Identification and Estimation of Demand Models with Endogenous Product Entry and Exit

Abstract

This paper addresses the endogeneity of firms' entry and exit decisions in estimating demand for differentiated products. Under standard conditions, the selection propensity score is insufficient to control selection bias, leading to the inconsistency of conventional methods of handling selection. We introduce a novel, straightforward two-step approach to estimate demand while accounting for endogenous product entry and exit. In the first step, our method estimates a nonparametric finite mixture model of product entry and exit that accommodates latent market types. Assuming a finite-dimensional support set for the latent variable does not introduce misspecification bias in the product entry model or the corresponding selection term used in demand estimation. In the second step, our method estimates the demand parameters while controlling for selection by incorporating product entry probabilities for each latent market type. We apply this approach to data from the airline industry, revealing that conventional methods to address selection bias underestimate demand price elasticities.