Variety and consumer demand in the retail food industry




Brand, James Michael

Journal Title

Journal ISSN

Volume Title



This dissertation addresses a number of questions which relate to a growing body of evidence for the rise of markups among large firms, and concentration in many industries, in the United States over time. Although existing evidence suggests that markups have been on the rise, there are a number of competing explanations for these trends which have differing policy implications. In this dissertation I argue that, in the retail food industry, the most likely cause of rising markups over time is the growing prevalence of “niche” goods which more closely match consumers’ idiosyncratic needs, and estimate a number of structural models of consumer demand in order to study the evidence for this hypothesis. The first chapter discusses the existing evidence that markups have been on the rise in the United States and introduces the theoretical framework and empirical evidence motivating the study of markups in the retail food industry. In this chapter I show that the number of goods carried by retail food stores in my data has grown significantly between 2006 and 2017, and that the characteristics of the offered goods have changed over time. Using a simple theoretical model, I argue that firms’ incentives to offer “niche” goods, which I define herein, grow as they become able to stock a larger portfolio of goods. I also introduce the data set which is used in all three chapters of this dissertation and briefly discuss the way in which I select the categories which are the focus of my empirical work in Chapters 2 and 3. The second chapter introduces three empirical models of consumer choice in nine retail food categories. The first is a traditional mixed logit model in the spirit of Berry, Levinsohn and Pakes (1995), which has been applied in many studies in industrial organization. The second is an approximation of this model, developed by Salanie and Wolak (2019), which allows consumer preferences to differ flexibly in every three-digit ZIP code in my data. The third model is a constant elasticity model of demand, which makes different assumptions than do mixed logit models and serves in part as a robustness check against their potentially strong assumptions. I estimate each of these models separately in nine large categories of products in 2006 and 2017 and demonstrate that each model implies that consumers have become significantly less price sensitive over time. A simple pricing rule demonstrates that the firms in my data may have been able to sustain larger markups over time solely due to the observed changes in the price sensitivity of consumers, absent any changes in firm pricing behavior. In the third and final chapter, I estimate two additional models of demand in order to determine the structural reasons that consumers have become less price sensitive. First, I demonstrate that an assumption restricting changes in unobserved product quality over time allows us to distinguish between the effects of rising horizontal differentiation and changes in the direct disutility of price due to, for example, changing consumer demographics and wealth. In a second model, I introduce a new scaling parameter which allows me to measure differences in consumers’ price sensitivity for newer and older goods (as measured by how long a good has been sold in a store in my sample) separately. Together, these models provide evidence that horizontal differentiation has increased significantly over time and that consumers are particularly insensitive to changes in the prices of newer goods. I take these findings as evidence that niche products do play a significant role in explaining the results of Chapter 2, as predicted by the model introduced in Chapter 1.



LCSH Subject Headings