Structural State Dependence and Consumers' Unobserved Heterogeneity: A Case Study of the U.S. Coffee Market

Abstract

In many markets of differentiated products, consumers show persistence in their consumption choices. Understanding how past consumption affects consumers' current choices and distinguishing different explanations are beneficial for price setting. In this paper, a discrete choice model is used to investigate the state dependence effect on the U.S. coffee market. Using a disaggregate data set that captures a detailed purchase history and demographic information for a massive number of consumers, I first estimate the spurious state dependence effect without assuming consumer heterogeneity. I then consider both structural state dependence and consumers' unobserved heterogeneity and distinguish those two effects by applying the discrete mass point algorithm semiparametrically. I further explore the potential correlated errors and endogeneity problem and compare the various methods to solve them. My findings show that structural state dependence and consumers' unobserved heterogeneity are the main explanations for the observed consumption persistence in the U.S. coffee market. My findings also suggest that there exist two types of consumers in the U.S. coffee market: one type treats ground coffee as normal goods and the other treats ground coffee as inferior goods. The structural state dependence is positive and significant for one type of consumers and negative for the other type of consumers.