Panel Data Discrete Choice Models of Consumer Demand

1. Introduction This chapter deals with the vast literature on panel data discrete choice models of consumer demand. The reason this area is so active is that very high quality data is available. Firms like Nielsen and IRI have, for over 30 years, been collecting panel data on households' purchases of consumer goods. This is known as ―scanner data,‖ because it is collected by check-out machine scanners. Available scanner data sets often follow households for several years, and record all their purchases in several different product categories. The typical data set not only contains information on the universal product codes (UPC) of the consumer goods that households buy on each shopping trip, but also information on several exogenous forcing variables, such as price and whether the goods were displayed or advertised in various ways. To my knowledge the first paper using scanner data to study the impact of price and other marketing variables on consumer demand was Guadagni and Little (1983) in Marketing Science. But few economists knew about scanner data until the mid to late 90s. Once they became aware of this treasure trove of data, they started to use it very actively. Today, estimation of demand models on scanner data has become a major part of the field of empirical industrial organization. Thus, the consumer demand literature based on scanner data is unusual relative to other literatures discussed in this Handbook in two respects. First, it remains true that the majority of work in this area is by marketers rather than economists. Second, this is an uncommon case where the ―imperial science‖ of economics (see, e.g., Stigler (1984)) has experienced a substantial knowledge transfer from another area (i.e., marketing). Furthermore, it should be noted that discrete choice models of consumer demand are also widely used in other fields like transportation research, agricultural and resource economics, environmental economics, etc.. Given that the literature on panel data models of consumer demand is so large, I will make no attempt to survey all the important papers in the field. Instead, I will focus on the main research questions that dominate this area, and the progress that has been made in addressing them. Thus, I apologize in advance for the many important papers that are not cited. The most salient feature of scanner panel data is that consumers exhibit substantial persistence in their brand choices. In the language of marketing, consumers show …