Large Deviation Approach for Moment Condition Models

Yale University, New Haven CT

Investigators

Abstract

This project applies large deviation theory to evaluate econometric methods for moment condition models. Large deviation theory enables us to explore the first-order but global asymptotic properties of econometric methods and provides a different viewpoint from the conventional econometric theory, which mainly focuses on local asymptotic properties. The project introduces the large deviation approach to three decision problems for moment condition models: parameter estimation and testing, set inference, and moment selection. The main purpose of this project is to find large deviation optimal econometric methods when existing methods are indistinguishable under the conventional econometric theory (parameter estimation and testing problems) or there is no available framework to evaluate econometric methods (set inference, and moment selection problems). As a broader impact, applications of those optimal methods to empirical economic analysis, such as analysis on returns to schooling and income dynamics, are beneficial for society by providing some new insights on policy studies.