Junrok Park

Research

Buyers-optimal Learning In Collective Purchase Problems (Submitted)

• Abstract: Multiple buyers negotiate with a monopolist over a public good—such as municipalities financing emission controls or residents funding flood barriers—without knowing their valuations. We examine how buyers can design information before bargaining. A benevolent expert commits ex-ante to an experiment, sending private signals to buyers; anticipating this, the monopolist selects an incentive-compatible, individually rational mechanism. We show that the equal-profit (EP) experiment, which makes the seller indifferent between selling and withholding supply at each signal profile, minimizes expected profit and maximizes buyer surplus. The expert uses signal correlation or interim-value interdependence to implement the EP design. In a commonvalue environment, we characterize the EP experiment for any group size and trace its properties as the number of buyers grows. We find interdependence essential: for large groups, the buyer-optimal design induces independent signals and grants each buyer a positive payoff. In addition, a partial centralization scheme—learning together while individually bargaining—outperforms both complete centralization and decentralization. (JEL D82, H41, C78)

Optimal Robust Double Auction (with Pavel Andreyanov and Tomasz Sadzik) (Reject-and-resubmit at American Economic Review)

• Abstract:  In a stylized exchange economy with single, continuous good and quasi-linear utilities, we propose a novel double-auction format featuring two (forward and reverse) clock auctions, Vickrey-style payments, and carefully designed per-unit taxes. In the spirit of Ausubel (2004), we show a sincere ex-post perfect equilibrium of the game and that the market-clearing price is the Walrasian equilibrium price in an economy with deformed utilities. Furthermore, we show how the auctioneer can adjust the clocks dynamically to minimize the informational spillover between the two auctions. Finally, we show that the said taxes can implement an optimal robust mechanism in the sense of ex-post IC and IR constraints but ex-ante objectives, such as efficiency or revenue, in private values setting similar to Lu and Robert (2001). Further tractability is achieved given quadratic utilities, allowing for comparisons with the nearly efficient robust mechanisms of Andreyanov and Sadzik (2021).

Optimality of Two-tier Quotas in Selection (with Byeong-hyeon Jeong)

• Abstract: We study an optimal selection problem, such as college admissions, hiring processes, and resource distribution, where the principal seeks to maximize utility by selecting a subset of objects from different groups. The principal relies on an informed agent to assess unobservable attributes that affect the decision. We consider mechanisms in which the principal commits to a selection rule, and the agent reports the attributes of objects. We show that the optimal mechanism can be implemented through quota-based mechanisms with at most two tiers, where the principal sets quotas for objects to be selected with certainty (first tier) and via lottery (second tier), delegating the selection of objects for each quota to the agent. A single-tier quota suffices to achieve the first-best outcome when preferences are aligned. Our results demonstrate that quota-based mechanisms are not only optimal but also simple and tractable, providing a framework for characterizing optimal mechanisms across various real-world applications.