Published gender pay ratios by group compare women in a group against men in the same group, so each figure has a different denominator. Asian women earn about 79 percent of Asian men's median, White women about 83 percent of White men's, Hispanic women about 87 percent of Hispanic men's, and Black women about 92 percent of Black men's. A high ratio can reflect low male earnings in that group rather than high female earnings, which is the standard misreading.
Read the denominator first, every time
The single most important thing about these figures is that each one uses a different comparison group. When a source says Black women earn 92 percent, it means 92 percent of what Black men earn. When it says Asian women earn 79 percent, it means 79 percent of what Asian men earn. The denominators are different populations with different median earnings.
That makes the four percentages non-comparable with each other, even though they are almost always printed as a list that invites exactly that comparison. Ranking them produces a conclusion that reverses the underlying earnings picture. A narrow ratio can mean the men in that group are also paid relatively little, so the group with the highest ratio may be the group with the lowest earnings on both sides. The presentation format is doing the damage rather than the data.
This is not a subtle statistical point. It is the difference between “this group of women is doing well” and “this group of men is doing badly”, and the same number is consistent with both. A figure without its denominator stated is not usable, and a great deal of published commentary omits it.
What the figures are
Using median weekly earnings from the household survey, and comparing each group of women to men in the same group, the figures run as follows. Asian women sit at roughly 79 percent, White women at roughly 83 percent, Hispanic women at roughly 87 percent, and Black women at roughly 92 percent. Four different comparisons against four different reference points, presented as one list.
Set against a single common denominator — all men, or White men — the ordering changes substantially, because the groups have very different median earnings to begin with. Both presentations are entirely legitimate and they answer different questions. The within-group version asks about the gap between women and men who share a group. The common-denominator version asks about position in the overall earnings distribution, which is usually what a reader assumed they were being told.
Which version a source is using ought to be stated explicitly, and frequently is not. When it is missing, the within-group version is the more common of the two. You can usually detect it from the shape of the numbers: the figure for Black women will be the highest of the four rather than among the lowest, which is the signature of a within-group comparison.
Age changes the picture more than most factors
The overall gender ratio conceals a strong age pattern that is more informative than most of the demographic breakdowns. Among workers aged 16 to 24 the ratio sits around 94 percent — close to parity. From 25 to 34 it is about 89 percent. From 35 onward it settles into a band of roughly 77 to 83 percent and stays there.
That shape tells you the gap is largely a second-half-of-career phenomenon rather than something applied at entry. Young workers of both sexes start in similar positions and the divergence appears over the years when promotion, specialization and career interruption all concentrate. Any explanation of the headline figure has to account for that timing, and explanations resting on entry-level treatment do not.
The gap is wider in higher-paying occupations
The other pattern worth knowing runs directly against most people’s intuition. Gender pay gaps tend to be wider in higher-earning occupations than in lower-earning ones rather than narrower. That surprises almost everybody, and the reason it happens is structural rather than cultural.
The mechanism is not mysterious. In occupations with a narrow earnings distribution — where a defined rate or a tight band does the pricing — there is very little room for two people in the same role to be paid differently. In occupations with wide distributions, where bonus, commission, client book, specialization and negotiation all move the number, the room for divergence is correspondingly large. The same forces that let a high earner earn substantially more than their peer also let gaps open up.
The practical consequence is worth carrying into a job search. Somebody entering a high-paying field with a wide earnings distribution should expect pay-setting there to be less standardized and more individually negotiated than in a field with a tight band. That is a structural feature of the occupation rather than a personal difficulty, and knowing it in advance is most of what you can do about it.
What no published US source gives you
It is worth being clear about the limits here, because a great deal of published writing implies a precision that does not exist. The main occupational wage survey carries no breakdown by sex or race at all. It is a survey of establishments rather than of people, so those characteristics are never collected in the first place — this is a structural gap rather than an omission anybody could fix by asking.
That means there is no published source giving you the median pay for women in a specific occupation in a specific metropolitan area. The demographic figures come from a household survey with a much smaller sample, so they support national and broad occupational comparisons and not fine-grained local ones. Anybody presenting a metro-level gap for a specific job is estimating rather than reporting.
Nor is there a published intersection of every dimension at once. Figures by sex exist, and figures by group exist. Figures by sex, group, occupation, age and location simultaneously do not, because the sample would not support them at any useful confidence. Anybody presenting one has estimated it, and the estimate should be labeled as such.
Using these figures well
Quote them with their denominators attached, always. “Black women earn 92 percent of what Black men earn” is a sentence that cannot be misread; “Black women earn 92 percent” is one that will be.
Use them to understand patterns rather than to establish anything about a particular employer or a particular person’s pay. They describe distributions across a national workforce, and the distance between that and a claim about one workplace is exactly as large as it was for the headline figure.
And when you meet a ranking of groups by pay ratio, check whether the denominators are the same before drawing any conclusion from the order. If they are not, the ranking is measuring something other than what it appears to measure.
This is general information about what these statistics show rather than legal advice. A question about your own pay relative to a colleague is a legal question with its own tests and time limits, and it is covered separately in this section.
Common questions
What do the group pay ratios compare?
Women in a group against men in the same group. Each figure has a different denominator, which is why they cannot be ranked against each other.
What are the figures?
Roughly 79 percent for Asian women, 83 percent for White women, 87 percent for Hispanic women and 92 percent for Black women — each against men in the same group.
Why is the highest ratio not the best outcome?
Because a narrow ratio can mean the men in that group are also paid relatively little. The same number is consistent with women doing well and with men doing badly.
How does the gap change with age?
About 94 percent at 16 to 24, about 89 percent at 25 to 34, then a band of roughly 77 to 83 percent from 35 onward. It is largely a second-half-of-career phenomenon.
Is the gap bigger in low-paying jobs?
No, the reverse. It tends to be wider in higher-earning occupations, because wide earnings distributions leave more room for two people in the same role to be paid differently.
Can I get a gap figure for my occupation and city?
No. The main occupational wage survey collects no breakdown by sex or race, and the household survey that does has too small a sample for fine-grained local figures.
Do figures exist for every combination of factors?
No. Figures by sex and by group exist; figures by sex, group, occupation, age and location simultaneously do not, because the sample would not support them.
How should I quote these?
With the denominator attached. 'Black women earn 92 percent of what Black men earn' cannot be misread; 'Black women earn 92 percent' will be.