Ulric B. and Evelyn L. Bray Social Sciences Seminar

Nested logit models represent consumers as agents that choose sequentially over product groups, hence allowing for flexible substitution patterns across products. Assuming knowledge of these nest has proven problematic in many applications. We make use of the panel structure of consumer choice data, where there are many consumers and relatively few products, to estimate both the nested structure as well as the structural parameters. We propose a two-step estimation strategy where in the first step we use clustering methods to classify products, and in the second step we estimate the model conditional on the estimated nest structure, as in Bonhomme, Lamadon, Manresa (2019).

Written with Milena Almagro