Employer Diversity-Consciousness Selection Effects in Blind-Hiring Field Experiments

What the evidence says

The evidence on this topic comes from a single tier-A randomized experiment in veterinary medicine (Begeny 2020), which tested how managers evaluate identical employee profiles that differ only in the name signaling gender. The controlling variable is not diversity training or hiring procedure but a belief: whether the manager thinks gender discrimination is "no longer an issue" in their profession. Among the managers who hold that belief, assigning a male name to an otherwise identical profile increases competence evaluations by 0.48 points on a 7-point scale and increases recommended salary by £2564.23 (about $3475, an 8% pay gap), relative to the female-named version of the same profile. Among managers who reject that belief, the same manipulation produces no measured effect on competence evaluations (θ = −0.13, not significant).

The mechanism the study establishes is a moderated-mediation chain: competence evaluations strongly predict how managers say they would treat an employee overall (B = 0.77), and the competence evaluations themselves are shifted by the interaction between the male/female name condition and the manager's belief that discrimination is solved (B = 0.22). That biased evaluation carries through into treatment recommendations — including supervisory opportunities and promotion advice — via an indirect effect of 0.36. In other words, the belief that a bias problem no longer exists does not eliminate the bias; in this sample it is the condition under which the bias appears in evaluation and then propagates into recommendations for advancement and pay.

This belief is not rare among managers: 44.5% of the sampled veterinary-medicine managers scored above the midpoint of a 1–7 scale indicating they believed gender discrimination was no longer an issue in their profession, and men were more likely to hold this belief (M = 4.75) than women (M = 3.56).

The claims here say nothing about blind or anonymized hiring procedures themselves — the experiment manipulates whether a name signaling gender is visible to a named-profile evaluator, not whether identifying information is withheld across a full hiring process. There is also no claim measuring actual hiring decisions, only performance-review-style competence and salary/treatment recommendations, and no claim covering any profession other than veterinary medicine or any country other than the one associated with the £ figures.

Where the evidence splits

No contradictions in the current sources. The apparent split between managers who show bias (0.48-point competence gap, £2564.23 salary gap) and managers who show none (θ = −0.13, ns) is not an unexplained disagreement between sources — it is the same study's own stated moderator: the effect appears only among managers who believe gender discrimination is no longer an issue, and is absent among managers who reject that belief.

Thresholds and dose-response

The one source on this page defines its moderating belief by a scale-midpoint cutoff: managers "scoring above the scale's midpoint (neither agree nor disagree)" on the 1-7 gender-discrimination-belief item are classified as believing discrimination is no longer an issue. At and above that threshold, the biased evaluation and pay-recommendation effects appear (0.48-point competence gap, £2564.23 salary gap); below it, no significant competence-evaluation difference is measured (θ = −0.13, ns).

Open empirical questions

• Does the belief-moderated bias in evaluation, pay, and treatment recommendations persist when measured in actual hiring decisions rather than stated recommendations?
• Does the effect, or its size, replicate in professions other than veterinary medicine, or outside the sampled country?
• Were managers aware their salary and treatment recommendations would be implemented, and does that awareness change the size of the effect?
• Does the 44.5% prevalence of the "discrimination is no longer an issue" belief, and the gender gap in holding it, replicate in other manager populations?