Cartels present one of microeconomics' most enduring paradoxes. Theory tells us that firms colluding to restrict output and raise prices can capture monopoly rents that vastly exceed competitive profits. Yet empirical evidence reveals that cartels are remarkably fragile institutions, with the average detected cartel surviving roughly five to seven years before collapsing under its own internal contradictions.

This tension between theoretical attractiveness and practical instability is best understood through the lens of repeated game theory. The folk theorem establishes that cooperation can be sustained as a subgame perfect equilibrium when participants are sufficiently patient and possess credible punishment strategies. But the theorem's elegance obscures the demanding informational and structural conditions required for collusion to persist in real markets.

Understanding cartel fragility matters beyond academic interest. Competition authorities worldwide deploy increasingly sophisticated screening tools grounded in this theoretical literature, and leniency programs explicitly exploit the strategic vulnerabilities that mechanism design predicts. By examining the conditions under which collusive equilibria break down, we gain insight into both optimal antitrust enforcement and the deeper question of when self-enforcing agreements can substitute for formal contracts in economic life.

The canonical analysis of collusion begins with infinitely repeated Bertrand or Cournot competition. In the stage game, firms have a dominant incentive to deviate from any tacit agreement, since undercutting a collusive price yields immediate windfall profits. The folk theorem demonstrates that this myopic temptation can be overcome when firms place sufficient weight on future payoffs and can credibly threaten punishment.

Formally, sustainability requires that the discounted present value of collusive profits exceed the one-shot deviation gain plus the discounted continuation value under punishment. Letting δ denote the discount factor, π^c the collusive payoff, π^d the deviation payoff, and πⁿ the Nash reversion payoff, the incentive compatibility constraint reduces to δ ≥ (π^d − π^c) / (π^d − πⁿ).

This threshold tightens considerably when we relax the assumption of perfect monitoring. Green and Porter's seminal contribution introduced demand uncertainty, where firms cannot directly observe rivals' output and must infer cheating from noisy price signals. The optimal collusive scheme then involves reversionary punishment phases triggered by price drops, even when no actual deviation has occurred—an unavoidable cost of imperfect monitoring.

The detection probability enters multiplicatively with discount factors, meaning that even modest improvements in monitoring technology can destabilize previously sustainable agreements. Abreu, Pearce, and Stacchetti extended this framework to characterize the entire set of perfect public equilibria, showing that the efficiency frontier of sustainable collusion shrinks as monitoring deteriorates.

Critically, these conditions are not just stricter than they appear—they are inherently fragile. The discount factor incorporates not merely time preference but also the probability of continued interaction, market exit, and regulatory intervention. Any shock to expected longevity directly undermines the collusive equilibrium.

Takeaway

Self-enforcing agreements depend on a delicate balance between patience and punishment. When either weakens, even optimally designed collusion unravels from within.

Real-world markets systematically violate the symmetry assumptions that make textbook collusion appear robust. Cost asymmetries among cartel members generate divergent preferences over the collusive price, since low-cost firms prefer aggressive output expansion while high-cost firms favor restricting supply. Bargaining over the division of cartel rents becomes correspondingly contentious.

Capacity heterogeneity compounds these tensions. A firm with substantial excess capacity faces a larger deviation gain—it can flood the market profitably—while simultaneously offering a weaker punishment threat if other firms are capacity-constrained. The Compte, Jenny, and Rey framework demonstrates that capacity-constrained firms find collusion harder to sustain precisely when it would be most profitable.

Demand volatility introduces a second-order problem beyond mere monitoring noise. Rotemberg and Saloner showed that during demand booms, the one-shot deviation temptation grows faster than the discounted continuation value, forcing cartels to adopt countercyclical pricing. Cartel prices must actually fall during booms to maintain incentive compatibility—a counterintuitive prediction with empirical support.

Entry threats further erode sustainability. Even modest probabilities of entry reduce the effective discount factor by truncating the expected horizon of collusive rents. Cartels frequently invest in entry deterrence mechanisms, but these investments themselves create asymmetries and free-riding problems among incumbents.

Finally, asymmetric information about costs and demand creates ex ante bargaining frictions. Mechanism design results indicate that achieving efficient collusion under asymmetric information generally requires side payments, which are both legally hazardous and difficult to enforce without leaving evidence. The very features that make collusion valuable—information asymmetries that limit competition—simultaneously undermine the cartel's internal coordination.

Takeaway

Heterogeneity is the silent solvent of collusion. The more firms differ in costs, capacities, or expectations, the more the collusive contract must do work that no informal agreement can sustain.

Modern antitrust enforcement increasingly relies on economic screens—statistical tests designed to identify markets where collusive conduct is plausible. Structural screens examine market conditions conducive to collusion, including concentration, symmetry, demand stability, and entry barriers. Behavioral screens scrutinize observed market outcomes for patterns inconsistent with competitive equilibrium.

Harrington's work on cartel detection identifies several robust empirical signatures. Price variance typically declines during cartel periods relative to competitive benchmarks. Market shares stabilize as firms cease competing for marginal customers. Bid patterns in procurement auctions exhibit telltale rotation or complementary bidding structures that pure competition would not produce.

Leniency programs represent perhaps the most successful application of mechanism design to antitrust. By offering full immunity to the first cartel member to defect and provide evidence, authorities transform the cartel's internal coordination problem into a prisoner's dilemma with a dominant strategy. The mere existence of credible leniency programs raises each member's belief that others may defect, accelerating the unraveling that theory predicts.

Optimal enforcement design must balance deterrence and detection. Setting penalties too high relative to detection probability creates risk that no cartels form but those that do are exceptionally stable, since members face catastrophic punishment for defection. Calibrating sanctions against detection probability, as Becker's framework suggests, requires careful empirical estimation of the underlying parameters.

Recent advances apply machine learning to bid data and price series, identifying anomalies that traditional screens miss. These methods raise the equilibrium detection probability, further compressing the parameter space within which collusion remains incentive compatible. The theoretical literature thus provides not merely a description of cartel fragility but an operational blueprint for accelerating it.

Takeaway

Good enforcement design does not merely punish cartels—it weaponizes their internal incentive structure against them, turning members into informants by construction.

The microeconomic theory of collusion delivers a striking verdict: cartels are not stable institutions temporarily disrupted by enforcement, but inherently unstable arrangements whose collapse is a matter of when, not whether. The folk theorem's permissive conclusions about sustainable cooperation depend on assumptions that real markets systematically violate.

This perspective has profound implications for competition policy. Rather than treating each cartel as a discrete enforcement problem, authorities can design institutions that exploit the structural vulnerabilities theory identifies. Leniency programs, screening algorithms, and information disclosure requirements all leverage the same incentive logic that makes self-enforcing agreements fragile.

The deeper lesson extends beyond antitrust. Wherever economic agents attempt to sustain cooperation without formal contracts—whether in international agreements, informal labor arrangements, or platform governance—the same destabilizing forces operate. Understanding cartel collapse illuminates a general truth about the limits of self-enforcement in economic life.