Academics examined correlation scatter plots and made judgments of group bias. Half of the subjects judged bias in salaries from graphs that showed salaries plotted against merit with a separate correlation ellipse for each of two groups. The others judged test bias from graphs that showed job performance vs test scores. Relative positions of the centroids for the two groups and the within-group correlation were systematically varied. Judgments of bias were not consistent with either "forward" or "reverse" regression definitions of bias. Judged bias is not a monotonic function of the group difference in salary between persons of equal merit nor is it a monotonic function of the group difference in merit between persons of equal salary. Instead, judgments in both tasks were consistent with a difference of differences model; for example, when the group difference in standard deviation units is equal on both salary and merit, judges say the situation is unbiased. The results are consistent with a one-mediator model of group equity, which assumes that both measures are imperfectly and equally correlated with the mediator.
|Original language||English (US)|
|Number of pages||13|
|Journal||Organizational Behavior and Human Decision Processes|
|State||Published - Apr 1986|
ASJC Scopus subject areas
- Applied Psychology
- Organizational Behavior and Human Resource Management