Measuring social progress requires the right metric. The broadening and narrowing of wage gaps, that women and Black men earn 87 cents to every dollar a white man earns, for example, are often cited in discussions about the persistence of employer discrimination in the workforce. Wage gaps are calculations currently based only on average wages, and it’s a metric University of Toronto Researcher Rahul Deb and his co-authors have found makes discrimination too easy to dismiss. Instead, Deb et al propose, calculations should be based on a model that considers the entire wage distribution – instead of just the average or mean — to better isolate the effects of discrimination on compensation of diverse groups of people.
“When people talk about wage discrimination, especially in the popular press, they only refer to average wages that are reflected in wage gaps,” explained Deb, a Professor with the Department of Economics. “Any economist will tell you that there are many reasons why average wages could differ and one of those reasons, of course, is discrimination. In our study, we compare male white and male Black workers in the United States. We examine demographic buckets made up of people who are either white or Black, but who otherwise look the same in terms of the demographic information we have for them.”
Based on averages, Black men make less than white men, even if they are demographically identical in other ways. The crux of the investigation Deb and his team undertook was to examine whether the numbers could be interpreted as reliable evidence of discrimination.
“It’s possible that, even when studying data about two different racial groups of workers who have the same demographic profiles, that their wages are higher or lower in consideration of their productivity,” Deb said. “In a world where we have a lot of wage inequality, it’s easier to dismiss the influence of discrimination if we think that higher productivity workers are paid proportionally more relative to lower productivity workers.”
According to Deb, the average wage data could be interpreted as meaning that white workers are more productive and that leads to higher average wages. But to dismiss discrimination as a factor, would require some groups to have unnaturally high productivity.
“It is possible that among the two groups, the likelihood that the share of white workers that have unnaturally high productivity could be bigger than the share of Black workers who have unnaturally high productivity, even if, on average they’re the same,” said Deb. “When you average, you’re averaging across the entire population distribution in terms of productivity. When wage gaps are calculated through averages, you’re capturing people who have extremely high and extremely low productivity. If we just reward white CEOs with high salaries, that show up in terms of differences in wages between white and Black workers.”
In their test of the wage discrimination model using US Census Data, Deb and his colleagues compared the wage distribution of white men with high school education with that of Black men with at least some college. Their test reasonably assumes that more educated Black workers have higher average productivity than their white counterparts. They show that the less educated white men make higher wages at all quantiles of the wage distribution.
“If you look at wage distribution across different racial groups that are otherwise demographically similar in terms of age and education, based on their productivity, then you get a completely different set of results,” Deb said.
Deb’s model of wage distribution shows that, without discrimination, it is not possible for white workers with lower average productivity to make higher wages across all quantiles of the wage distribution. In other words, the results show that employers discriminate against Black workers.
The results of the study have several implications, including the potential development of a tool based on the model that would allow human resource leadership to enter in worker data and plot wage distribution among workers of similar education and experience doing the same job in the same place.
“If you’re an HR professional or a journalist, it’s been very easy to just look at average wages and say there’s a wage gap,” Deb explained. “Our model is equally easy to apply because we’re saying, don’t average it, plot the cumulative distributions of wages, which show the fraction of people who make less than a certain wage. It’s very easy to test in practice. All you need to do is just plot the wages and look at it the results on a rough graph to see if one line lies above the other.”
The main benefit of the wage distribution model enables employers to use more of the data they have captured about their own employees and to compare it to large, national datasets, like census data. The results of better statistics could lead to better monitoring of social progress and improve workplace culture.
“In order to fix discrimination, you need to know the source of that discrimination,” Deb said.
The working paper, Statistical Discrimination and the Distribution of Wages, by Prashant Bharadwaj Rahul Deb Ludovic Renou was published by the National Bureau of Economic Research in June 2024.
Return to the Department of Economics website.
Scroll more news.