The pay gap between women and men, figure by figure
Three different figures circulate for France, and all are accurate. They measure nested gaps whose decomposition is the real subject.
The three measures
The total earnings gap compares all labour income received in a year. It is the highest figure, because it incorporates differences in working time and in duration of employment over the year.
The full-time equivalent gap neutralises duration. The gap at comparable job additionally neutralises sector, occupation, experience and establishment size. Each neutralisation reduces the figure.
None of these figures is wrong, and none is the only relevant one. Each answers a distinct question: how much is received in total, how much an hour of work earns, how much a given job is paid.
What the reduction means
Moving from one measure to another does not make part of the gap disappear: it shifts it onto the neutralised variables. Part-time work and occupational segregation become the subject instead of being the residual.
The comparable-job figure is therefore the smallest and the least informative about the phenomenon as a whole, although it is most often cited as a measure of discrimination. The opposite argument can also be made: it is the only one that isolates a difference in treatment in an identical situation.
The decomposition method
The standard technique, formulated independently by Ronald Oaxaca and Alan Blinder in 1973, separates the share of the gap explained by observable characteristics from the residual share.
That residual share is frequently presented as discrimination. That is an abusive reading: it contains everything that was not measured, and the control variables may themselves result from discriminatory mechanisms. It may therefore overstate as well as understate discrimination, depending on the variables chosen.
What longitudinal data bring
Work on the effect of parenthood, notably by Henrik Kleven and colleagues, follows the same people before and after a birth and finds a sharp and lasting break in mothers' earnings.
That approach sidesteps part of the control variable problem, since it compares a person with themselves, and it locates a substantial share of the gap around one precise event.
The work of Claudia Goldin, awarded the Sveriges Riksbank Prize in 2023, offers a complementary reading: some jobs reward availability and long hours heavily, and the gap widens when one of the parents reduces that availability.
Mandatory publication
Several countries require firms to publish gap indicators. France introduced an index in 2019, combining several indicators including the pay gap at comparable job.
Available evaluations, in France as in the United Kingdom, find modest effects on measured gaps, and flag scope for reporting firms to optimise the indicators. A composite indicator can show a good score while leaving the total earnings gap untouched.
Key points
Three figures coexist, from total earnings to the comparable-job gap, and all are accurate.
Neutralising a variable shifts the gap onto that variable instead of removing it.
The residual share of a decomposition contains everything unmeasured, not only discrimination.
This article describes documented mechanisms and disagreements. It recommends no position, no party and no vote, and commits to no reading other than that of the work it cites.
And you, where do you stand?
The Civix test situates your convictions on six dimensions and matches you with one of the twelve profiles, without ever telling you what to think.
Start my test