Home / Events / Sharp and Robust Estimation of Partially Identified Discrete Response Models
Statistics Seminar

Sharp and Robust Estimation of Partially Identified Discrete Response Models

With Tatiana Komarova (University of Cambridge)

Sharp and Robust Estimation of Partially Identified Discrete Response Models

Semiparametric discrete choice models are point identified with continuous covariates and may become partially identified with discrete covariates. Classical estimators, such as maximum score (Manski (1975)), lose their desirable properties without point identification. They may not be sharp, converging to outer regions the identified set (Komarova (2013)), and in many discrete designs, weakly converge to random sets. They lack robustness as their distribution limit changes discontinuously with model parameters. We propose a new class of estimators based on the quantile of a random set, which are both sharp and robust, also applicable to single-index and discrete panel data models.

(joint work with S.Khan and D. Nekipelov)

Add to your calendar or Include in your list