Title: Robust inference in differentiated products demand models
Authors: Stephane Auray - Ensai (France)
Nicolas Lepage-Saucier - CREST-ENSAI (France)
Pujee Tuvaandorj - CREST-ENSAI (France) [presenting]
Abstract: Robust inference in random coefficient logit models for differentiated products demand is studied. The model is subject to two different irregularities that may lead to a failure of standard inference: (i) the variance of the random coefficients is often close to zero (implying little variation in tastes), which leads to the boundary parameter problem; and (ii) the strength of the available instruments is often put in doubt, which may cause weak identification. We construct test statistics that simultaneously overcome both types of irregularities. The test statistics are asymptotically pivotal i.e., their asymptotic distribution does not depend on unknown parameters irrespective of the identification strength of the instruments and the degree of heterogeneity. We evaluate the performance of the inference procedures through simulations and present an application to the U.S. automobile market.