|
Abstract:
|
In the literature on estimating random coefficients discrete choice models with aggregate data, typically price is the only strategic variable to firms, the model does not incorporate any form of dynamics in brand choice, and consumer heterogeneity is modeled using a unimodal parametric distribution. We adapt Bayesian semiparametric methods to a structural random coefficients discrete choice model with advertising goodwill effect, learning, state dependence, and unobserved heterogeneity. By semiparametric we mean consumer heterogeneity is modeled using the recent order-based dependent Dirichlet Process suggested by Griffin and Steel; this approach helps recover a richer variety of heterogeneity distributions while allowing the nonparametric distribution to depend on consumers' covariates. The model is applied to analyze demand for ready-to-eat breakfast cereal in selected U.S. cities.
|
