M. Rotundo et Pr. Sackett, Effect of rater race on conclusions regarding differential prediction in cognitive ability tests, J APPL PSYC, 84(5), 1999, pp. 815-822
The possibility of predictive bias by race in employment tests is commonly
examined by across-group comparisons of the slopes and intercepts of regres
sion lines using test scores to predict performance measures. This research
assumed that the criteria, primarily supervisory ratings, were unbiased. H
owever, a concern is that the apparent lack of differential prediction in c
ognitive ability tests may be an artifact of the predominant use of perform
ance ratings provided by supervisors who are members of the majority group;
a criterion that is potentially biased against members of the minority gro
up. We posited that ratings by a supervisor of the same race as the employe
e being rated would be less open to claims of bias. We compared ability-per
formance relationships in samples of Black and White employees that allowed
for between-subjects and within-subjects comparisons under 2 conditions: w
hen all employees were rated by a White supervisor and when each employee w
as rated by a supervisor of the same race. Neither analysis found evidence
of predictive bias against Black employees.