Dual-anonymization Yields Promising Results for Reducing Gender Bias: A Naturalistic Field Experiment of Applications for Hubble Space Telescope Time

Anonymizing personal information, such as gender, about applicants reduced gender bias, particularly in male reviewers, who tended to rate female applicants significantly worse than
male applicants.

• When applicant information was not blinded, female applicants were significantly less likely to be selected for funding and HST access than male applicants.
• Proposals by female applicants were significantly more likely to be accepted under the double-blind review process than when applicant information was not blinded.
• Adopting the double-blind review process successfully eliminated gender bias exhibited by male reviewers toward female applicants.
  • When applicant information was not blinded, gender bias emerged when male reviewers rated female applicants significantly worse than they rated male applicants. However, this discrepancy was no longer the case under the double-blind review process.

Researchers argue that these results have important practical implications for gender equity in all areas of science and academia. Double-blind review processes could be adopted not just for funding and resource applications, but also for conference presentations, publications, and employment.

Using archival data from the Hubble Space Telescope Time Allocation Committee (HST TAC), researchers examined 16 cycles of 15,545 applicants seeking funding and access to the HST. In an attempt to reduce bias against women, the TAC had transitioned from a single-anonymous review system (in which the reviewers were kept anonymous) to a dual-anonymous review system (in which both the reviewers and the applicants were kept anonymous). Different versions of the dual-anonymous system had been implemented previously, but the commonality was that the name and identity of the Principle Investigator (PI) of the grant had been either obscured or removed. The researchers examined the success rate of male and female applicants under the two systems to examine the impact of bias in each version.

The archival data allowed researchers to compare each reviewer’s ratings across the applications they reviewed, therefore
allowing them to control for variation in the reviewers’ ratings (that is, some reviewers tending to give higher scores than others). The strengths of this study were its longitudinal design, its large sample size, and the nature of the quasi-field experimental data being obtained from a national agency.