Published online by Cambridge University Press: 06 July 2010
Appreciating the opportunity
Figure 8.1 portrays the price history of two yogurt brands in a US city during the 1980s. There was little price differentiation between the brands during the 126-week period in the graph, and also clearly little price variation within each brand. In addition, product characteristics and marketing activities (e.g., advertising, promotions) varied little during this period. A choice model estimated from these scanner panel data for these brands would indicate that the only significant explanatory attributes were the brand-specific constants and past behaviour (i.e., some form of behavioural inertia; see Guadagni and Little 1983).
A choice model cannot detect the effect of price and other marketing attributes in such data for the simple reason that the parameters of any statistical model are driven by the degree of variation in each attribute, not by the absolute levels of the attributes represented in the data. Hence, if there is little variation in an explanatory attribute, its parameter cannot be estimated.
The situation depicted in figure 8.1 is quite common, and is much more the rule than the exception in competitive markets. Indeed, think about how in your own experience the prices of many products/services move up or down at a market level: credit card interest rates, home mortgage rates, fuel, etc. This happens because competition tends to result in similar products with similar marketing activities.