ESSAYS ON PHARMACEUTICAL MARKETS

Abstract

This dissertation consists of three distinct essays on the studies of pharmaceutical markets using discrete choice models.

The first chapter considers a simple framework to model the choices of multiple goods. In a standard discrete choice model, an agent is assumed to choose one object from a finite set of alternatives. In practice, it often happens that an agent can choose more than one object. The traditional approach of re-defining choice sets has shortcomings. Usually, we don't have enough variables to identify the parameters associated with the goods combinations and the likelihood function involves high dimensional integrals. In this paper, I propose an alternative model where a choice of several products is assumed to be the result of a sequence of simpler single product choices. The resulting likelihood is easy to evaluate and the model sheds light on the product complementarity and consumers' satiation patterns. An empirical study of prescription drug choices is presented.

The second chapter studies the evolution of market demands after the entry of new products into the prescription ED drug market. I propose a dynamic discrete choice model to investigate the choice and learning process about the new products. The model also explicitly takes into account that prescription decisions are potentially made jointly by physicians and patients. A new solution algorithm is proposed. Using a micro-level dataset, I find results that are informative for understanding consumers' learning processes and for understanding specific features of this pharmaceutical product market. Specifically, I find that: 1) Patients play an important role in drug choices. 2) Drug promotions generate rather precise information about new drug quality. 3) If drug advertising is prohibited, the diffusion of new drugs will be slowed down.

The third chapter is a more detailed investigation on the direct (non-informational) effects of physician-targeting promotions in the mature diabetes drug market and cholesterol-reducing drug market. I find that prescription decisions are more likely to be influenced by physician-specific heterogeneous preferences than by promotional activities. This paper also proposes a bi-linear probability model to explicitly model the patient-level heterogeneity. The major results are unchanged after controlling for patient heterogeneity.