The Flexible Inverse Logit (FIL) Model

Abstract

Measuring substitution patterns across differentiated products is at the heart of many empirical studies. Most of the approaches used in applied work, including the leading approach pioneered by Berry, Levinsohn, and Pakes (1995) (BLP), impose distributional and functional form assumptions that may restrict substitution patterns in the space of product characteristics. In this paper, I propose a flexible approach that avoids making arbitrary assumptions on the way substitution patterns depend on product characteristics. I show that my approach yields substitution patterns that BLP cannot recover, including complementarity in demand. I further show that the inverse demand model on which I rely is consistent with utility maximization by heterogeneous consumers. This makes my approach useful to address many economic questions of interest, including the measurement of market power and merger evaluation.