Gender differences in the intention to study math increase with math performance

The underrepresentation of women in math-intensive STEM fields (mathematics, computer science, physical sciences, etc.) remains a significant concern. This disparity is not only a matter of fairness but also contributes to gender wage inequalities and potential skill shortages. While substantial research exists on gender gaps in STEM, it has largely overlooked the interaction between math ability and the gender gap in educational and career intentions. This study aims to address this gap by analyzing how the gender gap in intentions to pursue math-related studies or careers evolves across the spectrum of math abilities. This approach offers three main advantages: 1) It clarifies whether the gender gap is primarily a concern among high-ability students, directly impacting worker shortages and wage inequality. 2) It investigates whether the disparity in intentions results in a gender gap in math performance among students who actually pursue math-related fields, leading to underrepresentation of women among the most capable students. 3) Examining the gender gap along the ability distribution allows for more targeted policy interventions to reduce the disparity. While prior research has touched on this area, this paper leverages a much larger dataset and employs more sophisticated analysis to draw robust conclusions.

Existing literature extensively examines gender gaps in STEM and math-related studies. However, much of this research fails to fully consider how the gender gap interacts with math ability. Studies have highlighted various factors, including gender differences in comparative advantages in math versus reading, math-related self-concepts, and interest in math. Previous work also showed that high-performing girls are more likely to excel in reading compared to boys, potentially influencing their choices away from math-intensive fields. The role of math self-concept and interest in math has also been investigated as potential contributing factors. While there are studies investigating gender differences across different ability levels, they often focus on college students, potentially introducing selection biases. This study adds to the existing literature by providing a more comprehensive analysis using a broader dataset, focusing on 15-year-old students to minimize selection effects and exploring the interaction between math ability and gender in shaping intentions to pursue math.

This study uses data from the 2012 Programme for International Student Assessment (PISA 2012), a large-scale international survey encompassing 61 countries and over 251,000 15-year-old students. PISA 2012 assessed students' math skills through tests and questionnaires, including a series of five questions designed to gauge their intentions to pursue math-related studies or careers. A binary variable was created, indicating "strong intentions" for those who responded positively to all five questions. Math performance was standardized within each country to have a mean of 0 and a standard deviation of 1, allowing for cross-country comparisons. Linear probability models were used to analyze the relationship between math intentions, math ability, and gender, incorporating control variables for other cognitive abilities (reading and science) and socioeconomic factors. Robustness checks were conducted using alternative measures of math intentions, different statistical models (logit, probit), and a non-standardized math performance measure. To further investigate the relation between intentions and actual choices, data from a French high school dataset containing students' intentions to enrol in a scientific track, along with their actual enrolment, was used. The study also leverages the High School Longitudinal Study (HSLS:09) from the United States to assess actual enrollment in math-related courses. Finally, data from the Programme for the International Assessment of Adult Competencies (PIAAC) were analyzed to examine the relationship between math ability, occupation, and gender among adults, providing insights into the long-term consequences of earlier educational choices. In all analyses, appropriate weights were employed to ensure the representativeness of the samples.

The key findings of the study are as follows: 1) A positive and linear relationship was observed between math performance and the probability of intending to pursue math-related studies for both boys and girls. 2) However, this relationship was significantly stronger for boys than for girls. The difference in the slopes of the relationship between math performance and intention to pursue math, when comparing boys and girls, was statistically significant. A one-standard deviation increase in math performance increased the probability of boys intending to pursue math by 5.4 percentage points, while for girls the increase was only 3.3 percentage points. 3) The gender gap in intentions to pursue math was minimal (1.8 percentage points) among students with low math abilities but grew steadily with increasing math ability, reaching 8 percentage points among the highest-performing students. 4) This widening gap is reflected in the ratio of boys to girls with strong math intentions, increasing from roughly 1 among low performers to 1.5 among high performers. 5) Globally, and within both OECD and non-OECD countries, the gender gap in math performance was considerably larger among students intending to pursue math than in the general student population. 6) In the majority of the countries studied, the representation of girls among the top-performing math students was lower among those intending to pursue math compared to the overall population. 7) Further analysis in France, using data linking high school students' intentions to enroll in a science track and actual enrollment, confirmed the observed pattern in the PISA data, showing a similar relationship between math ability and study choices, with girls exhibiting a weaker relation than boys. Similarly, results from HSLS:09 in the US showed a consistent pattern in actual enrollment in math-related high school courses, with the gender gap increasing with math ability. 8) Finally, the analysis of PIAAC data showed a similar trend amongst adults, with the probability of working in a math-related occupation or intensively using math skills at work increasing with math ability, but significantly more so for men compared to women. The gender gap in math performance was also higher among this group. 9) The observed patterns were not primarily explained by gender differences in comparative advantages in reading relative to math, math self-concept, declared interest in math, or even socio-economic background (though SES explained approximately half of the effect).

The study's findings address the research question by demonstrating a clear interaction between gender, math performance, and the intention to pursue math-related studies. The increasing gender gap in intentions among higher performers highlights a critical concern: not only are fewer girls pursuing math-intensive fields, but this disparity is particularly pronounced among the most talented girls. This poses a significant problem for addressing underrepresentation in STEM. The fact that the observed patterns persist across various samples, methodologies, and even into the adult workforce, strengthens the robustness of the results. While factors like SES play some role, the inability to fully account for the findings through traditional explanatory factors suggests the influence of other mechanisms, most likely gender stereotypes. The persistence of the stereotype that math is not for girls may discourage high-performing girls from choosing these fields, contributing to a self-fulfilling prophecy. This suggests that interventions should go beyond simply improving girls' math skills and should target the underlying societal biases and stereotypes that discourage high-achieving girls from pursuing math-related careers.

This study demonstrates that the gender gap in intentions to study math significantly increases with math performance, with high-performing girls showing less inclination towards math than their male counterparts. This disparity, not fully explained by known factors, likely results in a lower representation of high-achieving girls in math-related fields. Future research could explore interventions aimed at high-performing girls to counteract these trends. More broadly, policymakers need to consider the whole ability distribution when designing interventions to improve gender equality in STEM.

A primary limitation of this study is the reliance on self-reported intentions rather than actual enrollment or career choices. While the study employs data from France and the US to investigate the relationship between intentions and actions, this association is not thoroughly explored across all countries. Moreover, the study cannot definitively identify the precise mechanisms driving the observed patterns, although gender stereotypes and socio-economic factors are potential contributors. Further research is needed to delve deeper into causal mechanisms and test the effectiveness of specific policy interventions.