VoxEU Column Gender Labour Markets

Gender and task rewards in exams: Implications for performance and labour outcomes

15 Feb 2025

Women’s under-representation in high-paying occupations remains a critical issue, with recent debates focusing on whether gender differences in behaviour contribute to these gaps. Using data from university admission exams in Brazil, this column examines gender performance differences under varying task reward structures and assesses their impact on labour market outcomes. Higher exam rewards widen the gender performance gap, with women underperforming relative to men. While performing well in high-reward exam tasks positively correlates with wages, gender disparities in exam behaviour do not explain the gender wage gap.

Bruna Borges

Postdoctoral Fellow Sao Paulo School of Economics

Fernanda Estevan

Associate Professor Sao Paulo School of Economics

Although women have made significant progress in education and labour market outcomes, they remain under-represented in high-paying jobs (Bertrand 2018). Achieving leadership positions often demands exceptional performance in competitive environments and prioritising high-reward tasks over less lucrative ones.

Prior research highlights gender differences in competitiveness, confidence, risk-taking, and women’s underperformance on high-stakes tasks as possible explanations. Much of this literature focuses on real-world settings, such as exams. In a 2016 Vox column, Rey Biel and Iriberri (2016) summarise their study on a two-stage mathematics competition in Madrid, Spain, emphasising how competitive environments can amplify gender disparities and potentially explain the ‘glass ceiling’ phenomenon (the invisible barrier that prevents women from advancing to an organisation’s top positions). In another Vox column, Calsamiglia et al. (2016) discuss how students in Spain respond differently to exam pressure, with girls generally outperforming boys in low-stakes tests but losing their advantage as the stakes rise. Whether gender differences in exam performance can explain gender disparities in the labour market remains an open question (Blau and Kahn 2017).

In a paper with Louis-Philippe Morin ((Borges et al. 2025), we analyse data from a large Brazilian university admission exam to investigate how variations in exam rewards influence gender performance gaps. This comprehensive dataset enables us to explore the underlying mechanisms driving these disparities. Importantly, the study tracks test takers’ labour market outcomes for up to 14 years after the exam, providing valuable insights into whether the observed gender exam performance differences have long-term labour market consequences.

Data and methodology

We use admission exam data from a selective Brazilian university, Universidade Estadual de Campinas (UNICAMP), to verify how the female–male performance gap changes in parts of the exam that count relatively more toward the final admission exam score. The exam has two phases that test applicants on various high school subjects. In Phase 1, all subjects are equally weighted, whereas in Phase 2, subjects designated ‘priority’ carry double the weight based on the applicant’s chosen major. This setup allows us to compare how men and women react when facing high versus low task rewards.

Our study compares applicants’ performance in high-reward (priority) versus low-reward (non-priority) subjects during Phase 2 of the UNICAMP exam to assess how increased rewards influence gender performance gaps. By incorporating controls for subject, subject-gender, and individual fixed effects, it is possible to isolate the impact of the reward structure. To address potential selection biases – such as applicants’ choice of majors and their associated priority subjects – the regression includes additional controls, such as prior performance in Phase 1. To ensure that gender-specific performance gaps are not driven by how men and women select priority subjects, the paper tests for such gaps in future priority subjects during Phase 1, where rewards are equal. The analysis reveals no evidence that women perform worse in subjects that later become priorities, supporting the validity of the empirical strategy.

Gender gaps in performance under high rewards

Women’s relative performance declines significantly in priority subjects compared to non-priority ones. This gender performance gap, equivalent to 9% of the within-applicant standard deviation, is particularly pronounced among higher-performing individuals, potentially affecting admissions to competitive majors.

There is evidence that differences in exam strategies contribute to the observed gaps. Women are more likely to omit questions in high-reward priority subjects, reflecting a 13% increase relative to the mean, whereas men are more inclined to attempt answering but providing an answer that is incorrect. This pattern aligns with findings by Rey-Biel and Iriberri (2016), who show that women tend to increase the number of omitted questions as competitive pressure intensifies. Moreover, this omission pattern appears in male-dominated subjects, such as physics and math, suggesting the influence of gender-stereotyped beliefs (Coffman 2014, Bordalo et al. 2019, Exley and Kessler 2022).

In Phase 2, women distribute their efforts more evenly across and within each subject. This balanced strategy contrasts with that of men, who concentrate more intensively on priority subjects to maximise their weighted scores. Thus, the data indicate a distinct approach to test-taking among genders. This pattern helps explain the observed performance gaps between genders in high-reward tasks and highlights the underlying differences in strategic behaviour when facing competitive pressure.

University admission

We also analyse the impact of the gender performance gap on university admissions using a counterfactual analysis. Our study simulates admissions by adjusting female applicants’ scores in priority subjects to eliminate the observed gender gap while accounting for the potential trade-offs in non-priority subjects. Closing the gender performance gap would modestly increase the overall admission rate for women by 1.1%, and decrease it for men by 0.6%. However, the effects are more pronounced in competitive majors like medicine and economics. This suggests that addressing performance gaps in high-reward subjects could have meaningful implications for gender parity in access to competitive university programmes.

Labour market implications

Applicants who perform better in priority subjects earn higher wages on average in the labour market. To demonstrate this, we combine UNICAMP administrative data with Brazil’s formal labour market records, tracking wages and employment for up to 14 years after the exam. We then present wage regressions controlling for ‘relative priority performance’ (performance on more heavily weighted priority subjects relative to non-priority subjects) and other factors such as ability and (intended) major.

Figure 1 shows that the gender performance gap observed in high-reward exam tasks does not account for the gender wage gap. In contrast, a measure of general academic ability, specifically the standardised end-of-high-school exam (ENEM) score, explains a significant portion (28%) of the gender wage gap.¹ Furthermore, nearly two-thirds of the gender wage gap can be attributed to differences in chosen fields of study, emphasising the impact of educational pathways on career earnings.

Figure 1 Breakdown of factors reducing the gender wage gap

Notes: The gender wage gap is computed using a sample of applicants to UNICAMP from 2001–2004 who attended both P1 and P2 for admission and subsequently worked in the formal labor market between 7 and 12 years after the admission exam (N=29,906). The dependent variable is the logarithm of the average annual wage observed over this period. In the first bar, we show estimates for the raw wage gap, controlling only for admission exam year fixed effects. In the second bar, we add a control for relative priority performance, measured as the difference between the applicant’s average residuals in priority and non-priority subjects, following the specification in column (7) of Table 2 in Borges et al. (2025), excluding the interaction term ‘Female × Priority’. In the third bar, we control for exam year fixed effects and academic ability, proxied by the individual’s normalised ENEM scores (mean zero, standard deviation one, by year). In the fourth bar, we include exam year fixed effects and fixed effects for the applicant’s intended major based on the first-choice major declared during the UNICAMP registration process.

Thus, our findings indicate that while excelling in high-reward tasks is linked to higher wages, addressing broader issues like major selection may be more critical for reducing gender disparities in earnings.

Conclusion

Our work provides evidence that women and men respond differently to increased rewards. We propose potential mechanisms through which this difference may occur and examine whether it has implications in the labour market. In our analysis, we observe that both men and women take identical exams but face varying rewards for their performance on different questions.

Our study reveals that women perform worse than men on high-reward tasks during competitive exams. Additionally, women are more likely to omit questions in high-reward situations, with this omission pattern particularly evident in male-dominated subjects. Finally, while the ability to prioritise high versus low-reward questions is associated with higher wages, the study raises questions about whether the observed gender differences in exam performance translate into meaningful effects in the labour market.

Editors’ note: This column is published in collaboration with the International Economic Associations’ Women in Leadership in Economics initiative, which aims to enhance the role of women in economics through research, building partnerships, and amplifying voices.

References

Bertrand, M (2018), “Coase lecture: The glass ceiling”, Economica 85(338): 205–31.

Blau, F D and L M Kahn (2017), “The gender wage gap: Extent, trends, and explanations”, Journal of Economic Literature 55(3): 789–865.

Bordalo, P, K Coffman, N Gennaioli and A Shleifer (2019), “Beliefs about gender”, American Economic Review 109(3): 739–73.

Borges, B, F Estevan and L P Morin (2025), “Exam Reward Structure, Gender Performance Gaps, and Labor Market Outcomes”, available at SSRN, accepted at the Journal of Labor Economics.

Calsamiglia, C, N Iriberri and G Azmat (2016), “Under pressure: Gender differences in responding to exam stakes”, VoxEU.org, 22 January.

Coffman, K B (2014), “Evidence on self-stereotyping and the contribution of ideas”, Quarterly Journal of Economics 129(4): 1625–60.

Exley, C L and J B Kessler (2022), “The gender gap in self-promotion”, Quarterly Journal of Economics 137(3): 1345–81.

Rey Biel, P and N Iriberri (2016), “Gender gap in a two-stage maths competition”, VoxEU.org, 23 October.

Footnotes

Although women have made significant progress in education and labour market outcomes, they remain under-represented in high-paying jobs (Bertrand 2018). Achieving leadership positions often demands exceptional performance in competitive environments and prioritising high-reward tasks over less lucrative ones. Prior research highlights gender differences in competitiveness, confidence, risk-taking, and women’s underperformance on high-stakes tasks as possible explanations. Much of this literature focuses on real-world settings, such as exams. In a 2016 Vox column, Rey Biel and Iriberri (2016) summarise their study on a two-stage mathematics competition in Madrid, Spain, emphasising how competitive environments can amplify gender disparities and potentially explain the ‘glass ceiling’ phenomenon (the invisible barrier that prevents women from advancing to an organisation’s top positions). In another Vox column, Calsamiglia et al. (2016) discuss how students in Spain respond differently to exam pressure, with girls generally outperforming boys in low-stakes tests but losing their advantage as the stakes rise. Whether gender differences in exam performance can explain gender disparities in the labour market remains an open question (Blau and Kahn 2017). We analyse data from a large Brazilian university admission exam to investigate how variations in exam rewards influence gender performance gaps (Borges et al. 2025). This comprehensive dataset enables us to explore the underlying mechanisms driving these disparities. Importantly, the study tracks test takers’ labour market outcomes for up to 14 years after the exam, providing valuable insights into whether the observed gender exam performance differences have long-term labour market consequences. Data and methodology We use admission exam data from a selective Brazilian university, Universidade Estadual de Campinas (UNICAMP), to verify how the female–male performance gap changes in parts of the exam that count relatively more toward the final admission exam score. The exam has two phases that test applicants on various high school subjects. In Phase 1, all subjects are equally weighted, whereas in Phase 2, subjects designated ‘priority’ carry double the weight based on the applicant’s chosen major. This setup allows us to compare how men and women react when facing high versus low task rewards. Our study compares applicants’ performance in high-reward (priority) versus low-reward (non-priority) subjects during Phase 2 of the UNICAMP exam to assess how increased rewards influence gender performance gaps. By incorporating controls for subject, subject-gender, and individual fixed effects, it is possible to isolate the impact of the reward structure. To address potential selection biases – such as applicants’ choice of majors and their associated priority subjects – the regression includes additional controls, such as prior performance in Phase 1. To ensure that gender-specific performance gaps are not driven by how men and women select priority subjects, the paper tests for such gaps in future priority subjects during Phase 1, where rewards are equal. The analysis reveals no evidence that women perform worse in subjects that later become priorities, supporting the validity of the empirical strategy. Gender gaps in performance under high rewards Women’s relative performance declines significantly in priority subjects compared to non-priority ones. This gender performance gap, equivalent to 9% of the within-applicant standard deviation, is particularly pronounced among higher-performing individuals, potentially affecting admissions to competitive majors. There is evidence that differences in exam strategies contribute to the observed gaps. Women are more likely to omit questions in high-reward priority subjects, reflecting a 13% increase relative to the mean, whereas men are more inclined to attempt answering but providing an answer that is incorrect. This pattern aligns with findings by Rey-Biel and Iriberri (2016), who show that women tend to increase the number of omitted questions as competitive pressure intensifies. Moreover, this omission pattern appears in male-dominated subjects, such as physics and math, suggesting the influence of gender-stereotyped beliefs (Coffman 2014, Bordalo et al. 2019, Exley and Kessler 2022). In Phase 2, women distribute their efforts more evenly across and within each subject. This balanced strategy contrasts with that of men, who concentrate more intensively on priority subjects to maximise their weighted scores. Thus, the data indicate a distinct approach to test-taking among genders. This pattern helps explain the observed performance gaps between genders in high-reward tasks and highlights the underlying differences in strategic behaviour when facing competitive pressure. University admission We also analyse the impact of the gender performance gap on university admissions using a counterfactual analysis. Our study simulates admissions by adjusting female applicants’ scores in priority subjects to eliminate the observed gender gap while accounting for the potential trade-offs in non-priority subjects. Closing the gender performance gap would modestly increase the overall admission rate for women by 1.1%, and decrease it for men by 0.6%. However, the effects are more pronounced in competitive majors like medicine and economics. This suggests that addressing performance gaps in high-reward subjects could have meaningful implications for gender parity in access to competitive university programmes. Labour market implications Applicants who perform better in priority subjects earn higher wages on average in the labour market. To demonstrate this, we combine UNICAMP administrative data with Brazil’s formal labour market records, tracking wages and employment for up to 14 years after the exam. We then present wage regressions controlling for ‘relative priority performance’ (performance on more heavily weighted priority subjects relative to non-priority subjects) and other factors such as ability and (intended) major. Figure 1 shows that the gender performance gap observed in high-reward exam tasks does not account for the gender wage gap. In contrast, a measure of general academic ability, specifically the standardised end-of-high-school exam (ENEM) score, explains a significant portion (28%) of the gender wage gap. Furthermore, nearly two-thirds of the gender wage gap can be attributed to differences in chosen fields of study, emphasising the impact of educational pathways on career earnings. Figure 1 Breakdown of factors reducing the gender wage gap Notes: The gender wage gap is computed using a sample of applicants to UNICAMP from 2001–2004 who attended both P_1 and P_2 for admission and subsequently worked in the formal labor market between 7 and 12 years after the admission exam (N=29,906). The dependent variable is the logarithm of the average annual wage observed over this period. In the first bar, we show estimates for the raw wage gap, controlling only for admission exam year fixed effects. In the second bar, we add a control for relative priority performance, measured as the difference between the applicant’s average residuals in priority and non-priority subjects, following the specification in column (7) of Table 2 in Borges et al. (2025), excluding the interaction term ‘Female × Priority’. In the third bar, we control for exam year fixed effects and academic ability, proxied by the individual’s normalised ENEM scores (mean zero, standard deviation one, by year). In the fourth bar, we include exam year fixed effects and fixed effects for the applicant’s intended major based on the first-choice major declared during the UNICAMP registration process. Thus, our findings indicate that while excelling in high-reward tasks is linked to higher wages, addressing broader issues like major selection may be more critical for reducing gender disparities in earnings. Conclusion Our work provides evidence that women and men respond differently to increased rewards. We propose potential mechanisms through which this difference may occur and examine whether it has implications in the labour market. In our analysis, we observe that both men and women take identical exams but face varying rewards for their performance on different questions. Our study reveals that women perform worse than men on high-reward tasks during competitive exams. Additionally, women are more likely to omit questions in high-reward situations, with this omission pattern particularly evident in male-dominated subjects. Finally, while the ability to prioritise high versus low-reward questions is associated with higher wages, the study raises questions about whether the observed gender differences in exam performance translate into meaningful effects in the labour market. Editors’ note: This column is published in collaboration with the International Economic Associations’ Women in Leadership in Economics initiative, which aims to enhance the role of women in economics through research, building partnerships, and amplifying voices. References Bertrand, M (2018), “Coase lecture: The glass ceiling”, Economica 85(338): 205–31. Blau, F D and L M Kahn (2017), “The gender wage gap: Extent, trends, and explanations”, Journal of Economic Literature 55(3): 789–865. Bordalo, P, K Coffman, N Gennaioli and A Shleifer (2019), “Beliefs about gender”, American Economic Review 109(3): 739–73. Borges, B, F Estevan and L P Morin (2025), “Exam Reward Structure, Gender Performance Gaps, and Labor Market Outcomes”, available at SSRN, accepted at the Journal of Labor Economics. Calsamiglia, C, N Iriberri and G Azmat (2016), “Under pressure: Gender differences in responding to exam stakes”, VoxEU.org, 22 January. Coffman, K B (2014), “Evidence on self-stereotyping and the contribution of ideas”, Quarterly Journal of Economics 129(4): 1625–60. Exley, C L and J B Kessler (2022), “The gender gap in self-promotion”, Quarterly Journal of Economics 137(3): 1345–81. Rey Biel, P and N Iriberri (2016), “Gender gap in a two-stage maths competition”, VoxEU.org, 23 October.