See my extended RESEARCH STATEMENT here.
UNDER REVIEW
“Reputation, Referrals, and Artificial Intelligence”, with Yi Chen and Mark Satterthwaite.
- PAPER
- SLIDES from the Antitrust@Cornell25 Workshop at Cornell University, 05/2025
- SLIDES from the Theory Workshop at the Department of Economics at the University of Toronto, 03/2025
Abstract. We study a dynamic model of referrals among experts that are horizontally specialized in treating heterogeneous problems. Demand for an expert’s services increases in their reputation, a decaying record of their historic success rate in solving problems. Experts refer mismatched problems, trading off today’s revenue with a higher demand tomorrow. In the unique steady state of Markov equilibria of the economy, the market generally neither supports efficient referrals nor optimal specialization, a problem potentially exacerbated by the adoption of AI. Our results suggest that regulation should allow for referral alliances and partnerships among small numbers of experts with complementary skills.
Keywords. Referrals, Reputation, Dynamic Games, Markov Equilibrium, AI.
JEL codes. C73, D83, I11, J44, L84.
“Managerial Poaching and Talent Reallocation”, with Yi Chen, Fabiano Dal-Ri and Daniela Scur.
- PAPER
- SLIDES from the NBER Summer Institute 2025 Personnel Economics at the Royal Sonesta, Cambridge, 07/2025
- SLIDES from the NBER Organizational Economics Working Group Spring 2025 at the Royal Sonesta, Cambridge, 04/2025
- SLIDES from the Empirical Management Conference 2024 at HBS, 11/2024
Abstract. This paper presents a model of employee poaching with asymmetric employer learning. Firms poach managers not only due to their track record but also for their personnel-specific information about workers. In equilibrium, more productive firms poach managers, whose compensation increases in the quality of their information about workers. While poaching reassigns more able workers to more productive firms, efficiency does not obtain due to information frictions. Drawing on the universe of contracts in Brazil’s formal labor market, we test implications of our model and show they are consistent with manager and worker movements and their compensation histories.
Keywords. Poaching, asymmetric learning, managerial compensation, Brazil.
JEL codes. C78, D21, J24, J30, M51.
“Optimally Informative Rankings and Consumer Search”, with Maarten Janssen, Marcel Preuss and Cole Williams.
Abstract. This paper investigates the optimal information policy of an online platform (or multi-product firm) when ranking products in response to a consumer search query. The informativeness of rankings ranges from full information to full obfuscation, and consumers learn their match values with the products by engaging in costly sequential search. Invoking continuous match value distributions allows us to establish a novel result about consumer search. While consumers buy products with high match values and continue searching when they encounter low match values, they abort search without buying a product for intermediate ones. For a large class of distributions, the optimal strategy of a platform maximizing the probability of the consumer buying a product is to provide either full or no information at all. As a result, platform and consumer welfare are either fully aligned or at odds with each other.
Keywords. Sequential search, search platforms, learning, ranking informativeness.
JEL codes. C72, D11, D21, D83.
“Vertical Differentiation, Branding, and Product Confusion”, with Chenyang Li, Sherif Nasser and Christian Schmid.
(Reject & Resubmit at Marketing Science, new version coming soon!!!)
Abstract. The decision whether a multi-product firm offers its goods under a joint or separate brands is essential for its success. When selling vertically differentiated products, it needs to consider pricing, cannibalization, and—when branding jointly—product confusion, i.e., consumers associating characteristics with the wrong product. The extent of confusion crucially depends on the accuracy of available information. We study the branding and pricing problem of a firm selling vertically differentiated products to a mix of naive and sophisticated consumers. The analysis accounts for the spillover arising from product confusion. Our findings suggest that joint branding is optimal when the spillover is either high or low but not in between. When low, firms jointly brand to save the cost of erecting a second brand. In contrast, when spillover effects are high, the firm brands jointly since it is inherently more profitable even if building additional brands is free. In between, firms opt for separate branding despite the additional cost of building more brands. We also find that a higher fraction of sophisticated consumers does not necessarily push the firm towards joint branding. Moreover, firms are inclined to jointly brand more similar products, and, if doing so, decrease price dispersion.
Keywords. Vertical differentiation, product confusion, branding, cannibalization, information provision.
JEL codes. D21, L12, L23, M21, M31.
WORK IN PROGRESS
“Education Signaling and Employer Learning Heterogeneity”, with Yuhan Chen and Michael Waldman.
Abstract. We investigate the implications of heterogeneous employer learning on education signaling and workers sorting across industries. In the equilibrium of our model, higher-ability workers join industries with faster employer learning speeds, resulting in a matching distortion of workers and industries. In addition, our results are robust to varying degrees of asymmetric employer learning, and establish that industry choice itself serves as a signal of worker ability. Finally, our theoretical approach suggests a novel perspective on a heretofore neglected labor market puzzle, i.e., why few of the richest individuals have obtained higher degrees of education.
Keywords. Education signaling, employer learning, industry choice, labor market sorting.
JEL codes. C78, D82, I26, M51.
“Eavesdropping and Innovation”, with Yi Chen and Jorge Lemus.
Abstract. Innovation often requires completing a series of steps, some of which are unmarketable and unpatentable. The first firm to complete all these steps profits either from a first-mover advantage or from patented components of the final product. Less innovative firms, unable to complete some of these steps, may resort to espionage to learn the missing steps and advance the development of their final product. We show that larger market rewards (e.g., stronger patents) or more efficient experimentation can harm innovation under espionage. We also investigate the role of a rival’s acquisition, third-party hackers, and different espionage methods.
Keywords. Espionage, innovation, research, development, experimentation, patents.
JEL codes. C73, D83, O31, O34.
“Machine Learning, Regularization, and Reporting Bias”, with Yi Chen, Kai Du and Zhe Wang.
“Digitization, Artificial Intelligence, and Equilibrium Multiplicity in Modern Labor Markets”, with Kevin Boudreau and Cole Williams.
“Algorithm Transparency in Search Markets”, with Raphael Boleslavsky and Mehdi Shadmehr.
“The Disruption of Attention Platforms in the Era of Generative AI”, with Chenyang Li, Xun Wu and Fei Xiao.
PUBLICATIONS
“Search Platforms: Big Data and Sponsored Positions”, with Maarten Janssen, Marcel Preuss and Cole Williams.
(Conditionally accepted at Economic Journal)
- PAPER
- SLIDES from the Queen’s University Microeconomics Workshop, 03/2025
- SLIDES from the Columbia-Duke-Berkeley-MIT-Northwestern IO Theory Conference at Duke University, 11/2024
Abstract. We study a search platform ranking firms’ products across sponsored and organic positions, accounting for the incentives of both firms and consumers. To characterize an optimal ranking when the number of firms is large, we formulate a Mixing Principle for Consumer Search, adapting tools from the social learning literature. The platform assigns the products it deems best to sponsored positions and obfuscates the content of organic positions subject to consumers’ participation constraints. Obfuscation serves to maximize the platform’s revenue from both sponsored position auctions and commission fees. Our results allow us to analyze the welfare effects of sponsored positions.
Keywords. Consumer search, search advertising, obfuscation, mixing lemma.
JEL codes. D44, D83, L86, M37.
“The Organization of Innovation: Incomplete Contracts and the Outsourcing Decision”, with Sean Nicholson, June Pan, Michael Waldman and Lucy Xiaolu Wang.
American Economic Journal: Microeconomics (forthcoming).
Abstract. Why do firms outsource research and development (R&D) for some products while conducting R&D in-house for similar ones? An innovating firm risks cannibalizing its existing products. The more profitable these products, the more the firm wants to limit cannibalization. We apply this logic to the organization of R&D by introducing a novel theoretical model in which developing in-house provides the firm more control over the new product’s location in product space. An empirical analysis of our testable predictions using pharmaceutical data concerning patents, patent expiration, and outsourcing at various stages of the R&D process supports our theoretical approach.
Keywords. R&D, patenting, outsourcing, cannibalization, innovation, vertical integration.
JEL codes. D23, L24, L65, O32.
“Strategic Wage Posting, Market Power, and Mismatch.”
Journal of Labor Economics (forthcoming).
Abstract. This paper analyzes the effects of market power on the performance of matching markets with slot-specific transfers. I find that strategic wage setting (wage posting) in such markets does, in general, not result in assortative matching since firms with higher quotas—and thus relatively more market power—pay on average lower wages than their competitors due to a lack of within-firm rivalry. If highly productive firms command market power, the resulting welfare loss may be substantial. Moreover, firms typically do not gain from wage discrimination. Thus, strategic wage setting is compatible with the absence of entry-level wage variation within firms.
Keywords. Multi-unit auctions, wage posting, strategic bidding, many-to-one matching, within-firm rivalry
JEL codes. C78, D44, D47, J44, L11.
“Actions and Signals”, with Michael Waldman.
Journal of Economic Theory 230: 106100, 2025.
Abstract. For many signaling applications, the real-world scenario fails to match the model setup. Actions are often unobservable but combine with the sender’s type to determine an observable signal. Introducing a theory of generalized signaling, we show the nature of equilibria can fundamentally differ from the canonical signaling model. In fact, we find both over- and under-investment arise in important signaling environments. Our model extends to multiple actions and signal jamming environments with incomplete sender information. We present results for equilibrium construction and show that uninformed sender behavior may be more efficient than a fully informed sender’s.
Keywords. Generalized signaling, signal jamming, under-investment.
JEL codes. C72, D82, D83, I26.
“Online Advertising, Data Sharing, and Consumer Control”, with Justin Johnson and Marcel Preuss.
Management Science 70(4): 7984 – 8002, 2024.
Abstract. Invoking a property rights approach, we examine how competition between advertising exchanges influences the targeting options that these exchanges make available to advertisers. When advertisers have strong property rights over data regarding consumers’ active purchase interests, competition between ad exchanges leads to too little sharing of data. This may harm consumers, who receive too few pertinent ads, and advertisers themselves can also be harmed due to a situation resembling a prisoner’s dilemma. We find that reallocating property rights to consumers, i.e., giving them the right to opt out of tracking may also benefit consumers who allow tracking, by altering the incentives of ad exchanges to offer improved targeting options. In addition, we show that initiatives by Apple and Google to limit third-party tracking and to introduce alternative tracking systems such as Google’s Topics, might benefit consumers by weakening the data property rights of advertisers. Because more data is shared by default under such systems, this can be true even if these systems are less accurate than third-party tracking systems.
Keywords. Online advertising, data sharing, privacy, retargeting, cross-targeting, property rights
JEL codes. D18, D23, O34, L86, M37.
“Branding Vertical Product Line Extensions”, with Christian Schmid.
Journal of Economics & Management Strategy 33(4): 909 – 936, 2024.
Abstract. Firms that sell vertically differentiated products infrequently roll out multiple products at the same time. In fact, it is often a firm already selling a well-established product that decides to expand up- or downwards when such an opportunity arises. A critical decision in this scenario is whether to introduce the new product under an existing brand. In this paper, we develop a game-theoretic model in which firms expand their product line to cater to a different customer segment, choosing their branding strategy, new product quality, and prices. We find that the firm’s optimal branding strategy depends on both the vertical direction of the expansion and the level of competition, and identify a novel interaction effect between these factors. In particular, firms engaged in direct competition employ branding as a commitment device to soften quality competition. When these firms extend their product line upwards, this creates a misalignment between firms’ actions and consumer preferences. We also derive conditions under which firms, against conventional wisdom, choose to differentiate their products more when selling them under the same brand. Finally, we characterize the welfare effects of branding in this setting, and argue that our findings are consistent with observations from the car industry.
Keywords. Vertical differentiation, brand management, product line, car industry
JEL codes. D21, L13, L23, M21, M31.
“The Strategic Decentralization of Recruiting”, with Yi Chen and Zhe Wang.
Journal of Economic Theory 209: 105639, 2023.
Abstract. We propose a model of strategic delegation in professional labor markets in which big and fringe firms compete for heterogeneous workers. In this context, big firms decide whether to exercise their market power to suppress wages (wage suppression) or to delegate hiring to divisions, thereby committing to bidding more fiercely for more skilled workers (talent acquisition). This reduces the incentive for other firms to go toe-to-toe with the decentralizing firm. In equilibrium, a big firm commanding labor market power opts for either of these two strategies. We find that the presence of a big firm in the labor market is detrimental to social welfare only if its size exceeds the tipping point beyond which it ceases to decentralize. Moreover, our model is compatible with recent studies that negatively link labor market concentration and labor share. Finally, the introduction of a wage floor encourages the big firm to decentralize, increasing match quality, such that social welfare may increase despite a drop in employment.
Keywords. Strategic delegation, market power, recruiting, decentralization, labor share, wage floor
JEL codes. C78, D44, J31, J42.
“Self-Reported Signaling”, with Michael Waldman.
American Economic Journal: Microeconomics 15(3): 78 – 117, 2023.
Abstract. In many real-world settings, an action that affects the value of a product or service is self-reported rather than publicly observable. We investigate self-reporting when the action serves as a signal. In our model, a sender chooses an action, and then sends a message concerning the action to multiple receivers. Receivers then bid for the sender’s service after deciding whether to audit the sender. We find that self-reporting can reverse the standard result in signaling models that there is overinvestment in the signal, and that the possibility of misrepresentation may in fact improve welfare given self-reported signaling.
Keywords. Signaling, lying, misrepresentation, resume padding, college admissions.
JEL codes. D82, J31, L15.
“Strategic Games beyond Expected Utility”, with Klaus Ritzberger.
Economic Theory 48: 377 – 398, 2011.
Abstract. This paper argues that Nash equilibrium is a solution where all strategic uncertainty has been resolved and, therefore, inappropriate to model situations that involve ambiguity. Instead, to capture what players will do in the presence of some strategic uncertainty, takes a solution concept that is closed under best replies. It is shown that such a solution concept, fixed sets under the best reply correspondence, exists for a class of games significantly wider than those games for which generalizations of Nash equilibrium exist. In particular, this solution can do without the expected utility hypothesis.
Keywords. Ambiguity, fixed sets under the best reply correspondence, Nash equilibrium, non-expected utility.
JEL codes. C6, C72, C79, C81.
THOMAS JUNGBAUER 