Today is Equal Pay Day, identified each year by the National Committee on Pay Equity to illustrate the point in the year at which a woman’s earnings “catch up” to a man’s earnings during the previous year.

Though it can be calculated a couple different ways, the gender pay gap most commonly refers to the difference in median earnings between a man and a woman who held a full-time job for an entire year. Currently, it’s estimated to be just under 80 percent using Census data.

Equal Pay Day is always greeted by a flurry of critics who claim that the gender pay gap is a “myth” or that it is based on “bogus statistics.” The Washington Post’s Fact Checker even gave President Obama a rating of “two Pinocchios” for quoting published statistics from the Census Bureau. It’s possible that no other statistic produced by a federal statistical agency is as adamantly denied as the gender pay gap. How can this be?

How occupational segregation fits in

Many skeptics fail to appreciate the role of occupational segregation, or the tendency of men and women to be disproportionately employed in different occupations, in determining the wages of men and women.

Almost all critics of the pay gap argue that you have to make statistical adjustments to the published pay gap data to properly understand the disadvantage that women face in the labor market. These critics argue that women and men are different in several ways, from their educational background to their personal preferences for work, and those differences have to be held constant to understand what kind of pay disparities women face.

The critics are not entirely wrong on this point. We want to compare apples with apples and oranges with oranges, and we need to understand what factors explain the gender pay gap.

In a comprehensive look at its components, economists Francine Blau and Lawrence Kahn attribute much of the recent narrowing of the pay gap to women’s significant gains in educational attainment and work experience.

They also find that much of the remaining pay gap is due to the types of occupations in which men and women typically work. Occupations are distinguished by gender for many reasons, ranging from social norms that constrain employers’ and employees’ decisionmaking to personal preferences (which may be grounded in social norms) to discrimination from employers.

In another famous paper, economist Claudia Goldin shows that when wide gender pay gaps persist within occupations, it’s often because of differences in work hours and schedules.

Most critics of the pay gap agree with economists that much of the gap is related to occupational differences and occupational characteristics. However, they inappropriately conclude that because the pay gap can be explained by occupational differences, the pay gap is not real or is not evidence of labor market discrimination.

The problem with this argument is that if you statistically control for the occupations men and women hold, you remove one of the most important mechanisms driving down women’s wages relative to men.

It would be akin to calling the achievement gap between black and white students a myth because those gaps are much smaller if you only compare white and black students within the same school or the same school district. Few policy analysts would accept this argument against the achievement gap, because differences between school environments are one of the driving forces behind the achievement gap. You can’t separate a child’s education from his or her educational performance.

Yet the gender pay gap is frequently dismissed by statistically removing the chief determinant of wages in a market economy: the type of work you do.

In the 1970s and 1980s, advocates of “comparable worth” policies explicitly connected the gender pay gap to pay differences between female- and male-dominated occupations otherwise comparable in productive value.

They specifically demanded equal pay for work of equal value—not just equal pay within the same occupations—because they argued women were often paid less by segregating occupations and paying higher wages in male-dominated occupations.

During that period, Blau and Barbara Bergmann separately developed economic models showing how the segregation of women into lower-paying firms or occupations produced pay gaps.

Where do we go from here?

Thanks to the work of Goldin and others, researchers have a renewed focus on occupational differences as the heart of the gender pay gap and not a reason to dismiss the pay gap. But clear policy solutions are difficult to formulate.

Although vigorous enforcement of antidiscrimination laws and promotion of women in male-dominated fields is critical, substantial practical problems are associated with implementing true comparable worth policies. Without wage rates set in the market, there’s no reliable way to provide real-time values for different jobs to assess their comparable worth.

And, as Goldin emphasizes, some of the most promising solutions may not have a role for public policy, including changes in the workplace to allow for flexible scheduling and to limit the rewards for working long hours that typically privilege men.

Perhaps most fundamentally of all, we can expect the gender pay gap to persist as long as employers’ and employees’ ideas about gender norms and occupations persist.