Essays on estimation of dynamic games

This thesis considers estimation of discrete choice stationary dynamic games. Chapter 1 shows that when payoffs are linear in the parameters value functions are linear in the parameters and the equation system characterizing the Markovian equilibrium is linear in the parameters. This formulation allows us to estimate the model using Least Squares. We derive an optimal weight matrix for the Least Squares estimator and show that the efficient estimator is a Generalized Least Squares estimator. Chapter 2 shows that when time invariant unobservables are present the efficient estimator is a Generalized Fixed Effects estimator. Time invariant unobservables can be correlated with observed states. We do not need to impose any distributional assumption on time invariant unobservables. Our estimators have a closed form solution. In Chapter 3 we apply the framework developed in Chapters 1 and 2 to analyze the effects of the privatization of public banks on financial development. We build a dynamic entry game to analyze the Brazilian banking market. We show that profits of private banks are positively affected by the number of public branches operating in Brazilian isolated markets. The spill-over generated by public banks is quantified based on a dynamic oligopoly model. A counterfactual in which public banks are privatized is examined. It shows that the number of active branches operating in the long-run in a small market drops significantly