Statistical Discrimination, Stereotyping, and Evaluations of Worker Productivity
*Part of TESS 2004 Telephone Survey
Sample size: 1017
Field period: 11/11/2004- 01/27/2005
This project seeks to test the hypothesis that evaluations of worker productivity are biased by cultural stereotypes regarding race and gender.
Theories of statistical discrimination argue that employer preferences for white male employees are an unfortunate but rational response to “noisy” labor markets in which information about worker productivity is costly (Bielby and Baron 1986). However, these theories are predicated on the assumption that individuals make unbiased judgments of productivity (Arrow 1973). Research from social psychology indicates that unbiased judgments of female and minority competence and performance are rare. Few empirical investigations of statistical discrimination theories exist (Oettinger 1996), and those that do rely on cross-sectional survey data.
Use of an experimental design improves on existing research in this area by avoiding selection bias, and also allows us to make stronger claims about causality than would be possible using cross-sectional survey data (Aronson, Ellsworth, and Carlsmith 1989).
H1: Women will receive lower evaluations of productivity than men.
H2: African Americans will receive lower evaluations of productivity than whites.
H3: The effects of race and gender will be mediated by ratings of status.
Apparent race (African American or white) and gender (male or female) of the target of evaluation in a vignette describing work behavior.
Measures of perceived productivity, value to the company, work effort, competence, deservedness of a raise, and measures of status (extent to which the target is perceived to be respected, honorable, prestigious, intelligent, knowledgeable, and capable).
8. Additional Information: (any additional information that is essential for readers to understand how the study was conducted; please limit to 1000 characters including spaces).
Preliminary findings suggest a positive effect of race and gender, with African Americans and women being rated higher than white men, and a negative interaction of race and gender, such that African American women are rated lower than white women or African American men. However, the findings vary widely across demographic categories. Perhaps most importantly, these findings largely disappear when the data set is restricted to individuals who are currently employed in the paid labor market. Further analysis is required before making conclusions from the data. An important question for future research is how employers, rather than all employed workers, might evaluate the vignettes.
Further analysis is required before drawing conclusions from the data.