Master Thesis
Part of the MSc in Economics: Behavioural Economics and Game
Theory
-
The effect of gender identity norms and
traditional gender role ideologies on the
gender gap in competition, moderated by the
willingness to improve
Name: Sanne Eitjes
Student Number: 10421785
Supervisor: Mw. dr. G. (Giorgia) Romagnoli
Submission date: 15 August 2018
STATEMENT OF ORIGINALITY
This document is written by student Sanne Eitjes who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
ABSTRACT
This paper examines the underpinnings of the gender gap in competition by looking at two potential factors of nurture. First, it is argued that social gender identity, activated through priming, influences preferences for competition. For females, gender priming resulted in a significantly higher likeliness to compete whereas for males the opposite result was found, but only when competing against other men. Second, it is argued that individuals born 1980 are more likely to adopt the idea that female performance is inferior to male performance, due to holding traditional gender role ideologies. It is found that men born 1980 and earlier show a higher level of satisfaction when comparing their performance from a real-effort task to fellow men than to women. Another goal of this paper is to test for a potential moderating effect for the willingness to become better on the gender gap in competition. This was tested by giving the participants the option to do a real-effort task for a second time within a usual competitive environment. No moderating effect for self-improvement was found.
TABLE OF CONTENTS
1. INTRODUCTION ... 1
2. LITERATURE REVIEW ... 4
2.1 – GENDER AND LABOR MARKET OUTCOMES ... 4
2.2 – GENDER DIFFERENCES IN COMPETITION ... 5
2.3 – SOCIAL IDENTITY THEORY ... 7
2.4 – GENDER ROLE IDEOLOGIES ... 8
3. HYPOTHESES AND RESEARCH CONTRIBUTION ... 10
3.1 – HYPOTHESES ... 10 3.2 – RESEARCH CONTRIBUTION ... 12 4. METHODOLOGY ... 13 4.1 – DATA COLLECTION ... 13 4.2 – EXPERIMENTAL DESIGN ... 14 4.2.1 – GENERAL INFORMATION ... 14 4.2.2 – PRIMING PROCEDURE ... 14
4.2.3 – STAGE 1, PART 1: ABSTRACT REASONING COMPETITION ... 15
4.2.4 – STAGE 1, PART 2: FEEDBACK AND ASSIGNMENT TO REFERENCE GROUPS ... 17
4.2.5 – STAGE 2: RE-ENTERING THE COMPETITION ... 18
4.2.6 – STAGE 2: FINAL QUESTIONNAIRE ... 19
5. EXPERIMENTAL RESULTS ... 20
5.1 - PARTICIPANTS AND SUMMARY STATISTICS ... 20
5.2 – GENDER-GAP IN COMPETITION... 20
5.3 – HYPOTHESES TEST ... 21
6 – DISCUSSION AND CONCLUSION ... 31
6.1 – SUMMARY OF FINDINGS ... 31
6.2 – LIMITATIONS ... 32
6.3 - CONCLUSION AND FUTURE RESEARCH ... 34
REFERENCES ... 35
APPENDIX A ... 38
1. INTRODUCTION
Despite the important strides made to diminish the gender income inequality, there still exists a gender wage gap of roughly 14% in OECD countries (OECD, 2018). Following the definition of the OECD (2018), the gender wage gap is defined as “the difference between median earnings of men and women relative to median earnings of men”.
As top-level jobs commonly require workers to be competitive, the gender gap in competitiveness is frequently used to explain the occurrence of the gender wage gap and the differences in labor market outcomes between men and women. So far, almost all experimental studies find that men are more willing to compete than women (Niederle & Vesterlund, 2007), except for studies that were conducted in a matrilineal society (Gneezy et al., 2009), in a patriarchal society in transition (Dariel et al., 2013) and in a setting of self-competition (Apicella et al., 2017).
A natural question that arises is whether these differing attitudes of males and females towards competition are correlated with biological factors (factors of nature) or whether they are correlated with psychological and socio-psychological factors (factors of nurture). Gneezy at al. (2009) show that while males are more competitive inclined than females among the patriarchal Maasai of Tanzania, females choose a competitive environment more often among the matrilineal Khasi of India. These contrasting behaviors highlight the importance of nurture in explaining the differing attitudes of males and females towards competition. The goal of this paper is to explore the underpinnings of the gender gap in competition and find ways to increase the number of women entering competitive environments. First, there is tested to what extent preferences towards competition are influenced by socially constructed gender norms and holding traditional gender role ideologies. This is central to the idea that preferences towards competition are shaped by factors of nurture. Second, it is tested in what ways information about relative performance in a real-effort task and information about the gender of co-competitors can aspire people to compete and get better.
A first potential factor of nurture that potentially can declare differing attitudes towards competition between men and women is the prevalence of social gender norms on what is acceptable for men to do and what is acceptable for women to do. Akerlof and Kranton (2000) suggest a model where one’s identity directly is part of the utility function. Under this model, economic decisions are partly explained by the need to behave according to one’s sense of belonging. Hence, deviating from behavior that is expected within a certain social category decreases utility. Another potential factor of nurture that potentially can declare differing attitudes towards competition between men and women is the holding of traditional gender role ideologies. According to Trelaeven (2015), individuals born 1980 and earlier (also referred to as the Baby-Boomers and generation) hold more traditional gender role ideologies than individuals born 1981 and later (also: generation Y and Z).
First, it is argued that the gender gap in competitiveness is part of one’s gender identity. In an online experiment, half of the participants was exposed to a priming stimulus to make the gender identity salient, before making the decision to enter a competitive environment. It is argued that females are less likely to enter a competitive environment under gender priming, for the reason they are triggered to behave according to female-specific (docile, non-competitive) roles. On the other hand, it is argued that males are more likely to enter a competitive environment under gender priming, for the reason they are triggered to behave according to male-specific (confident) roles. It is found that females primed with their gender, are significantly less likely to enter a competitive environment. In addition, males who were primed with their gender and competing against a single-sex group, are found to be significantly more likely to enter a competitive environment.
Second, there is tested whether information about relative performance and information about co-participants’ gender influences preferences towards competition. It is argued that people generally prefer to compete against other women, as they are influenced by generally held gender-conditioned ideas about behavior of others (Holm, 2000). No significant effect for a preference to compete against women is found. Also, it is argued that beliefs about gender roles influence expectations about performance. According to Buchanan (2014), individuals holding more traditional gender role ideologies are more likely to adopt the idea that female performance is inferior to male performance. In the same online experiment,
participants completed an abstract reasoning test of 10 problems and asked to state their level of satisfaction with their score both before and after receiving information about their co-participants’ gender and performance. Men born 1980 and earlier are found to be significantly more responsive to receiving information about their relative performance compared to fellow men than to fellow women. However, this did not result in a lower preference to compete against other men.
Third, it is tested in what ways information about relative performance in the abstract reasoning test and information about co-participants’ gender can aspire people to improve their personal score. A unique aspect of the experimental design is that the participants were given the option to complete the 10 abstract reasoning problems for a second time. Due to only having a small chance on winning money in case of winning the competition, it can be reasoned that the choice for retaking the test is partly guided by the intrinsic motivation for the willingness to improve the personal score. An interesting question to ask is whether adding an element of self-improvement to a usual competitive setting (other-competition) can moderate the gender gap in competition. However, no moderating effect for self-improvement is found.
Results of this paper suggest that the decision to avoid competition for women and to compete more for men are part of socially constructed gender norms and not guided by lack of ability. The result that men born 1980 and earlier are more responsive to male performance suggests that they may be prone to the idea that female performance is inferior to male performance. Both results found might be important in declaring differing labor market outcomes between men and women.
The remainder of this paper is organized as follows. The next section provides a literature review. Section three presents the hypotheses. Section four describes the experimental design. Section five presents the experimental results. Section six summarizes, discusses and concludes.
2. LITERATURE REVIEW
2.1 – GENDER AND LABOR MARKET OUTCOMES
In labor economics, both traditional and alternative explanations are distinguished to understand differences in labor market outcomes between men and women. In the influential chapter in the Handbook of Labor Economics, where Antonji and Blank (1999) summarize the literature on gender inequalities, two traditional main sources are discussed for the existence of the gender wage gap, which are human-capital accumulation and discrimination. Gender discrimination, regarded as a demand-side explanation, is defined as a situation in which evenly productive men and women are being paid an unequal wage (Azmat & Petrongolo, 2014). In order to correctly measure gender discrimination in the labor market, the challenge is to find a way to accurately measure levels of productivity. Commonly methods used to measure the “discrimination residual” is by decomposing wages differentials between men and women into a gender gap which can be declared by observable worker differences in characteristics and into a gender gap which cannot be declared by any kind reason and therefore can be attributed to factors of discrimination (Azmat & Petrangelo 2014). Differences in accumulation of human capital in both pre-labor market entry (type of education) and post labor market entry (accumulated work experience) is regarded as a supply-side explanation for explaining different labor market outcomes between men and women (Marianne, 2011). Antonji and Blank (1999) review results of research into these two traditional sources in their work but conclude that sizeable parts of the gender wage gap are left unexplained.
To explain the residual gender gap, a more recent strand of literature concentrates on alternative supply-side explanations to declare different labor market outcomes between men and women. These alternative explanations are related to different preferences between men and women and brought into economic research due the increasing influence of psychology and experimental literature (Marianne, 2011). In particular, social preferences and preferences for risk and competition seem to differ systematically between men and women. Empirical evidence has shown these different preferences to play a role in explaining the gender wage gap and differences in occupational choice between men and women (Kleinjans, 2009).
2.2 – GENDER DIFFERENCES IN COMPETITION
Top-level jobs often are known for being highly competitive. Within these jobs, the difference between winners and losers is large. Winners are mostly disproportionally rewarded while losers are left empty-handed (Marianne, 2011). A proposed reason for why women are underrepresented in high-earning, high profile jobs is that men and women differ in their attitudes towards competition. Mainly, attitudes seem to differ because many women simply tend to shy away from competitive environments, even when they know they belong to the most capable, and because women systematically under-perform compared to men in competitive environments (Marianne, 2011).
Niederle and Vesterlund (2007) are one of the first who find evidence for the existence of different attitudes towards competition between men and women. In their experiment, 40 men and 40 women were asked to do a real-effort task during five minutes under different compensation plans. In the first compensation plan, the payoff of participants only depended on their own performance: each participant was paid a fixed piece rate of 50 cents for every problem solved correctly. Subsequently, participants performed in a tournament with two men and two women. Now, only the participant who solved the largest amount of problems was paid $2 per correct answer. Men slightly outperformed women in both compensation plans, however the differences were not significant in each case. After having performed both in the piece rate condition and the tournament, participants selected which of the two they desired on a successive task. Here, a large and significant gender difference in the entry of the tournament condition was observed: 35% of women as opposed to 73% of men entered the tournament condition. Niederle and Vesterlund (2007) find that this gender gap in tournament entry can be attributed to greater male overconfidence in their abilities relative to female overconfidence and to the different attitudes to competition associated with each gender. On the other hand, the experiment provides little evidence for the role for either risk of feedback aversion having an influence on the decision to enter a tournament.
The finding of Niederle and Vesterlund (2007) is replicated many times in several subsequent studies. Healy and pate (2011), who test for the effect of team competition on the gender gap in competition, find that females have a preference for competing in teams. When looking at the individual setting, a large gender gap in competition is found: 81% of the males, as
opposed to 28%, prefer to compete as an individual. Buser et al. (2014) find that the willingness to compete relates to educational choices: in their study conducted the Netherlands, they find replicate the finding that boys are more competitive than girls (49% vs. 23%). Also, they find that people who choose a more prestigious academic study track (mostly math- and science- intensive), are more likely to be competitive. Reuben et al. (2015) find that the willingness to compete relates to occupational choices: individuals, who are more competitive and overconfident expect to earn a significantly higher future earnings. So far, almost all previous studies find that men are more willing to compete than women. The limited number of studies that find the opposite result have in common that they were conducted in either a matrilineal society (Gneezy et al., 2009) or in a patriarchal society in transition (Dariel et al., 2017). The finding that the gender gap in competition is reversed in a matrilineal society (Gneezy et al., 2009) shows that preferences for competition are shaped by factors of nurture. Gneezy et al. (2009) conduct a simple experimental task of throwing a tennis ball into a bucket and compare attitudes towards competition between the matrilineal Khasi in India and the patriarchal Maasai in Tanzania. They find that Khasi women select into a competitive environment more frequently than Khasi men, as opposed to Maasai women, who select into a competitive environment at half of the rate of Maasai men. Andersen et al. (2013) complement the findings of Gneezy et al. (2009) and find that competitive attitudes are developed at a young age for Maasai Men (between 7 and 15 years), but not for young Maasai women. More recently, Dariel et al. (2017) test for gender differences in competitive attitudes among students from the United Arab Emirates, and find no difference in preference for competition. The United Arab Emirates is known as a typical patriarchal society but has recently implemented various initiatives that aim to empower women and provide equal rights to employees. Their research considers both competition in mixed-sex and single-sex groups.
Besides Dariel et al. (2017), two other studies so far explored whether attitudes towards competition are influenced by knowing co-participants’ gender. First, Datta Gupta et al. (2013), who conducted a study a France, find experimental evidence that knowing the gender of co-competitors directly affects men’s decisions to compete, as men prefer to compete against women over competing against men. Sutter and Glätzle-Rützler (2014) find no
significant differences in likeliness to sort into a competitive environment between a mixed-sex and single-mixed-sex treatments.
Two other studies develop experiments to explore whether a gender difference in willingness to compete exists when competing against oneself (self-competition). Apicella et al. (2017) cannot reject the hypothesis that gender differences in the willingness to self-compete differ. Similarly, Carpenter et al. (2017) find that women are more likely to select into intrapersonal competition than interpersonal competition.
2.3 – SOCIAL IDENTITY THEORY
According to Akerlof and Kranton (2000, 2010), an individuals’ social identity, which also includes gender identity, could be an important factor in declaring economic choices, such as occupational sorting and labor force participation. The concept of social identity refers to one’s feeling of belonging to one or multiple social categories (Marianne, 2011). Each of these social groups is linked with a unique role identity (Eagly and Karau, 2002). A role identity covers behavioral norms on how people in social category should behave (Eagly & Karau, 2002).
Lately, economists have become interested in the question to what extent the notion of identity plays a role in declaring various aspects of economic decision making (Akerlof & Kranton, 2000; Chen and Li; 2009). Akerlof and Kranton (2000) are the first to suggest a model where one’s identity directly is part of the utility function. Under this model, economic decisions are partly explained by the need to behave according to one’s sense of belonging. The most relevant social identity related to this the paper is the concept of gender identity, where “men” and “women” are the most common social categories. The two categories are encompassed with a unique role identity which cover behavioral prescriptions. According to psychologists, women are generally expected to be generous and docile, while men are expected to be self-assertive and confident (Eagly, 1987). If there is not behaved according to these prescriptions, utility is expected to decrease (Akerlof & Kranton, 2000).
A logical question that arises is whether gender identity norms play a role in declaring gender differences in psychological traits, such as attitudes towards altruism, negotiation, risk and competition. For example, some argue that women are generally expected to be risk-averse
according to female gender norms, while being a risk-taker is seen as the norm for males. In an experiment, Eckel and Grossman (2002) ask men to estimate the level of risk aversion of women and find that men assume that women are even more risk-averse then they actually are. The idea is that above expectations are part of socially constructed gender identity norms (factors of nurture), instead of being a reflection on innate behavioral differences (factors of nature).
An often-used method to make one’s identity salient is by exposing participants to a priming stimulus. So far, a lot of studies have shown that identity priming is a successful tool for influencing behavior: First, Shih et al. (1999) find that Asian-American perform better in a mathematics test when they were reminded of their ethnicity. At the same time, they performed worse when made aware of their gender identity. Boschini et al. (2009) find that economic preferences are influenced by gender identity with women becoming more generous in the dictator game, but only when they are assigned to mixed-gender groups. Cadsby et al. (2013) experimentally test for gender differences in competition and draw on social identity theory. They find that female MBA students who are primed with their professional identity are significantly more likely to compete than female MBA students who were primed with their gender/family identity. Also, Drupp et al. (2017) find that the gender gap in risk-taking is moderated when the professional identity is made salient.
2.4 – GENDER ROLE IDEOLOGIES
Another culturally defined construct that potentially can declare gender differences in psychological traits, such as attitudes towards competition, is the concept of gender role ideology. The concept of gender role ideology refers to an individual’s view toward the roles men and women should occupy in the society (Somech & Drach-Zahavy, 2016). Three types of gender role ideologies are distinguished: traditional, transitional and egalitarian. According to traditional gender norms, women should be responsible for taking care of the family, whereas men should be responsible for earning money (livelihood). According to egalitarian gender norms, men and women are equally responsible to taking care of the livelihood and family. Transitional gender norms lie in between the perceptions of the traditional and egalitarian roles (Somech & Drach-Zahavy, 2016).
As an explanatory variable, different types of gender role ideologies have been used to predict the differences in labor market outcomes between men and women. Trelaeven (2015) tests for differences in beliefs about gender roles between different generations and finds that generation Y and generation Z (born 1981 and later) significantly hold more progressive beliefs about gender equality than generation X and the Baby-Boomers (born 1980 and earlier). Buchanan (2014) finds that beliefs about gender roles influence expectations about performance. In particular, it is argued that individuals holding more traditional gender role ideologies, are more likely to adopt the idea that female workplace performance is inferior to male performance. Firestone et al. (1999) find that holding traditional gender-role ideology leads to lower observed occupational earnings for both men and women.
3. HYPOTHESES AND RESEARCH CONTRIBUTION
3.1 – HYPOTHESES
The main interest of this study is to test whether (1) social identity norms and the concept of gender role ideologies are responsible for declaring gender differences in attitudes towards competition, to test whether (2) information about relative performance and about co-participant’s gender influences preferences towards competition and to test whether (3) information about relative performance and about co-participant’s gender can aspire people to improve their score.
The current study draws on the idea that social identity theory and on the concept of gender role ideologies play a role in declaring gender differences in attitudes towards competition. To test for this presumption, half of the participants is exposed to a priming stimulus to make the gender identity salient. The main prediction is that males show a higher likeliness to enter a competitive environment under gender priming, whereas females show a lower likeliness to enter a competitive environment under gender priming. In addition, it is predicted that participants born 1980 and earlier (generation X and Baby-Boomers) are more responsive to receiving information about their relative performance compared to fellow men than to fellow women. The responsiveness measure used in the experiment is ∆s1. The testing of this expectation, embodied in hypothesis H2a and H2b, is central to the idea that individuals born 1980 and earlier hold more traditional gender role ideologies and are therefore more likely to adopt the idea that female performance is inferior to male performance (Buchanan, 2014). Building on hypotheses 1 and 2, there is examined whether giving participants information about their co-participants’ gender influences the likeliness to enter a competitive environment. For both men and women, the likeliness to enter a competitive environment is expected to be higher when competing against other women. A possible explanation for this
1“Difference-in-satisfaction score”: All participants in the experiment were asked to indicate their
level of satisfaction with their personal score before (“satisfaction Before” or !") and after
(“Satisfaction After” or !#) receiving information about their relative performance compared to either
a mixed-sex or single-sex group (“satisfaction Before” or !"). Difference-in-satisfaction score (∆s) is
is that they might be influenced by generally held gender-conditioned ideas about behavior of others (Holm, 2000). On top of that, this effect is expected to be stronger for men and women born in 1980 and earlier (Baby-Boomers and generation X). This is rooted in the idea that people born in those generations are affected by the belief that male performance is worth more than female performance. Due to holding egalitarian gender role beliefs, participants born in generation Y and Z (1981 and later) are not expected to value male and female performance in a different way, and thus expected to be unresponsive to information about co-participants’ gender when making the decision to enter a competitive environment. Finally, the experiment examines whether giving participants information about relative performance and about co-participant’s gender can aspire them to improve their score in an abstract reasoning test, when giving them to option to take an exact same test for a second time. It is argued that women born 1980 and earlier show a lower willingness to become better as compared to women born 1981 and later. This is rooted in the idea that a higher level of ∆s raises confidence and confidence is expected to be positively correlated with the willingness to become better. Also, it is argued that individuals’ holding the idea that they are good in solving abstract reasoning problems, show a higher willingness to improve. As for women born 1981 and later it is expected that they do not make a distinction between male and female performance, they are expected to adopt the idea that they are either as good or better in solving abstract reasoning problems than men. Therefore, the willingness to improve is expected to be higher for them. Since women born 1980 and earlier are expected to be less confident about their performance in general (hypothesis 2a and 2b), it is hypothesized that they show a lower willingness to become better as compared to women born 1981 and later. For males, the opposite explanation holds. Therefore, men born 1980 and earlier are expected to show a higher willingness to become better as compared to men 1981 and later.
The main hypotheses are formulated as follows:
H1a. Males are more likely to enter a competitive environment under gender priming H1b. Females are less likely to enter a competitive environment under gender priming H2a. Males born 1980 and earlier are more responsive to receiving information about their relative performance compared to a single-sex group than to a mixed-sex group
H2b. Females born 1980 and earlier are more responsive to receiving information about their relative performance compared to a mixed-sex group than to a single-sex group.
H3a. Females are more likely to enter a competitive environment when being assigned to a single-sex group. This effect is stronger for females born 1980 and earlier.
H3b. Males are less likely to enter a competitive environment when being assigned to a single-sex group. This effect is stronger for males born 1980 and earlier.
H4a. Women born 1980 and earlier show a lower willingness to become better as compared to women born 1981 and later.
H4b. Men born 1980 and earlier show a higher willingness to become better as compared to men 1981 and later.
3.2 – RESEARCH CONTRIBUTION
The current study differs from previous experimental studies in the following ways: first, there is tested for potential effects of socially constructed gender identity norms on preferences for competition, similar to Cadsby et al. (2013), but using the priming method of Boschini et al. (2009). Boschini et al. (2009) use a very simple, but successful priming method, as they find that economic preferences are influenced by gender identity with women becoming more generous in the dictator game, but only when they are assigned to gender groups. Most previous studies on attitudes towards competition only consider mixed-sex groups. In this study, attitudes towards competition are considered in both mixed-mixed-sex and single-sex group, to test for any potential relationship between social identity gender norms, holding traditional gender role ideologies and knowing co-participants’ gender on the likeliness to compete. Lastly, the current study distinguishes itself from the rest by using an alternative measure of competition. Where almost all previous studies measure competition using the competition measure of Niederle and Vesterlund (2007), in this study a new measure is introduced. The decision to enter a competitive environment is both a matter of willingness to win and willingness to improve. By adding an element of willingness to improve, there can be tested for any potential moderating effects of self-improvement on the gender gap in willingness to compete against others.
4. METHODOLOGY
In the following, a detailed description is provided of the experiment conducted in this study, including a motivation for all choices made.
4.1 – DATA COLLECTION
The experiment of this study was conducted online. The main reason for conducting it online and not in the lab was to acquire a larger and more diverse subject pool. Participants were recruited by sending the experiment to family, friends and acquaintances of the author through Facebook and e-mail in first place. 48 people completed the experiment in this way. Participation was completely voluntary in this group. To complement the dataset, another 152 participants were recruited through Amazon Mechanical Turk (MTurk). Mturk is an online marketplace where requesters for human intelligence can hire workers to complete surveys and experiments (also: Human Intelligence Tasks or HITs) in return for money. According to Buhrmester et al. (2011), data generated by Mturk population is at least as reliable as data acquired via traditional ways. In addition, they find that unrealistic compensation rates do not influence data quality, but only slow down the pace of data collection. Therefore, the compensation rate for MTurk workers was set to $0,50. Another reason for the low compensation rate was to make data from MTurk workers as comparable as possible to data from participants who were recruited on voluntary base. The main motivation to minimize the extrinsic reward in general was to ensure that decisions made in the actual experiment were guided by a participants’ intrinsic motivation to improve their personal score.
4.2 – EXPERIMENTAL DESIGN
The online experiment was programmed in Qualtrics Survey Software. The experiment consisted of two treatments, a gender priming treatment and the neutral priming treatment. Both treatments had two stages. In figure 1, an overview is provided of the flow of the different treatments. The exact experimental instructions can be found in the appendix A.
Figure 1. Experimental flow of different treatments
4.2.1 – GENERAL INFORMATION
All participants started the experiment on the instructions page, where they were given general information. Participants were told that they were about to enter an experiment which is about solving abstract reasoning problems and that the experiment was conducted as part of a MSc Thesis at the University of Amsterdam. At the bottom of this page, all participants had to concede with the payment/ privacy and deception conditions before continuing to the real experiment.
4.2.2 – PRIMING PROCEDURE
Subsequently, Qualtrics randomly assigned each participant to one of the two treatments of the experiment. The two treatments were identical, except for the way in which participants were asked to state their gender. In the gender priming treatment, participants were asked to state their gender right before they received the instructions of stage 1. The purpose of the gender priming treatment was to make their gender identity salient. This way of gender priming is copied from the experimental design of Boschini et al., (2009). The idea to prime participants with their gender before entering a competition is based on the article of Gneezy et al. (2009), who confirm the conjecture that preferences for competition are shaped by
factors of nurture. The other half of the participants was assigned to the neutral priming treatment. In this treatment, the gender question was not a separate question, but part of the larger demographic questionnaire. Besides from the way the participants were asked to state their gender, the two treatments were identical and each consisted of each two stages.
4.2.3 – STAGE 1, PART 1: ABSTRACT REASONING COMPETITION
In stage 1, participants had to participate in an abstract reasoning competition. Stage 1 started with an instructions page. Here, the participants were told that they had to solve as many out of 10 abstract reasoning problems as possible (also referred to as task 1). In addition, they were informed that at the end of the abstract reasoning task, they would be randomly assigned to a group of 10 people (also: reference group) aged between 18-65 years who all completed the same test before2. Participants were told that they would be
competing against these 10 people, and that they would earn a lottery ticket3 if they solved
either a greater amount or an equal amount of problems as the best person in their group. In previous prominent literature in this field, it has been shown multiple times that women are substantially less likely to enter a winner-takes-all tournament than men (e.g. Niederle & Vesterlund, 2007; Healy & Pate, 2011; Datta Gupta et al., 2013). For this reason, it was chosen to reward participants according to a similar principle. Usually, a winner-takes-all payment structure only rewards the best person of the group, and randomly rewards one person in case of a tie. In order to trigger participants to do well and not to discourage them to re-enter the competition in stage 2, the payment structure was adjusted by both rewarding solving the most problems as well as rewarding a tie.
To trigger the participants to put effort in the abstract reasoning competition, it was emphasized that abstract reasoning tests are a good measure of general intelligence and commonly used within selection processes of multiple well-known firms (such as Shell, KPMG, Deloitte, Microsoft, Ford Motor and more). The goal of emphasizing of this was to activate
2 The raw data on the 10 abstract reasoning problems was obtained from https://openpsychometrics.org. The 10
problems were part of a larger experimental IQ test of 25 problems in total. In total, 400 participants
participated in this test. Out of these 400 people, 100 males and 100 females were randomly selected who were aged between 18-65 years. On top of that, these 100 males and females were randomly divided into 10 groups. Each female group contained 10 females, each male group contained 10 males and each mixed-gender group contained 5 males and 5 females.
3 Lottery tickets could be used in the lottery that took place on the 15st of August, 2018. Out of all obtained
the belief that is possible for everyone to score well in the abstract reasoning test. The instructions page ended with an example question of a simple abstract reasoning problem. Before starting the abstract reasoning competition, participants were asked to fill in a demographic questionnaire. These questions asked about age, gender, level of education, income and ethnicity. For the participants in the gender priming condition, the question about gender was left out, as they were asked for it before. Lastly, they were asked to fill in two control questions in order to ensure that the nature of the abstract reasoning test was well understood. After this, participants started with task 1. Figure 2 shows a screenshot of the first abstract reasoning problem of task 1. The remaining nine questions were similar but became more difficult as the test approached. For the reason it could not been tracked down how much time the participants in the original test had to complete these 10 questions, it was chosen to not impose a time limit for the participants in this experiment.
4.2.4 – STAGE 1, PART 2: FEEDBACK AND ASSIGNMENT TO REFERENCE GROUPS
After completing task 1, participants immediately received feedback on their score by informing them on how many problems they solved correctly out of 10. Additionally, they were asked to indicate how satisfied they were with this score on a 7-points scale. Subsequently, all participants were randomly assigned to either a mixed-gender group (with 5 males and 5 females) or a single-sex group. Participants were provided with information about how their reference group performed in the test. This information was about the gender composition of the group, the numbers of answered correctly by each of the 10 group members (shown in a graph), the highest score, the average score and average age of all the group members. Next, participants were asked to take the information of their group into consideration, and to indicate their level of satisfaction with their personal score on a 7-points scale again. This exact question was worded in the following way: “The above information on the performance of your group members provides an indication on how well the average (male/female4) population performs in an abstract reasoning test. Comparing
yourself to the 10 people5 above, indicate again how satisfied you are with having solved X out
of 10 problems in the abstract reasoning test”. The average score, average age and highest score of all the 30 reference groups can be found in appendix B.
Depending on whether the participant managed to solve a greater or an equal amount of problems as the best person in their group, the experiment continued differently. ‘Winners’ were congratulated and assigned a unique lottery number. ‘Losers’ were informed that they did not solve enough problems to earn a lottery ticket. To keep participants motivated, all participants were given general feedback on their score. Feedback could be derived from a scoring table. The main purpose of providing participants with general feedback was to encourage them to re-enter the competition in stage 2 by giving them tips which could be brought in practice immediately. The main message that was spread through the feedback is the reminder that the abstract reasoning test is an assessment of general intelligence, and not of prior knowledge.
4 In mixed-gender groups: average population; in single-sex male groups: average male population; in single sex
female groups: average female population
4.2.5 – STAGE 2: RE-ENTERING THE COMPETITION
In stage 2, participants could decide the way to continue the rest of the experiment their selves by choosing between two possible paths. Figure 2 shows a screenshot of the two paths.
Figure 3. Screenshot of the two possible paths that could be taken in stage 2
In path 1, the experiment continued by re-entering the abstract reasoning competition. Participants were informed that they would be taking the exact same abstract reasoning test again and that the rules of the competition were the same as in stage 1: they would be assigned to a random group of 10 people again and had to solve a greater or equal amount of problems as the best person in the group to win the competition. Winners were assigned a(n) additional lottery ticket. In addition: the advantages of taking path 1 were clarified: the opportunity to improve the personal score (1) and the opportunity to win a(n) (additional) lottery ticket (2). In order to help the participants to improve their score, they were told they would be provided with short instructions on how to solve a complex abstract reasoning problem before taking the test. Instructions were provided for two reasons: first, to trigger participants to re-enter the competition. Second, to measure the effort put into reading the by asking them two control questions about this in the final questionnaire.
In order to measure whether the willingness to re-enter the competition and the willingness to improve is influenced by the gender composition of the reference group, participants were informed they would be assigned to a group with the same gender composition as in stage 1. Therefore, everyone assigned to a single-sex group in stage 1 was informed he/she would be randomly assigned to a single-sex group again after task 2. Similarly, everyone assigned to a mixed-sex group in stage 1 was informed they would be competing against a mixed-group again in stage 2. The opportunity to let participants choose whether to re-enter the competition or not, gives useful insights to what extent information about gender composition influences likeliness to enter a competitive environment. In order to separate
between willingness to compete and willingness to improve the personal score, two control questions were added to the final questionnaire.
In summary, the rest of stage 2 continued as follows: after having received the instructions on how to do well in an abstract reasoning test, participants in path 1 re-entered the exact same abstract reasoning competition as in stage 1. After this, they were informed on their score and asked to state their level of satisfaction with this score on a 7-point scale. Subsequently, they were assigned to a random reference group with the same gender composition as in stage 1. On the next page, ‘winners’ were congratulated and assigned a(n) additional lottery ticket. ‘Losers’ were informed they did not solve enough problems. In addition, general feedback could be derived from the same feedback table that was provided in stage 1. Finally, everyone was forwarded to the final questionnaire. Participants in path 2 were forwarded to the final questionnaire immediately, without taking the abstract reasoning test again.
4.2.6 – STAGE 2: FINAL QUESTIONNAIRE
Both treatments ended with a final questionnaire. Self-reported levels of effort and enjoyment in task one and task two were collected using a five-point scale. Next to this, participants were asked to indicate how they perceived the level of the task, also using a five-point scale. As is has been shown that the type of task matters for measuring the gender gap in tournament entry (Niederle & Vesterlund, 2011), all participants were asked to state their belief about whether they thought men or women generally do better in the abstract reasoning test they just completed. For participants who re-entered the competition, the level of effort put in reading the instructions was measured by letting them answer two control questions. Finally, to measure whether re-entering the competition was a matter of willingness to improve or willingness to compete, participants were asked in two separate questions to indicate to what extent the possibility to improve their personal score and the possibility to win a(n) additional lottery ticket had triggered them to choose for path 1 on a five-point scale.
5. EXPERIMENTAL RESULTS
5.1 - PARTICIPANTS AND SUMMARY STATISTICS
In total, 200 participants completed the experiment. Out of these 200 people, 20 were excluded. They were excluded for two different reasons: first, for being older than 66 years or younger than 18 years. Second, for answering the question on how hard they worked for task 2, while not having participated in this task. This behavior shows that this question (and probably more) was answered without reading it properly or taking it seriously. After filtering out those people, the final sample consisted of 180 useful observations. Of them, 100 were male and 80 of them female. Of the 100 men, 54 and 46 were assigned to a mixed-and single-sex reference group. Of the 85 women, 45 were assigned to a single-single-sex group and 35 to a mixed-sex group. The average age was 35.47 years. 68 belonged to the Baby-Boomers and generation X (born 1980 and earlier) and 112 belonged to generation Y and Z (born 1981 and later). 18.33% participants had a student status and the average net monthly income was 2606 EURO. Many of the participants owned a tertiary degree (78.8%) and labeled as higher educated throughout the experiment. The majority of the participants was Caucasian (61.11%), followed by Asians/ Pacific Islanders (28.33%), African Americans/ Blacks (6.11%), Hispanics/ Latinos (6.11%), other (2.78%) and Native Americans (1.67%). Lastly, 41.6% of the participants had participated in an economic experiment before.
5.2 – GENDER-GAP IN COMPETITION
In the experiment, 36% of the males and 42.5% of the females continued to task 2, and therefore chose to enter a competitive environment. The hypothesis that this gap significantly differs from 0 cannot be confirmed (p=0.3723, two-sided two-sample t test), also not after controlling for performance. This result is interesting, as it contradicts to the result found in almost all previous studies: that men are more willing to compete than women. In the following section, hypotheses tests are conducted to test for the role of gender priming, knowing the co-participants’ gender and holding traditional role ideologies on the willingness to enter a competitive environment. At the end of the next section, there is tested for a moderating effect of willingness to become better on the gender gap in competition.
5.3 – HYPOTHESES TEST
Result 1 Overall, females are less likely to enter a competitive environment under gender
priming than under neutral priming.
Result 2 Overall, males are not less likely to enter a competitive environment under gender
priming. However, if assigned to a single-sex group, males are more likely to enter a competitive environment under gender priming.
The willingness to enter a competitive environment is measured by looking at the likeliness of continuing to task 2 in the experiment. Figure 4 shows the mean probabilities to continue to task 2 by gender and by priming treatment. At first view, the nature of the priming seems to influence the likeliness for continuing to task 2 for both gender groups: first, under neutral priming, 45.55 % of the females chooses to continue to task 2 as compared to 40.4% under gender priming. On the contrary, under neutral priming, 32.7% of the males chooses to continue to task 2 against 39.6% under gender priming.
Figure 4. Likeliness to continue to task 2 by gender and by priming treatment.
Notes. The vertical axis measures the average probability of continuing to task 2. The bars show the average values to continue to task 2 of men and women by priming treatment.
Table 1 presents the marginal effects of the probit estimates of the determinants for the likeliness to continue to task 2. The dependent variable estimates the probability of continuing to task 2 and is a dummy: it takes value one if a participant chose to re-enter the competition in stage 2, and 0 if the participant chose to continue directly to the questionnaire. Model (1) and model (3) estimate the probability of continuing to task 2 for women (1) and men (3) controlling for the type of priming. From these models, it can be concluded that women are on average 5 percentage points less likely to continue to task 2 of the experiment under gender priming. Next to that, men are on average 6.8 percentage points more likely to continue to task 2 of the experiment under gender priming. As expected, “Gender Priming’’ has a negative sign for women and a positive sign for men. However, both effects are not significant.
Model (2) and (4) additionally control for the gender composition of the reference group (single-sex reference group), for being born 1980 and earlier (old), for relative performance compared to the average score in the group6, for interaction variables and for some
additional variables which are expected to influence the decision to continue to task 2. The effect of gender priming becomes significant (p<0.10) for women after adding the control variables. Women, being primed with their gender, are 33 percentage points less likely to continue to task 2. Therefore, the hypothesis that females are less likely to enter a competitive environment under gender priming is confirmed. For males, the effect for gender priming itself is not significant. When testing for a combined effect of gender priming and being assigned to a single-sex group, males seem to be 33 percentage points more likely to continue to task 2 (p<0.05). This could imply that both being assigned to a single-sex group and being primed with gender have a strengthening effect in triggering males to behave according to male-specific gender norms, such a being competitive and confident.
6 Relative performance task 1 (compared to average) = &'()* ,-./ 0
Table 1. Determinants of continuing to task 2 (average marginal effects probit estimates) Women Men Mean (1) (2) Mean (3) (4) Dependent variable 0.425 (0.50) 0.36 (0.48) Gender priming (Dummy) 0.59 (0.50) -0.05 -0.33* 0.48 (0.50) 0.07 -0.16 Single-sex (dummy) 0.56 (0.50) -0.04 0.46 (0.50) -0.14 Old (dummy) 0.35 (0.48) -0.16 0.40 (0.49) -0.02 Relative performance 112.78 (43.15) -0.05* 113.98 (44.86) -0.0046*** Single-sex * Gender priming 0.31 (0.47) 0.30 0.21 (0.41) 0.33** Single-sex * old 0.19 (0.39) -0.02 0.18 (0.39) -0.15 Old * Gender priming 0.1875 (0.39) 0.37* 0.21 (0.41) 0.12 Relative performance * Single sex 66.51 (68.45) 0.0001 50.89 (63.22) -0.0004 Men do better (dummy) 0.23 (0.42) -0.02 0.33 (0.47) 0.23*** Won from group
(dummy) 0.24 (0.43) 0.19 0.25 (0.44) 0.30* Level of task 3.64 (0.86) -0.11 3.52 (0.85) -0.07 Enjoyment of task 2.74 (1.11) 0.12** 3.05 (1.14) 0.096** Higher Educated 0.76 (0.43) 0.21* 0.81 (0.39) -0.022 N 80 80 80 100 100 100 Pseudo R² 0.0018 0.2734 0.0039 0.2203
Notes. Dependent variable is a dummy indicating the likeliness to continue to task 2 (1 = continue, 0 = not continue). Table presents means and average marginal effects of probit estimates. Gender Priming is a dummy
and takes 1 for being assigned to the gender priming treatment. Single-sex is a dummy and takes 1 for being assigned to a single-sex group. Old is a dummy and takes 1 for being born in 1980 and earlier. Relative performance6 is expressed as percentage. Men do better is a dummy and takes 1 for the belief that men do
better in an abstract reasoning test. Won from group is a dummy and takes 1 if won from the reference group in task 1. Level of task varies takes 1 for very easy and 5 for very hard. Enjoyment of task takes 1 for none at all and 5 for a great deal. Higher Educated is a dummy and takes 1 for owning a tertiary degree. Standard errors in parentheses. Significance: *p<0.01; **<0.05; ***p<0.01
Result 3 Overall, females are not more likely to enter a competitive environment when being assigned to a single-sex group. This effect is not stronger for females born 1980 and earlier.
Result 4. Overall,males are not less likely to enter a competitive environment when being assigned to a single-sex group. This effect is not stronger for males born 1980 and earlier.
From table 1, it can be concluded that generally women are not more likely to continue to task 2 when being assigned to a single-sex group. In addition, it is hypothesized that females born 1980 are even more likely to continue to task 2 when being assigned to a single-sex group due to holding more traditional gender role ideologies. However no significant effect for the interaction effect between “single-sex” and “old” is found. Throughout this paper, it is argued that information about relative performance compared to either a single-sex or mixed-sex group influences the likeliness to enter a competitive environment. Therefore, there is tested for a potential interaction effect between being assigned to a single-sex group and relative performance. However, relatively scoring higher compared to a single-sex group does not result in a higher likeliness to continue to task 2 as compared to relatively scoring higher compared to a mixed-sex group for women.
As expected, for males, the effect of being assigned to a single-sex group is negative, but not significant. Therefore, it can be concluded that men are not sensible to the gender of the co-participants when deciding to enter a competitive environment. This result is remarkable, given the highly significant effect (p<0.01) for holding the belief that men do better in the abstract reasoning test. In addition, is it hypothesized that males born 1980 and earlier are even less likely to continue to stage 2. However, no significant effect for the interaction effect between “single-sex” and “old” is found. Similar to women, no significant effect for an
interaction effect between being assigned to a single-sex group and relative performance is found.
Additional findings
Females (males), who rate their enjoyment in the abstract reasoning test with one additional point, are 12 (9.6) percentage points more likely to continue to task 2. This effect is significant at the 5% level. Furthermore, males who hold the belief that men do better in solving abstract reasoning problems than women, are 23% percentage points more likely to continue to task 2 (p<0.01). Higher educated women are more likely to continue to task 2 (p<0.10). The higher the relative performance, the lower the likeliness to continue to task 2 for both men and women. Furthermore, women born 1980 and earlier and primed with their gender are 37% points more likely to continue to task 2 (p<0.10).
Result 5. Overall, males born 1980 and earlier are more satisfied when comparing their score
to a single-sex group.
Result 6. Overall, females born 1980 and earlier are not less satisfied when comparing their score to a single-sex group.
Table 2 provides summary statistics of the “difference-in-satisfaction score” (∆s), which is defined as ∆s = $#− $". Here, $# is referred to as the level of satisfaction after that the
receiving information about the gender composition and performance of the reference group. $" is referred to as the level of satisfaction before receiving this information. Levels of
satisfaction were measured on a 7-point-scale. The table also summarizes mean performance compared to the average7 and highest score8 of the reference group. Relatives scores are expressed as a percentage, where 100 is the baseline and means that the score was exactly the same as the reference score.
7 Relative performance task 1 (compared to average) = &'()* ,-./ 0
12*)-3* .'()* )*4*)*5'* 3)(67∗ 099%; 8 Relative performance task 1 (compared to highest score) = &'()* ,-./ 0
To test whether people born before 1980 respond differently to male and female performance, an OLS regressions are run. Results are presented in table 3. Models (1) till (4) test for the effect of gender composition on ∆s males born 1980 or earlier. Model (5) till (8) test this for females. Of course, there is controlled for relative performance in all the models, as this is expected to be the main determinant for ∆s. In all models, robust standard errors are used to control for heteroscedasticity. This does not affect the coefficients, but only inflates the standard errors.
Table 2. Summary statistics
Generation Y and Generation Z (born 1981 or later)
Baby Boomers and Generation X (born 1980 or earlier) Mixed-sex group (n=54) Single-sex group (n=58) Mixed-sex group (n=35) Single-sex group (n=33) ∆s = .-− .C 1.38 (0.35) 1.46 (0.41) 0.51 (0.98) 0.67 (1.63) Score task 1 5.09 (1.58) 5.24 (1.84) 5.69 (1.67) 5.45 (1.41) Relative performance task 1
(compared to average) 108.87 (43.81) 113.73 (50.87) 118.04 (40.13) 115.56 (35.51) Relative performance task 1
(compared to highest score)
67.34 (24.40) 76.12 (32.97) 73.72 (23.03) 79.16 (24.38) Notes. Table presents means of ∆s, absolute performance and relative performance by generation and different gender composition conditions (mixed-sex and single-sex). Standard errors are in the parentheses.
For men born 1981 or earlier, performance significantly influences ∆s positively in all models, which is in line with rational behavior. No significant effect on ∆s for being assigned to a single-sex group is found. However, a positive significant effect (p<0.10) for the interaction effect between relative performance and assignment to a single-sex group is found. This implies that the effect of a 1 percentage increase in relative performance on ∆s is greater for males assigned to a single-sex group than males assigned to a mixed-sex group. Therefore, we can confirm the hypothesis that males born 1980 and earlier are more satisfied when comparing their score to a single-sex group.
Table 3. Determinants of ∆s (OLS estimates) for Baby-Boomers and generation X (born 1980 and earlier) Baby-Boomers and generation X
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Assigned to single-sex group (dummy) -1.72 (1.13) -1.81 (1.16) -1.02 (1.30) -2.09 (1.34) Relative performance 1 (compared to average) 0.013** (0.004) 0.008*** (0.003) 0.005 (0.006) 0.002 (0.007) Relative performance 2
(compared to highest score)
0.02*** (0.007) 0.013** 0.013 (0.012) 0.002 (0.01) Single-sex * Relative performance 1 0.019*
(0.01)
0.006 (0.012)
Single-sex * Relative performance 2 0.029*
(0.016) 0.02 (0.020) Constant -1.18*** (0.41) -0.79** (0.37) -1.45*** (0.48) -0.79* (0.43) 0.34 (0.67) 0.83 (0.80) -0.06 (0.85) 0.94 (1.49) N 40 40 40 40 28 28 28 28 R² 0.1453 0.2581 0.1764 0.2712 0.0212 0.0439 0.0584 0.0439 Notes. Dependent variable is ∆s = !"− !$. Table presents coefficients of OLS regressions. “Assigned to single-sex group” is a dummy that takes value 1 for participants
assigned to a single-sex group. Relative performance 1 and 2 are in percentages. Relative performance 1 measures relative performance compared to average score. Relative performance 2 measures relative performance compared to highest score. Interaction variables are added to test for any interaction effects between relative performance and gender composition of reference group. Robust standard errors in parentheses. Significance: *p<0.01; **<0.05; ***p<0.01.
For females, no significant effect for on ∆s for being assigned to a single-sex group is found.
However, regression results seem to be of poor quality. This could be due to the small sample size (n=28). First, R-squared values are low. Second, relative performance does not significantly influence ∆s in any of the 4 models, which does not satisfy rationality. Also, all other coefficients are mostly highly insignificant. Therefore, results from these regressions should be handled with caution.
Result 7 Overall, women born 1980 and earlier do not show a lower willingness to become
better as compared to women born 1981 and later.
Result 8 Overall, men born 1980 and earlier do not show a higher willingness to become
better as compared to men born 1981 and later.
The decision whether or not to re-enter the competition in stage two is both a matter of the willingness to improve the personal score and the willingness to compete. A possible explanation for finding that gender priming (for men and women) and information about co-participants’ gender (for men) does not affect the likeliness to re-enter a competitive environment, could be that the decision to continue to task 2 is driven by the intrinsic motivation to improve the personal score, instead of the willingness to compete for the lottery ticket.
Of the 180 participants, 70 chose to continue to task 2. For these 70 people, the possibility to improve their score (mean: 3.67, SD: 1.08) was more important than the possibility to win an additional lottery ticket (mean: 3.24, SD: 1.02). However, the difference is not significant (p=0.0178).
Table 4 presents OLS estimates of the determinants of the self-reported measure “willingness to improve”. It is hypothesized that women born 1980 and earlier are less eager to become better in the abstract reasoning test as compared to women born 1980 and later, for the reason they are more prone to gender stereotyping. Also, when testing for a combined effect of being assigned to a single-sex group and being born 1980 and earlier, the willingness to improve is not influenced. An interest of this paper is to test whether information about relative performance and about co-participant’s gender can aspire people to improve their
Table 4. Determinants (OLS estimates) ofself-reported measure: “willingness to improve” Men Women (1) (2) (3) (4) Gender Priming (dummy) 0.56 (0.43) 0.018 (0.76) Single sex (dummy) -0.025 (0.004) 0.28 (1.14) 1.07 (0.95) 1.57 (1.16) Old (dummy) 0.43 (0.64) 0.15 (0.76) Relative performance 0.0003 (0.77) 0.011** (0.0046) -0.12 (0.008) -0.0007 (0.007) Gender Priming * Single Sex -0.14 (0.73) -0.69 (0.95)
Old * Single sex 0.13
(0.80)
-0.46 (0.97)
Old * Gender Priming -0.80
(0.70) 0.67 (0.92) Relative performance * Single sex 0.003 (0.007) 0.003 (0.01) -0.11 (0.08) -0.011 (0.007) Men do better (dummy) -0.002 (0.38) -0.21 (0.62) Won from group
(dummy) 1.45 (0.94) -0.13 (0.66) Level of task 0.41** (0.21) 0.39 (0.23) Enjoyment of task 0.19 (0.17) 0.52** (0.19) Higher Educated -0.19 (0.55) -1.19 (0.69) Constant 3.62*** (0.47) 0.59 (1.20) 3.82*** (0.72) 1.69 (1.25) N 36 36 34 34 Prob > F 0.8329 0.0066 0.1848 0.0012 R² 0.0305 0.4076 0.1378 0.5021
Notes. Dependent variable is the self-reported measure “Willingness to improve” and takes 1 for none at all and 5 for a great deal. Table presents coefficients of OLS regressions. Gender Priming is a dummy and takes 1 for being assigned to the gender priming treatment. Single-sex is a dummy and takes 1 for being assigned to a single-sex group. Old is a dummy and takes 1 for being born in 1980 and earlier. Relative performance6 is
score. From model 1, it can be concluded that women do not differ in their willingness to become better when receiving information about their relative performance in a mixed-sex group as compared to receiving information about their relative performance in a single-sex group
For men, it is hypothesized that being born 1980 or earlier leads to a higher willingness to become better, as they are expected to be more prone to gender-stereotyping. No difference in the willingness to become better is found between males born 1980 and earlier and 1981 and later. Also, no significant effect is found for a combined effect of being assigned to a single-sex group and being old. In addition, men do not differ in their willingness to become better when receiving information about their relative performance in a mixed-sex group as compared to receiving information about their relative performance in a single-sex group
One should be cautious about drawing any conclusions, as regression results might be of poor for the reason stated preferences are used when measuring the willingness to improve. A possible limitation of using stated preferences is that they are vulnerable to lack of attention (Buurman et al., 2012).
reasoning test. Won from group is a dummy and takes 1 if won from the reference group in task 1. Level of task varies takes 1 for very easy and 5 for very hard. Enjoyment of task takes 1 for none at all and 5 for a great deal. Higher Educated is a dummy and takes 1 for owning a tertiary degree. Robust standard errors in parentheses. Significance: *p<0.01; **<0.05; ***p<0.01.
6 – DISCUSSION AND CONCLUSION
6.1 – SUMMARY OF FINDINGS
The main interest of this study is to test whether (1) social identity norms and the concept of gender role ideologies are responsible for declaring gender differences in attitudes towards competition, to test whether (2) information about relative performance and about co-participant’s gender influences preferences towards competition and to test whether (3) information about relative performance and about co-participant’s gender can aspire people to improve their score.
First, it is tested whether socially constructed gender norms and the concept of gender role ideologies influence the likeliness to enter a competitive environment for both men and women. It is found that women generally are less likely to enter a competitive environment under gender priming. For men, no effect for gender priming itself is found. However, when looking at the combined effect for being assigned to a single-sex group and gender priming, males are significantly more likely to opt for the competitive environment. From above results, it can be concluded that both men and women are influenced by generally held gender-stereotyping ideas when deciding to enter a competitive environment.
Second, it is tested whether information about relative performance and about co-participants’ gender influences preferences towards competition for both men and women. The hypothesis that men and women generally prefer to compete against other women in not confirmed. Throughout the paper, it is argued that participants born 1980 and earlier (Baby-Boomers and generation X) hold more traditional gender role ideologies than participants born 1981 and later (generation Y and Z) and therefore more likely to adopt the idea that female performance is inferior to male performance. For them, it is argued that the willingness to compete against other women is even stronger. However, neither for women nor for men born 1980 or earlier the likeliness to compete against other women seems to be stronger. To complement the above finding it is tested whether people born before 1980 (generation X and Baby-Boomers) respond differently to male and female performance. It is hypothesized that both men and women born 1980 and earlier are more responsive when
