The Value of Hiring Through Employee Referrals

The question

Employee referrals reliably produce better hires than non-referral applicants — but why? Better people overall, or a better match to the specific job?

The method

Personnel records for hundreds of thousands of applicants and workers across nine large firms in three industries (call centers, trucking, high-tech) [1], comparing referred and non-referred hires on productivity, turnover, and profit, plus four separate tests designed to tell “selection” (referrals bring better-matched people in the door) apart from “treatment” (referring coworkers coach or monitor them afterward) [2] (Abstract; §1).

The findings

Referred and non-referred workers look almost identical on schooling, cognitive ability, and personality (Table 1) [3] — referrals aren’t simply funneling in generally superior candidates. But referred workers quit noticeably less often in all three industries (Table 2) [3], and do measurably better on the rare, high-stakes outcomes that matter most: fewer preventable trucking accidents (Table 4) [3] and more patent applications at the high-tech firm (Table 5) [3]. Four separate tests all point the same direction — referred workers’ performance doesn’t drop after the referring coworker leaves; referral bonuses tied to tenure don’t change quitting behavior around the bonus threshold; having more on-the-job friends (as opposed to being referred) doesn’t predict better outcomes; and referred workers already score higher on pre-hire tests and interviews, before any coaching could occur (§5) [3]. The paper’s conclusion: referrals work through selection — a better job match — not through coaching or friendship after the hire. Referrals also aren’t uniform: people tend to refer others like themselves, including in bad behavior (workers who’ve had accidents refer other workers who go on to have accidents) [3], so referrals from a firm’s own best performers are the most profitable, while referrals from below-median performers can be worse than no referral at all (§7; the paper’s Proposition 1 formalizes why a firm’s lower hiring bar for referred applicants doesn’t erase their quality advantage) [3].

The limits

The turnover and productivity findings are demonstrated across nine firms and three industries, but the deepest test — matched data on who specifically referred whom — exists only for one firm’s trucking subsample of roughly 900 drivers (§3) [3], so the homophily and referrer-quality finding rests on a narrower slice of the data than the headline referred-vs-non-referred comparison.

• When you’re deciding whether to build a referral program at all, versus relying on job postings and recruiters: the paper’s finding is that referred workers aren’t smarter or more skilled on paper (Table 1) [3] — the gain comes from a better job-specific match and lower turnover, so a referral program is a selection tool, not a talent-upgrade tool.
• When referral bonuses flow to everyone equally regardless of who’s making the referral: the data says referrals from your best performers are the most profitable, and referrals from below-median performers can underperform non-referred hires (§7) [3] — an undifferentiated bonus is leaving that signal on the table.
• When a referred hire isn’t working out and you’re tempted to fault the referring employee for “vouching wrong”: the mechanism the paper identifies is a better initial job match, not ongoing coaching (§5) [3] — the referring employee’s job was mostly done at the point of introduction, and a later stumble isn’t evidence they misjudged the person.