Selling Information in Games with Externalities
Thomas Falconer, Anubhav Ratha, Jalal Kazempour, Pierre Pinson, Maryam Kamgarpour · 2025
Read Selling Information in Games with Externalities by Thomas Falconer, Anubhav Ratha, Jalal Kazempour, Pierre Pinson, Maryam Kamgarpour (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-00405.txt.
Introduction
Selling data to a rival can earn money, yet it can also sharpen the rival's next move. Thomas Falconer and co-authors study that trade-off in a 2025 game theory paper, titled Selling Information to a Competitor in the text itself.
The buyer picks from a menu of data products, and its appetite depends on what it already knows in private. The seller loses in proportion to how much the data lifts the buyer's chance of a correct move. The authors map the best menu for two buyer types and for a smooth range of types. Selling can still pay, but holding back wins when the loss tops what the buyer will pay. Tougher rivalry trims the gains.
Catalog record
| CREATOR | Thomas Falconer, Anubhav Ratha, Jalal Kazempour, Pierre Pinson, Maryam Kamgarpour |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 86 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-00405
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Selling Information in Games with Externalities Thomas Falconer, Anubhav Ratha, Jalal Kazempour, Pierre Pinson, Maryam Kamgarpour Abstract A competitive market is modeled as a game of incomplete information. One player observes some payoff-relevant state and can sell (possibly noisy) messages thereof to the other, whose willingness to pay is contingent on their own beliefs. We frame the decision of what information to sell, and at what price, as a product versioning problem. The optimal menu screens buyer types to maximize profit, which is the payment minus the externality induced by selling information to a competitor, that is, the cost of refining a competitor’s beliefs. For a class of games with binary actions and states, we derive the following insights: (i) payments are necessary to provide incentives for information sharing amongst competing firms; (ii) the optimal menu benefits both the buyer and the seller; (iii) the seller cannot steer the buyer’s actions at the expense of social welfare; (iv) as such, as competition grows fiercer it can be optimal to sell no information at all. 1 Introduction Two key trends mark the rise of today’s digital economy: (i) the collection of vast amounts of data from our ever-more digitized lives; and (ii) the advancement of computing algorithms, hardware, and platforms that can transform this data into actionable insights. Many firms are therefore striving to deploy state-of-the-art machine learning models to improve their value propositions. Consequently, the demand for data is growing at an unprecedented rate, to the extent that it has even been dubbed the world’s most valuable commodity—the “oil” of the digital age—by The Economist 2017 . This raises an important question: If data is a commodity, then how should we value it? Unlike material commodities, data is an intangible asset, the very definition of which is a matter of debate. In fact, data is often viewed as an ephemeral entity that can be processed into subjective information, so it’s value is not an intrinsic property, but dependent on when, how, and by whom it is actually used ( Floridi 2002 ; Floridi 2009 ) . We adopt this view by modeling a buyer as a decision-maker under uncertainty that seeks additional information to refine their beliefs about some state of the world. In turn, a monopolistic seller owns data relevant to the buyer’s decision, and offers (possibly noisy) messages thereof in exchange for money, for who the value of this information depends on their prior beliefs. We explore what information, and at what price, the seller should offer to maximize profit. We assume that both parties compete in a downstream market, so the seller has information relevant to each of their payoffs. This is motivated by growing concerns over the relationship between data and the distribution of market power in many industries, insofar that competing firms face asymmetric access to information. For example, in June 2021, the EU launched an antitrust investigation into whether Google distorted competition by restricting third party access to user information for advertising, reserving exclusive use for itself ( European Commission 2021 ) . Central to this debate is whether firms should share information to benefit social welfare, or whether they can sell their information for profit. However, packaging and pricing information to monetize it within competitive environments is not a simple task, as the seller needs to consider not only the buyer’s utility, but also any externalities, both positive or negative, they may induce by sharing information with competitor. We model a competitive market environment as a game of incomplete information, where the players face a common fundamental uncertainty associated with a payoff-relevant state of the world. Each player also observes a private signal that determines their beliefs about the fundamental uncertainty, meaning they are also subject to strategic uncertainty with respect to the signals received by the other players. Players then simultaneously select an action to maximize their expected payoffs. We extend this framework by including an interim step where one player (the seller) observes the realized state before actions are chosen. The buyer’s beliefs, and thus their valuation for information, is unknown to the seller, who designs and offers a menu of communication rules, which are likelihood functions which prescribe a distribution over messages the buyer could receive from the seller conditioned on the realized state. It can therefore be characterized by a degree of obfuscation of the true data. This setup is adapted from those seen in information design literature, where a social planner commits to a communication rule to influence the behavior of players in a game ( Bergemann & Morris 2019 ) . This is equivalent to selecting the Bayes correlated equilibrium that maximizes the planner’s objective ( Bergemann & Morris 2016 ) . In our case, communication rules have an associated price and the problem becomes one of joint information and mechanism design, in which the seller must elicit the buyer’s valuation for information. 1.1 Contributions Before discussing related works, we summarize our contributions and outline of the paper: • In Section 2 , we formalize a two-player game of incomplete information that we use to model a competitive market. We adopt a simplified setup where both action sets and the state of the world are binary, and consider the case wherein each player has a dominant strategy to match their action with the state. As such, the buyer’s private type is one-dimensional and their valuation is both piecewise linear and independent of the seller’s action. Since the communication rule purchased influences the buyer’s decisions, the seller has (anti-) coordination incentives if their expected payoff (decreases) increases with the probability that the buyer also chooses the correct action. Although we focus on competitive environments with anti-coordination incentives, we show that our setup extends to the opposite case which could be viewed as a game with strategic complements instead of strategic substitutes. • In Section 3 , we outline the aspects that constrain the menu offered by the seller. Based on the principle that information is only valuable insofar as it changes the buyer’s action, the seller’s messages are viewed as action recommendations. We establish in Definitions 3.4 and 3.5 that a menu is incentive compatible if the buyer is best off truthfully reporting their type and obediently following the recommended action. • In Section 4 , we adapt the Myersonian auction format to provide novel insights into selling information to a competitor (see Proposition 4.3 ). To study the nuances of this product versioning (or second-degree price discrimination) problem, we first consider just two buyer types, ranked according to their valuation of the fully informative communication rule (see Corollary 4.4 ). We show that the seller can reduce the information rent of the less informed type by offering partial information depending on the type congruence—whether or not they would take different actions given their prior beliefs—, which is consistent with previous findings ( Bergemann et al. 2018 ) . However, unlike this prior work, we show that competition between the buyer and the seller induces an externality in the form of a cost of refining a competitor’s beliefs. This makes the optimal menu a function of both the intensity of competition and the seller’s own beliefs. For instance,