TL;DR
This paper investigates the "law of one price" in random assignment games, a model for two-sided markets. They show that in large, randomly generated markets with bounded productivities, workers tend to earn roughly equal wages, and firms earn similar profits, suggesting a convergence towards a single price for labor.
What Problem Does This Paper Address?
The "law of one price" is a fundamental economic principle stating that identical goods should sell for the same price. In labor markets, this implies that equally skilled workers should earn the same salary. However, real-world labor markets exhibit significant wage dispersion, even among workers with similar skills. This paper explores whether a simplified model of a two-sided market, specifically the assignment game with random productivities, can shed light on the sources of this wage dispersion and whether a "law of one price" can emerge as an approximation under certain conditions. They aim to understand how market size and the nature of productivity distributions affect wage equality and profit distribution.
Methods
The authors build upon the Shapley-Shubik assignment game model, which involves firms hiring workers with associated productivities and transferable utility (allowing for flexible salary negotiations). Their key innovation is to introduce random, i.i.d. productivities drawn from a bounded distribution. They then analyze the core of these randomly generated markets. The core represents a set of stable allocations where no coalition of agents can improve their outcome by deviating. The paper's main theoretical contribution is proving that for large markets with bounded, continuous, and strictly positive productivity distributions, the core is "narrow." This means that within any core allocation, the wages of workers are very close to each other, and the profits of firms are also very close. They use graph-theoretic arguments and probabilistic bounds to establish that the difference in wages and profits is on the order of $O(\log(n)/n)$, where $n$ is the market size. They also explore the implications of unbalanced markets (where the number of firms and workers differs) and the impact of unbounded productivity distributions.
Datasets
Not discussed in the paper. The paper focuses on theoretical analysis and simulation results, not on empirical datasets.
Key Insights and Ideas
- Convergence to Wage Equality: Even with some heterogeneity in firm and worker preferences, large random assignment markets with bounded productivities tend to equalize wages. This suggests that the "law of one price" is a reasonable approximation in such idealized settings.
- The Core as a Benchmark: Analyzing the core of the assignment game provides a robust benchmark for understanding the range of possible outcomes and the degree of wage dispersion.
- Impact of Market Size: The convergence to wage equality is a phenomenon that becomes more pronounced as the market size ($n$) increases.
- Unbounded Distributions Matter: The theoretical results for bounded distributions do not hold for unbounded distributions (like the exponential distribution). Unbounded distributions can lead to significant wage dispersion due to "outlier" agents with extremely high or low productivities.
- "Good" vs. "Bad" Outliers: Unbounded distributions can create "good" outliers (highly productive agents who command a large share of the surplus) and "bad" outliers (highly unproductive agents whose low reservation value can effectively set a floor for wages).
Key Results
- Theorem 1: In a random assignment game with $n$ firms and $n$ workers, where productivities are i.i.d. from a bounded, continuous, and strictly positive distribution, with high probability, any core allocation satisfies:
- The difference between any two firms' profits is $O(\log(n)/n)$.
- The difference between any two workers' salaries is $O(\log(n)/n)$.
- Balanced Markets: In balanced markets (equal number of firms and workers) with uniform distributions, the workers' share of the total surplus is $O(\log^2(n)/n)$ under the firm-optimal core allocation.
- Unbalanced Markets: Even slightly unbalanced markets tend to give almost all the surplus to the "short side" of the market.
- Unbounded Distributions: For exponential distributions, wage dispersion can be significant, with workers' salaries differing by $O(\log(n))$.
Limitations and Future Work
- Bounded Distribution Assumption: The main theoretical results rely on the assumption of bounded productivity distributions. The paper acknowledges that this is a strong assumption and that unbounded distributions lead to different outcomes (significant wage dispersion).
- "Approximate" Law of One Price: The paper establishes an approximate law of one price, where differences are of order $O(\log(n)/n)$. While small for large $n$, this difference is not zero, indicating that some wage dispersion persists.
- Future Work:
- Providing tighter bounds on wage dispersion.
- Further analytical study of unbounded distributions, particularly the exponential distribution, where they conjecture workers' share of surplus is $\Theta(n \log(\log(n)))$.
- Investigating correlated random matrices and the impact of interaction variables (complementarities) in production functions.
- Extending the analysis to markets with non-separable production functions.
Summary updated