Pricing Public Goods for Private Sale

Michal Feldman, David Kempe, Brendan Lucier, Renato Paes Leme

Abstract

We consider the pricing problem faced by a seller who assigns a price to a good that confers its benefits not only to its buyers, but also to other individuals around them. For example, a snow-blower is potentially useful not only to the household that buys it, but also to others on the same street. Given that the seller is constrained to selling such a (locally) public good via individual private sales, how should he set his prices given the distribution of values held by the agents? We study this problem as a two-stage game. In the first stage, the seller chooses and announces a price for the product. In the second stage, the agents (each having a private value for the good) decide simultaneously whether or not they will buy the product. In the resulting game, which can exhibit a multiplicity of equilibria, agents must strategize about whether they will themselves purchase the good to receive its benefits. In the case of a fully public good (where all agents benefit whenever any agent purchases), we describe a pricing mechanism that is approximately revenue-optimal (up to a constant factor) when values are drawn from a regular distribution. We then study settings in which the good is only "locally" public: agents are arranged in a network and share benefits only with their neighbors. We describe a pricing method that approximately maximizes revenue, in the worst case over equilibria of agent behavior, for any $d$-regular network. Finally, we show that approximately optimal prices can be found for general networks in the special case that private values are drawn from a uniform distribution. We also discuss some barriers to extending these results to general networks and regular distributions.

Introduction

Standard economic models of pricing typically assume that a product’s utility is confined to the individual purchaser. However, many real-world goods exhibit significant positive externalities. A neighborhood Wi-Fi router, a shared piece of laboratory equipment, or a snowblower all provide utility to a "local" network of individuals beyond the owner.

When a seller offers such a (locally) public good through private sales, they face a unique challenge: the free-rider problem. Rational agents, knowing they can benefit from a neighbor's purchase, are incentivized to wait rather than buy. This creates a complex strategic environment where the seller must set prices that maximize revenue despite agents' attempts to avoid purchasing. This paper models this scenario as a two-stage Bayesian game and provides robust pricing strategies that guarantee high revenue even in the worst-case equilibrium.

The Pricing Game Model

The interaction is modeled in two distinct stages:

  1. The Seller's Move: The seller announces a price $p$ (or a vector of prices) for the good, knowing the distribution $F$ from which agent valuations are drawn.
  2. The Agents' Move: Agents, situated as nodes in a graph $G$, observe the price and their own private valuation $v_i$. They then play a simultaneous game to decide whether to purchase.

An agent's decision is governed by a threshold strategy: they purchase if their private value exceeds the price divided by the probability that none of their neighbors purchase. Because multiple Nash equilibria can exist—some yielding high revenue and others very low—the authors focus on finding prices that perform well in the worst-case equilibrium.

Key Theoretical Insights

The Single-Buyer Reduction

In a "globally public good" scenario (represented by a clique where everyone benefits if anyone buys), the authors find that agents strategize such that, in expectation, only one person purchases. This allows the multi-buyer problem to be reduced to a single-item auction. By applying Myerson’s Lemma and the Prophet Inequality, the authors demonstrate that a single uniform price can guarantee at least a constant fraction (approximately 1/4) of the optimal revenue achievable by any complex mechanism.

The Hipster Game Equivalence

An intriguing mathematical finding is the equivalence between the Public Goods game and what the authors call the "Hipster Game." In the Hipster Game, a good has extreme negative externalities—an agent only derives value if no one else in their neighborhood owns it. Despite the opposite nature of the externalities, the equilibrium structure and expected revenue calculations for both games are mathematically identical.

Results for Networked Environments

$d$-Regular and General Networks

For networks where every agent has exactly $d$ neighbors, the authors identify a uniform pricing strategy that depends only on $d$ and the valuation distribution $F$. This strategy guarantees a constant fraction of the optimal revenue in the worst-case equilibrium.

For general, arbitrary networks, the problem becomes significantly more difficult. The authors prove that approximating the worst-case revenue for a given price is NP-hard to within a factor of $n^{1-\epsilon}$. This suggests that for complex social topologies, finding the absolute optimal price is computationally intractable.

The Uniform Distribution Heuristic

Despite the hardness results for general graphs, a remarkably simple rule emerges when agent valuations are drawn from a uniform distribution $[0, 1]$. In this specific case, the revenue equation can be decoupled from the network structure. The authors show that simply setting a price of $p = 1/2$ guarantees the seller at least an $e/4$ fraction of the optimal worst-case revenue, regardless of the network's shape.

Conclusion and Future Directions

This research provides a foundational framework for pricing in the presence of network externalities. It demonstrates that while the strategic landscape of free-riding is complex, simple uniform pricing strategies are often robust and approximately optimal.

Future research may expand these findings by considering "imperfect" public goods—where the purchaser retains a higher utility than their neighbors—or by exploring sequential purchase models where the seller approaches buyers one by one. Additionally, applying this framework to social policy, such as setting fines or subsidies to encourage the provision of public services, remains a promising area for further study.

Summary updated

Abstract

We consider the pricing problem faced by a seller who assigns a price to a good that confers its benefits not only to its buyers, but also to other individuals around them. For example, a snow-blower is potentially useful not only to the household that buys it, but also to others on the same street. Given that the seller is constrained to selling such a (locally) public good via individual private sales, how should he set his prices given the distribution of values held by the agents? We study this problem as a two-stage game. In the first stage, the seller chooses and announces a price for the product. In the second stage, the agents (each having a private value for the good) decide simultaneously whether or not they will buy the product. In the resulting game, which can exhibit a multiplicity of equilibria, agents must strategize about whether they will themselves purchase the good to receive its benefits. In the case of a fully public good (where all agents benefit whenever any agent purchases), we describe a pricing mechanism that is approximately revenue-optimal (up to a constant factor) when values are drawn from a regular distribution. We then study settings in which the good is only "locally" public: agents are arranged in a network and share benefits only with their neighbors. We describe a pricing method that approximately maximizes revenue, in the worst case over equilibria of agent behavior, for any $d$-regular network. Finally, we show that approximately optimal prices can be found for general networks in the special case that private values are drawn from a uniform distribution. We also discuss some barriers to extending these results to general networks and regular distributions.
BibTeX
@inproceedings{feldman2013pricing,
  author = {Michal Feldman and David Kempe and Brendan Lucier and Renato Paes Leme},
  title = {{Pricing Public Goods for Private Sale}},
  booktitle = {EC},
  year = {2013},
  url = {https://openreview.net/forum?id=CdUeNAV3RW},
}

Loading related content...

Figure