Abstract: Around the world, women are in general underrepresented in STEM majors in higher education. Leveraging economy-wide data on Taiwan's centralized university admissions and a matching model, this paper decomposes the STEM gender gap into factors of both demand and supply sides. We find that either the demand factors, i.e., gendered preferences, or the supply factors, i.e., gender-specific labor market outcomes after graduation, contribute to a relatively small fraction of the overall STEM gender gap. The majority of this gap results from the disproportionately low participation of female applicants in university entrance exams for STEM-related subjects, which are critical for admission to STEM majors. Counterfactual simulations show that some commonly proposed policies can be effective to improve female representation in STEM majors, particularly for the applicants who have taken STEM subject exams. Our findings underscore the importance of earlier-stage interventions, such as encouraging female students to build more STEM-related readiness before applying to university.

Abstract: This paper uses data from coordinated application and admissions systems in Australia, Brazil, Chile, China, Finland, Greece, Spain, Sweden, Uganda, and Taiwan to document differences in gender representation among talented students applying to STEM majors. These ten settings are very different in size, economic development, culture, gender norms, and geographic location. However, in all of them, university admission decisions rely on algorithms that allocate students to specific college-major combinations based on their academic performance when applying to university. We focus on students scoring in the top 10% of the university admission exam and show that female representation among STEM-major applicants varies from 22% in Taiwan to 46% in Sweden. In the contexts we study, these differences can be driven either by gender gaps in academic performance at the time of application or by gender gaps in the programs that these top-scoring students rank in their application lists. While we find some significant variation in female representation among top 10% scores -- 32.3% in Uganda to 65.6% in Sweden -- we find a remarkably stable gender gap in applications to STEM across settings -- between 22 and 29 percentage points in all education systems, but China and Australia, where it reaches 37% and 16% respectively. These results indicate that i.) closing gaps in academic performance is not enough to eliminate inequality in college trajectories across gender groups and ii.) the gender gap in major choices does not significantly vary with economic development.

Abstract: The uneven spatial distribution of schools hinders high-achieving students, particularly the socioeconomically disadvantaged, from accessing elite educational resources because of long commutes and associated costs. We study this problem within the context of COMIPEMS, a centralized system of public high school admissions in Mexico City's metropolitan area. We estimate student preferences by assuming that the observed matching is stable in the equilibrium. Our preference estimates show that students face a quality-proximity trade-off which is heterogeneous by program track (academic versus non-academic). Counterfactual distance-based subsidies would significantly increase high-achieving students’ admission to elite programs, especially in the academic track. In contrast, prioritization policies alone have negligible effects on assignment outcomes without addressing the demand side or locations of supply. Our findings call for increasing the supply of elite programs in remote places within this centralized assignment system.

Abstact: This paper studies how the design of centralized assignment rules shapes both student allocation and school quality. We examine this question in the context of the 2025 reform in Mexico City and its metropolitan area, which eliminated exam-based assignments for non-elite schools administered by COMIPEMS. We assemble administrative applicant-level data to capture demand, and standardized test archives to construct a school quality measure. We then develop a structural model of demand and supply within the assignment mechanism. On the demand side, we impose the Stability assumption on the observed matching for precise estimates of student preferences. On the supply side, schools choose quality to balance enrollment incentives against the cost of maintaining quality. Estimates show that elite schools are most attractive on average. Schools raise quality when enrollment increases. While elite schools maintain a certain level of quality independent of demand pressure, non-elite schools change quality only when enrollment is responsive. Counterfactual reveals that the 2025 reform shifts a share of high-achieving students toward non-elite schools but reduces overall enrollment. Elite schools lower quality as demand pressure weakens, whereas non-elite schools improve quality as a response to stronger cohorts. This paper highlights both the equity gains and the capacity challenges of the reform.