Consumer learning and heterogeneity: Dynamics of demand for prescription drugs after patent expiration

This paper introduces an empirical demand model with aggregate learning and consumer heterogeneity in price sensitivity. The model is applied to the prescription drug market after patent expiration during the 80s, when the diffusion of generic drugs is fairly slow and consumer price sensitivity is known to be heterogeneous. I also introduce a new estimation approach to take the price endogeneity problem into account for this class of models. My approach is to estimate the demand model jointly with a pseudo-pricing policy function, which is a reduced-form function of observed and unobserved state variables. My estimation and counterfactual exercises demonstrate the value of estimating such a structural demand model, e.g., it allows one to learn whether consumers have optimistic or pessimistic prior, how aggregate learning affects the product diffusion rates in different latent segments of the population, and quantify the role of learning in explaining new product diffusion. I also find evidence that the brand-name price elasticities of demand (evaluated at the observed prices) are often less than one and increase over time, suggesting that brand-name firms might set their prices lower than what they would do if they were myopic, in order to slow down the learning process for generic qualities. But such an incentive might diminish over time as the uncertainty slowly resolves.