Blind Hiring Statistics: What the Research Actually Shows (2026)

Blind hiring is one of the most-cited and least-checked topics in recruitment. The same statistics circulate for years, sources unnamed, caveats dropped. This page collects the key numbers with their actual sources — including the famous ones that don't hold up as well as the slide decks claim. Cite accordingly.

The callback studies — the strongest evidence

The most robust research design in this field: send identical CVs, vary only the name, count the interview invitations.

• ~50% more callbacks for white-sounding names. Bertrand & Mullainathan's landmark field experiment sent nearly 5,000 CVs to real job ads in Boston and Chicago; identical CVs with white-sounding names (Emily, Greg) received about 50% more callbacks than those with African-American-sounding names (Lakisha, Jamal). (NBER Working Paper 9873, 2003/2004)
• Higher-quality CVs didn't close the gap. In the same study, improving credentials increased callbacks for white-named CVs substantially more than for African-American-named ones — the penalty compounds, not shrinks, with quality.
• The pattern replicates internationally. Correspondence studies in the UK and Europe repeatedly find ethnic-minority applicants must send meaningfully more applications to receive the same number of callbacks as majority-group applicants with identical CVs (e.g. the Oxford/Nuffield GEMM study reported ~60% more applications needed in the UK).

Why this matters for blind hiring: the name alone moves interview odds by tens of percent on otherwise identical CVs. Removing it during screening removes that specific effect — this is the part of blind hiring the evidence supports most strongly.

The orchestra study — famous, and contested

• The claim: Goldin & Rouse analyzed 14,000+ auditions at major US orchestras (1970–1996) and reported that screened ("blind") auditions increased a woman's probability of advancing from preliminary rounds by ~50%, explaining 25–46% of the rise in female orchestra membership. (American Economic Review, 2000; Harvard GAP summary)
• The caveat: later statistical reviews (notably Gelman, 2019) showed the headline results rest on small samples and mixed-sign coefficients — several key estimates are not statistically significant. The study is suggestive, not the slam-dunk it's quoted as.

Practical takeaway: if you cite one number in a client deck, cite the callback studies, not the orchestra.

The government trial that surprised everyone

• Blind screening reduced female shortlisting in one large trial. The Australian government's BETA unit ran a randomized trial ("Going blind to see more clearly", 2017) with over 2,000 public servants: de-identifying CVs slightly decreased the rate at which women and minority candidates were shortlisted — because reviewers had been actively favoring them when identity was visible. Blinding removed that positive weighting too.

Why this matters: blind screening removes identity effects in both directions. If your team currently applies affirmative consideration at screening, blinding will neutralize it — decide which policy you actually want before you deploy.

Adoption and context numbers

• ~44% of organizations use AI in resume screening (SHRM research, widely cited 2025–2026) — which interacts with blind hiring: an AI screener sees whatever the CV contains, including the identity signals humans were blinded to, and whatever candidates hid in the file.
• Application volumes in the hundreds per posting (Employ Inc. data reports ~250+ applications per job in 2026) — manual, careful, bias-aware review of every CV is arithmetically impossible; process design is where fairness now lives.
• Public-sector and RPO tenders increasingly mandate anonymized submission — for agencies, blind-ready CV formatting is becoming table stakes rather than a differentiator.

How to use these numbers honestly

• Cite Bertrand & Mullainathan for why names bias screening — it's the strongest result in the field.
• Mention the orchestra study only with its caveat; savvy clients know the critique.
• Use the BETA trial to set expectations: blind hiring standardizes decisions; it doesn't guarantee any particular demographic outcome.
• Measure your own pipeline before/after — your data beats any citation. Implementation steps in our blind hiring guide.