Incorporating Demographic Variables in Brand Choice Models: An Indivisible Alternatives Framework

Incorporating demographic variables in brand choice models is conceptually appealing and has numerous managerial benefits. Retailers and brand managers can assess geodemographic variations in demand and marketing mix response in order to implement micromarketing strategies. For example, a retailer planning to locate a new outlet can get some sense of the differences in demand patterns and price and promotion sensitivities in the new trading area in order to make initial stocking, inventory, pricing, and promotion decisions. For existing outlets, retailers can fine tune the assortment and merchandising activities in a category to match local market conditions. Similarly, a packaged goods manufacturers would benefit from getting a sense of whether they should stress promotional activity for one part of a brand's product line in some retail trading areas, and other parts of the brand's line in other retail trading areas. Unfortunately, a general finding across existing studies is that the impact of demographic variables on brand choice is neither strong nor consistent. These findings are puzzling given that one would expect certain demographic variables, such as income, to have some influence on brand choice behavior. Moreover, Hoch et al. (Hoch, S. J., B. D. Kim, A. J. Montgomery, P. E. Rossi. 1995. Determinants of store-level price elasticity. J. Marketing Res. 32(February) 17–29.), using store level data, find that a relatively small set of demographic variables is much more influential than competitive variables in explaining differences in price sensitivity across retail trading areas. In light of such contradictory findings, there is a need for a better conceptualization of the role of income and other demographic variables in household purchasing, and this is the primary objective of this study. This paper presents a microeconomic-based framework called the “indivisible alternatives” (IDA) framework to model household brand choice. A key aspect of the IDA framework is that it explicitly models alternatives in the market as being indivisible, while past work has either explicitly or implicitly assumed perfect divisibility. Indivisibilities force a household's selection of a brand size and its level of category expenditure into a single joint decision. Hence, demographic variables that influence category expenditure (e.g., income) also influence the choice of a brand size. In addition, since alternatives in the category come in several discrete sizes, indivisibilities introduce differences in holding costs in the choice of a brand-size combination. Consequently, demographic factors that influence holding costs through consumption rate differences (e.g., household size) impact the choice of a brand size in a product category. In this manner, indivisibilities in market offerings naturally lead to differences in choice behavior across households that can be linked to demographic factors in a logical fashion. In empirical applications to two different scanner panel data sets (ketchup and ground coffee), the proposed framework compares favorably with two benchmark specifications in terms of both goodness of fit and predictive validity. The results from these product categories indicate that a household's price sensitivity is inversely related to its income level, and that factors such as household size and seasonality, which are likely to influence consumption rates, make households more or less willing to buy larger package sizes. Income elasticity estimates from the model confirm that, ceteris paribus, households with lower incomes will have a higher propensity to purchase private labels and generic brands, and a lower propensity to purchase national brands, compared with households with higher incomes. The micromarketing potential of the IDA framework is explored in market simulations conducted for two different zip codes within a metropolitan area that vary with respect to income levels and other demographic factors. The results indicate that for the product categories studied, there are substantial differences across market areas in their response to the retail promotion of very large and very small package sizes within the same brand's product line. The differences in response suggest that it may be beneficial to customize promotional programs to market areas based on their underlying demographic composition. Finally, our findings suggest a link between household income and deal proneness that is conditional on other demographic variables and the expenditure required for the brand size in question. Specifically, we find that smaller households with lower income levels are more likely to respond to promotions for smaller, less expensive items within a category, while larger households with higher incomes are more likely to respond to promotions for larger, more expensive items within a category.