The Role of Referrals in Immobility, Inequality, and Inefficiency in Labor Markets

Lukas Bolte; Matthew O. Jackson; Nicole Immorlica

arxiv: 2012.15753 · v2 · submitted 2020-12-22 · 💰 econ.GN · physics.soc-ph· q-fin.EC

The Role of Referrals in Immobility, Inequality, and Inefficiency in Labor Markets

Lukas Bolte , Nicole Immorlica , Matthew O. Jackson This is my paper

Pith reviewed 2026-05-24 14:14 UTC · model grok-4.3

classification 💰 econ.GN physics.soc-phq-fin.EC

keywords referralslabor marketsinequalityimmobilityhomophilysocial networksaffirmative actionalgorithmic fairness

0 comments

The pith

Job referrals improve hiring matches and productivity but increase inequality and reduce mobility when networks exhibit homophily.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a model showing that referrals create more hiring opportunities and better worker-firm matches, raising overall productivity. At the same time, workers with few employed connections receive fewer referrals and face worse outcomes, which widens inequality. Homophily in social networks compounds this by channeling referrals within groups, limiting movement across social or economic strata over time. The authors derive conditions under which more even distribution of referrals can shrink inequality gaps while raising future productivity and mobility. They then use the framework to trace short-run and long-run welfare effects of interventions such as affirmative action and algorithmic fairness rules.

Core claim

Referrals lead to more opportunities for workers to be hired, which lead to better matches and increased productivity, but also disadvantage job-seekers with few or no connections to employed workers, increasing inequality. Coupled with homophily, referrals also lead to immobility. Conditions exist under which distributing referrals more evenly reduces inequality and improves future productivity and mobility.

What carries the argument

A model of referral-based hiring within homophilous social networks that tracks how referral flows affect match quality, inequality, and intergenerational mobility.

If this is right

Referrals raise aggregate productivity through improved matches.
Workers lacking employed connections receive fewer opportunities, widening inequality.
Homophily channels referrals inside groups and produces immobility across periods.
Even distribution of referrals can reduce inequality and raise future productivity and mobility under stated conditions.
Policies such as affirmative action and algorithmic fairness alter short-run and long-run welfare through their effects on referral flows.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Network position becomes an independent driver of labor-market outcomes that persists even after individual productivity is accounted for.
Interventions aimed at network formation itself could interact with referral policies in ways the model does not yet explore.
The same referral process may generate different inequality-mobility trade-offs when labor markets vary in the share of jobs filled by referrals.

Load-bearing premise

The model assumes a specific form of homophily in the underlying social network and a particular process by which referrals are generated and used.

What would settle it

Longitudinal data on worker networks, referral receipt, and wage or occupational mobility that shows no rise in group-level immobility or inequality despite high homophily would falsify the central mechanism.

Figures

Figures reproduced from arXiv: 2012.15753 by Lukas Bolte, Matthew O. Jackson, Nicole Immorlica.

**Figure 2.** Figure 2: Referrals for a group happen according to a Poisson distribution: [PITH_FULL_IMAGE:figures/full_fig_p022_2.png] view at source ↗

**Figure 3.** Figure 3: The blue line is with myopic firms. The orange line is when firms can redraw from [PITH_FULL_IMAGE:figures/full_fig_p025_3.png] view at source ↗

**Figure 4.** Figure 4: A graphical representation of the pool with and without the policy. [PITH_FULL_IMAGE:figures/full_fig_p050_4.png] view at source ↗

**Figure 5.** Figure 5: Pool 1 Consider the firms hiring workers in A from pool 1. The distribution of values in set A is constant by construction; and, by assumption, a firm at first hiring such worker can replace the worker with a worker with value ˜v 0 2 ≥ v˜2 under λ 0 and with a worker with value ˜v2 under λ. Thus, average production after firing of firms hiring workers in A is larger given λ 0 . Consider the firms hiring wo… view at source ↗

**Figure 6.** Figure 6: The parameter values are: nH = nL = 1 so that n = 2. We plot the mass of high-value workers employed in steady state as a function of the degree of value-homophily, α. this example the mass of high valued workers translates directly into the overall productivity. 13 [PITH_FULL_IMAGE:figures/full_fig_p058_6.png] view at source ↗

read the original abstract

We study the consequences of job markets' heavy reliance on referrals. Referrals lead to more opportunities for workers to be hired, which lead to better matches and increased productivity, but also disadvantage job-seekers with few or no connections to employed workers, increasing inequality. Coupled with homophily, referrals also lead to immobility. We identify conditions under which distributing referrals more evenly reduces inequality and improves future productivity and mobility. We use the model to examine the short and long-run welfare impacts of policies such as affirmative action and algorithmic fairness.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This model links referrals and homophily to productivity gains alongside inequality and immobility, then derives conditions under which referral redistribution improves both, though the results track the specific network and matching assumptions closely.

read the letter

The key things here are that referrals raise hiring odds and match quality but, with homophily, also lock in low mobility for unconnected workers and widen gaps, and the authors map out when spreading referrals more evenly can cut inequality without sacrificing long-run productivity. They then use the setup to compare affirmative action and algorithmic fairness rules on short- and long-run welfare. That combination of static efficiency, dynamic mobility, and policy application is the main addition over earlier referral papers. The framework stays tractable enough to produce clear thresholds, which is useful for thinking about interventions even without calibration. The derivations appear to deliver the stated conditions on redistribution, so the central logic holds together on its own terms. The main soft spot is that the immobility and inequality effects, plus the policy ranking, depend on the exact homophily kernel and the referral-offer protocol. Changing either can alter or reverse the qualitative results, and the abstract gives no sign of analytic or numerical checks that isolate this dependence. If the full paper does not supply those checks, the policy conclusions travel less far than they first appear. This is aimed at labor economists who already work with network models of hiring. A reader who wants one setup that tracks productivity, inequality, and mobility together will get something concrete from the conditions they identify. It is worth sending to referees because the question is policy-relevant and the model is coherent enough to repay detailed review even if the robustness section needs work.

Referee Report

2 major / 1 minor

Summary. The paper develops a theoretical model of labor markets in which referrals through social networks increase hiring opportunities and improve match quality/productivity, but combined with homophily generate inequality (disadvantaging unconnected workers) and immobility. It derives conditions under which more even referral distribution reduces inequality while raising future productivity and mobility, and uses the framework to evaluate short- and long-run welfare effects of affirmative action and algorithmic fairness policies.

Significance. If the results are robust, the paper offers a useful framework for analyzing efficiency-equity trade-offs in network-based hiring and for ranking policy interventions. The explicit identification of threshold conditions on referral distribution is a constructive contribution. However, the absence of robustness checks on the imposed homophily kernel and referral protocol limits the strength of the conclusions.

major comments (2)

[Model section] Model section (homophily and referral protocol): the qualitative claims on immobility, inequality, and the ranking of even-distribution policies are shown only for a specific functional form of the homophily kernel and a particular referral-offer/acceptance rule. The manuscript supplies no analytic derivation or numerical sensitivity exercise demonstrating that these results survive changes to either functional form, yet altering the kernel or protocol can eliminate or reverse the reported effects on mobility and the policy welfare ordering.
[Welfare and policy section] Welfare and policy section: the short- and long-run welfare comparisons for affirmative action and algorithmic fairness are derived under the baseline network-formation rule; because the immobility and inequality magnitudes are not shown to be robust to alternative homophily strengths or referral-matching specifications, the policy conclusions rest on the same untested functional choices.

minor comments (1)

Notation for network statistics and referral probabilities should be defined once in a dedicated table or appendix to improve readability across equilibrium derivations.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. The concerns about robustness to alternative specifications of the homophily kernel and referral protocol are well-taken, and we address them point by point below, outlining how we plan to strengthen the paper.

read point-by-point responses

Referee: [Model section] Model section (homophily and referral protocol): the qualitative claims on immobility, inequality, and the ranking of even-distribution policies are shown only for a specific functional form of the homophily kernel and a particular referral-offer/acceptance rule. The manuscript supplies no analytic derivation or numerical sensitivity exercise demonstrating that these results survive changes to either functional form, yet altering the kernel or protocol can eliminate or reverse the reported effects on mobility and the policy welfare ordering.

Authors: We chose the specific homophily kernel and referral protocol to permit closed-form solutions while capturing the essential feature that connection probabilities decline with social distance. The core mechanisms generating immobility and inequality—differential referral access due to homophily—hold under any kernel satisfying this monotonicity property, which includes several common alternatives. Nevertheless, we agree that explicit sensitivity analysis would improve the paper. In revision we will add numerical exercises that vary the homophily strength parameter and consider an alternative acceptance rule (e.g., random acceptance conditional on meeting a minimum productivity threshold). These checks will confirm whether the qualitative directions of the effects on mobility, inequality, and the ranking of even-distribution policies remain intact. revision: partial
Referee: [Welfare and policy section] Welfare and policy section: the short- and long-run welfare comparisons for affirmative action and algorithmic fairness are derived under the baseline network-formation rule; because the immobility and inequality magnitudes are not shown to be robust to alternative homophily strengths or referral-matching specifications, the policy conclusions rest on the same untested functional choices.

Authors: The policy welfare comparisons are presented as illustrations within the baseline framework rather than universal rankings. Once the numerical robustness checks described above are incorporated, we will qualify the policy results by noting the range of homophily and protocol parameters under which the short- and long-run welfare orderings continue to hold. This will make the policy implications more precise without altering the main qualitative insights. revision: partial

Circularity Check

0 steps flagged

No significant circularity in theoretical model derivations

full rationale

The paper is a self-contained theoretical model deriving labor-market outcomes from explicit assumptions on network homophily and referral generation/acceptance protocols. No equations or claims reduce by construction to fitted parameters, self-citations, or renamed inputs; the immobility/inequality results follow directly from the stated primitives without the forbidden patterns of self-definition or load-bearing self-citation chains. External benchmarks (standard network-formation models) are not required for the internal logic, confirming the derivation chain is independent.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract only; cannot enumerate free parameters, axioms, or invented entities from the full model. The abstract implies reliance on network homophily and referral generation rules as modeling choices.

axioms (1)

domain assumption Social networks exhibit homophily in connections
Invoked to generate immobility from referrals.

pith-pipeline@v0.9.0 · 5624 in / 1192 out tokens · 19526 ms · 2026-05-24T14:14:10.002769+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Belief Aggregation under Costly Information
econ.TH 2026-06 unverdicted novelty 5.0

A costly-information model provides an epistemic foundation that rationalizes linear, geometric, power, and multiplicative belief pooling and links the choice of rule to welfare losses and equilibrium prices.

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages · cited by 1 Pith paper

[1]

Chapter 22 - Labor Markets and Referrals,

German translation: Campus, 1975; Japanese translation: 1977; Spanish translation: Ed- itorial Critica, 1979; Yugoslav translation: Cekade, 1984. Expanded edition with an annex ”On Economic Inequality after a Quarter Century” [jointly with James Foster], Oxford: Clarendon Press; New York: Oxford University Press, 1997. Topa, Giorgio, “Chapter 22 - Labor M...

work page 1975
[2]

By Lemma 6, production before ﬁring is lower under ˜ v′ 1 than under ˜v1 as ˜v <˜v1 < ˜v′

Total production must be higher under ˜v′ 1 than under ˜v1 as is evident from (12). By Lemma 6, production before ﬁring is lower under ˜ v′ 1 than under ˜v1 as ˜v <˜v1 < ˜v′

work page
[3]

Hence, it must be that ˜v′ 2 < ˜v2 for total production to be the same under the two thresholds. But then (4) cannot be satisﬁed for both ˜v1 and ˜v2 and ˜v′ 1 and ˜v′ 2 as the both terms on the right-hand side are larger given ˜v1 and ˜v2 than given ˜v′ 1 and ˜v′ 2 whereas ˜v1< ˜v′ 1 by assumption. Lastly, let us show that ˜v2 is unique. By the uniquenes...

work page
[4]

We know that ˜v′ 1 > ˜v1

and (˜v1, ˜v2) respectively. We know that ˜v′ 1 > ˜v1. To reach a contradiction suppose that ˜v′ 2≥ ˜v2. We study how production after ﬁring diﬀers given λ and λ′ with λ,λ′ > 0. To this end, we partition the set of workers into several subsets whose aggregate composition of workers stays constant given λ and λ′. We then track the extend to which each of t...

work page
[5]

inbreeding-bias

Due to uniform sampling from pool 1, it must be that the same fraction of workers in both rectangles is hired from the aforementioned sets. We graphically represent this fact by placing the rectangles side-by-side and, given λ′, horizontally slice through them (the top dashed lines) with the top area representing the set of unemployed workers and the bott...

work page

[1] [1]

Chapter 22 - Labor Markets and Referrals,

German translation: Campus, 1975; Japanese translation: 1977; Spanish translation: Ed- itorial Critica, 1979; Yugoslav translation: Cekade, 1984. Expanded edition with an annex ”On Economic Inequality after a Quarter Century” [jointly with James Foster], Oxford: Clarendon Press; New York: Oxford University Press, 1997. Topa, Giorgio, “Chapter 22 - Labor M...

work page 1975

[2] [2]

By Lemma 6, production before ﬁring is lower under ˜ v′ 1 than under ˜v1 as ˜v <˜v1 < ˜v′

Total production must be higher under ˜v′ 1 than under ˜v1 as is evident from (12). By Lemma 6, production before ﬁring is lower under ˜ v′ 1 than under ˜v1 as ˜v <˜v1 < ˜v′

work page

[3] [3]

Hence, it must be that ˜v′ 2 < ˜v2 for total production to be the same under the two thresholds. But then (4) cannot be satisﬁed for both ˜v1 and ˜v2 and ˜v′ 1 and ˜v′ 2 as the both terms on the right-hand side are larger given ˜v1 and ˜v2 than given ˜v′ 1 and ˜v′ 2 whereas ˜v1< ˜v′ 1 by assumption. Lastly, let us show that ˜v2 is unique. By the uniquenes...

work page

[4] [4]

We know that ˜v′ 1 > ˜v1

and (˜v1, ˜v2) respectively. We know that ˜v′ 1 > ˜v1. To reach a contradiction suppose that ˜v′ 2≥ ˜v2. We study how production after ﬁring diﬀers given λ and λ′ with λ,λ′ > 0. To this end, we partition the set of workers into several subsets whose aggregate composition of workers stays constant given λ and λ′. We then track the extend to which each of t...

work page

[5] [5]

inbreeding-bias

Due to uniform sampling from pool 1, it must be that the same fraction of workers in both rectangles is hired from the aforementioned sets. We graphically represent this fact by placing the rectangles side-by-side and, given λ′, horizontally slice through them (the top dashed lines) with the top area representing the set of unemployed workers and the bott...

work page