Do reference checks and background checks add predictive value?

The honest answer: reference checks add little to your ability to predict who will perform best, and background checks add essentially none in a predictive sense — but both have real value for verification and red-flag screening. Do not confuse the two jobs.

What the evidence says. The most-cited meta-analytic estimate is Aamodt & Williams (2005): an uncorrected validity of about .18, rising to a corrected ~.29, for references/letters of recommendation predicting job performance. Older meta-analytic work (Hunter & Hunter, 1984, carried into Schmidt & Hunter, 1998) put reference checks at about .26. These are low relative to a structured interview (.42) or a work sample (.33) — and they are inflated by the same reliability problem that plagues all reference data: two referees rating the same person agree only weakly, which caps how much a reference can predict anything. Note also that Sackett et al. (2022) did not revise the reference-check estimate, because the original studies lacked the data to recompute it; so even the modest numbers above rest on old evidence, and the ~.29 figure itself traces to an unpublished conference paper rather than a peer-reviewed journal — which is why this note is labelled directional.

Why references predict so weakly. Candidates pick their own referees, referees are reluctant to say anything negative (partly for fear of defamation claims), and the resulting ratings have little range. A glowing reference tells you the candidate can find someone to say nice things — not much more.

What they are genuinely good for.

• Verification. Confirming dates, titles, and that the person did the job they claim. Résumé inflation is rampant — HireRight’s 2017 Employment Screening Benchmark Report found 85% of employers caught applicants misrepresenting information on résumés or applications, up from 66% five years earlier — and a verification call catches it.
• Red-flag screening. A hesitation on “would you rehire?”, a refusal to confirm employment, or a pattern of the same problem across multiple references is a meaningful negative signal — even though references are poor at distinguishing good from great.
• Background checks (criminal, credential, credit where lawful) similarly screen out specific risks and support a defence against negligent-hiring claims; they do not forecast performance.

How to use them well. Treat references as a late-stage confirmation step on a candidate you already intend to hire on the strength of higher-validity methods — not as the thing that ranks your finalists. Ask the same structured, job-relevant questions of each referee, and weight verified facts over opinions.

For an Ontario SMB, the privacy, consent, and human-rights limits on what you may ask referees and what background information you may collect are governed by law — including consent requirements and the limits in the Ontario Human Rights Code — and are covered in the Compliance cluster; do not run criminal or credit checks without confirming you are permitted to. The dollar logic of avoiding a bad hire sits in the numbers cluster.