Investigating the effects of marketing variables and unobserved heterogeneity in a multinomial probit model

The authors investigate the effects of marketing variables and household heterogeneity in a multinomial probit model of brand choice. The household specific vector of intrinsic brand preferences and price sensitivity is assumed to be distributed across households following some underlying probability distribution. This unknown underlying distribution is approximated by a discrete distribution with a finite number of supports in the estimation. The method of simulated moments is used in the estimation of the model parameters. The empirical application of the methodology using scanner panel data on the purchases of saltine crackers enables one to (a) study the effects of marketing variables on brand choice probabilities while allowing for similarities across choice alternatives, (b) recover differential patterns of competition across brands, (c) decompose the pattern of inter-brand competition into a component due to perceived similarities across choice alternatives and a component due to heterogeneity, and (d) determine the extent of variation in price sensitivity across households for the product category under investigation.