Essays on the Econometrics of Dynamic Discrete Choice

Abstract

Disentangling the causal effect of policy from that ofbehavior---i.e. controlling for selection---is a foundational empirical challenge in economics. Dynamic discrete choice models are a structural approach that posits that selection is driven by forward-looking, optimizing decision makers. The resultant econometric problem is to recover the structural parameters that characterize the model. This dissertation contributes tothe econometrics of dynamic discrete choice models in several directions. Chapter two shows identification of dynamic discrete choice models with continuous permanent unobserved heterogeneity. That is, the model allows for infintely many persistent unobserved differences between decision making agents. The previous literature allowed for only finitely many types of persistent unobserved differences. The third chapter provides a hypothesis test for animportant modeling assumption, that of ‘homogeneity’. Commonly, it is assumed that behavior is sufficiently similar across time or markets that data can be pooled across these dimensions. However, this assumption may fail in the presence of a structural break or multiple equilibria. The chapter proposes a hypothesis test to evaluate whether the homogeneity assumption holds in the data. As an approximate randomization test, the hypothesis test is valid even without a large sample. The fourth chapter provides a computationally advantageous estimator for dynamicdiscrete choice models. The estimator is based on the observation that dynamic discrete choice models possess a multiple index structure. The chapter shows that index sufficiency can be used to construct a set of equality constraints, which restrict the parameter of interest to belong to a subspace of the parameter space. The chapter proposes an estimator that imposes these restrictions, thus attaining computational gains by reducing the effective dimension of the parameter space.