Heterogeneous Consumer Search and Social Learning: Empirical Studies

University Of California-San Diego, La Jolla CA

Investigators

Abstract

This pair of research projects explores two aspects of consumer demand that can have important implications for market efficiency: heterogeneous consumer search and social learning. Both projects make use of new datasets that are uniquely well-suited for the questions to be addressed. The first project examines the nature and intensity of consumer search in retail markets for prescription drugs. Retail prices for prescription drugs vary widely across pharmacies even within the same local market, and previous research has shown that patterns in price dispersion are consistent with economic models of consumer search: for instance, prices tend to be much less disperse for medications that are purchased repeatedly (for which the incentives to price-shop are strongest) than for one-time medications. This project will investigate data on over 200,000 prescription purchases in two small cities. For each transaction, the data identify the actual price and payment type (cash, third-party insurance, or Medicaid) as well as prescription characteristics. Since data are available on prices and quantities at all competing pharmacies, information about the extent of consumer search is revealed in the observed sensitivity of market shares to relative price differences. The research will provide answers to important questions about price dispersion and consumer search: How intense is consumer search in these markets (i.e., what fraction of consumers chooses to price-shop before purchasing?) Is variation in search intensity across drugs consistent with the predictions of economic theories of consumer search? What do the estimated search intensities reveal about the underlying distribution of search costs? The second project concerns the impact of social learning on individuals' choices of health insurance plans. Employees of large firms typically choose from a set of health insurance options sponsored by their employer; since the choice of a health plan can be a complex decision, individuals may attempt to learn about plans' relative qualities from co-workers-either through direct communication or by drawing inferences based observed choices. The causes and consequences of this kind of learning have been formalized and discussed in the theoretical literature, but little effort has been made to measure the importance of social learning empirically. This study will use data on the health plan choices of the employees of a major university to obtain quantitative estimates of the effects of social learning. Information about employee demographics--and especially the departments in which they work-will be used in identifying the communication channels through which learning occurs. Stated very simply, if individuals' decisions are significantly more similar within departments than they are across departments, this can be taken as suggestive of a learning or "herding" effect. This is the basic idea underlying the project's principal approach, which is to estimate an econometric model of health plan choice that incorporates the influence of neighbors' decisions on the decision of the individual. The empirical results will yield important insights into the nature and extent of learning, the implications for demand analysis and health plan competition, and the magnitude of potential welfare losses due to imitative behavior.