Gender pay gap statistics: the definitions

The gender pay gap is usually quoted as a single ratio, but the federal statistics behind it come in several forms with different definitions, and the number depends on which one is used. This page sets out the definitions. It does not interpret them.

The Census annual ratio

The most cited figure is the Census Bureau’s female-to-male earnings ratio for full-time, year-round workers, published each September in the annual income report from the Current Population Survey’s Annual Social and Economic Supplement. It divides the median annual earnings of women who worked full time (35 or more hours a week) for the full year (50 or more weeks) by the same median for men. The 2025 report, covering 2024 earnings, put the ratio at 83.9 percent, meaning the median woman in that population earned about 84 cents for each dollar earned by the median man. Earnings here are total money earnings across all jobs, including self-employment income.

The ratio is a comparison of two medians. It is not an average of individual gaps, and it is not restricted to men and women in the same occupation, employer or hours.

The BLS weekly ratio

BLS publishes a parallel figure from the monthly CPS: the ratio of median usual weekly earnings of women to men among full-time wage and salary workers. It appears quarterly in the weekly earnings release and annually in “Highlights of Women’s Earnings,” with detail by age, race, occupation, industry, education and state. Because it is weekly rather than annual and excludes the self-employed, it differs from the Census ratio by a few percentage points in a given year; it is usually slightly higher. Within the report, the ratio by occupation shows how much of the overall gap sits between occupations and how much within them.

Hourly comparisons

Neither headline series is hourly. Full-time men work more hours per week on average than full-time women, so a weekly or annual ratio includes an hours difference that an hourly ratio does not. Researchers compute hourly ratios from CPS microdata; BLS does not publish one as a headline. OEWS, the source of the occupation tables on this site, collects no information on sex and cannot be used for a gap at all.

Unadjusted and adjusted

The published ratios are unadjusted: they compare all women with all men in the stated population. An adjusted gap controls for characteristics such as occupation, industry, hours, experience and education, and is always smaller, because part of the raw difference reflects where women and men work. How much smaller depends on which controls are used, and the federal agencies do not publish an official adjusted figure. Both numbers are informative; they answer different questions, and a report that quotes one as if it were the other is misreading it.

Medians, means and the distribution

A ratio of medians describes the middle of each distribution. A ratio of means would be pulled by the top, where men are more heavily represented, and would be lower. Ratios computed at other percentiles differ again. The Census and BLS reports both publish distributions, not only medians, for readers who want the shape.

Change over time

The Census series runs back to 1960 and shows the ratio rising from about 60 percent to the low 80s, most of the change occurring between 1980 and 2000, with slower movement since. Year-to-year changes of a point or less are within sampling error and the Census Bureau says whether a change is statistically significant. Composition matters too: a recession that removes low-wage jobs changes both medians without changing anyone’s pay; see wage growth and the ECI for the same problem in another series.

On this site

The trends section planned for Wave 2 will carry the Census and BLS ratios with their definitions and release dates. The state pages carry OEWS wages, which have no sex dimension. Readers looking for an occupation’s gap will find it in the BLS report’s occupation table, not here.