Source-linked AI summary
A Short-term Intervention for Long-term Fairness in the Labor Market
Lily Hu, Yiling Chen
TL;DR
The paper asks how to address persistent group disparities when input-output individual fairness may leave them intact. It models a parity-constrained entry stage followed by performance-based hiring, and finds a stable group-symmetric equilibrium that can Pareto-dominate unconstrained strategies under particular conditions.
Problem
Input-output individual fairness can leave labor-market disparities intact because workers’ investments are choices shaped by unequal group standing and hiring incentives.
Source: §1 · pp. 1–2Method
The paper models a dual labor market with statistical parity in the TLM and reputational, best-response hiring in the PLM.
Source: §1 · p. 2Results
Under particular labor-market conditions, the TLM constraint produces group-symmetric outcomes that Pareto-dominate group-blind and statistically discriminatory hiring.
Source: §1 · p. 2Takeaways & Limitations
Procedural fairness in hiring alone is insufficient for addressing unequal opportunities and qualifications, while the proposed intervention can become unnecessary once group equality is restored.
Source: §1 · p. 2The model simplifies a more complex and heterogeneous real labor-market pipeline, and the range of possible policy interventions is broader than those considered.
Source: §4 · p. 8
Abstract
from arXiv · showhide
The persistence of racial inequality in the U.S. labor market against a general backdrop of formal equality of opportunity is a troubling phenomenon that has significant ramifications on the design of hiring policies. In this paper, we show that current group disparate outcomes may be immovable even when hiring decisions are bound by an input-output notion of "individual fairness." Instead, we construct a dynamic reputational model of the labor market that illustrates the reinforcing nature of asymmetric outcomes resulting from groups' divergent accesses to resources and as a result, investment choices. To address these disparities, we adopt a dual labor market composed of a Temporary Labor Market (TLM), in which firms' hiring strategies are constrained to ensure statistical parity of workers granted entry into the pipeline, and a Permanent Labor Market (PLM), in which firms hire top performers as desired. Individual worker reputations produce externalities for their group; the corresponding feedback loop raises the collective reputation of the initially disadvantaged group via a TLM fairness intervention that need not be permanent. We show that such a restriction on hiring practices induces an equilibrium that, under particular market conditions, Pareto-dominates those arising from strategies that statistically discriminate or employ a "group-blind" criterion. The enduring nature of equilibria that are both inequitable and Pareto suboptimal suggests that fairness interventions beyond procedural checks of hiring decisions will be of critical importance in a world where machines play a greater role in the employment process.
1 INTRODUCTION
The paper argues that unequal hiring outcomes can persist under input-output individual fairness because investment choices and hiring incentives reinforce group differences. It models a labor-market-specific parity intervention that can produce stable group equality and, under particular conditions, better outcomes than unconstrained rational hiring.
1 INTRODUCTION
- Statistical-discrimination equilibria can reinforce unequal outcomes as firms’ group-specific thresholds shape workers’ investment choices and confirm prior beliefs.Workers in the unfavored group lower investment while those in the favored group continue investing at high levels.Source: §1 · pp. 1–2
- Input-output fairness based on observed investments does not resolve disparities when investment is a strategic choice shaped by labor-market incentives.Even local procedural fairness with respect to investment choices can coexist with globally disparate outcomes.Source: §1 · p. 2
- The proposed intervention uses labor-market feedback to move asymmetric equilibria toward a stable, self-sustaining group-equitable equilibrium.The paper argues that procedural fairness in hiring alone is insufficient to address unequal opportunities and qualifications.Source: §1 · p. 2
- The model constrains hiring to statistical parity in the TLM, then allows PLM firms to use best responses while individual reputations affect group reputation.A worker’s individual reputation summarizes prior job performance; group reputation measures the share of members producing good outcomes.Source: §1 · p. 2
- Under particular labor-market conditions, TLM parity yields group-symmetric outcomes that Pareto-dominate group-blind and statistically discriminatory hiring.The intervention need not be permanent once group equality is restored and statistical parity becomes firms’ natural hiring outcome.Source: §1 · p. 2
1.1 Related Work
- The model assumes inherent equality between groups, attributing observed differences in investment and outcomes to unequal societal standing rather than fundamental individual differences.The paper also relates its two-stage setup to fairness approaches that first preserve statistical parity and then treat similar agents similarly.Source: §1.1 · pp. 2–3
- The paper frames hiring as reputational and incorporates a dynamic group reputation parameter into individuals’ costs, while defining individual reputation as a total history of previous outcomes.Unlike Levin, collective reputation also explicitly impacts future generations’ investment costs.Source: §1.1 · p. 3
2 MODEL
The model separates labor-market entry from permanent employment so that statistical parity can be imposed at entry while firms later use reputational hiring. Worker investment, performance histories, and group reputation connect the stages over time.
2 MODEL
- The model uses a Temporary Labor Market (TLM) with parity-constrained hiring followed by a Permanent Labor Market (PLM) where firms use best-response hiring.Workers flow between stages, and individual reputations in the PLM produce externalities for collective group reputations that shape pre-market investment decisions.Source: §2 · p. 3
2.1 General Setup
- The setup keeps group shares and ability distributions stable, while group reputation measures the share of members producing good outcomes over a lagged time window.Workers invest before entry based on expected wages, and wages vary with the proportion of good workers in the PLM.Source: §2.1 · p. 3
2.2 Temporary Labor Market
- Before TLM entry, workers invest in human capital under group-sensitive costs; investment is an imperfect signal that increases the probability of qualification.TLM hiring depends on observable investment and group membership, and workers denied entry are excluded from skilled-job candidacy.Source: §2.2 · pp. 3–4
2.3 Permanent Labor Market
- In the PLM, workers choose effort that affects observable good or bad outcomes, and firms use recent outcome histories to identify workers who consistently exert effort.High effort costs depend on qualification, while individual reputations summarize the proportion of good outcomes in a worker’s recent history.Source: §2.3 · p. 4
- The paper defines long-term fairness as group equality in labor-market outcomes, including job and wage prospects at a fair steady state.The model does not assert that outcomes along the path to the symmetric equilibrium satisfy fairness notions.Source: §2.3 · p. 4
3 RESULTS
The section characterizes labor-market equilibria under a TLM statistical-parity constraint and compares them with unconstrained hiring strategies. The constraint yields a stable symmetric steady state and, under unsaturated skilled-worker demand, a Pareto-dominant outcome relative to group-blind and statistical-discriminatory hiring.
3.1 Equilibrium Strategies and Steady-States
- In the PLM, firms use a reputation threshold that decreases over time, while workers exert high effort only when its wage return covers the effort cost.Repeated firm-worker interactions produce a recursive relationship in the proportion of good-outcome workers in each group.Source: §3.1 · p. 6
- Under TLM statistical parity, the model has a unique stable symmetric steady state with equal group reputations after convergence time T and a unique stable wage.The result applies under the described market conditions and when ℓ proportion of workers enter the TLM.Source: §3.1 · p. 6
- Parity maintains equal underlying ability distributions and proportional group representation, constraining reputational feedback as group outcomes and reputations converge.As the system approaches the symmetric steady state, the severity of the TLM constraint slackens until it recedes into disuse.Source: §3.1 · p. 6
3.2 Comparative Statics with Unconstrained Hiring Strategies
- With unsaturated PLM demand, the TLM constraint yields a symmetric steady state that Pareto-dominates the asymmetric equilibria under group-blind and statistical-discriminatory hiring.The comparison is stated for skilled-worker demand at w = w̄.Source: §3.2 · p. 7
- Group-blind hiring can sustain divergent group reputations because it preserves neither equal ability distributions nor proportional representation.With unsaturated PLM demand at w = w̄, the constrained equilibrium makes group B workers better off without worsening group W outcomes.Source: §3.2 · p. 7
- Statistical-discriminatory hiring does not ensure parity and can sustain asymmetric equilibria in which firms’ stricter standards reinforce lower investment in the lower-standing group.These effects mirror the Pareto-dominated outcomes under group-blind hiring.Source: §3.2 · p. 7
4 DISCUSSION
The discussion frames labor-market inequality as a dynamic problem shaped by historical differences in resources, opportunities, and investment choices, not only by hiring procedures. It argues that fairness should be assessed over long-run outcomes while recognizing limits in the model and in the policy options examined.
4 DISCUSSION
- Historical differences in resources and opportunities shape workers’ choices and outcomes, so fair hiring procedures alone may not address labor-market inequality.The discussion describes this as development bias and contrasts dynamic labor-market analysis with criteria focused only on qualifications.Source: §4 · p. 7
- A fairness intervention can aim for permanent group egalitarianism and may be economically and socially optimal, even though outcomes on the path to symmetry are not guaranteed to be fair.The discussion emphasizes steady-state outcomes and fairness as a long-term project.Source: §4 · p. 7
- The model’s reputation measure uses workers’ prior permanent-market outcomes, a form of history that fits employers’ increased access to worker data.The discussion leaves the design of systems that decide when to use group-membership considerations to future work.Source: §4 · p. 7
- The labor-market pipeline is more complex and heterogeneous than the model, and other interventions may also produce group-egalitarian outcomes.Comparing their costs and efficiencies, and their welfare effects, remains an open question for theoretical and empirical work.Source: §4 · p. 8
APPENDIX
The appendix establishes convergence under statistical-parity hiring and contrasts it with group-blind hiring, where asymmetric outcomes may persist. It also shows that, under specified market conditions, the symmetric outcome can Pareto-dominate asymmetric outcomes under group-blind or statistically discriminatory hiring.
APPENDIX
- A firm reputation threshold near pH induces high effort when workers can afford it, while recent performance estimates individual reputation.The argument invokes the law of large numbers and assumes workers have incentives to pursue skilled work when individually rational.Source: APPENDIX · p. 8
- The TLM constraint preserves the groups’ ability distributions and parity in skilled-job offers, linking group reputation to the proportion of qualified workers.With proportional representation, instantaneous group reputation scales with the group’s proportion of qualified workers.Source: APPENDIX · pp. 8–9
- The convergence proof composes a contracting labor-flow map with a weakly contracting investment map to form a contraction.The recursive self-map therefore has a unique fixed point.Source: APPENDIX · p. 9
- From unequal initial group reputations, statistical-parity hiring converges to equal group outcomes and a unique equilibrium wage.The proof establishes a unique fixed point, with group reputations converging to equality.Source: APPENDIX · pp. 8–9
- Group-blind hiring does not ensure statistical parity, and its reputation feedback need not contract, so asymmetric equilibria can persist.The appendix describes such equilibria as maintaining distinct group investment costs and unequal reputations.Source: APPENDIX · p. 9
- With unsaturated PLM demand and a wage that cannot rise, the fair equilibrium Pareto-dominates group-blind outcomes: some qualified group B workers gain access, while group W workers are no worse off.Group-blind thresholds can exclude high-ability group B workers who would be hired under the fair regime.Source: APPENDIX · p. 9
- Under statistical discriminatory hiring, asymmetric equilibria can also arise and are Pareto-dominated by statistical-parity hiring.The proof compares the groups’ incentive-compatibility constraints and applies the same lost-efficiency argument.Source: APPENDIX · p. 9