AbstractsEconomics

The first chapter studies semiparametric point identification and estimation of complete information entry games and proposes a root-n consistent estimator. The proposed method focuses on a two-player entry game using an example of discount retailers, where the potential profit of one retailer depends on the actions of its competitor, and the unobserved heterogeneities of the two retailers can be correlated. These two features lead to two challenges in identification and estimation: multiple equilibria and endogeneity. To address these two challenges, this paper provides a new identification and estimation strategy under a symmetry condition on unobservables. This new identification procedure requires neither an equilibrium selection rule of multiple equilibria nor parametric distributional assumptions on unobservables to solve the endogeneity problem. It also requires a weaker support condition than that in the existing literature. Following the identification argument, this paper proposes a semiparametric two-step estimation procedure using plug-in kernel estimators. Given the symmetry assumption, this paper shows that the proposed estimator is root-n consistent, unlike existing estimators for this model. The second chapter considers a Monte Carlo simulation study for complete information entry games. The purpose of this study is to provide evidence consistent with the root-n consistency of the semiparametric estimator proposed by Zhou (2014a) and to compare this proposed estimator with an existing parametric estimator. The results are consistent with the proposed estimator being root-n consistent, as predicted by Zhou (2014a). In addition, the parametric estimator outperforms the semiparametric estimator with lower biases and variances when the model is correctly specified. When the model is incorrectly specified, the parametric estimator is inconsistent, while the semiparametric estimator is consistent. The third chapter applies existing parametric estimation methods and a new semiparametric estimation method by Zhou (2014a) to entry games of discount retailers. Using data on Kmart's and Walmart's entry decisions in 1997 across counties in the U.S., this paper finds that, with a caveat for the possible misspecification of the latent function, semiparametric and parametric estimators give similar estimates. This result informally suggests that normality seems to be a reasonable approximation for the distribution of unobservables in the discount retailing industry.