lifescore Edge5 min read

Do job referrals widen opportunity gaps?

A randomized employee program reduced attrition and labor costs, while observational employer data link referred hires to lower recruiting costs; experiments show incentives and referral constraints change who gets put forward, but eventual hiring or promotion gaps remain unresolved.

Read the evidence

August 4, 2026

A network of paths converges on a threshold, with one connected signal passing through and an adjustable second gate beyond.

A referral is not simply a recommendation. It is a gate into a network.

That gate can help an employer find people who stay longer. It can also narrow the candidate pool to people who already resemble those inside. The strongest evidence does not justify a verdict that referrals are good or bad. It shows which parts of the system change the result.

Referred hires often look valuable—but selection remains

Burks and colleagues analyzed personnel data from nine large firms across call centers, trucking and high technology.

Referred applicants were more likely to be hired and accept an offer. Once hired, they performed similarly on most measures and were less likely to quit. The firms also earned more worker-level profit in the call-center and trucking settings because recruiting and turnover costs were lower. Referred truckers had fewer accidents; referred high-technology workers produced more patents.

That is a meaningful business case. It is not a randomized trial of receiving a referral. People who can attract a referral, the employees who choose them and the jobs where firms use the channel can differ before hiring begins.

The referral premium may reflect information and matching, selection or both.

Incentives can change where a referrer searches

A randomized experiment in Kolkata shows that referrals are not fixed outputs of a social network.

Adult participants could refer someone for temporary work. Some received a fixed finder fee. Others were paid according to the referred worker’s performance.

Under the high-stakes performance contract, participants were about seven percentage points less likely to refer relatives and eight points more likely to refer coworkers. Higher-ability participants also selected stronger performers under the incentive; lower-ability participants did not show the same improvement.

The reward changed which relationship became useful. But this was one-shot temporary work, not a conventional employee referral program, and it was not a test of gender or racial access.

Referral constraints and referrer gender changed who appeared

A recruitment experiment for enumerator work in Malawi directly tested the gender composition of referrals.

Women made up 30% of unrestricted referrals, compared with 38% of the conventional applicant pool. Men referred men 77% of the time. Women referred women 43% of the time.

Yet the experiment also changed the request. When participants were asked to refer a woman, men on fixed fees produced candidates with statistically similar qualification rates to men assigned to refer men. That equality did not hold under performance pay: male referrals in the male-only arm qualified at 62%, versus 41% in the female-only arm. Performance pay did not shift male referrers toward more women or improve female-referral quality.

The result is not referrals exclude women. It is more precise: unrestricted referrals contained fewer women, driven by men referring women only 23% of the time. The study did not map participants’ networks or identify whether the pattern came from network scarcity, preferences, information or social incentives.

A real employee program improved retention indirectly

A 13-month randomized program across 238 stores and more than 10,000 workers in one European grocery chain offers a different result.

The referral program reduced attrition by about 15% and lowered labor costs. But larger bonuses increased referral hires while reducing their retention advantage. At the largest bonus, only 5% of hires came through referrals.

Much of the retention benefit was indirect: nonreferred workers stayed longer too.

This is strong evidence that a referral program can change a workplace beyond the people it directly hires. The study did not report race or gender composition, so it cannot establish whether those gains widened or narrowed demographic access.

Same-gender referral patterns do not settle segregation’s direction

In a study of 453 Stockholm business students, 73% referred someone of their own gender. Most close friends were also same-gender, and half of referrals named a listed friend.

The pattern is evidence of network homophily. It cannot estimate a hiring-gap effect: only 13 unique referrals applied, and none was hired.

The study’s longer-run direction is theoretical. Only 13 of 313 unique referred students applied and none was hired, so it could not estimate a hiring-gap effect. Whether the pattern raises or reduces segregation depends on gender-specific referral rates, job context, take-up and hiring.

What changes the access effect

Design question	What the evidence shows	What remains unknown
Who can refer?	Employees, applicants and outside contacts search different networks	Which design generalizes across employers
What earns the reward?	Performance incentives can shift search away from relatives toward coworkers	Whether the shift broadens protected-group access
Is the request unrestricted?	Unrestricted referrals contained fewer women in the Malawi experiment	Whether network composition, preferences, information or social incentives caused the pattern
How large is the bonus?	Larger bonuses raised referral hires while reducing their retention advantage	The best incentive outside the tested program
What counts as success?	Retention and labor cost can improve	Whether access, pay and promotion improve too

Primary sources

Independent editorial summary. The authors and institutions are not affiliated with LifeScore.

The lifescore take

Do not ask whether referrals work. Ask what the referral system selects. A network can carry information that a résumé does not. It can also hide people who have no path into that network. Incentives, eligibility and the wording of the request all change which contacts become candidates. The useful audit follows the full route: referral, application, evaluation, hire, retention and progression. A win at one stage does not certify fairness at the next. ## What this evidence does not answer - It does not establish that every referral program widens inequality. - It does not show that referred workers are inherently better workers. - It does not make employee, applicant and student referrals equivalent. - It does not infer gender or racial outcomes from studies that did not measure them. - It does not identify one optimal bonus or referral constraint. - It does not prove that changing who is referred changes who is hired. - It does not substitute employer retention for equal opportunity.

Article link