The Huffington Post’s Jillian Berman cites two studies she claims debunks Mark Perry’s and my Wall Street Journal article, which casts doubt on the so-called “gender pay gap” supposedly causing women to earn just 77 cents for each dollar received by men. One of these studies, from the St. Louis Fed, actually supports our claims. The title, “Gender Wage Gap May Be Much Smaller Than Most Think” kind of gives it away.
But Berman cites another study from the American Association of University Women that’s more interesting. The AAUW study, which is based on data from the Department of Education’s “Baccalaureate and Beyond” dataset, claims that–even among new college graduates where marriage and children play a smaller role – women earn 6.6% less than men. Now, that’s the unexplained gender difference which could be attributable to labor market discrimination – or it could be attributable to a number of other factors.
In any event, since it’s being cited, I figured I’d take a closer look. The study is not bad, but I can see a number of ways its results might be skewed:
Occupation and college major controls: The AAUW study has broad controls for occupation and college major. Say, one occupational group is “Math, computer, and physical science” and one college major is “Social sciences.” But it’s certainly possible – probable, I think – that within those broad groups men cluster into higher-paying jobs/majors. Say, within social sciences men are more likely to major in economics and women in psychology.
College GPA: The AAUW study controls for undergraduate GPA, which makes sense. Within any given college major, higher-GPA students are likely to earn more in the workforce. But across college majors it’s a different story, because lower-paying majors – say, education or English – award much higher average GPAs than higher-paying majors such as math or economics. Something like SAT scores, where everyone takes the same test, would probably be a better control.
Geography: The AAUW study controls only for very broad region (Northeast, Midwest, etc.). But Texas A&M economist Lori Taylor shows that tight geographic controls are important: two predominantly female professions – teaching and nursing – exist in pretty much every community in America, including low-cost rural areas. Other white collar professions are clustered in suburbs and cities where the cost of living, and wages, is higher. Without good geographic controls, you’re comparing disproportionately female individuals from lower cost of living areas to disproportionately male workers in high-cost areas.
To check on all of these factors, I downloaded recent data from the Census Bureau’s American Community Survey. The ACS has very detailed data on occupations, college majors, and local region.
To keep things simple (this is a blog post, folks) I restricted my analysis to individuals aged 23-27 who were white/native-born/non-Hispanic (to get around race/immigration issues), unmarried with no kids in their household, worked full-time the previous year, worked in the private sector, and worked between 35-40 hours per week. I controlled for detailed college major, occupation, industry, and for hours worked (in the 35-40 range I restricted myself to).
Results: a female pay penalty of 1.5%, a result that is small and easily explainable by factors other than discrimination – say, men are more likely to bargain over wages, or men work in jobs with greater financial or physical risk. In other words, at a time when men and women are most similar – fresh out of college, no kids, and so on – the labor market doesn’t seem to discriminate. It’s only when men and women become more different in terms of things the labor market actually values – say, years of work experience – that pay differences get bigger. The logical conclusion is that it’s differences in these pay-related factors, not labor market discrimination, that accounts for the so-called “gender pay gap.”
Follow AEIdeas on Twitter at @AEIdeas.