We study a regulation in Chile that mandates warning labels on products whose sugar or caloric concentration exceeds certain thresholds. We show that consumers substitute from labeled to unlabeled products—a pattern mostly driven by products that consumers mistakenly believe to be healthy. On the supply side, we find substantial reformulation of products and bunching at the thresholds. We develop and estimate an equilibrium model of demand for food and firms' pricing and nutritional choices. We find that food labels increase consumer welfare by 1.8% of total expenditure, and that these effects are enhanced by firms' responses. We then use the model to study alternative policy designs. Under optimal policy thresholds, food labels and sugar taxes generate similar gains in consumer welfare, but food labels benefit the poor relatively more.

This paper analyzes the effects of lowering the legal blood alcohol content limit for drivers from 0.05 to 0.03 grams of alcohol per deciliter of blood (g/dL) and increasing license suspension periods for offenders. We take advantage of a rich data set of administrative records that allow us to identify direct measures of accidents involving alcohol including fatalities and injuries. Results show a significant decrease of 32% in alcohol-related car accidents right after the law was approved but the effects moderate over time (15% after three years). There is also a significant reduction in injuries (31% right after the approval and 11% after three years) but no statistically significant effects on deaths. Complementary analysis of blood samples shows that the law had an effect on blood alcohol content (BAC) of male drivers up to the 90th percentile of the BAC distribution.

We study the consequences of affirmative action in centralized college admissions systems. We develop an empirical framework to examine the effects of a large-scale program in Brazil that required all federal institutions to reserve half their seats for socioeconomically and racially marginalized groups. By exploiting admissions cutoffs, we find that marginally benefited students are more likely to attend college and are enrolled at higher-quality degrees four years later. Meanwhile, there are no observed impacts for marginally displaced non-targeted students. To study the effects of larger changes in affirmative action, we estimate a joint model of school choices and potential outcomes. We find that the policy has impacts on college attendance and persistence that imply a virtually one-to-one income transfer from the non-targeted to the targeted group. These findings indicate that introducing affirmative action can increase equity without affecting efficiency.

We study how colleges adjust tuition in response to student financial aid, emphasizing the role of targeting. Our framework highlights two forces: a direct effect that increases tuition, and a composition effect that can reduce tuition when aid targets more price-sensitive students. Using a reform in Brazil's student loan program, we document both effects empirically and estimate an equilibrium model to evaluate counterfactual policies. We find that merit-based targeting raises tuition by 2.2%, while need-based targeting lowers it by 0.8%, reflecting that lower-income students are more price-sensitive. These price effects have a strong impact on enrollment decisions.

We study the rapid expansion of Brazil's private online higher-education sector and its effects on market structure and college enrollment. Exploiting regional and field-specific variation in online education penetration, we find that online programs expand access for older students but divert younger students from higher-quality in-person programs. Greater competition lowers tuition prices but also reduces the supply of in-person degrees. Using an equilibrium model of college education, we show that in the absence of online programs, total enrollment would be 14 percent lower, while in-person enrollment would rise by 33 percent. On net, aggregate labor-market value added declines by 1.4 percent. Online education raises value added for older students, who benefit from increased access, but lowers it for younger students, who shift toward lower-return online options. Counterfactual policies that restrict online enrollment to older cohorts could increase value added for younger students without reducing gains for older cohorts.

We study the aggregate and heterogeneous effects of a front-of-package labeling policy implemented in Chile. We find that consumers reduced their sugar and caloric intake by 9% and 6%, respectively. On the demand side, labels prompt consumers to substitute within categories rather than switching between categories. Within-category responses are more pronounced when labels provide new information. On the supply side, we observe bunching at regulatory thresholds, with substantial heterogeneity across categories, consistent with differing costs of product reformulation. We conclude that considering policy-response heterogeneity is key for effective policy design.

Women make up only 35% of global STEM graduates, a share unchanged for a decade. Using administrative data from ten centralized university admissions systems, we provide the first cross-national decomposition of the STEM gender gap into a pipeline gap (access and preparedness) and a choice gap (application decisions). The pipeline gap varies widely—from female disadvantage in Uganda to advantage in Sweden—yet the choice gap is strikingly consistent: even among top scorers, women are 25 percentage points less likely than men to apply to STEM. This stability across diverse contexts points to structural forces beyond local conditions.

This paper studies how schools respond to financial incentives. Governments can penalize institutions with high dropout or loan default rates, and these institutions can respond by increasing quality or changing the selection of students. We study the predictions of the model using a 2017 reform in Brazil, which made schools pay a fee for students receiving federal student loans that dropped out or defaulted. Consistent with the predictions of the model, we find that schools more reliant on government aid reduced dropout rates, primarily by increasing quality. We build an equilibrium model to illustrate the trade-off faced by policymakers, and use our estimates to study optimal policy. We find that schools should pay for approximately half of all dollars charged off in default.

In markets with existing private options, the optimal level of public provision may require balancing a tradeoff between reducing private options' market power with the possibility of crowding out high-quality products. We study the equilibrium effects of public education provision in the Dominican Republic, where the government aimed to increase the number of public school classrooms by 78% over a four-year period. We use an event study framework to estimate the effect of a new public school on local outcomes. We estimate that despite increasing local students' hours of instruction, a new public school does not have an effect on local students' test scores. But this null result hides considerable changes in students' schooling options. We find that a new public school increased public sector enrollment significantly. As public enrollment increased, a large number of private schools closed while the surviving schools lowered prices and increased school quality. To study whether different levels of public provision may have had non-zero effects on student achievement, we specify and estimate an empirical model of demand (students choosing schools) and supply (schools choosing whether to enter, stay open, and what price to charge). We estimate that a larger increase in public provision would have maximized student learning by further reducing private schools' market power without crowding out the highest quality private schools. Due to equilibrium competitive effects, we find that the optimal level is non-monotonic in the quality of the increased public schooling.

AI, and especially Generative AI, are profoundly affecting higher education, challenging longstanding philosophical, pedagogical, cultural, and economic assumptions and reshaping its core practices. Much current research is trying to understand and assess these impacts; however, future trends and developments could be even more disruptive for both faculty and students. This paper provides a longer-term perspective in the form of a horizon scan, using the IDEA protocol with a group of 27 education and AI experts from the UK and USA to identify and evaluate future issues that may significantly impact the sector but have yet to reach prominence in discussions within the sector. It identifies 14 issues (from an initial longlist of 72) with the highest potential impact over both the near-term (such as data privacy, shadow curricula, and personalized education) and the longer-term (such as AI learning companions, the dematerialization of universities, and the need to prepare for existential risk). It also identifies important cross-cutting themes that arise from these issues: reflecting on the real value of education, understanding social aspects of learning, and the social and political context of AI deployment. The paper seeks to support educators in moving from a reactive approach to handling AI impacts towards a more proactive approach of planning for possible future social and technological transitions and providing students with the skills and resources needed to navigate them.