Gender segregation – the tendency of men and women to work in different kinds of jobs – is an enduring problem in the United States. Because jobs dominated by women tend to be paid less than those dominated by men, segregation contributes to gender inequality. Despite progress over time, the rate of desegregation has slowed in recent decades, and segregation remains a major contributor to the gender pay gap.
However, gender segregation may be worse than we thought. It turns out that the way we typically measure segregation – using occupations – conceals gender segregation based on job titles. Therefore, we know very little about how men and women might be segregated into different job titles within occupations, especially over time and at the national level.
My recent article in the American Sociological Review shows that occupations hide significant levels of gender segregation and that job title desegregation may be slowing relative to occupational desegregation.
More generally, I show how the way we measure occupations obscures important patterns related to inequality and stratification.
What’s in an occupation?
Occupations are central to social science. We use occupations to understand social mobility, the rise of income inequality, and the future of work. But the way we measure occupations conceals important patterns.
We can think of occupations as categories of work containing unique jobs. In most of the large, nationally representative surveys that researchers rely on, we can see a respondents’ occupation, but not their job. For example, we might be able to see that someone is a “physician or surgeon,” but we do not know that person’s actual job title or tasks. This helps protect the individual’s anonymity, but it limits what we know about their work.
Most surveys use a list of occupations designated by the U.S. Census Bureau. The assumption is that the jobs within occupations are mostly similar, but researchers have very little information about how similar they really are and what may be hidden.
I was able to investigate this issue by using a unique dataset from the General Social Survey, a nationally representative survey of U.S. residents from 1972 to 2018. In addition to traditional data on occupations, I was able to access underlying data about respondents’ actual job names and descriptions of their work. These data had never been analyzed before.
Using computational text analysis, I could then investigate how similar jobs and tasks are within occupations.
I found lots of variation. While some occupations are very cohesive (i.e., the jobs within them are very similar based on the words respondents say), it turns out that many occupations contain a lot of diversity in terms of job titles and tasks.
Not only that, but occupations are becoming substantially less cohesive and more internally varied over time.
I also found interesting patterns of occupational variation. Higher-status occupations tend to have a lot of task variation but more similar job titles. Gender diversity is associated with more variation in terms of both job titles and tasks.
What does all this mean? When researchers analyze standard survey data, they are missing out on a lot of potentially important information about the diversity of workers’ jobs and tasks.
The case of gender segregation
One reason that this within-occupation variation matters is that it affects our understanding of gender segregation. We know that men and women tend to concentrate in different occupations. For example, in the GSS data, chief executives are 81% men and childcare workers are 97% women.
But we know less about how men and women might tend to have different job titles within the same occupation. Occupations that appear to be well integrated by gender (such as postsecondary teachers, which are 52% men), may actually have substantial job title segregation.
Using the GSS data on job titles, I was able to calculate the level of job title segregation and compare it to our traditional occupational segregation measures. I found that job title segregation is significantly higher than occupational segregation.
For example, in the period from 2010 to 2018, 52% of men or women would have to switch occupations in order for there to be no segregation (i.e., an equal distribution of men and women across occupations). But when we look at job titles, that number increases to 72%.
Furthermore, job title segregation may be a stubborn problem. Over the past decade, the level of job title segregation has plateaued. Though more data are needed to confirm this pattern, it suggests a troubling possibility whereby improvements in occupational segregation may mask stagnating segregation at the job title level.
What else might occupations hide?
I use the case of gender segregation, but a similar story may be true for race, class, and other worker characteristics. Even within the Census’s detailed occupations, I found striking levels of variation and segregation. This segregation is typically rendered invisible with standard survey data.
The consequences of this invisibility are brought into sharp relief when we consider how the U.S. Equal Employment Opportunity Commission monitors the demographic composition of firms. EEO-1 surveys collect data from employers on the gender and racial composition of just 10 large occupational categories.
The issue of variation within occupations is likely to become more problematic as the labor market becomes more complex and job titles proliferate. (In some companies, employees are even offered the opportunity to write their own job titles.) In this context, we need to continue to pay attention to how job title differentiation might influence stratification and inequality.
Occupations are necessary and useful concepts, and many occupations are good representations of the work that incumbents do. But – as with any categories – it is critical to understand what goes into the aggregation process and to shed light on hidden forms of inequality.
Read more
Ananda Martin-Caughey. “What’s in an Occupation? Investigating Within-Occupation Variation and Gender Segregation Using Job Titles and Task Descriptions” in American Sociological Review2021.
A free, pre-publication version of the article is available on SocArXiv.
Image: boaphotostudio via pixabay
No Comments