TL;DR
This paper explores how companies competing in multiple markets simultaneously can achieve better collusion than in a single market, even with imperfect information about rivals' actions. They introduce a novel "generalized EV" (gEV) strategy that uses simple, two-state automata and "mild punishment" to sustain higher profits, demonstrating that coordinated behavior across markets can be more effective than previously thought.
What Problem Does This Paper Address?
Imagine two companies, like ride-sharing services, competing in many different cities (markets). In each city, they face a "prisoner's dilemma" situation: it's best for each to undercut the other (defect), but if they both cooperated (e.g., kept prices high), they'd both be better off. When these games are repeated over time, cooperation can emerge. However, a key challenge is that companies can't perfectly observe each other's every move (private monitoring, like not knowing exact pricing, but maybe getting a noisy signal from sales data).
Previous research suggested that competing in multiple markets simultaneously didn't necessarily help firms collude more effectively than in a single market, especially under perfect monitoring. This paper tackles this discrepancy by investigating if multimarket contact can lead to better collusion under private monitoring, and if so, how. They aim to find simple strategies that achieve this, moving beyond overly complex theoretical equilibria.
Methods
The core of the paper lies in developing and analyzing a new class of strategies called generalized EV (gEV). These are two-state automaton strategies, meaning each player's behavior is governed by a simple state machine with two states (e.g., "cooperate" and "punish"). This simplicity is crucial because real-world firms can't coordinate complex strategies without communication.
A key innovation is the concept of "mild punishment": instead of punishing a defector in all markets, the strategy only punishes in a subset. This ensures that even during punishment, the firm doesn't fall below a minimum profit threshold. The paper also introduces "partial indifference", a condition that simplifies the analysis by making a player indifferent to certain actions in a subset of markets. This allows them to derive specific payoff bounds and construct equilibria that outperform simpler strategies like the basic EV strategy.
The analysis relies on the belief-free equilibrium concept, which is standard in game theory with imperfect monitoring. This concept ensures that strategies remain optimal even if players have imperfect beliefs about their opponent's history.
Datasets
Not discussed in the paper. The paper is theoretical and focuses on game-theoretic models.
Key Insights and Ideas
- Multimarket contact can enhance collusion under private monitoring: Contrary to some prior findings, this paper shows that competing in multiple markets can indeed lead to more cooperative outcomes than in a single market.
- Simplicity and "mild punishment" are key: The most effective strategies are not overly complex and involve punishing defections in a limited way, preserving some cooperation. This reflects real-world constraints on coordination.
- "Partial indifference" is a powerful analytical tool: This condition simplifies the complex problem of finding equilibria in multimarket games, allowing for the derivation of optimal payoffs.
- The structure of information matters: The way players receive noisy signals about each other's actions is critical. The paper highlights how specific signal patterns can be leveraged for better collusion.
- Cooperation in "A markets" is crucial: The gEV strategy designates a subset of markets ("A markets") where cooperation is always maintained, even during punishment phases in other markets. This sustained cooperation is vital for achieving higher payoffs.
Key Results
The paper constructs a generalized EV (gEV) strategy that forms a belief-free equilibrium in repeated multimarket games with private monitoring. This gEV strategy achieves higher per-market equilibrium payoffs than simpler benchmark strategies like the standard EV strategy or the simplified EV (sEV) strategy. The paper provides theoretical bounds for these payoffs and identifies conditions (specifically, "partial indifference") under which these optimal payoffs are achieved. The results demonstrate that the effectiveness of collusion is enhanced by the number of markets and the specific structure of the strategy, particularly the sustained cooperation in a designated subset of markets.
Limitations and Future Work
The authors acknowledge that their analysis focuses on a specific class of strategies (two-state automata) and that characterizing all possible equilibria in such games is extremely difficult. A significant area for future work is extending the analysis to asymmetric markets, where market conditions and payoffs might differ. This would involve not only determining optimal strategies but also allowing firms to strategically choose which markets act as "A markets" (for sustained cooperation) and which act as "B markets" (for punishment). Additionally, the paper suggests exploring the computational complexity of finding these equilibria and verifying their optimality against a broader range of strategies.
Summary updated