Hirotaka Ita

Barcelona School of Economics · Economics · 2026–27 job market

Fields

Econometrics

Job market paper

Asymptotic Minimax Optimality of Set Estimation

Non-convex identified sets of parameters arise naturally in econometric models through nonlinear relationships and structural identifying restrictions. Yet existing efficiency theory for set estimation relies on convexity, leaving open the fundamental question of how accurately these sets can be estimated from data. This paper develops a general theory of asymptotically efficient set estimation that accommodates a wider class of identified sets. I first derive the local asymptotic minimax efficiency bound under risk functions based on the Hausdorff metric. A key ingredient in the analysis is a limit experiment framework tailored to the local geometry of the identified set. I then establish the limiting distribution of a set estimator based on a criterion function. Unlike existing approaches that rely on support functions, the analysis preserves the full geometry of non-convex sets. The limiting distribution motivates a bootstrap-based confidence set, whose finite-sample performance is examined through Monte Carlo simulations. An application to optimal top income taxation with behavioral agents illustrates the empirical relevance of my framework.