The unadjusted gap compares all women's earnings with all men's and captures occupational sorting, hours and seniority. The adjusted gap compares people matched on those factors and asks whether like is paid like. The adjusted figure is always smaller, and it is smaller precisely because it has removed the things the unadjusted figure exists to show.
Why the adjusted figure is smaller
Because controlling for occupation, industry, hours and experience removes exactly those differences from the comparison. That is the point of the calculation, not a flaw in it.
The disagreement people have is usually not about arithmetic. It is about whether the sorting itself — who ends up in which occupation and at which level — should be treated as a given or as part of what is being measured.
What each one is good for
Unadjusted answers: how do women and men fare in this economy overall? Right for policy questions about occupational segregation, caring responsibilities and hours.
Adjusted answers: does this employer pay comparable people comparably? Right for a legal claim, a pay audit, and your own situation.
The trap in adjusting
Control for the wrong variable and you assume away the thing you were testing. Controlling for job level hides the possibility that people were promoted unequally. Controlling for prior salary imports whatever discrimination shaped that salary — which is precisely why several states now forbid its use.
A well-built pay analysis says which variables it controlled for and why, and a figure quoted with no such statement should be treated as decorative.
Both can be true at once
An employer can pay every comparable pair identically and still show a large unadjusted gap, because its senior roles are held disproportionately by one group. That is a real finding about that employer and it is not an equal pay violation.
What to ask when you see a figure
Adjusted or unadjusted, and for what. Those five words settle most arguments in this subject before they start.
Common questions
Which figure is the real one?
Both, for different questions. Unadjusted describes the labor market; adjusted asks whether comparable people are paid comparably.
Why is the adjusted gap smaller?
Because it removes occupation, hours, experience and level from the comparison — which is its purpose rather than a defect.
What is the danger in adjusting?
Controlling for the wrong variable assumes away what you were testing. Controlling for prior salary imports whatever shaped that salary.
Can an employer have no adjusted gap and a large unadjusted one?
Yes, if senior roles are held disproportionately by one group. That is a real finding and not an equal pay violation.
What should I ask about any published figure?
Adjusted or unadjusted, and for what. Most disagreements dissolve once that is answered.