Pay, and how it is set · 3.7

The pay gap, and what each measurement of it means

The pay gap, and what each measurement of it means. What is actually the case, and how it compares with what is repeated.

For a practical software perspective on this topic, this Monitask guide provides a concise companion guide.

Two numbers are in circulation, they differ by a factor of three or more, both are correct, and they answer different questions. Almost every argument about the pay gap is an argument between people quoting different ones.

The unadjusted gap

The difference between the median or mean earnings of all women and all men, with no controls. In most high-income countries it sits in the low to mid teens for hourly pay and considerably higher for annual earnings, because women work fewer paid hours on average.

It answers: how much less do women earn. That is a real question about the distribution of income and it is the number most reporting requirements collect.

The adjusted gap

The difference remaining after controlling for occupation, sector, hours, experience and firm. It is much smaller, typically low single digits.

It answers: are women paid less for the same work in the same place. Also a real question, and a different one.

Why controlling is not obviously the right move

Because occupation is partly the thing being explained. If women are concentrated in lower-paid occupations, and if part of the reason those occupations are lower paid is that women are concentrated in them, controlling for occupation removes the phenomenon from the measurement.

Adding controls always shrinks the gap. Whether each control clarifies or conceals depends on whether it is a confounder or a mechanism, and that is a judgement about the world, not a statistical decision.

What the research has converged on

That most of the remaining gap in countries with good administrative data opens at the birth of a first child, and does not close afterwards. Women's earnings drop sharply and recover partially; men's do not move.

This has been shown with population registers in several countries and is one of the more robust findings in the field. It relocates the question from pay-setting to the division of care, which is where the entry on care work picks it up.

The reporting requirements

The United Kingdom has required employers above a size threshold to publish gender pay gap figures since 2017. The European directive extends comparable duties across member states.

Publication alone changes little. Where effects have been found they are modest, and the Danish evidence described in the transparency entry suggests compression rather than uplift.

Hourly, weekly, annual

Three different gaps, increasing in that order, because hours differ. Quoting the annual gap while discussing pay rates, or the hourly gap while discussing lifetime income, is the commonest sleight in this subject and it is usually accidental.

Pension outcomes use a fourth measure again, accumulated over a career, and produce the largest gaps of all.

Beyond gender

Ethnicity pay gaps are measured in fewer countries, published by fewer employers, and in several cases are larger than the gender gap. Where both are measured, the intersection is larger than either.

Reporting requirements have generally started with gender because the categories are simpler to record, which is an administrative reason rather than a substantive one.

How to read a figure

Ask whether it is adjusted, and if so for what. Ask whether it is hourly, weekly or annual. Ask whether it is mean or median. Those three questions determine most of the difference between numbers that appear to contradict each other and do not.

The two misuses

Quoting the adjusted gap as though it showed the problem is small, when it was constructed by removing most of the problem. And quoting the unadjusted gap as though it showed women paid less for identical work, which it does not measure.

Both are common, both are made by people arguing in good faith, and both are avoided by saying which number you are using and what it contains.

The firm effect

A substantial part of the gap operates through which firms people work at rather than through pay within a firm, because higher-paying employers hire proportionately fewer women in several countries.

That mechanism is invisible to any measure computed inside a single employer, which is most of what pay gap reporting collects, and it means a set of individually clean employers can still produce a large aggregate gap.

The part-time interaction

Because part-time work carries an hourly penalty and is done disproportionately by women, part of the hourly gender gap runs through part-time status rather than through gender directly.

Controlling for it therefore removes a mechanism rather than a confounder, which is the same problem as controlling for occupation and is made less often.

For broader context, consult OECD gender wage-gap indicator.