“Employee referrals make a candidate several times more likely to be hired.”

The peer-reviewed literature on referrals is real and reasonably strong. The Burks and colleagues study uses personnel data from nine large firms across three industries, which is an unusually good dataset for a question like this.

What that study measures is post-hire: productivity, turnover, wages and profit per worker. Referred workers were similar on most productivity measures, substantially less likely to quit, and more profitable overall.

The widely quoted multipliers describing how much likelier a referred applicant is to be hired come from recruiting software vendors reporting their own funnel data, not from this literature.

Those funnel ratios are real numbers about real applications, but they are ratios between two very different populations. Referred applicants are pre-filtered by an employee who knows both the person and the role, so the comparison is not between two otherwise identical candidates.

Vendor funnel reports and the academic literature answer different questions and are routinely quoted as though they answered the same one. We treat the disagreement as a category error rather than as a conflict to be resolved into a single number.

Our reading is that this is good advice attached to a bad number. Seeking a referral is worth doing; the multiplier quoted to justify it describes a selection effect as if it were a causal one.

The QJE finding is more interesting than the version in circulation, because it explains why employers push referral schemes: retention and profit per worker, not better performance.