Fairness at Equilibrium in the Labor Market

Lily Hu, Yiling ChenarXiv 2017 · OpenReview

Abstract

Recent literature on computational notions of fairness has been broadly divided into two distinct camps, supporting interventions that address either individual-based or group-based fairness. Rather than privilege a single definition, we seek to resolve both within the particular domain of employment discrimination. To this end, we construct a dual labor market model composed of a Temporary Labor Market, in which firm strategies are constrained to ensure group-level fairness, and a Permanent Labor Market, in which individual worker fairness is guaranteed. We show that such restrictions on hiring practices induces an equilibrium that Pareto-dominates those arising from strategies that employ statistical discrimination or a "group-blind" criterion. Individual worker reputations produce externalities for collective reputation, generating a feedback loop termed a "self-fulfilling prophecy." Our model produces its own feedback loop, raising the collective reputation of an initially disadvantaged group via a fairness intervention that need not be permanent. Moreover, we show that, contrary to popular assumption, the asymmetric equilibria resulting from hiring practices that disregard group-fairness may be immovable without targeted intervention. The enduring nature of such equilibria that are both inequitable and Pareto inefficient suggest that fairness interventions are of critical importance in moving the labor market to be more socially just and efficient.

Introduction

In the field of algorithmic fairness, a persistent tension exists between two primary philosophical frameworks: individual fairness (the requirement that similar individuals be treated similarly) and group fairness (the requirement that outcomes be balanced across demographic groups). Most existing research treats these as static problems, analyzing a single point in time. However, the labor market is inherently dynamic, characterized by feedback loops where past hiring decisions influence future worker behavior and firm perceptions.

This paper bridges the gap between these two camps by modeling the labor market as a dynamic system. The authors argue that the perceived trade-off between fairness and economic efficiency is a false dichotomy. Instead, they demonstrate that targeted group-fairness interventions can break cycles of statistical discrimination, eventually leading to a stable, long-term equilibrium that is both fairer and more economically efficient (Pareto-dominant) for all participants.

The Problem: The Self-Fulfilling Prophecy

The core challenge in labor market fairness is the "self-fulfilling prophecy" of statistical discrimination. If a firm believes a certain group is less productive—often due to historical biases or lack of data—they may set a higher hiring bar for members of that group. Workers within that group, observing that the return on their effort is lower than that of their peers, are disincentivized to invest in their own skills and human capital.

This creates a "trap" where discriminatory outcomes appear rational to both firms and workers in the short term, but lead to long-term stagnation and inequality. The authors posit that we cannot fix future outcomes without addressing the recursive feedback loops created by historical inequities.

The Dual Labor Market Model

To address this, the authors propose a theoretical, dual-stage labor market model designed to transition from group-level interventions to individual-level meritocracy:

  1. Temporary Labor Market (TLM): This serves as the entry point for workers. In this stage, firms are legally or algorithmically required to satisfy group fairness (statistical parity). They must hire a proportional number of workers from different groups, regardless of their initial "collective reputations."
  2. Permanent Labor Market (PLM): Workers transition to this stage after their performance in the TLM has been observed. Here, firms hire based on individual reputation—a history of the worker's actual performance and skill acquisition.

Key Model Components

  • Investment Costs: Workers decide whether to invest in human capital based on expected wage gains. Crucially, the cost of this investment is influenced by a group's "collective reputation," representing the social and economic externalities of belonging to a disadvantaged group.
  • Recursive Dynamics: The model is dynamic; the proportion of high-skill workers in the market at any time $t$ is a function of the wages and reputations established at time $t-1$.

Key Insights and Results

The Efficiency of Fairness

A common misconception in policy design is that fairness comes at the expense of efficiency. This research proves the opposite: because statistical discrimination leaves "untapped resources" (skilled workers in disadvantaged groups) on the sidelines, a fairness intervention moves the market to a Pareto-dominant state. In this state, the overall productivity of the market increases, making everyone better off than they were under the discriminatory equilibrium.

Reputation as an Externality

The model treats an individual’s success as a contribution to their group's "collective reputation." In a discriminatory system, a worker from a marginalized group faces a "negative externality"—they must exert significantly more effort to prove their worth compared to a worker from a favored group. The TLM intervention internalizes these externalities by equalizing the hiring bar.

Convergence and Symmetry

In Theorem 3.1, the authors prove that under the TLM fairness constraint, groups with unequal initial standings will gradually converge to the same reputation level. This leads to a group-symmetric steady-state equilibrium. Once reputations have equalized, the requirement for group-level constraints becomes unnecessary; individual-level fairness in the PLM naturally maintains the balance. This suggests that fairness interventions need not be permanent to be effective.

Failure of "Group-Blind" Strategies

The paper demonstrates that "group-blind" hiring—ignoring group membership entirely—often fails to rectify systemic inequality. Because group-blind strategies do not account for the differing investment costs and historical externalities faced by workers, they tend to reinforce the status quo. The authors show that these strategies are often less efficient than the proposed fair intervention because they continue to exclude high-ability workers from disadvantaged backgrounds.

Conclusion

The enduring nature of inequitable equilibria suggests that the labor market cannot always correct itself. Targeted interventions, such as the Temporary Labor Market model, are essential to "jolt" the system out of inefficient, discriminatory cycles. By combining group-fairness constraints in the short term with individual-fairness rewards in the long term, society can move toward a labor market that is both more socially just and more economically productive.

Summary updated

Abstract

Recent literature on computational notions of fairness has been broadly divided into two distinct camps, supporting interventions that address either individual-based or group-based fairness. Rather than privilege a single definition, we seek to resolve both within the particular domain of employment discrimination. To this end, we construct a dual labor market model composed of a Temporary Labor Market, in which firm strategies are constrained to ensure group-level fairness, and a Permanent Labor Market, in which individual worker fairness is guaranteed. We show that such restrictions on hiring practices induces an equilibrium that Pareto-dominates those arising from strategies that employ statistical discrimination or a "group-blind" criterion. Individual worker reputations produce externalities for collective reputation, generating a feedback loop termed a "self-fulfilling prophecy." Our model produces its own feedback loop, raising the collective reputation of an initially disadvantaged group via a fairness intervention that need not be permanent. Moreover, we show that, contrary to popular assumption, the asymmetric equilibria resulting from hiring practices that disregard group-fairness may be immovable without targeted intervention. The enduring nature of such equilibria that are both inequitable and Pareto inefficient suggest that fairness interventions are of critical importance in moving the labor market to be more socially just and efficient.
BibTeX
@misc{hu2017fairness,
  author = {Lily Hu and Yiling Chen},
  title = {{Fairness at Equilibrium in the Labor Market}},
  year = {2017},
  eprint = {1707.01590},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1707.01590},
}
Figure