Directional Differentiability in Econometrics

University Of California-San Diego, La Jolla CA

Investigators

Abstract

This award to the University of California San Diego funds research to develop a new framework for statistical estimation of a broad group of economic models. The framework will be useful for situations that cannot be analyzed with current methods. The framework will also be useful for some kinds of statistical problems. The research aims to develop and employ a general asymptotic framework for conducting inference on non differentiable transformations of estimable parameters. The general framework will provide a unifying understanding of existing inferential methods developed for specific applications and will also ease the construction and analysis of new estimators and tests. The project will provide simple conditions that fully characterize when the standard bootstrap is consistent, will develop an alternative method that remains valid even when the standard bootstrap is inconsistent, and will study the local properties of the new tests.