On Consumer Heterogeneity in Food Demand Analysis: Application of Random Effects and Hierarchical Bayesian Models

Abstract

This paper explains how hierarchical Bayesian models with random effects of consumer heterogeneity are useful in econometric analysis of consumer demand, especially food demand, using scanner panel data. Hierarchical Bayesian models are found to be very popular in marketing research but, in contrast, hardly used in consumer demand analysis. One of the main reasons will be the way econometric theory considers random-effects models, which are not always consistent in estimation. As scanner panel data are more and more used, hierarchical Bayesian models will gradually become popular in food demand analysis, which is part of agricultural economics as well as part of consumer demand analysis and therefore flexible enough to incorporate popular methods of other disciplines such as marketing research. Hierarchical Bayesian models can estimate the random effects of individual consumers as well as avoiding the problem of overfitting, thereby increasing generalization ability or improving external validity.