This paper evaluates the causal impact of Sciences Po’s Conventions Éducation Prioritaire (CEP) affirmative action program on access to and success in elite higher education. Leveraging a novel instrumental variable strategy based on the random assignment of oral examiners with varying levels of grading leniency, we estimate the effect of CEP admission on students’ academic trajectories. Using linked administrative data covering all French high school graduates and detailed admissions records from Sciences Po, we find that CEP admission does not significantly increase the probability of completing a master’s degree (extensive margin), but it substantially shifts admitted students toward more selective and prestigious master’s programs (intensive margin). Our findings provide no evidence of mismatch and suggest that territorially targeted indirect affirmative action can improve access to elite academic pathways without compromising student outcomes.
Abstract: This paper studies the indirect (spillover) effects of Sciences Po’s Conventions Éducation Prioritaire (CEP), an institution-specific affirmative-action partnership with disadvantaged high schools. I link Ministry administrative records to newly available Sciences Po applications and admissions (separating CEP and BAC0 tracks) and exploit staggered school adoption, including the post-2020 expansion, using heterogeneity-robust event-study estimators with matched-student comparisons. Three results emerge. First, CEP sharply increases targeted applications to Sciences Po via the CEP track and raises Sciences Po admissions and enrollment; Regular Admission applications are unchanged. Second, CEP reallocatesstudents up the selectivity ladder: enrollment at elite Grandes Écoles (Top 10/20/30) rises, while total application numbers do not; applications shift away from Bachelor (and initially CPGE). Third, receiving-institution social composition becomes more advantaged, while baccalauréat performance is unchanged. Effects appear after 1–2 years and grow.
Abstract: This paper examines how the introduction of CEP students - who are more socio-economically and racially diverse - affects their more privileged peers at Sciences Po. Using a natural experiment from the random assignment of students into small groups (triplettes), we employ a difference-in-difference approach to assess changes in academic outcomes and social behaviors. I run a surveys to measure shifts in students’ social consciousness and friendship networks, identifying whether contact with CEP students increases awareness of inequality. This study aims to disentangle the effects of short interactions/social contact in class and the effects of friendships.
Abstract: Chlordecone, a pesticide that was found to be an endocrine disruptor and a potential carcinogenic, has been used in the French Caribbean islands for more than fifteen years after its ban by the WHO. This pesticide poisoning scandal has become associated with the exploitation and mistreatment of the French Caribbean population. I first use health survey data, to identify how the health behavior of the French Caribbean population has changed after the scandal's revelation. Then, to study how this impacted the population's mistrust in the government, I focus on Covid outbreaks. Indeed, during the massive vaccination campaigns, the French Caribbean territories stood out as outliers as the islands' inhabitants refused to get vaccinated, and up until today, covid vaccination rates remain disproportionately low in those islands, compared to mainland France and other French territories.
Abstract: This paper studies social networks in the presence of both conformity and homophily. Individuals can signal their private type by taking public actions, which can lead to segregation. Agents care about others’ perceptions of their type and place more weight on the opinions of those who are more similar to them. Agents also hold in higher esteem those whom they perceive as more similar. I first analyze the Bayes/Nash equilibria for a given exogenous network and focus on how players' actions vary depending on the network degree distribution and clustering coefficient. Then I analyze under what conditions can full separation be sustained. Finally, I study a dynamic continuous time version of the model that allows for link formations and action adjustments.