Gender differences in competitive
environments
The impact of social status and stress
Thesis MSc. Business Economic – Managerial Economics and strategy 20 ECTS
July 2018
Barbara Schram – 11863218
University supervisor: Thomas Buser Facility of Economics and Business
2 Statement of Originality
This document is written by Barbara Schram 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.
3
Gender differences in competitive
environments
The impact of social status and stress
Abstract – A possible cause for the gender gap in top management positions are differences between men and women in their behaviour inside a competitive environment. Competition can be classified into two dimensions: the rivalry for
resources and the social status ranking. This study analyses the social status ranking
dimension of competition, which is a competitive environment where performance is ranked according to relative performance. The best performance receives a high rank and this yields in social status. In addition, people experience competition as stressful. An unexplored and extra explanation for the gender gap on the labour market may result from gender differences in stress. This study examines what gender differences exist along the social status ranking dimension of competition. Furthermore, the relationship between the self-reported stress level and performance is analysed. A laboratory experiment is conducted in a high school (N=259)which assesses behaviour differences in a controlled competitive setting. The data show no treatment effect of social status
ranking on participant’s performance. Neither men nor women experience an extra
stress level from the competitive setting compared to the baseline. The analyses show a positive correlation between stress and performance for women. Therefore, the results suggest that social status ranking of competition has no influence on the performance of high school students but stress can explain performance differences of female participants.
4 1. Introduction
Despite major improvements in women’s education level, high-ranking positions in the labour market remain largely favourable to men (Cook and Glass, 2014). In 2014 only 16% of the top leadership positions in The Fortune 500 was represented by women (Cook and Glass, 2014). A possible cause is discrimination of women in the labour market, which leads to different treatment of men and women with equal preferences and abilities (Anderson, et al. 2013). Another cause are differences between men and women regarding their competitive preferences and their behaviour inside a competitive environment. According to research of Niederle and Vestelund (2007), men are more competitively inclined than women. Their research shows that gender differences exist in willingness to enter a competitive environment. Men are more likely than women to self-select into an environment that involves competition. In addition, women may behave differently than men inside a competitive environment. As Gneezy, Niederle and Rustichini (2003) indicate, women are less effective than men in a competitive environment, even when women are able to perform similarly as men in a non-competitive environment. Hence, men outperform women in a non-competitive environment. This might reduce the possibility of success for women if they compete for new jobs or promotions.
Competitive environments are experienced by many people as stressful. The amount of perceived stress can be measured by the stress hormone cortisol in the human body (Dickerson and Kemeny, 2004). Different studies show that competitive environments increases the cortisol level of participants compared to non-competitive environments (Buser et al., 2016, Zhong et al., 2015, Buckert et al., 2015). Moreover, Bateup et al. (2002) indicate that women experience a greater increase in the stress-related hormone cortisol during competition compared to men. This implies that gender differences in stress may give an extra explanation about the gender gap on the labour market. Examining the relationship between stress and performance of men and women in a controlled competitive setting could provide new insight into the literature about gender differences in competitive environments.
Competition can be classified in two different dimensions, namely rivalry for resources and social status ranking. The rivalry for resources dimension is a competitive environment where the best performers obtain a prize, for example in a tournament.
5
Croson and Gneezy (2009) provided an overview of the existing literature about gender differences in a competitive environment. This literature focuses on one dimension of competition, the rivalry for resources dimension. However, a recent study by Schram et al. (2017) investigates the other dimension of competition, defined by social status
ranking. This is defined as follows: participant’s performance in a competitive
environment results in a ranking according to relative performance (Ball et al., 2001). The best performers get a high ranking and this yields in social status for these participants (Ball et al., 2001). By examining the differential gender impact of social
status ranking, a new insight into the literature about competition is explored. Schram
et al. (2017) published the first study in economics that isolates the effect of social status
ranking from the rivalry for resources dimension and assesses if there exists gender
differences in the social status ranking dimension of competition. This research shows remarkable performance differences. Men outperform women with the social status
ranking treatment, whereas there are no gender differences in performance without social status ranking. Therefore, the gender gap exists along a relatively unexplored
dimension of competition as well. The question arises which underlying mechanism can explain this change in performance with socials status ranking.
Firstly, this paper aims to give in-depth insight into the unexplored dimension of competition, social status ranking. The existence of performance differences between men and women are examined as a result of social status ranking. The research of Schram et al., (2017) is assessed on robustness, by examining a different population and a different task in the experimental design. Secondly, this paper focuses on the self-reported stress level of the participants. The influence of a competitive setting on stressful reactions is assessed. Thirdly, this study examines the correlation between stress and participant’s performance. The following two research questions are answered in this study: What are the gender differences in performance in the social
status ranking dimension of competition? Is there a relationship between stress and performance in a competitive setting with social status ranking?
This study conducted a controlled experiment in two high schools to examine the performance differences between a non-competitive environment in contrast to a competitive environment which results in a social status ranking. Performance was measured using a real-effort task (IQ-test) and the population consisted of high school
6
students aging from 14 to 18 years old. In addition, this study measured the self-reported stress level of the participants. The influence of the competitive environment on the stress level of the participants is explored. This study shows that men and women perform on average equally in the non-competitive baseline. By comparing the competitive treatment to the baseline, no behavioural differences were found. Consequently, it can be stated that no significant gender differences exist in the social status dimension of competition in high school. Furthermore, the stress level of participants was on a higher level after the IQ-test. This effect was present for both men and women. Accordingly, no significant gender differences in participant’s stress level were present in the competitive setting compared to the baseline. Lastly, a positive relationship between self-reported stress and performance was found in this study for female but not for male participants.
This paper is structured as follows: section 2 focuses on the literature review and places this research in its literary context. In section 3 the study’s methods are explained. Based on the literature and the methods, the theoretical predictions are formulated in section 4. Section 5 presents the results and discusses the empirical analyses. In section 6, this paper concludes with limitations as well as opportunities for further research.
2. Related Literature
This section starts with a discussion of the relevant literature on gender differences in a competitive environment. Secondly, the literature about relative performance and the
social status ranking dimension of competition is reviewed. The last section discusses
the relevant literature about stress and competition.
2.1 Competitiveness and gender differences
Part of the gender gap in high profile job may result from different tendencies of men and women to compete. Niederle and Vestelund (2007) examined in a controlled laboratory setting, if men and women with the same abilities differ in their preferences to be involved in a competitive setting. Men select the competitive environment twice as often as women, while performance of men and women is equal. This implies that women are averse to competition and men are more eager to competition. Consequently, the number of women who are active in lucrative jobs is limited. Balafoutas et al. (2012) confirm these findings when assessing the relationship between
7
distributional preferences and competitive behaviour. Distributional preference is the appearance that decision makers have a concern for the welfare of others (Balafoutas et al., 2012). A person’s well-being and behaviour does not only depend on their own payoff but also on the payoff of others. Men and women significantly differ in distributional preferences, overconfidence, and risk attitude (Croson and Gneezy, 2009; Barber and Odea, 2001; Adreoni and Vesterlund, 2001). Selecting a competitive environment is correlated with these factors. Hence, the gender gap in competitive environments is driven by traits that correlate with gender.
Another explanation for the gender gap in the labour market is that women behave differently than men inside a competitive setting. Gneezy, Niederdele, and Rustichini (2003) demonstrated this difference in behaviour using single-sex and mixed-sex tournaments. A competitive environment with mixed gender groups results in a significant increase in performance of male participants, but not of female participants (Gneezy et al. 2003). This suggests that men are more effective compared to women in a competitive environment. Moreover, the study shows that women increase performance in a competitive single-sex environment compared to the non-competitive setting. This implies that women are less effective in a competitive environment where men are involved. Moreover, Gneezy & Rustichini (2004) considered the performance of 10-year-olds in competitive and non-competitive environments in their research. A gender difference is identified by the comparison of performance in competitive and non-competitive environments. Performance in a competitive setting increases to a higher level for men compared to women. Hence, gender differences in behaviour in a competitive setting appear already at a young age.
Experimental studies that show differences between men and women in competitiveness can be linked to actual labour market outcomes. Delfgaauw et al. (2013) examined the gender gap in a competitive environment with a field experiment in a Dutch retail chain. They showed that competition positively affects performance if a sufficient percentage of employees have the same gender. This confirms the findings of Gneezy et al. (2003) that competition increases performance for both men and women in a single-sex competitive environment. Flory et al. (2010) examined how men and women are influenced by competition and uncertainty, using a natural field experiment on job-entry decision in sixteen major US cities. The results indicate that competitive organizations
8
significantly increase the gender gap in application probabilities. Women’s tendency to apply for these jobs is substantially lower than for men (Flory et al., 2010). Surprisingly, this study indicates that both men and women are in general averse to competitive environments, although women dislike competitive environments more than men. The studies indicate that gender differences in competitive environments are present in the labour market.
2.2 Explanations for gender differences in competitive environments
According to Croson, and Gneezy, (2009), gender differences in competition are caused by both nature and nurture aspects. On the one hand, Anderson et al. (2013) provide an in-depth view on the effect of nurture on competitiveness by conducting a field experiment with children. This research examined the effect of society on competitive behaviour by comparing children in a matrilineal society and children from an patriarchal society. While no gender differences in competitiveness exist in a matrilineal society, girls become less competitive around puberty in the patriarchal society (Anderson, et al 2010). This implies that the societal structure (nurture) is linked to the observed gender gap in competitive environments. Part of every culture in the world is to raise boys and girls in a different way (Crosos, et al. 2004). On the other hand, nature also plays a role in explaining gender differences in competitiveness. Colarelli (2006) indicates that psychological mechanisms influence competitiveness. There exists a relationship between testosterone and the tendency to be involved in competitive activities (Dabbs, 2000) Testosterone influences the tendency to compete and the intensity of competitive behaviour (Colarelli et al., 2006). On average, men produce five to seven times more testosterone (Bateup, 2002). This indicates that the gender gap in top management positions arises because of the greater importance of competition for men.
An economic explanation for the gender gap in competition is that men are generally more confident than women in expectations of success (Gneezy, Niederdele, and Rustichini, 2003). All people are overconfident with respect to their own ability, but men are more overconfident than women (Barber and Odean, 2001). Men stress the importance of their own ability when they are successful (Beyer, 1990). In contrast, women attribute success more externally or more to effort. Women do not take credit for good performances since they believe success is a result of external factors (Beyer,
9
1990). This results in men having higher expectations about their own success. Confidence in one’s own ability seems strongly related to performance (Chermers, Hu, and Garcia 2001). Hence, the feeling of lower confidence in women plays a role in explaining the gender gap in performance in a competitive environment (Gneezy, Niederdele, and Rustichini, 2003).
2.3 Social status and relative performance
Economic decisions depends on social interests and influences (Weiss and Fershtman, 1998). Recent literature (Ryan & Deci, 2001; Luttmer, 2005) shows that people care about the well-being of other people and about their own relative well-being. For instance, additional wage may not increase someone’s well-being if those in the comparison group achieve the same or a larger increase in wage (Dolan et al., 2008). Weiss and Fershtman (1998) described the role of social status in economic decisions. Social status is the ranking of individuals in a given society or group, based on their attributes, equity and actions (Weiss and Fershtman, 1998). This study reviews the literature in psychology and economics to describe the role of social status in economic analysis, and demonstrates that preferences for social status affect economic decisions in our society (Weiss and Freshtman, 1998). However, it is difficult to determine the actual effect of social status in modern society. Different studies show that sensibility towards status differ across gender. Men have a higher interest in social status than women (Mujcic and Frijters 2013, Carlsson et al. 2009), however some studies (Johansson-Stenman et al. 2002; Alpizar et al. 2005) show a opposite effect. Schram et al. (2017) explored the gender effect of social status ranking inside a controlled competitive environment. In this study, the rivalry for resources effect was isolated from the social status ranking dimension. The study shows that social status ranking results in performance differences between men and women. The study shows a positive effect of social status on performance of men, although no effect on performance of women is identified. This implies that men outperform women in a competitive environment when social status is involved.
An environment that includes information about relative performance is a positive stimulant to increase participants achievements (Azmat and Iriberri, 2010). Students on a high school received information for one year that informed them about their performance compared to the class average and their deviation from the average (Azamt
10
and Iriberri, 2010). The study shows that presenting individual information about relative performance increases the grades of students with 5 percent. The effect was significant for the whole population. This indicates that individual information of relative performance is a trigger for participants to increase their performance, which applies for both genders. According to Schram et al. (2017), when the relative performance becomes public information, there will be a social status ranking and a gender gap in performance will appear. Hence, information about relative performance will result in an increase of performance for both genders, although the information should be kept private (Schram et al. 2017, Azmat and Iriberri, 2010).
2.4 Stress and competition
Multiple studies explore the relationship between stress and competition. Zhong et al., (2015) observe higher response of the stress-related hormone cortisol for participants in the tournament compared to the non-competitive treatment. Competitive inclined participants experience higher stress response than less competitive participants in both the tournament and the non-competitive environment (Zhong et al., 2015). This study identifies a relationship between competition and stress responses. In addition, Buckert et al., (2015) examine whether competitive environments induce stress reactions. This study shows that compared to the control group, the competitive setting raises reactions of mild stress. According to these studies, a competitive environment results in a higher stress level of participants. Bateup et al. (2002) indicate that a competitive setting has a greater effect on the stress level of women than it does on that of men. Different studies indicate that evolutionary roots cause women to experience more stress than men during competition (Campbell, 1999; Tayler et al., 2002). This effect emerges in women to avoid risky actions that might have a negative effect on the health of their children. In addition, the study of Goette (2015) indicates that stress is a major regulator of individual competitiveness because stress affects self-confidence. Low-anxiety individuals become more self-assured from stress, although stress has the opposite effect on high-anxiety individuals (Goette, 2015). In addition, the study assesses the effects of gender on these results and indicates that the stress-effect on confidence is still significant. This indicates that males and females respond different to stress conditions (Goette, 2015).
11
Buser, Dreber, and Mollerstrom (2016) explored the relationship between stress and gender differences in competitive environments. The study assesses if there exists a relationship between willingness to be involved in competitive and stressful environments. The study shows that competing (in a tournament) increases stress. This supports the research of Buckert al al, (2015) and Zhong et al. (2015) discussed above. In addition, a causal effect of stress on entering a competitive environment appears for women, but not for men. This indicates that the level of stress has an impact on entering a competitive setting for women. Women with a high increase in cortisol are more likely to enter the competition and the other way around. However, the study indicates that stress reactions cannot explain the gender gap in willingness to be involved in a competitive environment, because cortisol reaction has no impact on the overall gender gap (Buser et al. 2016). Accoring to Shurchlov (2012), stress levels of competitive jobs are related to the gender gap in a competitive environment. This study indicates that stress levels are correlated with women’s performance and willingness to compete in a competitive environment. While women, compared to men, underperform in stressful competitive environments, women greatly increase their performance and their willingness to compete in low-pressure settings. In addition. the study investigates whether this relationship is present in the labour market. Gender differences exist for highly competitive jobs and the gender gap decreases for relatively less competitive jobs (Shurchlov 2012).
3 Methodology
For this research, a laboratory experiment was applied to a population of high school students. A similar experimental design as Schram, et al., (2017) was used in which it is possible to hold the rivalry of resources dimension constant and to vary the social
status ranking dimension. The extra element in this research compared to Schram et al.
(2017), is the measure of self-rated stress level of the participants. This was measured by using a questionnaire to identify a possible correlation between performance and stress.
Figure 1 represents a schematic view of the phases. All participants performed the same task and answered the same questionnaires. This section explains the phases of the experiment in more detail, as well as the experiment’s participants and the process.
12
Figure 1: Phases of the experimental design
3.1 Participants
The participants in this experiment were high school students from two different high schools in the Netherlands, namely ‘KSG Apeldoorn’ in Apeldoorn and ‘Staring’ in Lochem. The participants are VWO3, VWO4, VWO5, and HAVO4 students aging from 14 to 18 years old. These students are considered high educated children of Dutch society. Before the start of the experiment, the principals of both schools agreed to conduct the experiment. Likewise, all the mentors were informed. The experiment included 12 sessions in which 20 to 30 students were participating.
In high school, students receive grades which are in a sense a ranking of their relative performance. This ranking of relative performance may result in social status for students, which could affect performance. Therefore, conducting the research with this population may give an interesting insight into the effects of private and public information about relative performance of students.
3.2 Task
Performance was measured by using a real effort task. It consisted of 20 IQ-questions divided into four categories; analogies, spatial insight, digit series and figure series. The test is available in the appendix. This particular task was used for two different reasons. Firstly, as discussed in the literature review, Schram et al. (2017) show a gender gap in performance with social status ranking treatment, and the current study assessed whether this gender gap appears using a different task. Secondly, in economic experiments, it is common to incentivize the participants with money. However, in this
Phase Part Goal
(0) Instructions Explain the goal of the experiment
(I) Questionnaire Measure the self-rated stress level, baseline
(II) Treatment Receive an envelope with the treatment
(III) IQ-test (task) Measure the performance on the task (IV) Questionnaire Measure the self-rated stress level, treatment (V) Score category Receiving a note with score an score category (VI) Social status ranking Reading aloud score category
13
research students were not paid according to their performance. The experiment did, however, require a task that triggers the student’s attention. In general, people care about their IQ and how this is perceived by other people. Therefore, indicating in the instructions that the task was an IQ-test would stimulate the participants. The participants had 12 minutes to complete the task and the elapsed minutes were indicated to create extra stress for the participants. The scoring is as follows:
Correct answer: 10 points Unanswered question: 0 points
Wrong answer: -5 points
This way of scoring was set up to make the participants carefully think about how to answer the questions. This is used later in the analyses of the results to have the opportunity to measure performance of the participants in three different ways.
3.3 Score category
At the end of the session, every participant received a note with their score and their score category. Score categories were divided into four categories, as indicated in Table 1. Score categories were used instead of a real ranking (as in the design of Schram et al. 2017) in this experiment for two reasons. The first is a practical reason: on average a class consists of 25 students. The experiment was conducted with pen and paper and every test was checked manually in the same session (50 minutes). Correcting the test is time-consuming and no time was left during the session to rank out the whole class. Therefore, notes with score categories were prepared upfront and every participant was divided into a score category. Second, the students are aged 14 to 18 years old. Reading aloud the score of the IQ test to the whole class could make them feel uncomfortable. The score category is used to see the effect of social status ranking (explained in the next section), without reading aloud their exact score.
14 Table 1 Score category (points)
< 50 Points 50-100 Points 100-150 Points 150 > points
Notes The table indicates the experiment’s score categories. Participants can achieve a maximum of 200
points. The scores are equally divided into 4 categories. The participants receive a ‘score note’ which indicate their exact score and score category.
3.4 Treatment
The design consists of the baseline and one treatment. All participants were split in A players and B players. The A players were in the baseline and the B players were in the treatment. All players answered the same questionnaires, performed the same task and received a note with their score and score category from the IQ test (score note). The treatments differed in whether the score category is public or private information.
Baseline - The score and score category is private information. After receiving their
score note, the experiment ends for these players. Their score category is private information.
Status ranking treatment - The participants have to reveal their score category to the
other players of their session. Players have to read aloud their score category, one by one for all the participants of their session. Their score category becomes public information.
Every player in the treatment group read the same sentence aloud to create extra control to the experiment. Every participant in the treatment group had to pronounce the following sentence (translated to English):
15 3.5 Randomization
The randomization of this experiment was on an individual basis. The participants of every session randomly received an envelope with a player type, denoted by A or B, after the instructions. Inside the envelope, the details of the baseline/treatment was written down. Thereafter they opened the envelop and read whether their score category was public or private information. The participants were aware of the treatment before the start of the task. However, they were not aware of the difference between the A players and the B players.
3.6 Questionnaire
A short questionnaire was answered by the participants before the treatment was announced and before the start of the IQ-test. The questions are shown in Table 2. Since the focus of this research is on the participant’s performance of the task, the experiment required a short and not time or non-energy consuming questionnaire. The questionnaire consists of 9 questions to get an indication about the current state of mind of the participant.
Table 2 Questionnaire
How enthusiastic do you feel? How stressed do you feel? How motivated do you feel? How nervous do you feel? How confident do you feel? How happy do you feel? How irritated do you feel?
How eager do you like to be successful?
How worried are you about what other people think of you?
Notes The questionnaire consists of nine questions about the participant’s stress level and current state.
The participants answer this questionnaire two times to examine the treatment effect on a participant’s stress level
The questionnaire measures the self-reported stress levels of the participants on a scale of 1-7. The first questionnaire was used as a baseline and is answered before the participants knew the treatment. The same questionnaire was answered after completing
16
the task. The second questionnaire indicated the effect of social status ranking on the stress level (and other feelings) of the participant. Moreover, participants had to fill in their age and gender.
3.7 Pilot Experiment
A pilot session was organised with 10 participants at the University of Amsterdam, before running the experiment’s sessions at the high schools. After the pilot experiment, some in-depth questions to the pilot participants revealed what had to be adapted. Some questions that were perceived as too difficult/easy/unclear were adjusted. Subsequently, extra explanations and instructions were added to the IQ-test to avoid misunderstanding for participants. Furthermore, the questionnaire was adjusted. Extra stress-related questions, for example about nervousness and secureness, were included. This ensured that the questionnaire measures the stress level as well as the current state of the participants.
4 Theoretical predictions
The next section applies the theory discussed in section 2 to formulate the hypotheses of the experimental design. These hypotheses were tested in the results.
4.1 Men are more competitive inclined than women
The experiment’s task is an IQ-test that consists of a broad variation of questions (4 different categories). The task includes easy and difficult questions for every participant -i.e., some participants excel in spatial insight, other participants excel in analogies. The prediction is that the task is gender neutral and no stereotype treat could be identified. Without competition (in the baseline), no significant gender difference in performance is expected. The first hypothesis is formulated below:
Hypothesis 1. No gender differences in performance in the baseline can be identified
As discussed in section 2, literature shows that men embrace competition and perform better in a competitive environment (Gneezy, Niederdele, and Rustichini, 2003; Schram et al., 2017). Therefore, theory predicts that men will be positively stimulated by the competitive social status ranking treatment. In contrast, women are less effective in a
17
competitive environment which includes men. Theory expects social status ranking to have a negative effect on the performance of women. The study of Schram et al. (2017) shows the same results: with the social status ranking treatment, men outperform women. The second and third hypotheses are formulated:
Hypothesis 2. The average performance of men increases with the social status ranking
treatment compared to the baseline
Hypothesis 3. The average performance of women decreases with social status ranking
treatment compared to the baseline
4.2 Competition increases stress
A competitive environment increases the stress level of participants compared to a non-competitive environment (Buser et al., 2016). This study expects the non-competitive treatment (social status ranking) to be more stressful than the non-competitive baseline. It is supposed to be similar for men and women. However, the research of Bateup, (2002) shows that women, compared to men, experience a higher increase of the stress hormone cortisol in a competitive setting. It is expected that women experience a larger increase in their self-reported stress level arising from the social status ranking treatment compared to men. The fourth and fifth hypotheses are formulated below:
Hypothesis 4. The average stress level is higher in the social status ranking treatment
compared to the baseline
Hypothesis 5. Social status ranking treatment has a relative higher effect on the stress
level of women compared to men
The next part focuses on the causal relationship between stress and performance and whether the effect is different for men compared to women. Hanif et al., (2011) studied the effect of teacher stress and job performance. The findings reveal that there exists a negative relationship between teacher stress and job performance. This is supported by the research of Jamal (1984) who examined job stress and job performance among nurses. The empirical analysis of this study shows a negative linear relationship between
18
stress and job performance. This analysis investigates the influence of stress on participant’s performance. The sixth hypothesis is formulated as follows:
Hypothesis 6. Stress is negatively correlated with performance
5 Results
The presentation of the results focuses on gender differences in performance. In addition, the stress level and its relationship to performance is determined. Section 5.1 starts with the summary statistics of performance and stress. In sections 5.2 to 5.7 the results of the study are analysed and the hypotheses tested.
5.1 Summary statistics
This section starts with testing the experimental population and examining the mean variables between the baseline and the treatment. The summary statistics are classified into two parts. The first part analyses the data of participant’s performance, the second assesses the stress level of the participants.
5.1.1 Summary statistic randomization
Table 3 shows the descriptive statistics of the randomization of the experiment. Across the treatment and the baseline the number of observations are similar. The number of observations is not different for males and females. Moreover, there are no (gender) differences in the average age between the baseline and the treatment. This confirms the randomization of the experiment across treatments and gender.
Table 3Summary statistics, randomization
Baseline Treatment
Number of observations 131 128
Number of observation, male 63 56
Number of observations, female 68 72
Mean age male 15.9 15.6
Mean age female 15.6 15.6
Notes The table list the number of observations in the treatment compared to the baseline. In addition,
the number of observations per treatment are listed by gender. The mean age of the participants is also indicated.
19 5.1.2 Summary statistic performance
Table 4 illustrates the summary statistics of performance. It is measured in three ways: scores, mistakes, and unanswered questions (left open). The mean score of males and females is displayed in Table 4. The p-values in the last column are calculated with a t-test. No significant difference in performance between baseline and treatment was observed.
Table 4Summary statistics, performance
Baseline Treatment p-values
Mean score male 116.19 113.55 0.68
(4.27) (4.65)
Mean score female 115.42 116.69 0.81
(3.85) (3.75)
Mean mistakes male 3.28 3.44 0.73
(0.29) (0.32)
Mean mistakes female 3.04 3.18 0.74
(0.30) (0.27)
Mean left open male 3.43 3.33 0.81
(0.30) (0.33)
Mean left open female 3.91 3.62 0.46
(0.29) (0.27)
Notes The table indicates the mean values of performance of the participants. The p-values are
calculated with a t-test. The means are listed across gender and the baseline is compared with the treatment. In addition, performance is measured in three ways, in terms of score, number of mistakes and unanswered questions. The critical p-value is smaller than 0.05.
For instance, the mean score for a male was not significantly different in the treatment compared to the baseline (p-value is: 0.68). Performance was also measured in terms of number of mistakes and number of unanswered questions (left open). The p-values of Table 4 show the same result: no significant differences between the baseline and the treatment in the number of mistakes or unanswered questions were observed.
5.1.3 Summary statistic stress
Table 5 shows the summary statistics of participant’s self-reported stress level. As explained in the methods, participants answered the questionnaire before they learn their treatment and before the start of the task. This first questionnaire was used as a baseline
20
stress level and is named ‘stress before’. The stress level was measured again after performing the task, which is labeled as ‘stress after’. The p-values displayed in Table 5 were calculated with a t-test and show that participant’s stress level was not significantly different in the treatment compared to the baseline.
Table 5 Summary statistics, stress level
Baseline Treatment p-values
Mean stress before 3.85 3.98 0.48
(0.13) (0.12)
Mean stress after 4.87 4.94 0.64
(0.12) (0.11)
Mean stress difference -1.02 -0.98 0.85
(0.15) (0.14)
Mean stress before, male 4.15 4.20 0.83
(0.17) (0.19)
Mean stress before, female 3.58 3.80 0.34
(0.18) (0.14)
Mean stress after, male 4.90 4.92 0.58
(0.16) (0.18)
Mean stress after, female 4.83 4.96 0.58
(0.17) (0.14)
Mean stress difference, male -0.76 -0.74 0.95
(0.17) (0.21)
Mean stress difference, female -1.25 -1.15 0.74
(0.23) (0.19)
Notes The table shows the mean value of stress across gender and in the baseline compared to the
treatment. The p-values are from a t-test. Self-rated stress is measured using a questionnaire in the scale 1-7. Between parentheses are the standard errors.
5.2 Results
This section discusses the results and examine if the hypotheses formulated in section 4 were supported by the data of the experiment. The first part of the section focuses on the performance on the IQ-test in the treatment (social status ranking) compared to the baseline. The second part analyzes the self-rated stress level of the participants in the questionnaire.
21
5.2.1 What is the effect of social status ranking on performance?
Figure 2 illustrates the mean performance of the participants in the baseline compared to the social status ranking treatment. Performance is measured in this analysis by the mean score of the IQ test. The maximum score of the test is 200 points. The bars in Figure 2 are divided into male and female and the error bars observed in the figure are the 95% confidence intervals.
Figure 2: Performance (task)
The figure shows no gender differences in mean performance (score) in the baseline. Therefore, hypothesis 1 (No gender differences in performance in the baseline) is supported. The data show no gender differences in mean score in the con-competitive case (baseline). This implies that the task is gender neutral and no stereotype treat could be identified.
Result 1: In the non-competitive treatment (baseline), no significant gender differences in performance could be identified.
The mean performance (score on the IQ-test) was analysed to measure if a significant treatment effect in performance could be identified between the competitive treatment (social status ranking) and the baseline. Figure 2 illustrates additional results that are
116,2 115,4 113,5 116,7 0 20 40 60 80 100 120 140
Score baseline Score treatment
S
core
male female
22
identified: the social status ranking has no influence on the average performance for either male or female in this experiment. No differences were identified between the mean score in the treatment group and the baseline group.
This result is assessed in more detail by the regression output in Table 6. The social status ranking treatment and gender are regressed as an interaction variable. The following OLS-regression was set up to estimate the average treatment effect:
𝑦 = 𝛽0 + 𝛽1 𝑅𝐸𝐴𝐷 + 𝛽2𝐺𝐸𝑁𝐷𝐸𝑅 + 𝛽3 𝑅𝐸𝐴𝐷 ∗ 𝐺𝐸𝑁𝐷𝐸𝑅 (1)
The variable GENDER is a binary variable which equals 1 if the participant is male and 0 if the participant is female. The variable ‘READ’ is also a binary and indicates the treatment of the experiment. ‘READ’ equals 1 if participants are in the treatment and equals 0 if participants are in the baseline. The results of the regression are shown in Table 6.
Table 6OLS- Regression table performance
SCORE MISTAKES LEFT OPEN
(1) (2) (3) READ 1.26 0.14 -0.29 (5.38) (0.41) (0.40) GENDER 0.76 0.24 -0.48 (5.74) (0.42) (0.41) READ*GENDER -3.91 0.01 0.19 (8.29) (0.59) (0.59) Constant 115.42 3.04 3.91 (3.84) (0.30) (0.29)
Notes The participant is a male if GENDER=1 and female if GENDER=0. READ is the treatment
variable and READ=1 in the treatment and READ=0 in the baseline. The standard deviations can be observed in the parentheses. The columns are the dependent variables and the rows are the independent variables and the control variables. The critical value of the t-statistic is 1.96.
Table 6 shows the effect of social status ranking (READ) on average performance (SCORE) across males and females. The independent variable (y) is SCORE and is included in the first column in the table. The average score for a male in the treatment is for instance: READ + GENDER + READ*GENDER+ CONSTANT = 113,54. The
23
table indicates that the mean score for males is lower in the social status ranking treatment compared to the mean score for females. However, the interaction variable is not statistically significant (t-statistic is 0,47). The regression output shows the same result as in Figure 2. No significant treatment effect of social status ranking on average performance was identified for either males or females.
Result 2: Social status ranking has no significant effect on average performance (score)
for either male or female.
To summarize, participants show no difference in behaviour in the treatment compared to the baseline. This is the case for both males and females. Therefore, hypotheses 2 and 3 (the performance of men increases with the social status ranking treatment compared
to the baseline; the performance of women decreases with social status ranking treatment compared to the baseline) are not supported by the analyses.
5.3 Is there a gender gap in the number of mistakes or unanswered question?
Figure 3 illustrates the mean number of mistakes and unanswered questions in the IQ-test to investigate the treatment effect on performance in more detail. The black bars represent the baseline and the grey bars represent the treatment. The error bars in the figure are 95% confidence interval. The data are divided into male and female to investigate the gender gap in performance.
Figure 3 illustrates a decrease in the number of mistakes between the baseline and the treatment for males and females. The mean number of mistakes of male participants decreased with 9,1% in the social status ranking treatment compared to the baseline.
24
Figure 3: performance participants
This result is assessed in Table 6 with OLS-regression (1) and the number of ‘MISTAKES’ as the dependent variable (y). The interaction variable was insignificant which indicates that the treatment does not have a significant effect on the number of mistakes for both males and females. The analyses show no treatment effect on participant’s performance in number of mistakes.
Result 3: Social status ranking has no significant effect on participant’s performance
in number of mistakes.
Figure 3 demonstrates an increase in the average number of unanswered questions (LEFT OPEN) between the baseline and the social status treatment for both males and females. This indicates that participants have more unanswered questions on the IQ-test in the treatment compared to the baseline. Table 6 shows the unanswered (LEFT OPEN) questions as the dependent variable (y) in OLS-regression (1). The interaction variable of the number of left open questions is insignificant. Therefore, neither males nor females were more careful with filling in their questions in the treatment compared to the baseline. 3,3 3,4 3,0 3,9 3,4 3,3 3,2 3,6 0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5
male mistakes male leftopen female mistakes female leftopen
Que
sti
ons
25
Result 4: Social status ranking has no relationship on performance in terms of number
of unanswered questions.
In summary, the analyses reveal no treatment effect for either male or female on performance. Hypotheses 2 and 3 (the performance of men increases with the social
status ranking treatment compared to the baseline; the performance of women decreases with social status ranking treatment compared to the baseline) are not
supported by the data, no significant gender difference in performance in the social status ranking treatment compared to the baseline was identified. Possible explanations are discussed in detail in the upcoming subsection 5.4.
5.4 Why do the results show no significant treatment effect of social status ranking on performance?
The results of sections 5.2 and 5.3 indicate no gender differences in performance in high school with the social status ranking dimension of competition. This study does not support the findings of Schram et al. (2017). Possible explanations for the deviation of hypothesis 2 and 3 are discussed in the next section. This section examines the results of sections 5.2 and 5.3 and discusses possible explanations for the data not supporting these hypotheses. Based on the topics introduced in the literature review, the discussion is divided into three different categories: environment, population, and nurture.
5.4.1 Performance on school
In the last decade, a trend in education is identified in which women on average perform better than boys at schools (Duckworth & Seligman, 2006). As discussed in the related literature, women have lower expectations compared to men about their own abilities. However, school might be an environment where girls are confident about their (relative) performance. Confidence in your own ability is strongly related to educational performance (Chermers, Hu, and Garcia 2001). It is possible that women have the same or higher expectations about their success compared to men in this setting. This implies that females in high school may not be averse to competitive environments.
26 5.4.2 Population
Another possibility for the absence of results supporting the hypotheses is that ranking relative performance is common in high school. As discussed above, students receive many grades in school and this is in a sense a ranking of their relative performance. Students might feel comfortable to make a test and receive a relative rank. Consequently, the students do not care about revealing the relative performance to other participants in their session.
Another explanation is that participants know the other players of their session and already have an indication of the relative ranking before the start of the experiment. Participants might already have expectations about the performance of themselves compared to the group before they start the IQ-test. Consequently, participants will not be (de)motivated by the treatment, since an expectation about the outcome of the result is already existent.
The research is conducted within a population group with an average age of 16 years old. It might be the case that sensitivity for social status grows by age. The sensitivity of social status seems to change with age (Salmivalli, et al., 1996). However, related literature shows that competitive differences between men and women appear already at a young age. Children become aware that low status is a negative position among peers if they become older. This implies that the absence of effects of social status could be a characteristic of high school attendees.
5.4.3 Nurture on school
High school is an environment in which both boys and girls are stimulated to perform well. As discussed in the related literature, a possible explanation for the gender gap is that boys are raised different than girls i.e., the prototype is that girls play with dolls, boys play soccer games that include a competitive element. In the past decade, high schools have been environments where both boys and girls are stimulated to learn. This implies that a possible explanation for the absence of a gender gap in high school is that girls and boys are treated equally.
27
5.5 Is stress correlated with social status ranking?
Participants answered a questionnaire before and after taking the IQ-test in order to report the self-rated stress levels. The results of these questionnaires are analysed and discussed in the next sections.
5.5.1 Stress and social status ranking
This section examines if stress is correlated with social status ranking (the competitive situation) within this experiment. All participants filled in their stress level on a scale of 1 to 7 in a questionnaire. The participants did answer the questionnaire before the start of the task and the announcement of their treatment. This is illustrated in Figure 4 as ‘stress before’. After completion of the task, the participants filled in the questionnaire a second time. Before reading aloud their score category, the participants answered the questionnaire a second time. This is named ‘stress after’.
Figure 4: stress level participants
Figure 4 illustrates that the mean stress level of participants increases after completing the IQ-test. However, the increase in stress level is present for males and females in both the baseline and the treatment. The graph illustrates that the overall stress level after completing the task is similar between baseline and treatment. To investigate this relationship in more detail, the following OLS-regression was set up:
4,1 4,2 3,6 3,8 4,9 4,9 4,8 5,0 0,0 1,0 2,0 3,0 4,0 5,0 6,0
male baseline Male treatment female baseline female treatment
S
tre
ss l
eve
28
𝑦 = 𝛽0 + 𝛽1 𝑅𝐸𝐴𝐷 + 𝛽2 𝐺𝐸𝑁𝐷𝐸𝑅 + 𝛽3 𝑅𝐸𝐴𝐷 ∗ 𝐺𝐸𝑁𝐷𝐸𝑅 (2)
The regression output is shown in Table 7 and investigates the effect of the treatment compared to the baseline on the stress level. The first column shows the effect of the treatment on’ STRESS DIFFERENCE’ (STESS DIFFERENCE = STRESS AFTER – STRESS BEFORE) as the dependent variable (y). Gender is a binary variable and equals 1 if the participant is male and equals 0 if the participant is female. READ also is a binary variable and equals 1 if the participant is in the treatment and 0 if the participants are in the baseline. Table 7 indicates no significant difference in ‘STRESS DIFFERENCE’ in the treatment compared to the baseline. The regression shows no significant gender difference in stress level. Thus, participants did not experience extra stress from the treatment compared to the baseline.
Table 7 Regression table stress
Stress difference Stress before Stress after
(1) (2) (3) READ 0.10 0.22 0.12 (0.30) (0.23) (0.22) GENDER 0.50 0.56 0.07 (0.29) (0.25) (0.23) READ*GENDER -0.08 -0.17 -0.10 (0.40) (0.35) (0.33) Constant -1.26 3.58 4.83 (0.23) (0.18) (0.17)
Notes The regression table indicates the effect of the treatment on participant’s stress level. The
participant is a male if gender=1 and female if gender=0. Read is the treatment and read=1 in the treatment group and read=0 in the baseline. The standard deviations are displayed in the parenthesis. The columns are the dependent variables and the rows are the independent variables and the control variables. The critical value of the t-statistic is 1.96.
Result 5: The participants do not experience extra stress from the social status ranking
treatment compared to the baseline.
Table 7 indicates ‘stress before’ as the dependent variable (y) in the OLS-regression. ‘Stress before’ is the baseline stress level of the participants in this experiment. The interaction variable between the treatment and gender is insignificant. In the baseline,
29
participants show no gender difference in their stress level. ‘Stress after’ is indicated in the last column as the dependent variable (y). Table 7 shows the average treatment effect on stress level of the participants after performing the task. The variables are statistically insignificant and this indicates the same result as the ‘stress difference’ variable. The treatment does influence the stress level of the participants after completing the task.
To summarize, the participant does not experience extra stress from the treatment. Thus, hypothesis 4 (Participant’s stress level is higher in the social status ranking treatment
compared to the baseline) is not supported by the data. In addition, the analysis shows
no gender differences in stress experience. Therefore, hypothesis 5 (Social status
ranking treatment have a greater positive effect on the stress level of women compared to men) is not supported by the data.
Result 4 is in line with results 2 and 3: participants do not show different behaviour in the treatment compared to the baseline. It is not surprising that stress level is on average the same in the treatment compared to the baseline.
5.6 Is stress correlated with performance?
In this subsection, the correlation between stress and performance is examined. Score is considered as the dependent variable and STRESSDIFF (= stress after – stress before) is considered as the independent variable. the next OLS-regression was set up to determine the relationship:
𝑆𝑐𝑜𝑟𝑒 = 𝛽0 + 𝛽1 𝑆𝑇𝑅𝐸𝑆𝑆𝐷𝐼𝐹𝐹 + 𝛽2 𝐺𝐸𝑁𝐷𝐸𝑅 + 𝛽3 𝑆𝑇𝑅𝐸𝑆𝑆𝐷𝐼𝐹𝐹 ∗ 𝐺𝐸𝑁𝐷𝐸𝑅
+ 𝛽4 𝑆𝑇𝑅𝐸𝑆𝐵𝐸𝐹𝑂𝑅𝐸
The regression includes two control variables: GENDER, which is 1 if the participant is male and 0 if the participant is female. STRESS BEFORE is included as an indicator variable to control for baseline stress level. Table 8 shows the output of the regression. The first column of Table 8 regresses the effect of STRESSDIFF on score, without controlling for gender or baseline stress. Participants who experienced the task as stressful (high stressdiff) performed better (higher score) compared to participants who
30
did not experience the task as stressful. The correlation between stress and performance is statistically significant.
In the second column of Table 9, the gender effect of stress on performance is examined (score). Gender is a binary variable and equals 1 if the participant is male and equals 0 if the participant is female. Table 9 indicates that female participants who experienced the IQ-test as stressful (high stressdiff) had a better performance compared to female participants who did not experience the IQ-test as stressful. The effect is significant for female participants.1 For male participants, the effect is lower and not statistically significant. Thus, stress increases performance for female participants but not for male participants.
Table 8 Regression table stress
Score Score Score
(1) (2) (3) Stressdiff 3.11** 4.07** 4.33** (1.26) (1.61) (1.87) Gender -4.78 -3.411 (4.80) (4.88) Stressdiff*gender -2.27 -1.13 (2.66) (2.70) Constant 118.51 120.86 108.1 (2.39) (3.38) (19.01)
Control for gender no yes yes
Control for ‘stress before’ no no yes
Notes The regression table indicates the effect of the treatment on stress level. The participant is a male
if gender=1 and female if gender=0. Read is the treatment and read=1 in the treatment group and read=0 in the baseline. In parentheses are the standard deviations. (Stressdiff= ‘stress after- ‘stress before’) The columns indicate the dependent variable (score) and the rows are the independent variable and the control variables. The critical value of the t-statistic is 1.96.
The third column of Table 8 shows the regression that controls for baseline stress. Since stress level is limited to a range of 1-7 this is necessary to include. Participants who experience a higher level of stress before the task (for instance a 5) cannot report much increase in their stress level. The regression measures the effect of stress difference on
1 This effect is tested with a separate regression for male participants. This is not indicated in Table 8.
31
score more precisely. The last column controls for this effect using an indicator variable for the baseline stress level (stress before). The regression shows the same result: female participants that experience the task as stressful show a better average performance This effect is not observed for male participants.
Result 6: Female participants that experience the IQ-test as stressful show a better performance on average. This effect is not observed for male participants.
To summarize, hypothesis 6 (Stress is negatively correlated with performance) is not supported by result 6. A possible explanation is that female participants who cared more about their performance on the IQ test reported a higher stress level. This implies that participants with a high stress level try harder to perform on the IQ test. An analysis to test this effect is available in appendix 2.
The design of this research measures term stress. It is relevant to realize that short-term stress may have a different effect on performance than long-short-term stress. Stress on the labour market is usually long-term. Due to this different type of stress and the different context this research does not have implication for the labour market. Hence, stress of employees may have a different effect on performance. Moreover, reverse causality might be a applicable since stress influences performance, but performance may also influence stress. No causal relationship between stress and performance could be determined in the analysis.
6 Conclusion and recommendations
This study aimed to deepen the understanding of gender differences in the social status
ranking dimension of competition. The data do not show similar effects as the research
of Schram et al. (2017), since the treatment effect of social status ranking on performance was absent. In the non-competitive treatment (baseline) men and women performed equally on average. The analysis does not show behavioural differences between the competitive treatment and the baseline. A possible reason for the deviation in outcome to the research of Schram et al. (2017) lies in the population. High school students might be familiar with ranking based on performance since students frequently receive grades at school. To conclude, a high school is an environment where gender differences in competitiveness along the social status ranking dimension are not present.
32
The second part of this study focuses on the stress level of participants. Stress may give an extra explanation for the gender differences in competitive settings. In this study, the stress level of participants increases after completing the IQ-test. This effect is present for both men and women. Therefore, participants did not experience extra stress from the competitive setting compared to the baseline. Adjacent to the first result (no
behavioural differences between the social status ranking treatment compared to the baseline) it is not surprising that the data do not show a higher level of stress in the
social ranking treatment. In addition, high-stress levels that are identified in this analysis resulted in a higher score for female participants. Therefore, stress in this setting is positively correlated with performance for females. Thus, participants who experience a higher stress level, care more to perform well on the task. This resulted in participants with a higher stress level to have a better performance.
Due to the narrow scope and the limited amount of time to conduct the research a few limitations are identified. First, the participants only read aloud their score category and not their real rank. This may have driven the results since the effect of social status
ranking results from ranking performance. Computerization would have given the
opportunity to make an exact rank out of the score directly after the IQ-test. Second, it is common to use financial incentives in economic experiments. A financial incentive related to performance may give more insight into this experiment, because participants would care more to perform well on the task. No financial incentive has been used in this experiment. In addition, a population with employees instead of high school students would have given a deeper insight into the competitiveness of the labour market. Differences in competitiveness between men and women are present at young age. However, a population existing of people in different environments or age classes, would have given a better insight into the gender gap in for instance the labour market.
Further research is needed to determine the effect of social status ranking dimension in competitive environments. This research does not support the findings of Schram et al. (2017) and therefore the social status ranking dimension has to be explored in more detail. Girls perform equally compared to men in a competitive setting in high school, because they do not underestimate their performance. This is a relevant direction for further research, since a high school might be a setting where women feel comfortable
33
in competition. Finally, this study indicates that short-term stress increases performance. However, no conclusion could be made for other population groups for instance employees in the labour market are often exposed to long-term stress. Further research could investigate what effect long-term stress in a competitive setting would have on performance.
7 Appendix
7.1 Appendix 1 – Regression table correlation of stress on score for male
This appendix analyses if stress difference is significantly correlated with performance for male participants. The dependent variable is stress difference (stress after – stress before) and performance is measured as the score on the IQ-test. Table 9 shows that stress difference is not significantly correlated with the score for male participants.
Table 9 Regression table Stress male Score (1) STRESS DIFF 1.79 (2.22) Constant 116.09 (3.58)
Notes The regression table indicate whether stress is correlated with performance for male participants.
Score is the dependent variable and stress difference (stress after – stress before) is the independent variable.
7.2 Appendix 2 – Is stress correlated with care?
This appendix shows if participants that care more about the outcome of an IQ-test have a higher level of stress. The data of the questionnaire are analysed in more detail. As discussed above, the questionnaire includes the following question:
‘How eager do you like to be successful?’
Therefore, in this section, the correlation between the self-reported stress level and the eagerness of the participants to be successful is analysed. Gender is used as an interaction variable in this analyses and the following OLS-regress was set up:
34
𝑆𝑡𝑟𝑒𝑠𝑠𝑎𝑓𝑡𝑒𝑟 = 𝛽0+ 𝛽1𝐶𝐴𝑅𝐸𝐴𝐹𝑇𝐸𝑅 + 𝛽2𝐺𝐸𝑁𝐷𝐸𝑅 + 𝛽3𝐶𝐴𝑅𝐸𝐴𝐹𝑇𝐸𝑅 ∗ 𝐺𝐸𝑁𝐷𝐸𝑅
The regression output is shown in table 9. The table indicates a negative relationship between care and stress. This effect is significant and implies that participants who care more about the outcome of the test, have a decreased stress level.
Table 9 Regression table Care
Stress after (1) CARE AFTER -0.16** (0.68) GENDER -0.52 (0.36) Constant 5.43 (0.25)
Notes The regression table indicates whether care and stress are correlated in this analyses. Stress after
is the dependent variable in this analyses and care after is the independent variable in this analysis. Th
7.3 Appendix 3 – A question from every category (Task )
Evening - Diner
Morning - …….
a) Day soup b) Breakfast c) Coffee
d) Time of the day The answer is: b) Breakfast
0 1 3 6 10 .?.
a) 16 b) 15 c) 12 d) 6
35 The answer is: B
Veerbeeld vraag:
36 8. References
AbuAlRub, R. F. (2004). Job stress, job performance, and social support among hospital nurses. Journal of nursing scholarship, 36(1), 73-78.
Andersen, S., Ertac, S., Gneezy, U., List, J. A., & Maximiano, S. (2013). Gender, competitiveness, and socialization at a young age: Evidence from a matrilineal and a patriarchal society. Review of Economics and Statistics, 95(4), 1438-1443.
Azmat, G., & Iriberri, N. (2010). The importance of relative performance feedback information: Evidence from a natural experiment using high school students. Journal
of Public Economics, 94(7-8), 435-452.
Balafoutas, L., Kerschbamer, R., & Sutter, M. (2012). Distributional preferences and competitive behavior. Journal of economic behavior & organization, 83(1), 125-135.
Barber, B. M., & Odean, T. (2001). Boys will be boys: Gender, overconfidence, and common stock investment. The quarterly journal of economics, 116(1), 261-292.
Bateup, H. S., Booth, A., Shirtcliff, E. A., & Granger, D. A. (2002). Testosterone, cortisol, and women's competition. Evolution and Human Behavior, 23(3), 181-192.
Bertrand, M., & Hallock, K. F. (2001). The gender gap in top corporate jobs. ILR
Review, 55(1), 3-21.
Beyer, S. (1990). Gender differences in the accuracy of self-evaluations of performance. Journal of personality and social psychology, 59(5), 960.
Buser, T., Dreber, A., & Mollerstrom, J. (2017). The impact of stress on tournament entry. Experimental economics, 20(2), 506-530.
Buckert, M., Schwieren, C., Kudielka, B. M., & Fiebach, C. J. (2017). How stressful are economic competitions in the lab? An investigation with physiological measures.
