Essays on consumer search

This dissertation consists of three papers which are aimed at extending consumer search theory by relaxing some of the informational assumptions and by inquiring into the incentives of firms to share information. The first paper analyses firm's incentives to share information in search markets. Many benchmarks in over-the-counter markets, such as LIBOR, are formed based on the submission of banks' interbank interest rates. Benchmarks are important, as they help consumers to search and provide firms with a basis for price setting. This paper explores firms' incentives to contribute information about their costs for the purpose of benchmark formation. We find that in order to deter consumer search, firms find it optimal to share their private costs if the search friction is sufficiently high. Moreover, the reduction of search in these markets can dramatically decrease social welfare. The second analyses a model of oligopolistic competition in which consumers search without priors. Consumers do not have prior beliefs about the distribution of prices charged by firms and thus try to use a robust search procedure: they minimise the loss relative to the searcher, who knows the price distribution, in the worst case scenario. We show that the optimal stopping rule is stochastic and that for any distribution of search costs there is a unique market equilibrium which is characterised by price dispersion. Although listed prices approach the monopoly price as the number of firms increases, the effective price paid by consumers does not depend on the number of firms. The third paper considers a problem of sequential search, in which a searcher relies on her initial information to make inferences about the unknown distribution of alternatives. Bayesian search rules are good only for specific environments and are often difficult to characterize. In this paper, we propose a robust search rule that adapts to the information acquired while searching and performs well with respect to Bayesian optimal strategies for any priors. In a setting with binary-valued payoffs, we show that initial information and learning restrict the set of possible environments which gives rise to the adaptive rule with improved performance.