Economics Silvia Sarpietro - "Fixed Effects in the Tails (with Raffaella Giacomini, Annalisa Loviglio, and Yulong Want)"

Learning about the tails of unobserved heterogeneity is crucial for policy analysis, as many policy questions focus on extreme units, such as the most effective teachers or schools. Yet standard methods do not directly target the tails. We develop a method for ranking, estimation, and inference on fixed effects in the tails of the distribution of unobserved heterogeneity using micro-panel data. Unlike empirical Bayes methods, our semi-parametric approach expands fixed-effect estimates in the tails rather than shrinking them and avoids imposing assumptions on the full error distribution, while allowing arbitrary dependence between errors and fixed effects. Applied to Italian administrative school data, our estimates imply that replacing the bottom 5% of schools with average-quality schools would increase student achievement by more than 0.4 standard deviations, estimates that differ substantially from those obtained using existing methods and highlight the importance of credible tail inference.