Abstract

White workers in the United States earn almost 30 percent more per hour on average than Black workers, and this wage gap is associated with large racial differences in occupational assignments. In this article, we theoretically and empirically examine the Black-White disparity in occupations. First, we present a model based on Antonovics and Golan (2012) that relates occupational assignments to the incentives workers face while learning about their own unknown ability. Second, we document differences between Black and White workers in both the complexity of skills required in their initial occupations and the growth rates of this complexity over time. To do this, we match panel data from the National Longitudinal Survey of Youth 1979 with the Dictionary of Occupational Titles measures of occupational characteristics and find that, compared with White workers, Black workers start in occupations requiring less-complex skills, see slower growth in job complexity over time, and are relatively more likely to transition to jobs with lower complexity. Finally, we consider the relationship between our model and our empirical findings; for example, discrimination in hiring early in the career can have long-term consequences on the ability of Black workers to learn their best occupational match and explains part of their lower wage growth. We conclude with suggestions for policy and future research directions.

Limor Golan is an associate professor at Washington University in St. Louis and a research fellow at the Federal Reserve Bank of St. Louis. Carl Sanders is an assistant professor at Washington University in St. Louis.

INTRODUCTION

The labor market experiences of Black and White workers in the United States are dramati­cally different. A first-order difference is the well-documented racial wage gap: The average hourly wage for White workers is 30 percent higher than for Black workers. This racial wage gap has been shown to reflect differences in both socioeconomic backgrounds and discriminatory practices in the labor market and has created a sizable literature across multiple disciplines. But differences in earnings do not exhaust the racial differences in labor market experiences. In this article, we consider racial gaps in workers' occupations, that is, differences in what White and Black workers do rather than what they earn. Even the most basic descriptive statistics show large differences between Black and White workers in the types of work they perform. For example, Black workers were 12 percent of the working population in 2016 and made up 26 percent of the occupation "truck and tractor operators," while making up only 3 percent of "chief executives." White workers showed the opposite pattern, making up 45 percent of the working population in 2016 and 85 percent of chief executives.

Our focus on racial differences in occupations is consistent with recent work in empirical labor economics, which links differences in occupations and occupational mobility to workers' wage growth. It is often more informative to know someone's occupational title than their current wage: taking two young workers who each earn $15 an hour, the knowledge that one has the occupation "accountant" and the other "refrigeration mechanic" helps to make the (on average correct) prediction that the accountant will make significantly more than the mechanic 10 years later. Understanding the reasons Black and White workers take different occupations can provide insight into differences in wage levels and wage growth.

The primary contribution of this article is documenting and interpreting the differences in the relative occupational assignments of Black and White workers. In the first part of the study, we present an economic model to use as a framework for interpreting our empirical results on occupational choice, occupational turnover, and wage growth. In the second section, we document the racial gaps in occupational choice in representative U.S. data, looking both at aggregate trends in occupational choice within a worker's career and at occupation-to- occupation transition rates. Finally, in our empirical results, we compare the predictions of the model to our descriptive findings and discuss the implications of the underlying economic mechanisms for policy and future research.

Section 2 of our article gives a framework to interpret occupational mobility across races as the result of different economic circumstances. Our model is a learning model following Antonovics and Golan (2012) that is capable of generating occupational mobility and wage growth across the career. For simplicity, we present analysis of a two-period model. Each period, workers choose an occupation to maximize the expected present discounted value of lifetime income, but occupational choice is complicated by the fact that workers do not know their own ability and thereby their best occupational match. Working allows a worker to learn about his ability over time, but different jobs give information about the worker's skills at different rates.

The amount that workers learn about their skills depends on the intensity of their job. Different jobs require performance of tasks that require varying intensities of unobservable skills. The more output depends on the unobserved skills, the more information the job reveals about those skills. For example, workers learn more about their ability as a manager in management jobs. Information about skill levels may increase future earnings because it allows a better assignment of workers to jobs. Thus, workers experiment, forgoing expected current-period output in order to learn about their skills by taking jobs they would not take otherwise. Antonovics and Golan's (2012) results show that the optimal level of experimentation is initially small, increases as workers gain experience, and then declines as workers become increasingly certain about their skills.

Mapping the theoretical concept of "occupational intensity" to the data requires non- standard measures, since in Census-like data sets only the occupational title is recorded rather than any specific on-the-job activities. To overcome this data limitation, we use data on occupational characteristics from the Dictionary of Occupational Titles (DOT) merged with worker-­level panel data from the National Longitudinal Survey of Youth 1979 (NLSY79). We use the detailed occupation-level characteristics from the DOT to reduce the unordered list of occupational titles into a single-dimensional index that ranks occupations by the degree to which output depends on skills that are difficult to observe directly, e.g., creativity. This index ranks occupations with respect to the dependence of output on skills that are hard to observe, which we call "complexity" throughout for brevity.

In Section 3, we analyze the merged data sets and show that, as expected, the average White worker's first occupation tends to be more-complex than the average Black worker's first occupation. Moreover, over a career, the average White worker's occupational complexity grows faster than the average Black worker's. Further empirical analysis considers whether these differences are driven by rates of occupational switching, differences in the promotion rates of less- versus more-complex occupations, and the role of demographic characteristics and education in explaining these occupational gaps.

When we look more closely at the rates of occupational switching, we find that the slower pattern of occupational upgrading by Black workers relative to White workers is not driven by lower occupational mobility. Rather, Blacks are marginally more likely to switch occupations than Whites, but a greater proportion of Black occupational transitions are "downgrades," that is, switches toward occupations characterized by lower levels of complexity. Given an occupational switch, White workers make an occupational upgrade 54 percent of the time, compared with 51 percent for Black workers.

Regarding potential average demographic differences between races, we examine to what extent differences in workers' first occupations and the growth rate of occupational complexity are driven by race alone rather than other explanations. For example, Black workers may begin in less-complex occupations on average more often than White workers due simply to average differences in education levels. We find that racial differences in the speed of occupational upgrading persist even if we consider White and Black workers who are originally in the same occupation. This finding suggests that race-specific factors such as discrimination might partially explain differences in occupational transition rates.

Additionally, we consider the first job that workers have and find that controlling for measurable demographics such as education level and test scores does not eliminate the effect of race on the initial occupational assignment: Black workers with seemingly very similar skills as White workers tend to work in less-complex occupations. Our economic model suggests that if Black workers are discriminated against in hiring for high-complexity occupations, it can have long-term effects on their occupational complexity and wage growth relative to White workers.

From a policy perspective, one of the major issues surrounding racial labor market gaps is the question of mismatch. Workers may not always be well matched to their job or occupation, and if demographic background differences and racial discrimination make this problem more severe, there may be large productivity gains by improving the match of Black workers with their best occupations. In this article, we focus on the role of mismatch induced by informational frictions in the Black-White wage gap. Even if there were no barriers to hiring or finding jobs, workers and employers do not always have complete information about the worker's skills or the suitability of those skills to the tasks required by the job. Over time, as a worker comes to know his or her own skills and the employer observe the worker's performance, both the worker and the employer both could learn about the worker's ability and comparative advantage. Based on this information, workers change jobs and, over time, wages grow because workers work in occupations and jobs in which they are better matched. Policies that aim to improve the information available to both workers and employers could potentially reduce the costs of mismatch and, and if discrimination in occupational attainment leads to Black workers receiving less information about their skills over their careers, it may be necessary to target these policies to Black workers.

Read the full article.