The role of supervisors in the determination of wages and wage gaps

Abstract

This article considers labour market discrimination by supervisors as a potential contributor to racial and gender wage gaps. Empirical analysis reveals evidence that all workers, except Hispanic males, earn significantly higher hourly wages when working for a supervisor of the same race and sex as themselves. Furthermore, the results suggest that sex has a larger impact on wages than race for workers with white supervisors, while race has a larger impact on wages than sex for workers with minority supervisors. Based on past research, we theorize that the degree of labour discrimination workers face may also be dependent upon the location and size of the firm in which they are employed. However, decomposing the samples by firm location and size suggests that these two factors cannot adequately explain the observed matched supervisor–worker wage effects, which supports the notion that these wage effects are largely driven by factors other than supervisor discrimination.

Keywords:

• labour discrimination
• wage gaps
• supervisor
• race
• gender

JEL Classification:

• J15
• J16
• J31
• J71

Notes

¹ In our sample, 53% of employees are African American or Hispanic in the southern United States compared to 38% in the northeast, 34% in the west and 25% in the north central United States.

² The augmented data set accounts for the sex of the individual with authority over a worker’s supervisor. The authors theorize that even if a worker’s immediate supervisor has discriminatory gender preferences, those preferences may be nullified if their supervisor’s supervisor is not the same sex. For example, a female working under a discriminatory male supervisor may be spared from suffering labour market discrimination if her male supervisor has a female supervisor.

³ A similar strategy is employed in the racial profiling literature (e.g., see Close and Mason, 2006, 2007; Antonovics and Knight, 2009).

⁴ Scores on the AFQT have been the subject of much debate in the literature. Some see scores on the test as a measure of cognitive skills or natural ability (O’Neill, 1990; Farkas and Vicknair, 1996; Neal and Johnson, 1996), while Betts (1995) argues that the test is a measure of school quality, and Cordero-Guzmán (2001) suggests that test scores account for differences in family background characteristics and pre-employment resources and that the test is biased against minorities. We attempted to remedy this problem by following the technique proposed by Rodgers and Spriggs (1996, 2002) and obtaining an adjusted AFQT score; however, the results were not considerably different from those including the unadjusted AFQT score. Hence, we only report the results with the unadjusted AFQT score here.

⁵ Additionally, we considered bias due to selection into the labour force. However, the inverse mills ratio was insignificant in all of our regressions. Hence, we exclude the correction for labour force selection from the final results reported here. Our analysis shows that correcting for selection into different jobs also seems to correct for selection into the labour force.

⁶ An alternative way to deal with possible selection into jobs with different supervisors is to run a regression including fixed effects by individual. We run that specification (not included in this article and available upon request) and the results were similar. We find evidence that workers of the same race and sex as their supervisor earn significantly higher hourly wages than otherwise equal workers of a different race or sex.

⁷ The selection equations presented in Appendix 1 also show that individuals are significantly more likely to work under a supervisor of the same race and sex as themselves. Furthermore, individuals with high cognitive ability, as measured by AFQT scores, are less likely to work for African American or Hispanic supervisors. Individuals with more tenure are more likely to work for white male supervisors. Part-time workers are more likely to have a white or African American female supervisor and less likely to have a white male supervisor. Hence, in addition to controlling for supervisor and labour force selection, these selection equations also seem to assist in controlling for the occupational crowding discussed by Bergmann (1986).

⁸ For example, given the statistical significance of the squared tenure term, workers with a white male supervisor can expect about 8.6% higher hourly wages after working 1 year with a firm, other things equal. After 2 years of tenure, the impact falls to 7.4% and so on.

⁹ As stated previously, the African American and Hispanic supervisor categories are combined into a single minority supervisor category for these purposes because of small sample sizes when disaggregating by firm size and location.

¹⁰ Small firms are defined as those with fewer than 15 workers and large firms are defined as those with 15 or more workers. Current Federal Equal Employment Opportunity legislation does not cover firms with fewer than 15 employees. In other words, businesses with fewer than 15 employees are not covered by discrimination laws. This information is available at http://www.eeoc.gov/employees/coverage_private.cfm

¹¹ Cardoso and Winter-Ebmer (2010) also split their sample by firm size and analyse the impact of having a female supervisor on male and female wages in Portugal. They arbitrarily define small firms as those with 100 or fewer workers and large firms as those with more than 100 workers. Their analysis suggests that female workers profit more from a female supervisor in small firms than large firms.

¹² One caveat concerning the results in Table 3 which we must mention is that some of the regression results are obtained from a small number of observations, which may impact the preciseness of the regression estimates. In general, this is not a concern for workers with white male or white female supervisors but more problematic for workers with minority male and minority female supervisors, particularly among Hispanic workers. For example, as shown in Appendix 2, there are only eight Hispanic females who work for a minority male supervisor in a small southern firm. Furthermore, there are only five Hispanic males who work for a minority female supervisor in a small southern firm and two Hispanic males who work for a minority female supervisor in a small non-southern firm. In these few instances, the results should be interpreted with caution.