Uncertainty and Learning in Pharmaceutical Demand

Di erences in the e ectiveness and side e ects of di erent drugs introduces uncertainty into pharmaceutical demand. This in turn introduces dynamic aspects to drug choice: patients have an incentive to experiment to gain information useful for future choices. Exploiting a rich panel dataset on anti-ulcer drug prescriptions, the purpose of this paper is to measure the importance of uncertainty and learning for pharmaceutical drugs. To do so, we specify and estimate a dynamic structural model of demand for anti-ulcer drugs which embeds drug choice in a patient's multi-period expected utility maximization problem. Patients learn in Bayesian fashion about the e ectiveness of alternative drugs. Our results suggest that patients have considerable uncertainty as they begin treatment, but that it decreases quickly with drug use. Simulations imply increases in treatment costs due to uncertainty are modest, ranging from 1.5-4.2% depending on illness severity.