Measuring Changes in Brand Choice Behavior

The multinomial logit model is frequently used in marketing research to explain consumers’
brand choice decisions. In almost all applications of this model, the parameters of
the consumers’ utility function are assumed to be constant across time. In contrast to this
assumption, both marketing theory and statements from marketing practitioners suggest
the possibility of short-term fluctuations and long-term changes in consumers’ brand
choice behavior. In this paper, nonparametric brand-specific time-variable functions
replace the brand-specific constants usually found in brand choice models. I estimate the
model for panel data from two product categories and derive management implications.

Keywords: Brand Choice Model; Generalized Additive Model; Multinomial Logit, Paneldata, Estimation using R (S-Plus)