Physician learning and new drug diffusion

This paper studies the diffusion of new drugs by assessing the role of physicians' dynamic learning stemming from the effects of the two information channels: firms' marketing activities (e.g., detailing) and patients' experiences with their prescriptions. We use a physician-level dataset in the Erectile Dysfunction category that includes both prescription records and direct sales activities of firms. Empirical evidence for physicians' forward-looking behavior comes from the positive relationship between the size of a physician's patient base and the extent to which (s)he engages in early prescribing of the new drug. Our data also suggest that physicians with large patient bases strategically substitute patient experimentation for learning from detailing visits. We develop a dynamic Bayesian learning framework that considers such forward-looking behavior and also accounts for differences in learning across physicians.;We find that accounting for forward-looking behavior improves the model fit and that a static model provides incorrect inferences regarding the effects of detailing by ignoring strategic substitution across multiple information sources in physician learning. Estimation results indicate that both channels contribute to physician learning in the early period of a drug's life cycle and that patient feedback contributes more to a physician's learning than does firms' detailing. Given our finding of a significant difference between the two new drugs in the effectiveness of their detailing, we investigate the consequences of heterogeneous detailing strategies for the firms. Policy experiments based on the parameter estimates reveal that the new entrants' informative detailing activities benefit consumers via faster diffusion of the new drugs. From the firm's perspective, if Cialis, which has the highest mean efficacy, had followed Levitra's detailing strategy, it could achieve higher profits. Lastly, a cost-neutral reallocation of detailing across physicians shows that firms' volume-based detailing strategies are not necessarily optimal.