Blind/Anonymized Hiring Field Experiments (Identity Withheld)

What the evidence says

The strongest causal evidence on blind/anonymized hiring comes from a single quasi-experiment on orchestra auditions, where a physical screen conceals a candidate's identity — including gender — from the evaluating panel (Goldin 2000). Using individual fixed effects to track the same musicians auditioning with and without the screen, the study finds the screen increases the probability a woman advances from the preliminary round by about 11 percentage points, and increases the probability a woman wins the final round by about 33 percentage points among those who reached it. The mechanism the study attributes to these effects is concealment of visual gender cues from the panel during evaluation.

The same data show a second-order effect on the applicant pool itself: adopting blind auditions increased the number of women applying, but the additional applicants drawn in had lower average qualifications than existing applicants. This composition shift explains why an aggregate-level comparison (Table 4 in the source) can show a negative effect for women even though the within-individual estimate (Table 5) is positive — the aggregate result reflects a pool that got larger and, on average, less qualified, not a smaller effect of blindness on any given candidate's chances.

Combining both channels, the source estimates that the switch to blind auditions explains about one-third of the historical increase in the proportion of women among new orchestra hires, with another third attributed to the enlarged applicant pool described above. The source does not state what accounts for the remaining share of the historical increase. Separately, the same authors characterize the overall pattern as evidence that blind auditions "fostered impartiality" in hiring — but this is an inference from the advancement-outcome pattern rather than a direct measurement, and the source's own language ("the weight of the evidence suggests") flags it as an interpretation rather than a directly observed result.

Evidence on blind/anonymized hiring outside live auditions is much thinner and points a different direction. A review of blind resume field experiments — where race is withheld from resumes rather than identity being withheld from a live panel — finds results on minority hiring are "highly mixed," with some studies finding anonymization decreases minority hiring (Quillian 2021, citing Åslund & Skans 2012 and Behaghel et al. 2014).

None of the claims here address identity concealment carried through the full hiring process (beyond the audition or resume-screening stage alone, e.g., through interviews or reference checks), and none extend the orchestra findings to occupations outside classical music or the resume findings to protected characteristics other than race. That is an absence in the current evidence base, not a null result.

Where the evidence splits

Within the orchestra-audition data itself, the direction of the screen's effect is not consistent across stages: it increases advancement at the preliminary round (+11 points) and win probability at the final round (+33 points), but decreases advancement at the semifinal round (by 17–24 points). The source calls this semifinal result anomalous and notes its standard errors are substantial relative to the effect size, but does not offer an explanation for why the effect reverses sign mid-process. The split is unexplained by the source.

Within the blind-resume literature on minority hiring, results across studies are described as highly mixed, and at least part of the disagreement is attributed to a design factor rather than to the anonymization procedure itself: employers who volunteer to participate in these blind-hiring experiments may already have above-average concern about diversity, meaning the studies finding decreased minority hiring may reflect self-selection into the experiment (Bertrand & Duflo 2016) rather than a general effect of resume anonymization.

Thresholds and dose-response

No threshold established in the current sources.

Open empirical questions

• What explains the reversal at the semifinal audition round, where blind screening decreases advancement while it increases advancement at the preliminary and final rounds?
• Do blind/anonymized hiring effects extend to identity concealment carried through the full hiring process (past the audition or resume-screening stage alone, e.g., through interviews), or only to the single stage tested in each study?
• Does the applicant-pool composition effect found in orchestra auditions (larger but lower-average-quality pool) occur in blind resume screening as well?
• What is the effect of blind resume anonymization on minority hiring once employer self-selection into the experiment is accounted for or removed?