Identification of Discrete Choice Demand From Market Level Data

Joint with: Philip A. Haile, Yale University

We consider nonparametric identification of random utility discrete choice models of demand for differentiated products. We examine the case of market level data, i.e., observations of product characteristics, market characteristics, and market shares. Our representation

of preferences nests random coefficients discrete choice models widely used in practice in the literature on demand for differentiated products; however, our modelis nonparametric and distribution free. It allows for choice-specific unobservables, endogenous choice characteristics (e.g., prices), and high-dimensional taste shocks with

arbitrary correlation and heteroskedasticity. Using standard conditions from the literatures

on mulitnomial choice, nonparametric instrumental variables, and simultaneous equations, we demonstrate the identifiability of demand and of the full random utility

model.

For more information, contact Petra Todd.