Several women stage a protest in downtown Miami demanding equal pay for women.

The gender wage gap in the United States has been “been intensively investigated for a number of decades, but also remains an area of active and innovative research.” So starts a new working paper by Francine D. Blau and Lawrence M. Kahn, both economists at Cornell University. The paper reviews quite a bit of research on the topic and lays out some facts on the gap, its decline over the years, and its continued existence. Given that last point, it’s worth digging into the research and facts presented by Blau and Kahn to grapple with the difference in wages paid to similar men and women.

The most common presentation of the gender wage gap is the “unconditional gap”, or the gap between the earnings of the typical woman and the typical man. The U.S. Bureau of Labor Statistics reports that in 2014, for instance, the median weekly earnings of a woman working full-time were 83 percent of the weekly earnings of a man working full-time. But as the BLS points out, this comparison doesn’t take into account the number of factors that can affect this difference such as education levels, length of time in the labor force, occupations, and many others.

That’s where research like that surveyed by Blau and Kahn comes into play. These more robust studies account for these factors in understanding the wage gap. So here, in short, is what Blau and Kahn find.

The wage gap, while still around, has declined significantly since the 1950s. But the decline hasn’t been consistent over time. While the 1980s saw strong convergence and reduction in the gender wage gap, the rate of reduction slowed down over the following 20 or so years. What’s more, the increases that women saw overall in educational attainment and labor market experience were significant reasons for both the wage gap decline as well as women’s entrance into higher-paying occupations.

Yet while improvements in human capital—as some economists might call the improvements in education and work experience—used to be important in pushing down the wage gap, these factors don’t explain the current gap. In 1980, these factors explained about 27 percent of the gender pay gap; by 2010, they only explained 8 percent. Contrast that decline in explanatory power to the role of industry and occupational segregation. While these forces explained about 20 percent of the gender wage gap in 1980, they explained 51 percent of the gap as of 2010. The gap today is smaller, but more of it is explained by differences in occupational and industry employment between men and women.

The portion of the gap that cannot be explained, however, is still a significant portion of the gap—about 38 percent in 2010. There are numerous potential causes of this unexplained gap, one of which is discrimination, conscious or unconscious. At the same time, “compensating differentials” could also be part of this unexplained gap. In essence, women could be making less because they are less willing to take unpleasant jobs, which offer higher compensation to make up for those features. Such an effect lines up with research by Harvard economist Claudia Goldin on how the returns to long hours within some occupations (think corporate law) increases the gender wage gap as men are more willing to work those hours. Of course, discrimination can also be a reason why women are less likely to work at these jobs. It’s difficult to fully untangle discrimination from other explanations why there are human capital, industrial, and occupational differences between men and women.

The gender wage gap has declined quite a bit since the days of Don Draper. But to borrow the title of Goldin’s paper, the “convergence” has entered its last chapter, requiring more emphasis on understanding the causes of—and figuring out a solution to—the wage gap.