Skip to main content

Second Quarter 2021, 
Vol. 103, No. 2
Posted 2021-04-15

Understanding the Gender Earnings Gap: Hours Worked, Occupational Sorting, and Labor Market Experience

by Maria E. Canon, Limor Golan, and Cody A. Smith

Abstract

This article documents life-cycle gender differences in labor market outcomes using longitudinal data of a cohort of individuals from the National Longitudinal Survey of Youth 1979. As in other datasets, the gender earnings gap increases with age. We find that hours worked and labor market experience are the most substantial observable variables in explaining the gender pay gap. We also focus on patterns in occupational changes over the life cycle, as a large part of pay growth occurs when workers change jobs. We find that college-educated men, on average, move into occupations with higher task complexity. We further show that women are less likely to change occupations. Moreover, on average, pay grows when workers change occupations, but the growth is smaller for women. Finally, we discuss theories that are consistent with the patterns we document.


Maria Canon is a senior lecturer at Washington University in St. Louis. Limor Golan is a professor at Washington University in St. Louis and a research fellow at the Federal Reserve Bank of St. Louis. Cody A. Smith is a PhD candidate at Washington University in St. Louis.



INTRODUCTION

The labor force participation of women has increased substantially since the 1960s. At the same time, the gender earnings gap has declined from about 40 percent in the late 1960s to less than 28 percent in the early 1990s and has stopped converging since. Much of the gender earnings gap is explained by gender differences in labor force attachment and accumulated labor market experience. In particular, the gender earnings gap increases with age as the experience gap increases. One explanation for the remaining gender earnings gap is that many jobs disproportionately reward working long hours. In many jobs, the pay is nonlinear in hours worked and penalizes workers who choose to work fewer hours. This reward structure tends to affect women of childbearing years disproportionately and also affects their occupational choices. Furthermore, in most occupations, the representation of women at the top-paying jobs is low—even if at lower levels there are many women.

This article documents the evolution of the gender earnings gap over the life cycle using data from a cohort of men and women from the National Longitudinal Survey of Youth 1979 (NLSY79). The pattern is similar to the one documented in other datasets: The gender pay gap increases with age. To understand the factors affecting this pattern, we explore the role of occupations, hours worked, and work experience accumulated with age in the observed gender earnings. While previous literature analyzed occupational and labor force participation patterns, we document not only those but also task assignments and changes in task assignments over the life cycle. Previous literature did not analyze the evolution of task assignments and their changes over the life cycle in the context of the gender pay gap. Changes in occupational tasks and occupations are typically important to understanding the wage growth of workers over the life cycle. Our dataset allows us to account for both detailed labor supply history and heterogeneity in test scores and education.

Over time, as the earnings gap increases, the gap in weekly hours worked grows as well. This gap in hours is one of the factors that contributes to the increase in the earnings gap. We then study how occupations change over the life cycle. In particular, we explore whether the increase in the earnings gap is due to changes in occupational assignments and the evolution of the type of tasks workers are assigned. We also examine whether the occupational gap increases with age.

We define occupations along a unidimensional axis, measuring each occupation's demand for complex cognitive tasks. We focus on changes in nonroutine task complexity because over the life cycle workers typically transition into occupations with more cognitive task complexity and fewer motor skills requirements, which cause wages to increase (see Yamaguchi, 2012, for task analysis for men). We follow Antonovics and Golan (2012), who show that this transition is an important source of the increase in wage dispersion among white men, and Golan, Sanders, and James (2019) and Golan and Sanders (2019), who show that the transition is also an important source of the increase in the racial pay gap over the life cycle for men. We document that occupation matches and their pattern over the life cycle vary by education. As expected, workers with a college degree are matched with occupations of higher complexity than workers without a college degree.

We find that among workers with a college degree, men and women start in similar occupations. Over the life cycle, men surpass women in terms of the complexity of tasks performed. Among workers without a college degree, women start in higher-ranked occupations than men. Over the life cycle, these women remain ahead of these men. Thus, unlike the racial gaps and the increasing wage gaps for men, occupational task complexity may not explain much of the earnings gap for workers without a college degree.

The gender gaps above can be partly driven by differences in the observable and unobservable characteristics of men and women. We now analyze the role of differences between men and women based on observable characteristics in our data. A substantial gap in earnings remains after controlling for labor market experience, hours, Armed Forces Qualification Test (AFQT) scores, education, and occupation. We quantify the contribution of the different factors to the pay gap using the Blinder-Oaxca decomposition. Whereas a small difference in the earnings gap is explained by the compositional effects of college-educated men and women, the differences in hours and labor market experience account for the majority of the gender earnings gap for college- and noncollege-educated workers. Moreover, we find that in our sample, the increase in the earnings gap with age is associated with the increase in the labor market experience gap and the breaks in labor force participation for both college- and noncollege-educated workers.6 We discuss our findings in light of different explanations in the literature. The gaps in hours worked and as a result of experience accumulated may be a result of differences in preferences and roles that women play in caring for children. However, discrimination in the labor market and lack of opportunity and promotions may also lead to these choices. Gayle and Golan (2011) find evidence that while there are preference differences, discrimination plays an important role in the choices of hours worked and experience accumulated.

For college-educated women, we find that task complexity does not increase on average as much as it does for college-educated men (after the initial entry years).7 Job changes are an important factor contributing to wage growth over the life cycle. On average, workers' wages increase when they change jobs, regardless of whether they move to perform more- or less-­complex tasks (see Antonovics and Golan, 2012). We first document that women change occupations less often than men do. We next document that wage growth is lower for women when they change occupations. We then discuss theories that are consistent with the patterns we document. In particular, we discuss human capital and learning and sorting models. One theory is that of learning by doing (e.g., Jovanovic and Nyarko, 1997): Workers learn how to perform tasks and accordingly move up the job ladder, causing wages to increase over time. A second theory is search frictions in the labor market: Workers may not always find the jobs that best suit their qualifications initially, but over time they search and change jobs when they find better matches for their skill set. Lastly, workers do not always know which type of job matches their skills best and they learn about their own skills with experience (e.g., Jovanovic, 1979; Miller, 1984; and Antonovics and Golan, 2012, among others). Working fewer hours can reduce the amount of both learning by doing and learning about one's skills and therefore slow down the sorting of women into jobs that better suit their skill sets (e.g., Taber and Venjlin, 2016, and Lise and Postel-Vinay, 2020). However, discrimination may also imply that women are less likely to receive attractive offers and, therefore, are less likely to switch jobs. Family considerations may also affect women's likelihood of changing jobs if it requires moving.


Read the full article: https://files.stlouisfed.org/f...