Abstract
In large, random matching markets where firms and workers exhibit idiosyncratic match values, standard economic intuition might suggest significant wage dispersion. This paper investigates whether the "Law of One Price"—the principle that identical goods should command identical prices—holds in such environments. By analyzing the Assignment Game model with independent and identically distributed (i.i.d.) productivity values, we demonstrate that as markets grow, the range of stable wages and profits shrinks toward zero. This research establishes that even in the presence of individual heterogeneity, competition in large markets forces nearly identical payoffs for agents on the same side of the market, provided the productivity distribution is bounded.
Introduction
The "Law of One Price" (LOOP) is a cornerstone of economic theory, positing that in a frictionless, competitive market, identical goods must sell for the same price. In the context of labor markets, this implies that workers with equivalent skills should receive the same compensation. However, empirical data frequently reveals substantial wage dispersion among seemingly similar workers.
Economists often attribute this dispersion to "match-specific productivity"—the notion that the value created by a worker-firm pair depends on unique, idiosyncratic compatibility. This paper addresses a fundamental theoretical question: if every possible pairing in a market has a unique, random productivity value, does the Law of One Price collapse? Or does the scale of a large market eventually restore price uniformity through the sheer force of competition?
The Assignment Game Model
The authors utilize the Assignment Game framework, originally proposed by Shapley and Shubik (1971). This model describes a two-sided market (e.g., firms and workers) where agents can form pairs to generate surplus, which is then divided via money transfers.
The study considers a sequence of markets where the productivity of every firm-worker pair $(i, j)$ is an i.i.d. random variable $X_{i,j}$ drawn from a bounded distribution. The goal is to characterize the core of these games—the set of stable outcomes where no coalition of agents has an incentive to deviate from the current matching and payoff distribution.
Methodology
Theoretical Derivations
The core of an assignment game is often represented as a set of dual variables in a linear programming problem. The authors prove that this core becomes "narrow" as the number of agents $n$ increases. The mathematical proof relies on graph theory, specifically demonstrating that the "top-choice" edges—representing the most productive potential pairings—form an expander graph. By establishing the existence of short paths between any two agents within this graph, the authors show that their payoffs cannot diverge significantly without creating an unstable pair that would prefer to match with each other.
Numerical Simulations
To complement the theoretical proofs, the authors conducted simulations using synthetic data. They tested various productivity distributions, including:
- Uniform [0, 1]
- Exponential
- Weibull
These simulations were used to validate the predicted convergence rates and to investigate the boundaries of the theory, particularly in cases involving unbounded distributions where the standard convergence might fail.
Key Insights
The Narrow Core and Convergence
The most significant finding is the emergence of a "narrow core." In large markets, the difference between the highest and lowest possible stable wage for any given worker vanishes at a rate of $O(\frac{\log^2 n}{n})$. This suggests that as $n \to \infty$, the flexibility in bargaining disappears, and the market dictates a single price for labor.
Competition as a Leveler
While every firm may have a specific "favorite" worker, the density of the market ensures there are many "nearly-as-good" alternatives. This abundance of close substitutes prevents any individual worker from successfully demanding a wage significantly higher than the market average. Competition effectively smooths out the idiosyncratic differences between pairs.
The "Long Side" Penalty
In unbalanced markets (e.g., where the number of workers $n+1$ exceeds the number of firms $n$), the "long side" of the market loses almost all bargaining power. In these scenarios, the surplus is almost entirely captured by the "short side" (the firms), and the Law of One Price holds even more rigidly.
The Role of Boundedness
The convergence toward a single price depends critically on the productivity distribution being bounded. If the distribution has "heavy tails" (such as a Weibull distribution), "superstar" agents can emerge. These outliers are so much more productive than their peers that they can command unique, high payoffs, thereby breaking the uniformity of the Law of One Price.
Results
- Payoff Bounds: For bounded distributions, the difference between payoffs for agents on the same side of the market is at most $O(\frac{\log^2 n}{n})$ with high probability.
- Surplus Division: In a balanced market with $n$ agents on each side, the total share of the surplus going to the workers under a firm-optimal allocation is only $O(\log n)$. Given that the total surplus grows at $O(n)$, the workers' share becomes a vanishingly small fraction of the total value created.
- Simulation Validation: Numerical experiments confirm that these effects manifest in medium-sized markets (e.g., $n=100$ to $300$). For uniform distributions, wage differences approach zero rapidly, whereas they remain high for exponential (unbounded) distributions.
Limitations and Future Research
While this paper provides a robust theoretical basis for an approximate Law of One Price, several avenues for future work remain:
- Unbounded Distributions: The current theorem fails for distributions like the Exponential distribution. Characterizing the exact salary distributions in heavy-tailed scenarios remains an open challenge.
- Correlation and General Human Capital: The model assumes i.i.d. productivity. Real-world markets often feature "separable" quality, where some workers are objectively better for all firms. While the authors suggest the model can be extended to handle some degree of separability, non-separable complementarities require further study.
- Search Frictions: The model assumes a complete information environment. The introduction of search costs or information asymmetry may impede the convergence toward a single price, potentially explaining the wage dispersion observed in empirical data.
Summary updated