Observational Learning and Demand for Search Goods

In many differentiated good markets like music, books, and movies, the choice set of available products is overwhelmingly large, with many new products flowing into the market each week. Consumers are not aware of or are poorly informed about many of the available products. They learn about products and their preferences for them from the search and purchasing decisions of other consumers and through their own costly search. We use a variant of the sequential search models of Banerjee [7], Bikhchandani, Hirshleifer, and Welch [11] and Smith and Sorensen [34] to study market demand in these kinds of markets. We characterize the equilibrium dynamics of demand under different assumptions about the kind of information that consumers observe about the decisions of other consumers. Our main result consists of showing that long-run behavior in herding models can be invariant to the learning dynamics. We exploit this result to study the determinants of the likelihood that high-quality products end up with no sales and of the likelihood that low quality products end up with significantly positive sales. ∗Hendricks: Department of Economics, University of Texas, Austin, TX 78712. Sorensen: Graduate School of Business, Stanford University, Stanford, CA 94305. Wiseman: Department of Economics, University of Texas, Austin, TX 78712. We received very helpful comments from Daron Acemoglu, Nathan Hendren, Johannes Horner, Phil Reny, Lones Smith, and the participants at the microeconomics seminars at Brown University, Duke University, Hong Kong University, The Workshop on Industrial Organization at University of Santiago, Chile, and the “Theory in A2” Confererence at the University of Michigan.