(<0) indicates a risky (cautious) shift. 7In the analysis, degrees were coded on a scale from 0 to 4, reflecting 0 – no degree, 1 – high school degree, 2 – bachelor, 3 – master’s, 4 – PhD. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 6 Katharina Lima de Miranda et al. Fig. 1 Lotteries for group decision-making Notes: The figure displays the gains of each of the six lotteries of the experiment, with the orange side showing the high gain amount, and the pink side showing the low gain amount. 3. Hypotheses There is abundant evidence that women are more risk-averse than men in financial risk-taking (Charness & Gneezy, 2012; Croson & Gneezy, 2009), although the effects are sometimes small and task-specific (Filippin & Crosetto, 2016). For the task employed in the present paper, gender differences were consistently observed, such that we expect to see them also reflected in individual choices. Hypothesis 1: IC is higher for men than for women. In the group setting, we did not provide specific instructions to subjects on how to reach a group decision, leading to two potential scenarios for how decisions might be made. The first scenario assumes majority voting as the decision process. From the political economy literature, it is well-established that under certain conditions, in particular a one-dimensional spectrum and single-peaked preferences, the outcome of majority voting is determined by the median voter (Black, 1948). This has also been supported by experimental evidence (Ambrus et al., 2015). Assuming that women are generally more risk-averse than men, the median voter in femaledominated groups (FFF and FFM) is more likely to be female, leading to more cautious group https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 7 decisions. Conversely, in male-dominated groups (FMM and MMM), the median voter is likely to be male, resulting in riskier group decisions. The second scenario considers the possibility of a leading group member influencing the group’s decision. Literature suggests that women are less willing to take on leadership roles, especially when the leader’s gender is revealed to group members (Grossmann et al., 2015; Li et al., 2020). Women are most willing to lead in all-female or female-majority groups (Born et al., 2022). Consequently, the leader is more likely to be male, as the number of male group members increases, which should, according to Hypothesis 1, result in higher risk-taking of the group also in this scenario. Therefore, based on both scenarios, we hypothesize that: Hypothesis 2: GC increases with the number of male group members. The primary focus of this paper is to compare individual and group choices. As mentioned, group decisions do not always reflect the individual preferences of group members, and several explanations exist for that behavior. The conformity hypothesis states that individuals tend to align their preferences with those of the majority (Cialdini & Goldstein, 2004; Jagau & Offerman, 2018). This conformity can arise from two primary goals: (1) an accuracy goal, where individuals adjust their choices based on the belief that the majority is more likely to be correct, and (2) an affiliation goal, where individuals conform to the group to gain social approval and avoid negative judgment (Asch, 1956; Brown, 1965; Nordhøy, 1962; Stoner, 1968). In a group setting, this could lead to shifts toward either more cautious or riskier decisions depending on the majority’s preferences. Based on Hypotheses 1 and 2, we predict that in male-dominated groups, these majority preferences would align with those of the male group members, leading to a risky shift. Conversely, in female-dominated groups, the majority will reflect the female members’ preferences, resulting in a cautious shift. Another explanation for choice shifts is the diffusion of responsibility in the group context. When individuals make risky decisions, they bear the full responsibility for the outcome. However, in the group context, the perception of responsibility is diffused, and individuals may feel less accountable for the final decision.(Wallach et al., 1962,1964). The reduced sense of responsibility can make individuals more willing to take risks in the group setting, leading to riskier group decisions and explaining the phenomenon of a risky shift. While the diffusion of responsibility has been initially used to explain risky shifts, Eliaz et al. (2006) showed that the direction of choice shifts depends on the prevailing social norm within the group. If the social norm is cautious, the group is more likely to shift toward safer decisions; if the norm favors risk-taking, the group is more likely to shift to riskier choices. We hypothesize that the social norm in male-dominated groups is riskier than in female-dominated groups. Therefore, we expect a risky shift in male-dominated groups. A related theory is the risk-as-value hypothesis, which suggests that (moderate) risk-taking is a socially approved trait (Bauer & Turner, 1974; Vidmar, 1970). According to this hypothesis, individuals who perceive themselves as more risk-averse than others in the group may adjust their preferences to align more closely with the perceived cultural norm, which could lead to a general risky shift. However, previous studies with designs similar to ours have not observed a universal risky shift. Based on the literature that suggests that risk-taking is a cultural value more strongly associated with men (e.g., Daly & Wilson, 2001), we hypothesize that men particularly change their individual preferences toward higher risk-taking in group contexts, leading to a risky shift in group decision. Combining these theories, we propose the following hypothesis: Hypothesis 3: GC – IC is higher in malethan in female-dominated groups https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 8 Katharina Lima de Miranda et al. Fig. 2 Average group choices by group types Notes: The left figure displays the average group choice of female vs. male-dominated groups, whereas the right figure displays the average group choice with respect to the specific gender composition of the groups. The gray whiskers indicate the 95% confidence interval. 4. Results 4.1. Individual choice The average individual choice IC is 3.48 (SD =1.79). In line with Hypothesis 1, we find that men take more risk (M =3.97, SD =1.74) than women (3.02, SD =1.72). This difference is statistically significant (Wilcoxon rank sum test: z =5.884, p <.001).8There are no significant differences between the two treatments (GF and IF) in terms of IC (which holds for all subjects, for women only, and for men only9). 4.2. Group choice The average group choice is 3.58 (SD =1.73). In the first step, we compare the average group decisions between male-dominated and female-dominated groups and between group types, as shown in Fig. 2.10 8Table A3 in the Appendix states the number of observations per lottery and the frequencies. The same is given for male- /female-dominated groups and all group types. 9We do not find any treatment differences in terms of the average IC using a Wilcoxon rank sum test for all participants: z=1.087, Prob >|z| =.277; for women only: z =.580, Prob >|z| =.562; and for men only z =.851, Prob >|z| =.395. In addition, we do not find any significant treatment differences within the four group types for all participants, for women only and men only, except the FMM group testing for men only (here, men in the GF treatment show a higher average IC than in the IF treatment). Finally, within the respective treatments GF and IF, we do not observe significant differences between women’s IC across groups; the same holds for men. 10Again, we find no significant differences between the two treatments regarding GC. No treatment differences are found for all groups and group types separately. Wilcoxon rank sum test: All groups: z =−.109, Prob >|z| =.913; FFF: z =−.495, Prob >|z| =.621; FFM: z =−.418, Prob >|z| =.676; FMM: z =.402, Prob >|z| =.688; MMM: z =.284, Prob >|z| =.776. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 15 Fig. 4 Potential consensus-making mechanisms Notes: Differences between observed group choice and hypothetical group choice according to different consensus-making mechanisms by female and male-dominated groups. A positive number shows a risky, while a negative shows a cautious shift relative to the respective consensus-making mechanism. 4.4.1. Choice shifts relative to potential consensus-making mechanisms Next, we investigate choice shifts relative to the potential consensus-making mechanisms described above. We examine the differences between observed group choices and hypothetical group choices constructed for each group according to different consensus-making mechanisms, that is, GCg−GCc g, where c={Mean, Median, Min IC, Max IC, Low Majority, High Majority, Minimum Range Coalition, Leader}. Positive values indicate a risky shift relative to the potential consensus-making mechanism, while negative values indicate a cautious shift. Figure 4 shows that, for all mechanisms except the Leader scheme, choice shifts are larger for male-dominated groups compared to female-dominated groups. Specifically, male-dominated groups exhibit a larger risky shift under the Mean, Median, Min IC, and Low Majority schemes and a smaller cautious shift under the Max IC and High Majority schemes. In the Leader scheme, we find the opposite – a smaller cautious shift for female-dominated groups. Potential explanations for this result are discussed in the following subsection. On average, the median IC seems to align most closely with the observed group choices. Nevertheless, a risky shift for male-dominated groups can be observed. Table 6 shows that, for all consensus-making mechanisms except the Leader scheme, maledominated groups consistently demonstrate a significant additional risky shift compared to femaledominated groups. In the case of Max IC and High Majority Coalition, male-dominated groups exhibit smaller cautious shifts. In the Leader scheme, the results are qualitatively and quantitatively similar, but the sample size (N =39) is too small to detect a statistically significant effect. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 16 Katharina Lima de Miranda et al. Table 6 OLS regressions with the difference between actual and hypothetical group choice as dependent variable (group level) (1) (2) (3) (4) (5) (6) (7) (8) GC – avg. IC GC – median IC GC-min IC GC-max IC GC-low. maj. GC-high maj. GC-coalition GC-leader Male-dominated groups .562** .616*** .919*** .802*** .661*** .571** .676*** .658 (.22) (.22) (.23) (.25) (.22) (.23) (.23) (.62) Mechanism −.175*−.374*** −.427*** −.545*** −.248*** −.309*** −.362*** −.707*** (.09) (.07) (.08) (.09) (.08) (.09) (.08) (.17) Constant 2.385** 2.85*** 3.893*** 3.647*** 2.823*** 2.592** 2.963*** 3.042 (1.06) (1.06) (1.11) (1.21) (1.04) (1.11) (1.1) (2.74) No. of obs. 164 164 164 164 164 164 164 39 Adj. R2 .04 .141 .182 .204 .078 .077 .106 .274 Prob >chi2/F .052 0 0 0 .004 .005 .001 .007 Notes: Regression at the group level, standard errors are given in parenthesis, significance level: *p <.1, **p <.05, ***p <.1. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 17 Fig. 5 Simulated group choice according to potential consensus-making mechanisms Notes: The figure shows hypothetical group choices based on simulated groups based on the average ICs of men and women. The green bars represent the actual average group choices observed for female and male-dominated groups. The comparison highlights that male-dominated groups tend to have higher group choices than those predicted by any consensus-making mechanisms, whereas femaledominated groups tend to align with or fall below these predictions. To further rule out a purely mechanical effect, we simulate hypothetical groups based on the average ICs of men and women. Figure 5 illustrates this comparison, showing that the observed average GC in male-dominated groups exceeds all consensus-making mechanisms considered. In contrast, for female-dominated groups, the observed GC is equal to or below these mechanisms. This confirms that the effect extends beyond a purely mechanical effect. 4.4.2. Leader In treatment IF (79 groups in total), we included additional questions to identify whether group decisions were driven by a single individual. Table 7 shows that group leader were identified in many groups, with their occurrence evenly distributed across group types. In FMM, most leaders are male (nine men vs. two women), whereas leadership is evenly split between men and women in FFM (five men vs. five women). An important question is whether leaders exhibit different risk characteristics than non-leaders and whether these characteristics differ between male and female-dominated groups. Table 8 shows that leaders are generally more risk-taking than non-leaders. In male-dominated groups, this difference is marginally significant for the experimentally elicited risk preference (IC, row 1) and highly significant for subjective risk tolerance (row 2). Leaders in male-dominated groups also perceive themselves as more risk tolerant relative to others, that is, their subjective risk tolerance relative to others is, on average, higher than for non-leaders in these groups (row 3). For female-dominated https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 18 Katharina Lima de Miranda et al. Table 7 Distribution of leaders across group types Gender composition FFF MMM FMM FFM No leader 9 8 10 13 40 Leader 10 8 11(9/2) 10(5/5) 39 Notes: The table shows the number of groups, with and without leaders, sorted by group gender composition. Table 8 Comparison of individual risk preferences and subjective risk tolerance between leaders and non-leaders Male-dominated groups Diff. Female-dominated groups Diff. Non-leader Leader Non-leader Leader Individual choice (IC) 3.674 4.526 −.852*2.971 3.238 −.267 (1.888) (1.577) (.464) (1.701) (1.998) (.419) Subjective risk tolerance 3.185 2.526 .658*** 2.99 3.238 −.248 (.983) (.697) (.237) (.876) (.944) (.212) Subjective risk tolerance relative to others 3.154 2.737 .417*3.127 3.143 −.015 (.965) (.933) (.242) (.961) (.854) (.226) Recommendation risk-taking 1.848 1.789 .058 1.683 2.048 −.365*** (.592) (.535) (.147) (.544) (.384) (.125) Notes: Regressions at the group level, standard errors in parenthesis, significance level: *p <.1, **p <.5, ***p <.1. IC represents the individual choices in the incentivized risk experiment. Subjective risk tolerance, subjective risk tolerance relative to others and recommendation risk-taking are hypothetical self-assessments. Higher numbers indicated higher risk-taking (the original questions in the questionnaire are recoded to allow for an easier interpretation of results). groups, we do not find statistically significant differences in risk preferences or subjective risk tolerance between leaders and non-leaders.19 However, leaders in female-dominated groups believe that others should generally take less risk (row 4). 5. Conclusion The growing political focus on gender diversity has spurred studies examining its economic impact and the influence of an increasing proportion of women in decision-making bodies on decisionmaking. In this paper, we examine how gender diversity affects group decision-making, focusing on risk-taking, a key aspect of many economic and political decisions. Using a simple lottery choice experiment with monetary incentives, we find that risk-taking increases as the share of male group members rises, resulting in greater risk-taking in maledominated groups compared to female-dominated groups. The main result of this paper is genderspecific polarization in group decision-making. A clear choice shift pattern emerges when we account for the group composition: male-dominated groups exhibit risky shifts, taking more risks than the average and median individual preferences of group members would suggest. In contrast, female-dominated groups show no significant shift in risk preferences. To better understand the mechanisms behind these choice shifts, we examined eight potential consensus-making processes. This approach allowed us to rule out the possibility that the observed choice shift is due to a mechanical effect resulting from the underlying consensus-making processes. Our findings provide three key insights. First, while the median group choice is closest to the actual group choice, male-dominated groups display an additional risky shift compared to female-dominated groups. 19These results also hold if we only look at female group leaders. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 19 Second, the results for the majority mechanism align with theories such as the conformity and responsibility hypotheses, indicating that group members conform to the majority preferences. However, the majority preferences differ in male and female-dominated groups, leading to different choice shifts. Specifically, our findings support the assumption that men in male-dominated groups tend to adjust their choices to higher levels of risk when deciding in the group context. Third, leaders in male and female-dominated groups differ in their risk preferences and subjective risk tolerance. Leaders in male-dominated groups are more risk-taking than non-leaders, while we find no such difference in female-dominated groups. This result is consistent with the findings of Ertac and Gurdal (2012), who show that men are generally more willing to lead groups and that those who choose to lead tend to take more risks on behalf of the group than those who are not. For women, they find no differences. Additionally, we find that leaders in female-dominated groups generally recommend others to take less risk. This result aligns with our observation of a general trend of cautiousness in female-dominated groups. Furthermore, it supports the findings of Eckel and Grossman (2002), who show that women are stereotyped as more risk-averse than men and that beliefs about gender differences are greater than the actual gender differences in risk aversion. While our evidence is rather clear-cut, it may be sensitive to the elicitation method. Filippin and Crosetto (2016) showed that gender differences in risk-taking are particularly pronounced for elicitation methods involving a safe option, as in our study. Without such an option, gender differences are minimized, which could also reduce the effect of group composition on decisions. Additionally, the direction of group shifts seems to depend on the elicitation method. While we do not observe a general shift in our data, other studies with monetary incentives did. Pairwise choices, elicitation of willingness-to-pay, and the Holt-Laury method have generated cautious shifts in previous studies (Baker et al., 2008; Masclet et al., 2009; Pahlke et al., 2012; Shupp & Williams, 2008). In contrast, the investment game of Gneezy and Potters (1997) has generated risky shifts (Nieboer, 2015; Sutter, 2009). It remains an open question for future research to determine the impact of gender composition under these alternative elicitation methods. The behavior of gender-balanced groups, which was not considered in our study, also warrants further investigation. Our findings are important in several ways. First, we contribute to the literature on group decisionmaking by demonstrating that the gender composition of a group impacts decision-making under risk. Second, we contribute to the literature on choice shifts by showing that gender is a crucial determinant. The differences between groups of varying gender composition might help explain the mixed evidence on choice shifts in the literature. Finally, our results have important implications for policymakers. The systematic impact of gender composition on group decision-making suggests that political and firm-level actions to increase gender diversity can be expected to change decision-making under risk in the group context. In our experimental design with groups of three participants, a larger share of women reduced risk-taking in groups, but a majority of females was needed to significantly reduce the risk-taking of groups and prevent risky shifts. Choice shifts can be problematic in several ways. In particular, risky shifts can be expected to have undesired effects and potentially harm society. Risky shifts may be regarded as excessive risk-taking since the risk taken by the group exceeds that of the individual group members, creating a potentially undesired bias toward higher risk-taking. Consequently, this could induce decision-making bodies to take excessive financial risk, contribute to financial crises, or lead to risky, unethical decisions, resulting in a moral decline within companies or institutions (Armstrong et al., 2004). However, in certain contexts, higher risk-taking is desirable and can lead to better outcomes, such as greater creativity (Glover & Sautter, 1977) and innovation (Reynolds et al., 2009). In general, group polarization in decision-making may not be desirable in a multitude of situations as it can give rise to a detachment of a group from its’ context (den Nieuwenboer & Kaptein, 2008), and both excessive risk-taking and risk avoidance may be harmful to an organization or the society (Harjoto & Laksmana, 2018). https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 20 Katharina Lima de Miranda et al. Supplementary material. The supplementary material for this article can be found at https://doi.org/10.1017/eec.2024.4. Replication packages. A replication package for all results presented in this paper, including data and Stata code can be accessed here: https://osf.io/2wp4z/?view_only=fd895c321c564b169f4374338d34b774. Acknowledgements. We would like to thank participants at ESA in Berlin, the 12th Nordic Conference in Gothenburg, the Annual Meeting of the German Economic Association, the 20th Spring Meeting of Young Economists, and the International Meeting on Experimental and Behavioral Social Science for their helpful comments. Many thanks also to Julia van Hanxleden and Kateryna Ukrainets for their excellent research assistance. Funding statement. Financial support from the Kiel Institute for the World Economy is gratefully acknowledged. References Adams, R. B., & Ragunathan, V. (2015). Lehman sisters. FIRN Research Paper. Ahern, K. R., & Dittmar, A. K. (2012). The changing of the boards: The impact on firm valuation of mandated female board representation. The Quarterly Journal of Economics, 127(1), 137–197. Ambrus, A., Greiner, B., & Pathak, P. A. (2015). How individual preferences are aggregated in groups: An experimental study. Journal of Public Economics, 129, 1–13. Apesteguia, J., Azmat, G., & Iriberri, N. (2012). The impact of gender composition on team performance and decision making: Evidence from the field. Management Science, 58(1), 78–93. Ariely, D., Loewenstein, G., & Prelec, D. (2003). Coherent arbitrariness: Stable demand curves without stable preferences. The Quarterly Journal of Economics, 118(1), 73–105. Armstrong, R. W., Williams, R. J., & Barrett, J. D. (2004). The impact of banality, risky shift and escalating commitment on ethical decision making. Journal of Business Ethics, 53(4), 365–370. Asch, S. E. (1956). Studies of independence and conformity: I. A minority of one against a unanimous majority. Psychological monographs: General and applied, 70(9), 1–70. Bagues, M., & Campa, P. (2021). Can gender quotas in candidate lists empower women? Evidence from a regression discontinuity design. Journal of Public Economics, 194, 104315. Baker, R. J., Laury, S. K., & Williams, A. W. (2008). Comparing small-group and individual behavior in lottery-choice experiments. Southern Economic Journal, 75(2), 367–382. Baltrunaite, A., Bello, P., Casarico, A., & Profeta, P. (2014). Gender quotas and the quality of politicians. Journal of Public Economics, 118, 62–74. Bansak, C. A., Graham, M. E., & Zebedee, A. A. (2011). The effects of gender composition of senior management on the economic fallout. Applied Economics Letters, 18(16), 1603–1607. Bauer, R. H., & Turner, J. H. (1974). Betting behavior in sexually homogeneous and heterogeneous groups. Psychological Reports, 34(1), 251–258. Black, D. (1948). On the rationale of group decision-making. Journal of Political Economy, 56(1), 23–34. Bogan, V. L., Just, D. R., & Dev, C. S. (2013). Team diversity and investment decision-making behavior. Review of Behavioural Finance, 5(2), 134–152. Born, A., Ranehill, E., & Sandberg, A. (2022). Gender and willingness to lead: Does the gender composition of teams matter? Review of Economics and Statistics, 104(2), 259–275. Brown, R. (1965). Social psychology. New York: Free Press of Glencoe. Byrnes, J. P., Miller, D. C., & Schafer, W. D. (1999). Gender differences in risk taking: A meta-analysis. Psychological Bulletin, 125(3), 367–383. Castillo, M., Leo, G., & Petrie, R. (2015). Room composition effects on risk taking by gender. Working Paper, George Mason University, Department of Economics. Charness, G., & Gneezy, U. (2012). Strong evidence for gender differences in risk taking. Journal of Economic Behavior & Organization, 83(1), 50–58. Chen, J., & Houser, D. (2019). When are women willing to lead? The effect of team gender composition and gendered tasks. The Leadership Quarterly, 30(6), 101340. Cialdini, R., & Goldstein, N. (2004). Social influence: Compliance and conformity. Annual Review of Psychology, 55(1), 591–621. Croson, R., & Gneezy, U. (2009). Gender differences in preferences. Journal of Economic Literature, 47(2), 448–474. Cueva, C., & Rustichini, A. (2015). Is financial instability male-driven? Gender and cognitive skills in experimental asset markets. Journal of Economic Behavior & Organization, 119, 330–344. Daly, M., & Wilson, M. (2001). Risk-taking, intrasexual competition, and homicide. Nebraska Symposium on Motivation, 47, 1–36. Dave, C., Eckel, C. C., Johnson, C. A., & Rojas, C. (2010). Eliciting risk preferences: When is simple better? Journal of Risk and Uncertainty, 41(3), 219–243. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press Experimental Economics 21 den Nieuwenboer, N. A., & Kaptein, M. (2008). Spiraling down into corruption: A dynamic analysis of the social identity processes that cause corruption in organizations to grow. Journal of Business Ethics, 83(2), 133–146. Eckbo, B. E., Nygaard, K., & Thorburn, K. S. (2016). How costly is forced gender-balancing of corporate boards? European Corporate Governance Institute (ECGI)-Finance Working Paper, (463). Eckel, C. C., & Grossman, P. J. (2002). Sex differences and statistical stereotyping in attitudes toward financial risk. Evolution and Human Behavior, 23(4), 281–295. Eckel, C. C., & Grossman, P. J. (2008). Men, women and risk aversion: Experimental evidence. Handbook of Experimental Economics Results, 1, 1061–1073. Eliaz, K., Ray, D., & Razin, R. (2006). Choice shifts in groups: A decision-theoretic basis. American Economic Review, 96(4), 1321–1332. Ertac, S., & Gurdal, M. Y. (2012). Deciding to decide: Gender, leadership and risk-taking in groups. Journal of Economic Behavior & Organization, 83(1), 24–30. Ferreira, F., & Gyourko, J. (2014). Does gender matter for political leadership? The case of US mayors. Journal of Public Economics, 112, 24–39. Filippin, A., & Crosetto, P. (2016). A reconsideration of gender differences in risk attitudes. Management Science, 62(11), 3138–3160. Flory, J. A., Leibbrandt, A., Rott, C., & Stoddard, O. (2021). Increasing workplace diversity evidence from a recruiting experiment at a Fortune 500 Company. Journal of Human Resources, 56(1), 73–92. Gilley, K. M., Walters, B. A., & Olson, B. J. (2002). Top management team risk taking propensities and firm performance: Direct and moderating effects. Journal of Business Strategies, 19(2), 96–114. Glover, J. A., & Sautter, F. (1977). Relation of four components of creativity to risk-taking preferences. Psychological Reports, 41(1), 227–230. Gneezy, U., & Potters, J. (1997). An experiment on risk taking and evaluation periods. Quarterly Journal of Economics, 112(2), 631–646. Grossman, P. J., Komai, M., & Jensen, J. E. (2015). Leadership and gender in groups: An experiment. Canadian Journal of Economics/Revue canadienne d’économique, 48(1), 368–388. Harjoto, M., & Laksmana, I. (2018). The impact of corporate social responsibility on risk taking and firm value. Journal of Business Ethics, 151(2), 353–373. Harrison, G. W., Lau, M. I., Rutstr om, E. E., & Tarazona-Gómez, M. (2013). Preferences over social risk. Oxford Economic Papers, 65(1), 25–46. Hoogendoorn, S., Oosterbeek, H., & van Praag, M. (2013). The impact of gender diversity on the performance of business teams: evidence from a field experiment. Management Science, 59(7), 1514–1528. Jagau, S., & Offerman, T. (2018). Defaults, normative anchors, and the occurrence of risky and cautious shifts. Journal of Risk and Uncertainty, 56(3), 211–236. Johansen, M. A., & Sandnes, M. J. (2008). Gender equality in the boardroom: The Norwegian case of Rosa Parks? An event study [Unpublished master’s thesis]. The Norwegian School of Economics. Kanthak, K., & Woon, J. (2015). Women don’t run? Election aversion and candidate entry. American Journal of Political Science, 59(3), 595–612. Kofoed, M. S. (2019). The effect of same-gender or same-race role models on occupation choice evidence from randomly assigned mentors at West Point. Journal of Human Resources, 54(2), 430–467. Li, S., Chaudhuri, A., & Sbai, E. 2020. (Un)willing to lead? Men, women and the leadership gap. University of Auckland Working Paper. Masclet, D., Colombier, N., Denant-Boemont, L., & Lohéac, Y. (2009). Group and individual risk preferences: A lottery-choice experiment with self-employed and salaried workers. Journal of Economic Behavior & Organization, 70(3), 470–484. Matsa, D. A., & Miller, A. R. (2013). A female style in corporate leadership? Evidence from quotas. American Economic Journal: Applied Economics, 5(3), 136–169. Nieboer, J. (2015). Group member characteristics and risk taking by consensus. Journal of Behavioral and Experimental Economics, 57, 81–88. Nordhøy, F. (1962). Group interaction in decision-making under risk [Unpublished master thesis]. MIT School of Industrial Management. Nygaard, K. (2011). Forced board changes: Evidence from Norway. Norwegian School of Economics and Business Administration Discussion Paper, (5). Pahlke, J., Strasser, S., & Vieider, F. M. (2012). Risk taking for others under accountability. Economics Letters, 114(1), 102–105. Ranehill, E., & Weber, R. A. (2022). Gender preference gaps and voting for redistribution. Experimental Economics, 25(3), 845–875. Reynolds, D. B., Joseph, J., & Sherwood, R. (2009). Risky shift versus cautious shift: Determining differences in risk taking between private and public management decision-making. Journal of Business & Economics Research (JBER), 7(1), 63–78 Shupp, R. S., & Williams, A. W. (2008). Risk preference differentials of small groups and individuals. Economic Journal, 118(525), 258–283. https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press 22 Katharina Lima de Miranda et al. Stoner, J. A. F. (1961). A comparison of individual and group decisions involving risk [Unpublished master’s thesis]. Massachusetts Institute of Technology. Stoner, J. A. F. (1968). Risky and cautious shifts in group decisions. Journal of Experimental Social Psychology, 4(4), 442–459. Sutter, M. (2009). Individual behavior and group membership: Comment. American Economic Review, 99(5), 2247–2257. Vidmar, N. (1970). Group composition and the risky shift. Journal of Experimental Social Psychology, 6(2), 153–166. Wallach, M. A., Kogan, N., & Bem, D. J. (1962). Group influence on individual risk taking. The Journal of Abnormal and Social Psychology, 65(2), 75–86. Wallach, M. A., Kogan, N., & Bem, D. J. (1964). Diffusion of responsibility and level of risk taking in groups. The Journal of Abnormal and Social Psychology, 68(3), 263–274. Walls, M. R., & Dyer, J. S. (1996). Risk propensity and firm performance: A study of the petroleum exploration industry. Management Science, 42(7), 1004–1021. Cite this article: Lima de Miranda, K., Detlefsen, L., & Schmidt, U. (2025). Can gender diversity prevent risky choice shifts? The effect of gender composition on group decisions under risk. Experimental Economics, 1–22. https://doi.org/10.1017/eec.2024.4 https://doi.org/10.1017/eec.2024.4 Published online by Cambridge University Press