Competition and Efficiency in Congested Markets

Daron Acemoğlu, Asuman Ozdaglar

TL;DR

In most markets, competition is the gold standard for efficiency. However, this paper proves that in congested networks (like traffic or data routing), increasing competition can actually decrease social welfare. The authors establish a tight theoretical bound, showing that even in the worst-case scenario, a competitive market will still achieve at least 5/6 (about 83%) of the maximum possible social surplus.

What Problem Does This Paper Address?

Standard economic theory suggests that more competition leads to better outcomes. But congested markets—where one person’s use of a resource (a road, a data link) slows down everyone else—behave differently. These "congestion externalities" mean that individual users don't account for the delays they cause others.

While previous research explored "selfish routing" (where users choose paths without prices), this paper looks at what happens when profit-seeking companies own the links and set prices. The core question is: Does competition between these providers help or hurt the overall efficiency of the network?

Methods

The authors use a game-theoretic framework to model a two-stage "pricing-congestion game":

  1. Providers (Oligopolists): Multiple service providers own different links in a network and set prices to maximize their own profits.
  2. Users: A large population of "small" users chooses routes based on the effective cost, which is the sum of the price (toll) and the latency (delay). Users follow Wardrop’s principle, essentially a Nash equilibrium for traffic where no one can reduce their own cost by switching routes.

The study analyzes Subgame Perfect Equilibria, looking for both pure strategies (fixed prices) and mixed strategies (probabilistic pricing). They measure efficiency by comparing the "Social Surplus" (total willingness to pay minus total delay costs) of the equilibrium versus a "Social Optimum" designed by a perfect planner.

Datasets

As a theoretical paper in economics and operations research, this work does not use empirical datasets. Instead, it relies on mathematical proofs and stylized numerical examples (like the classic Pigou’s network) to demonstrate its findings.

Key Insights and Ideas

  • The Monopoly Paradox: Surprisingly, a monopolist owning the entire network achieves 100% efficiency. Why? Because the monopolist wants to maximize profit, they set prices that act exactly like a "Pigovian tax." This forces users to internalize the cost of the congestion they cause, leading to a socially optimal distribution of traffic.
  • The "Lock-in" Effect of Competition: When providers compete (an oligopoly), they lose this incentive to be efficient. If Provider A raises their price, some traffic moves to Provider B’s link. This extra traffic makes Provider B’s link more congested. Paradoxically, this makes Provider A’s link more attractive to the remaining users because their "outside option" (Provider B) just got slower. This gives providers more market power to distort prices away from the social optimum.
  • The 5/6 Bound: The authors prove that for any network of parallel links with linear latency functions, the efficiency will never drop below 5/6 of the optimum. This is a "Price of Anarchy" result for competitive pricing.
  • Complexity of Equilibrium: Pure strategy equilibria (where prices are stable) always exist for linear delays but can vanish if the delay functions are highly convex (i.e., if congestion spikes very suddenly). In those cases, providers might resort to mixed strategies, which can lead to even worse efficiency.

Key Results

  • Monopoly = Efficiency: In this specific model, a single provider reaches the social optimum.
  • Competition < Monopoly: Moving from a monopoly to a duopoly almost always reduces social surplus.
  • Tight Efficiency Bound: The 5/6 bound is "tight," meaning there are specific, simple network configurations where the efficiency is exactly 5/6, and it cannot go lower regardless of how many routes or competitors you add.
  • Mixed Strategy Risks: While pure strategies have a floor of 83% efficiency, mixed strategy equilibria can theoretically lead to "worst-case realizations" where efficiency becomes arbitrarily small.

Limitations and Future Work

  • Homogeneous Users: The model assumes all users have the same "reservation utility" (willingness to pay) and the same sensitivity to delay. In the real world, some people are in a rush and willing to pay more, while others are price-sensitive.
  • Parallel Links: The 5/6 bound is proven for parallel link networks. While many networks can be simplified this way, more complex topologies (like mesh networks) might yield different results.
  • Latency Assumptions: The core results assume that at zero traffic, there is zero delay ($l_i(0)=0$). Relaxing this assumption could change the characterization of the equilibrium prices.
  • Open Question: The authors leave the bound for average efficiency in mixed strategy equilibria as an open research question, though they conjecture it might also be 5/6.

Summary updated

Abstract

We study the efficiency of oligopoly equilibria in congested markets.The motivating examples are the allocation of network flows in a communication network or of traffic in a transportation network.We show that increasing competition among oligopolists can reduce efficiency, measured as the difference between users' willingness to pay and delay costs.We characterize a tight bound of 5/6 on efficiency in pure strategy equilibria.This bound is tight even when the number of routes and oligopolists is arbitrarily large.We also study the efficiency properties of mixed strategy equilibria.
BibTeX
@misc{acemoglu2005competition,
  author = {Daron Acemoğlu and Asuman Ozdaglar},
  title = {{Competition and Efficiency in Congested Markets}},
  year = {2005},
  doi = {10.2139/ssrn.676662},
  howpublished = {SSRN},
}

Loading related content...

Figure