Effects of Blind or Anonymized Resume Review on the Racial and Gender Callback Gap

What the evidence says

The claims here split into two bodies of evidence that measure different things: what happens to minority hiring when resumes are anonymized, and what produces the racial callback gap in ordinary (non-blind) correspondence studies. On anonymization itself, the evidence is thin and mixed. Quillian (2021) reports that results on minority hiring under blinded resumes are "highly mixed," citing Åslund & Skans (2012) and Behaghel et al. (2014) as studies that found decreases in minority hiring when resumes were anonymized — a direction of effect held with high confidence but no stated magnitude, so the size of that decrease is not established here. A separate strand of evidence offers a competing account of why those decreases appear: Bertrand & Duflo (2016), as described by Quillian, suggests employers who choose to participate in anonymization experiments are already unusually concerned with diversity, so the employer pool itself, not the anonymization treatment, may be producing the pattern. This is held with medium confidence on direction and low confidence on magnitude, and it is not quantified.

The second body of evidence, from Gallo (2024), does not test anonymization directly — it measures the racial and gender callback gap in standard correspondence studies where resumes carry racialized and gendered names. Black-sounding names receive callback rates 20–30% lower than white-sounding names, while female-sounding names show no reduction in callbacks. Gallo's own pooled estimate of this Black-name callback ratio (0.798) is smaller than a prior meta-analysis's estimate (0.68). This body of work also identifies a candidate mechanism operating through names rather than through the blinding process: perceived warmth and competence, as signaled by a name, positively predict callback rates (pooled r = 0.33, p = 0.03). Because that correlation is measured in studies where names are visible, it describes the mechanism anonymization would remove, not the effect of removing it.

Taken together, the sources establish a baseline racial gap in callbacks driven by name-signaled perception, and separately report inconsistent findings on whether removing those signals through anonymization changes minority hiring outcomes. No claim here quantifies how much of the 20–30% Black-name callback gap anonymization actually closes or preserves — that link is not measured in these sources.

Where the evidence splits

Whether anonymized resume review decreases minority hiring is disputed within the sources. Quillian reports that some studies (Åslund & Skans 2012, Behaghel et al. 2014) find decreases in minority hiring under blinded resumes, while a separate account (Bertrand & Duflo 2016) attributes at least some of those reported decreases to self-selection of diversity-conscious employers into anonymization experiments rather than to blinding itself. The design difference offered by the sources is selection into the experimental sample: employers who agree to anonymize may already differ from the general employer population in ways the studies do not fully control for.

The size of the Black-name callback gap also differs between Gallo's own pooled estimate and a prior meta-analysis. Gallo's sample-derived callback ratio (0.798) did not reach statistical significance (p = 0.07), while the prior, larger meta-analysis found a ratio of 0.68 at p < 0.001. The claim attributes this difference to sample size, describing the earlier study as the "larger meta-analysis," but does not state any other design or population difference between the two.

Thresholds and dose-response

No threshold established in the current sources.

Open empirical questions

• How much of the 20–30% Black-name callback gap measured in standard correspondence studies is closed by anonymized or blind resume review specifically?
• Does removing name-based signals through anonymization reduce the warmth/competence correlation (r = 0.33) that predicts callbacks, or does it operate through some other channel?
• What share of the reported decreases in minority hiring under blind review is attributable to employer self-selection into the experiments, as opposed to the blinding treatment itself?
• Does anonymized review change the callback gap for female-signaled names, given that Gallo's non-blind correspondence data show no reduction for female names?