Essays on the Econometrics and Empirics of Bayesian Games
Zhao, Jun
0000-0001-8846-347X
:
2021-05-17
Abstract
A Bayesian game is a game in which players have incomplete information about the other players. As a natural extension of games with complete information, Bayesian games are usually difficult to analyze both theoretically and empirically, due to the challenge in solving the equilibrium strategies and studying their properties. Examples of Bayesian games include the contest model and the auction model, whose structural analysis requires different econometric techniques. For auction models, the first order conditions are characterized as differential equations with well-defined boundary conditions, while for contest models, they are integral equations. Such discrepancy leads to case-specific identification and estimation strategies when studying the empirics of different Bayesian game models. This dissertation focuses on the empirical analyses of Bayesian games with applications like contest models and procurement auctions, and applies the state-of-the-art econometric techniques to study empirical questions arising from IO and political economy. The first chapter quantifies the influence of campaign spending on electoral outcomes in the U.S. Senate elections given endogenous entry, through constructing a two-stage game-theoretic model where candidates’ participation and spending behavior is modeled as an equilibrium result from a contest with incomplete information and endogenous entry, and specify voters’ decisions via a discrete choice model. We recover the private information held by the candidates but unobserved to econometricians and conduct counterfactual analyses to evaluate an alternative political institution. In the second chapter, we study nonparametric identification of Bayesian games within the private type paradigm and with an additive payoff function, where a part of the additive payoff function is unknown to the researcher. We impose the exclusion restriction in the form of an exogenous players' participation and establish the nonparametric identification results. The third chapter examines the possible bid-rigging behavior in procurement auctions. We propose a rank-based test for collusion that allows for affiliated private costs among bidders. The proposed test is adopted to analyze the possible bid-rigging occurrence in Japanese procurement auctions with the bridge construction projects. The results indicate evidence of prevalent collusion before the middle of year 2005.