Search for differentiated products: identification and estimation

TLDR

Consumer search decisions for differentiated products can be driven by low search costs or low tastes for already found products. The paper proposes an identification strategy to estimate search cost distribution amid unobserved tastes. The strategy uses conditional search decisions—search actions paired with prior product displays—and is applied to clickstream data from a hotel search platform. The estimated price elasticity of demand from the search model differs from the static model, highlighting bias from endogeneity of search‑generated choice sets.

Abstract

When consumers search for differentiated products, a given search decision can be explained either by low search cost or by low tastes for the set of products already found. We propose an identification strategy that allows to estimate the search cost distribution in the presence of unobserved tastes. The required data takes the form of conditional search decisions: observations of search actions combined with previously observed product displays. We develop an application using clickstream data from a hotel search platform. Estimates of price elasticity of demand in the search model differ from those in the static model, reflecting the bias due to endogeneity of search‐generated choice sets.