Shrinkage priors for high-dimensional demand estimation

Abstract Estimating demand for large assortments of differentiated goods requires the specification a system that is sufficiently flexible. However, flexible models are highly parameterized so estimation appropriate forms regularization to avoid overfitting. In this paper, we study Bayesian shrinkage priors pairwise product substitution parameters. We use log-linear as leading example. Log-linear by own and cross-price elasticities, total number elasticities grows quadratically in goods. Traditional regularized estimators shrink regression coefficients towards zero which can be at odds with many economic properties price effects. propose hierarchical extension class global-local commonly used modeling allow direction rate depend on classification tree. both simulated data retail scanner show that, absence strong signal data, estimates predictions improved imposing higher-level group rather than zero.