Hersh Chopra

This paper develops a theory of credential rationing when a privately informed sender relies on an intermediary to persuade receivers. The intermediary sells tests to the sender, who possesses imperfect private information about his quality and seeks to persuade the receivers to match with him. Screening the sender's private information distorts the allocation of tests and matches. To economize on information rents, the intermediary restricts the supply of high credentials by steering allocation toward less selective receivers. Furthermore, this rationing can shift matches toward the bottom receiver and reduce the top receiver's share relative to the no-screening benchmark.

We study non-cooperative information acquisition by bidders who will compete in an optimally-designed auction. The bidders simultaneously choose experiments privately revealing information about their respective willingnesses to pay. The auctioneer, knowing the experiments but not the realizations, designs an optimal auction against the resulting information structure. Bidders' best-responses are myerson-regular: they require no ironing. In the symmetric equilibrium we characterize, each bidder adopts an information structure that satisfies a somewhat stronger regularity property. Among symmetric profiles with this property we show that it is the unique equilibrium.

We study efficient and sender surplus-maximizing information design in selection markets, specializing our exposition to credit markets. A borrower (sender) designs information about the riskiness of his project and seeks to finance his project through a monopolist lender. We characterize the borrower-optimal learning when the lender's utility depends on both the demand of borrowers and on the risk composition of the borrowers. We use this characterization to describe the relation between the set of efficient equilibrium welfare, risk dispersion, and the shape of the repayment contract (securities).

We study efficient and consumer-surplus maximizing information policies in a bilateral trade setting where the buyer is initially privately imperfectly informed about his willingness to pay. We identify a canonical class of demand functions that can be implemented by information disclosures that are targeted based on the buyer’s initial private information. As an application we show that providing more information to the buyer can lead to higher market prices and a lower trade probability without affecting the consumer or producer surplus.

We study the design of mechanisms by an intermediary that generates information for a sender to persuade a receiver about an unknown attribute of the sender. The sender is initially privately, but imperfectly, informed about her attribute, and the receiver takes an action based on posterior beliefs about the sender’s attribute and the sender’s belief about the attribute. The mechanism generates information for the sender and also controls its disclosure to the receiver. The design of the optimal mechanism needs to screen the privately informed sender and thus confronts incentive-compatibility constraints. The mechanism also deals with obedience constraints, as the intermediary must generate just enough information to persuade the receiver. We characterize incentive-compatible mechanisms for a wide class of problems when the sender contracts with the intermediary. We use this characterization to study profit-maximizing mechanisms in three applications: the design of college-admissions tests, the optimal use of consumer data on a digital market platform, and the optimal design of credit rating schemes.

A sender with private information (high or low ability) tries to convince a receiver of having higher ability. A certifier offers a menu of Blackwell experiments and prices to screen the sender. The sender uses the experiment’s outcome to persuade the receiver to take a favorable action. This paper focuses on the equilibrium interaction in this certification game when the receiver can distinguish between outcomes of the experiment based on the hard information contained in the outcome. With binary ability, for a given experiment, the hard information of an outcome is the likelihood ratio of that outcome. The main result characterizes all possible equilibrium outcomes in terms of a convex combination of menus containing only simple experiments. Using this characterization, I show the existence of an equilibrium in which soft information overrules hard information; due to equilibrium self-selection of the sender, some outcomes whose hard information makes the receiver more pessimistic about the sender's ability end up persuading the receiver to choose the favorable action.

We study how a buyer and seller can independently and non-cooperatively choose to learn about their valuation and cost, respectively, to influence bargaining outcomes.

I develop a dynamic model to study the emergence of AI in markets that rely on learning-by-doing to train the human counterparts. I show that joint automation and productivity increases can be long-run welfare-reducing compared to interventions that automate only or increase productivity only. I then describe market mechanisms, such as reliance on humans to train AI can restore efficiency.

In this note, I describe how information can act both as a substitute and a complement when a decision maker anticipates receiving more information in the future.