Essays on the Econometrics of Discrete Games of Complete Information

Abstract

This dissertation considers questions in the econometrics of discrete games of complete information. Chapter 1 proposes an econometric framework that explicitly models the selection of equilibria as a discrete choice model. Such framework could achieve point identification. A simulation based approach is proposed to estimate the structural model. Chapter 2 discusses the relationship among the three types of partial identification approaches available in the literature. I show a modified version of bound estimation is equivalent to the sensitive analysis approach and the sharp identification approach in the sense that they draw inference on the same identified set. Chapter 3 studies identification and estimation of discrete games in large networks. Moment inequalities on choice probabilities of subnetworks are constructed to partially identify the model. Monte Carlo studies are conducted in Chapter 1 and 3 to evaluate the performance of approaches developed in this dissertation. Two applications are considered, Chapter 1 contains a study on the entry competition in the home improvement industry; Chapter 3 studies peer effects on smoking in friend networks.