Distributional Competition

This paper establishes the existence and characterizes the properties of symmetric equilibria in competitions where players choose performance distributions under convex costs, with applications to R&D, oligopolistic product design, and rank-order contests.

The Gist

Imagine a world where competition isn't just about how hard you push, but about the shape of your luck.

In most games we play, you choose a single action: you study for 5 hours, you bid $100, or you run a mile in 6 minutes. But in this paper, the author, Mark Whitmeyer, asks a different question: What if you could choose a whole "distribution" of outcomes?

Instead of deciding to run exactly 6 minutes, you decide to create a "cloud" of possibilities. Maybe you have a 10% chance of running 5 minutes, a 50% chance of 6 minutes, and a 40% chance of 7 minutes. The catch? Creating this specific cloud of possibilities costs money, and the cost depends on the entire shape of the cloud, not just the average.

Here is the paper broken down into simple concepts, using everyday analogies.

1. The Core Idea: Choosing Your Own "Cloud" of Luck

Imagine you are a manager trying to get the best return on an investment.

• The Old Way (Linear Cost): You pick a specific return (e.g., 5%) and flip a coin to see if you get it. The cost is just the average price of the coin flips.
• The New Way (This Paper): You design a complex portfolio. You care about the variance (how wild the swings are) and the tail risk (the chance of a total disaster). The cost of creating this specific portfolio depends on its global shape. Maybe it's cheap to have a high average, but very expensive to have a "fat tail" of disaster risks.

The paper studies what happens when multiple players do this simultaneously. They all try to shape their "cloud" of outcomes to beat the others, paying a price for the complexity of their cloud.

2. The Three Big Games Studied

The author applies this "cloud-shaping" logic to three specific real-world scenarios:

A. The "Tournament" (Rank-Order Contests)

Think of a race where only the winner gets a prize, or a promotion where only the top 3 get a raise.

• The Setup: Everyone picks a distribution of performance. The person with the highest number wins.
• The Finding: If you make the prize schedule more "inegalitarian" (e.g., the winner gets a massive bonus, and everyone else gets nothing), people don't just try harder; they take bigger risks.
• The Metaphor: Imagine a game show. If the prize is $100 for first and $0 for second, contestants might try a risky stunt that has a 1% chance of winning big and a 99% chance of failing. If the prize is $50 for first and $40 for second, they play it safe. The paper proves that making the prizes more unequal forces everyone to choose "riskier" clouds of outcomes, leading to more extreme results (both better and worse).

B. The "Patent Race" (Risky R&D)

Think of pharmaceutical companies racing to discover a new drug. The first one to find it gets the patent; the rest get nothing.

• The Setup: Companies choose a distribution of when they might discover the drug. They can aim for a steady, slow discovery or a "moonshot" strategy with a tiny chance of finding it tomorrow and a huge chance of never finding it.
• The Finding: Competition makes companies discover things too fast compared to what is best for society.
• The Metaphor: Imagine a group of people digging for gold. A wise planner would tell them to dig steadily to avoid wasting energy. But because they are competing, everyone starts digging frantically and chaotically, hoping to strike it first. The result? They find the gold sooner, but they waste a lot of resources doing it. The "equilibrium" (what happens naturally) is inefficiently fast.

C. The "Product War" (Price and Quality)

Think of companies selling smartphones. They choose a price and a "quality" (which is actually a random variable—maybe your phone works perfectly, or maybe it has a glitch).

• The Setup: Firms choose a price and a "cloud" of quality outcomes. Consumers buy the one with the best value (Quality minus Price).
• The Finding: In a market with many firms, prices usually drop to the cost of production (marginal cost). But this paper finds a twist: Prices only drop to zero profit if the products become identical.
• The Metaphor: Usually, we think "Bertrand competition" means if you have 100 sellers, they all slash prices to the bottom. This paper says: "Not so fast." If it's expensive to make a perfect product (the top of the quality cloud) compared to a bad one, firms will keep their prices high even with 100 competitors. They only drop prices to the bottom if they are forced to make products that are all exactly the same. If they can still differentiate their "clouds" of quality, they keep their markups.

3. The "No-Tie" Rule

A major technical hurdle in these games is what happens when two people get the exact same score (a tie). In real life, ties are messy.

• The Paper's Solution: The author proves that in a smart, rational equilibrium, nobody ever ties.
• The Metaphor: If you are playing a game where ties are broken by a coin flip, you will never choose a strategy that lands exactly on the same number as your opponent. You will always nudge your "cloud" slightly to the left or right to avoid the tie. The math shows that the "clouds" of all players will be perfectly smooth and spread out, with no clumps of probability at any single point.

Summary of the "Takeaway"

This paper provides a new toolbox for understanding competition. It moves beyond "how hard did you work?" to "what kind of risk profile did you choose?"

• In Contests: More unequal prizes = riskier, more volatile outcomes.
• In Innovation: Competition = inefficiently fast, chaotic discovery.
• In Markets: Prices only crash to the bottom if companies stop differentiating their products; otherwise, they can keep charging extra for "better" (though risky) quality clouds.

The author uses advanced math to prove these things exist and to show exactly how the "clouds" of outcomes look, but the core message is about how the shape of risk changes when we compete.

Technical Summary

Technical Summary: Distributional Competition

Problem and Framework
This paper studies a class of symmetric strategic interactions where $n \ge 2$ players choose probability distributions over a one-dimensional performance index $X = [0,1]$. Unlike standard models where agents choose deterministic actions and potentially mix over them, or where outcomes are generated by specific stochastic processes (e.g., diffusions), this framework allows players to select arbitrary distributions directly. The cost of generating a distribution is modeled via a general, convex cost functional $C(F)$, which can depend nonlinearly on the distribution's shape (e.g., its variance, tail risk, or aggregate intensity). Payoffs are determined by the realized performance profile, rewarding higher realizations, with ties broken uniformly.

The framework is designed to be agnostic regarding microfoundations, subsuming standard rank-order tournaments, all-pay auctions, and risky R&D competitions as special cases. It specifically addresses environments where the cost structure prevents the reduction of the problem to simple expected-cost randomization over deterministic actions.

Methodology
The analysis relies on the space of cumulative distribution functions (CDFs) equipped with the topology of weak convergence. The methodology proceeds in three stages:

1. Existence: The paper establishes the existence of a symmetric pure-strategy Nash equilibrium. Despite the payoff function being discontinuous at ties (where the winner changes abruptly), the author adapts results from Reny (1999). The key technical insight is that the discontinuities are "tightly structured" (occurring only at ties) and can be smoothed out via atomless deviations, satisfying the conditions for diagonal better-reply security and diagonal quasiconcavity.
2. Characterization via First-Order Approach: The paper derives necessary conditions for symmetric equilibria and for the solution to a symmetric social planner's problem. By utilizing the Gâteaux derivative of the cost functional, the author characterizes equilibria as distributions where the net marginal return (private benefit minus marginal cost) is constant on the support and maximal elsewhere. This extends the standard first-order approach to settings where the decision variable is a distribution rather than a scalar.
3. Specialized Applications: The general results are applied to three specific settings:
   • Rank-Order Contests: Where costs depend on an aggregate performance index (e.g., $C(F) = \Upsilon(\int \gamma dF)$).
   • Risky R&D: Modeled as a patent race where firms choose distributions over breakthrough times.
   • Oligopolistic Price and Quality Competition: Firms choose both a price and a distribution over product quality.

Key Results

• Existence and Structure: A symmetric pure-strategy equilibrium exists. Under natural conditions (specifically, a strict tie-disadvantage where escaping a tie yields a discrete payoff jump), symmetric equilibria are atomless on the interior of the support.
• Rank-Order Contests (Prize Inequality): For a broad class of convex costs, the paper analyzes the effect of prize inequality. If a prize vector $w$ majorizes a prize vector $v$ (i.e., $w$ is more unequal), the resulting equilibrium output under $w$ dominates the output under $v$ in the increasing convex order. This implies that more inegalitarian prizes lead to higher mean output but also riskier (more dispersed) output. This result holds even when the cost function is nonlinear in the distribution, contrasting with and complementing findings in literature that assumes linear costs or fixed means.
• Risky R&D (Efficiency): In a winner-take-all patent race, the equilibrium distribution of breakthrough times is shown to be first-order stochastically dominated by the planner's optimal distribution. This confirms that competition leads to inefficiently high effort and breakthroughs that occur "too quickly" from a social welfare perspective, as firms fail to internalize the business-stealing externality and duplication of effort.
• Price and Quality Competition: In an oligopoly where firms choose prices and stochastic quality, the paper investigates the limit as the number of firms grows large. It derives a necessary condition for prices to converge to marginal cost: the marginal cost of producing outcomes near the top of the quality support must converge to the marginal cost of outcomes near the bottom. If the cost function satisfies a "steepness" property (where marginal costs rise sufficiently fast), convergence to marginal cost pricing implies no product differentiation (the quality distributions collapse to a point). This effectively flips the standard Bertrand logic: marginal-cost pricing here implies perfectly homogeneous products.

Significance and Claims
The paper claims to provide a unified framework for analyzing strategic competition where agents have flexible control over the risk and shape of their output distributions. Its primary contributions are:

1. Generalization of Contest Theory: It moves beyond the standard "mixed strategy" benchmark (linear costs) to environments where the distribution's global shape directly affects costs and payoffs. This allows for the modeling of scenarios involving variance, downside exposure, and deadline effects that cannot be captured by simple randomization over deterministic actions.
2. Robust Comparative Statics: In the context of rank-order contests, it provides a stark conclusion regarding prize inequality: more unequal prizes increase both the mean and the riskiness of output, a result that holds across a broad class of cost functions.
3. Clarification of R&D Inefficiency: It offers a clean, reduced-form proof that competition in patent races leads to stochastically earlier (and thus socially excessive) breakthroughs, clarifying the mechanism of inefficiency without relying on specific search technologies.
4. New Insights into Oligopoly: It challenges the intuition that large markets always lead to product differentiation or that price convergence to marginal cost is compatible with heterogeneous products. Instead, it posits that under certain cost structures, the "Bertrand logic" is inverted: marginal-cost pricing forces products to become perfectly homogeneous.

The paper positions itself as complementary to existing literature on risk-taking in contests (e.g., Kim et al., 2023; Seel and Strack, 2013) and all-pay auctions, offering a distinct "reduced-form" approach where effort and risk-taking are intrinsically intertwined through the choice of distribution.