This paper develops a model for studying the problem of information intermediation faced by a platform that connects buyers and sellers. Buyers search for sellers in continuous time and are time-sensitive, while sellers have limited capacity for serving buyers and derive heterogeneous payoffs from being matched with different buyers. The platform controls the information the sellers observe about the buyers before forming a match. I show that full information disclosure is inefficient because of excessive rejections by sellers. When the platform observes the sellers’ preferences, there is a simple policy with partial disclosure that restores full efficiency. When seller preferences are unknown to the platform, I characterize the disclosure policy that maximizes the total surplus. In a setting with linear payoffs and a uniform distribution of seller attributes, I find that the optimal policy perfectly reveals low-cost buyers and pools high-cost buyers (upper-coarsening). With this policy, tighter constraints on sellers’ capacities or a higher buyer-to-seller ratio requires that less information be disclosed. For a general distribution of seller attributes, I develop an approach to solving the disclosure problem with heterogeneous and forward-looking sellers. I discuss several applications to the design of digital matching platforms.
The news topics discussed in traditional offline media like TV, radio and print newspapers differ from ones discussed in online media like blogs. The difference appears to be more evident in countries where traditional media is subject to censorship, heavily influenced by government, and, in short, not absolutely free. At the same time, blogosphere remains relatively free. In this paper, we provide a comparison between the topics discussed in online and offline media and develop an objective, automatically calculated index of media freedom. We find that the indices based on comparison of online and offline media are positively correlated with survey-based Freedom House indices for the sample of OECD countries and some developing countries. Our online-offline media index has an advantage over survey-based indices because it is computationally less costly. A theoretical model is built to illustrate how online media can be less biased than offline media when the government attempts to exert control over news sources in the country.
We analyze the incentives of online search intermediaries in environments where buyers must compete for limited supply (e.g. airlines, hotels). In our model, the intermediary manipulates the demand in two downstream product markets by choosing which product is the search default, and the cost of finding the alternative. In the absence of market power, the decentralized search decisions mimic those of a social planner with the same search technology, and thus a welfare-maximizing intermediary would set zero search costs. By contrast, an intermediary who maximizes seller revenue will optimally maintain positive search costs so that the default can be used to steer non-searchers to the market where they generate the most revenue. There may be no “right” default product: randomization may be used for both welfare and revenue maximization.
We consider a multi-object private values setting with quantity externalities. A value to a bidder from an object may depend on the total number of objects sold. For example, the likelihood a customer will respond to an advertisement is higher the fewer other advertisements are shown; a spectrum license is more valuable the fewer licenses are being allocated. In this setting we find the revenue-maximizing auction and the efficient auction. We show that both revenue-maximizing and efficient auctions have the property that the quantity of objects sold depends non-trivially on the whole profile of players’ valuations. That is, the quantity to sell is determined endogenously, within the auction. We demonstrate that auctions currently used for allocating advertising positions are suboptimal and offer simple designs that can implement (or approximate) optimal and efficient auctions under quantity externalities.
We study learning and information acquisition by a Bayesian agent whose prior belief is misspecified in the sense that it assigns probability zero to the true state of the world. At each instant, the agent takes an action and observes the corresponding payoff, which is the sum of a fixed but unknown function of the action and an additive error term. We provide a complete characterization of asymptotic actions and beliefs when the agent’s subjective state space is a doubleton. A simple example with three actions shows that in a misspecified environment a myopic agent’s beliefs converge while a sufficiently patient agent’s beliefs do not. This illustrates a novel interaction between misspecification and the agent’s subjective discount rate.