|By The Bureau of Labor Statistics
June 18, 2003
Although personal choices, occupational crowding, and discrimination, contribute to the gender gap, the higher share of women in an occupation is still the largest contributor.
According to the Bureau of Labor Statistics, women earned approximately 77 percent as much as men did in 1999.1 Although the existence of the gender pay gap is well documented, the factors that contribute to it are still debated. One such factor is the difference in the proportion of jobs held by women and men. However, understanding how occupational differences contribute to the gender pay gap is made more complicated by the fact that both men and women in predominately female occupations earn less than men and women in male dominated occupations.
An important question that has not been resolved is, why do predominately female occupations tend to pay less? A number of possibilities including worker characteristics, job characteristics, occupational crowding, devaluing by society of women, and discrimination have been posited.2
This article sheds some light on reasons for the gender earnings gap, focusing on the role that the share of women in an occupation plays. We utilize the methodology employed by George Johnson and Gary Solon to identify the sources of the relationship between wages and the share of women in an occupation.3 Johnson and Solon used Current Population Survey (CPS) data to estimate the relationship between wages and the concentration of females within occupations. They found that the relationship was negative, even after controlling for worker and job characteristics. Industry was found to have the largest effect on the relationship, primarily because predominately male industries, such as construction and manufacturing, pay higher wages.4
The data used in this article are from the 1989, 1992, and 1999 Outgoing Rotation Group Files of the CPS. The CPS has the advantage of including detailed occupational data on a large national sample of more than 50,000 households in addition to the more commonly found earnings, employment, human capital, and demographic information.5
One of the primary drawbacks of using the CPS for estimating the sources of the gender gap is that it does not include a measure for actual work experience. Because higher earnings are associated with more experience, and women tend to have less experience than men due to breaks in their labor force participation for childrearing and other reasons, this omission results in inaccurate estimates of the contribution of each characteristic to the wage gap.6
To approximate actual experience, researchers construct potential experience.7 This method likely overestimates the experience of women, as they are more likely than men to take time out of the labor force.8 However, it is well known that delayed marriages and lower fertility rates have contributed to a rise in women’s labor force participation, particularly among younger cohorts of women. Because of this pattern, the use of potential experience probably provides a better estimate of actual experience than it did in the past.
Summary and Conclusions
This article revisits the work of Johnson and Solon and others that estimate the portion of the gender gap that can be explained by the share of women in an occupation. Using microdata from the Outgoing Rotation Group files of the Current Population Survey, we first show that the share of women in an occupation is still one of the largest contributors to the gender wage gap. This occurs because women have a higher likelihood of working in female dominated jobs, which typically have below average wages. In fact, in 1999, our most recent data, the share of women is the largest contributor to the gender pay gap.
Without controlling for worker characteristics, we then estimate the size of the negative effect that employment in predominately female occupations has on the wages of men and women. We find that the relationship has weakened for both men and women since 1989. In 1999, a woman working in a predominately female occupation earned 25.9 percent less than a woman working in a predominately male occupation . The comparable figure for a man working in a predominately female occupation is 12.5 percent less than a man working in a predominately male occupation. One interpretation of these relationships is that they are the adverse consequences of occupational “crowding.” Limited occupational choice creates crowding in particular occupations, putting downward pressure on the wages of men and women in these occupations. The limited occupational choices could be due to discrimination. However, others might argue that the crowding reflects the personal choices of some men and many women who want or need a job that fits their family obligations. The wage and percent female relationship reflects the presence of a compensating differential.
If true, then the estimates need to be adjusted for not only labor supply factors that proxy for personal preferences, but also for labor demand and institutional factors (for example, government employment and unions) that are correlated with both wages and the share of women in an occupation. To remove the potential biases and identify their roles in explaining the wage-percent female relationship, we utilize the decomposition in Johnson and Solon that identifies the observable sources of the relationship between wages and the share of women in an occupation.15 In practice, this amounts to decomposing the difference between the relationship between wages and percentage of women that excludes worker characteristics and the relationship between wages and percentage of women that includes worker characteristics.
We find that, for men, the most important measurable factor in each year is the industry of employment as opposed to personal characteristics such as education, age, and region of residence. It is well known that particular industries pay more than others, and our results show that these industries, on average, have higher concentrations of men. The opposite occurs for women. Education and age are the most important factors for explaining the wage- and percent-female relationship. If education and age capture personal preferences that influence occupational choice, then solely focusing on expanding occupational choice will result in a small narrowing of the gender wage gap.16 This is because, as shown in our Oaxaca-Blinder decompositions, even after we add the concentration of women in an occupation to the model, the overall gender wage gap is still largely unexplained.
Download full article in PDF (49K)