A Bayesian Semi-parametric Approach for Random Coefficients Demand Framework

Abstract: 

This paper interprets and develops Berry et al. [1995] model in the Bayesian framework. First, we partition the consumer utility into two components: mean and heteroskedasticity. Our decomposition simplifies the interpretation of preference heterogeneity by linking it to the heteroskedasticity. Second, we model the mean utility nonparametrically by using a Dirichlet mixture model (DPM) while preserving the same consumer heterogeneity structure. This method identifies the importance of nonlinearity in the mean utility function. Third, we propose a new DPM framework to take account of discrete variables such as brand dummy variables. Fourth, we present a novel approach to utilize instrumental variables nonparametically in the presence of price endogeneity. Lastly, we apply the new model to Australian apple market to show its advantages.