This paper studies the design of higher education policies targeted at improving the recruitment of public servants. I leverage the introduction of a policy in Chile that aimed to raise teacher quality by combining financial incentives and admission standards. Exploiting the sharp assignment rule I estimate that, at the threshold, enrollment of high performing students at teacher colleges increased by 42%. For low-income students, two thirds of the increase is due to switching away from non-enrollment. The policy generated a positive composition effect of 0.25SD in the scores at the college entry exam, which led to an increase in 0.11SD in Teacher Value Added and 0.12SD in teaching skills. I embed the reduced-form results into a demand and supply model of higher education that incorporates a novel method for solving discrete-continuous games in large markets. Counterfactual policies lead to increases of up to 6.6% in the test scores of students enrolled at teacher colleges, and up to 20% in Teacher Value Added. Targeting the policy to low-income students yields further gains in Teacher Value Added at no additional cost. An alternative policy would need to increase by 35% the expected wages of graduates from teaching degrees to achieve similar gains.

This paper examines the effects of Free College policies on student enrollment and academic performance, with a focus on the 2016 Chilean reform that granted tuition-free higher education to students from the lowest five income deciles. Using a difference-in-differences approach, we find that Free College increased enrollment and persistence in higher education on the eligible but had modest effects on graduation and dropout rates. To disentangle the role of student effort from selection effects, we develop a structural model in which students choose effort levels in response to financial incentives. Our results highlight that Free College expanded access, in particular for low-achieving students. Despite the removal of academic progress requirements, we found no evidence of weakening performance.

This paper studies the roles of financial constraints and information frictions on enrollment and progression in higher education. We use Chilean administrative data, allowing us to link students to their parents and their high school peers. Our empirical strategy exploits the massive entry of private universities during the 1980s to instrument parental educational achievement, panel data methods to estimate peer influence, and the staggered rollout of Free College from 2016 that generated exogenous variation to out-of-pocket fees. Results show that subsidies increase university access, peers enhance enrollment and match quality, and parental exposure to university causally affects children's university enrollment. We use these findings to inform a dynamic structural model that quantifies these mechanisms and evaluates equity-oriented education policies.

While small-scale educational interventions provide valuable insights into effective investments, large-scale education programs may generate substantial general equilibrium (GE) effects that alter their ultimate impact on welfare and inequality. This paper analyzes the consequences of a large-scale college enrollment expansion in Chile, explicitly accounting for GE effects in both education and labor markets, to understand how firms adjust their demand for skill-biased labor in response to increased college graduate supply. The model enables us to decompose observed wage changes into individual returns to education and GE effects, revealing how firm responses shape education decisions and ultimately affect the distribution of earnings. By incorporating both household education choices and firm production decisions, this framework provides a comprehensive assessment of how large-scale education expansions affect welfare and distributional outcomes.

This project develops scalable, low-cost strategies to enhance early warning systems aimed at preventing school dropout in low- and middle-income countries. Despite the growing adoption of data-driven platforms that predict students' dropout risk and provide targeted guidance to school staff, real-world impact remains limited by low user engagement and minimal improvements in outcomes. We propose an integrated approach that strengthens both the targeting of at-risk students and the engagement of key stakeholders. By systematically combining school principals’ expert beliefs about students with machine learning predictions, we harness local knowledge to improve the accuracy of dropout risk assessments. Simultaneously, we foster parental involvement by inviting families of at-risk students to structured meetings with school leadership, sharing clear risk information and practical guidance to support continued enrollment. At-risk students are also offered tailored tutoring in foundational skills by trained educators.

We study the impact of a policy change in an online labor market requiring disclosure of projects’ experience requirements, which prompted roughly half of employers to stop revealing their budgets. Exploiting detailed data from the universe of auctions in PeoplePerHour—one of the largest online labor markets—we document that employers who conceal budgets after the policy systematically pay higher winning fees across all project types. Bid variance increases, indicating greater strategic uncertainty for workers. Worker win probabilities remain largely unaffected by the disclosure change, though we observe stronger price-based selection for low-value projects and more pronounced firm advantage in mid-value segments. To interpret these empirical patterns and evaluate counterfactual information policies, we build a structural beauty-contest auction model. Our framework explicitly captures the strategic interplay between employers and heterogeneous workers under incomplete information, enabling us to simulate the welfare and market impacts of alternative disclosure policies. The results highlight the multifaceted consequences of market design for both sides of online labor platforms, especially when buyers’ strategic concealment of information is endogenous.