In this lecture we consider the structural estimation of dynamic discrete games of incomplete information. We first introduce the behaviral framework using the dynamic entry/exit game analyzed in Aguirregabiria and Mira (ECMA, 2007), which involves describing the evolution of the state variables (which is affected by all players actions) and the player i’s (dynamic) utility maximization problem. Given this, we introduce the equilibrium concept, which is Markov Perfect Equilibrium (MPE) that both requires that all agents are behaving optimally (Bellman Optimality) conditional on other players actions, and that all players choices are mutual best responses (ensured by the Bayes-Nash Equilibrium Conditions). We then discuss how to solve for Markov Perfect Equilibrium and consider a variety of estimators (MPEC, NFXP, NPL, and 2-step methods).

Once all the machinery is up and running, we discuss the empirical results from A & M (2007), where the exit/entry model is estimated Chilean panel data from several retain industries.

Finally, we study the (statistical and computational) properties of the various estimators empirical based on a review of the Monte Carlo results from Egesdal, Lai, Su (QE, 2015)

Readings:

Material:

Slides

Video lecture

Lecture 19: Structural Estimation of Dynamic Games using MPEC, NPL and 2-step-PML