Quantifying Racial Disparities Using Consecutive Employment Spells

This paper develops a framework to quantify racial disparities in earnings and employment that are not plausibly due to differences in productivity. Over an employment cycle, employers learn about worker productivity and workers move to more productive and less prejudiced employers. I use implications of this behavior to match high-tenure Black and white workers on unobservables. I look at matched pairs who lose their jobs in a mass layoff. Gaps in earnings and separations between these workers in their next jobs are not plausibly due to differences in productivity. Using U.S. data, earnings differences between these matched workers are five log points, about a quarter of the racial earnings gap among high-tenure workers. Similarly, matched Black workers are more likely to separate.