Ahead overall, behind job by job
The Philippines publishes one of the largest gaps in women’s favour in the set. Over all employees in 2024 its hourly gender wage gap is minus 15.3%: the average woman earns 122.2 pesos an hour, the average man 105.9. Among 80 economies, only four report a figure further in women’s favour.
The same survey, cut into the nine job families the ILO classifies occupations into, says something else. Men are paid more in seven of the nine, and the median family’s gap is 6.3%. The two families where women lead are elementary occupations and clerical support; in skilled agricultural work the gap runs 39.4% the other way.
What moves between the two counts is not pay, it is who is doing which work. Filipino women in paid employment are concentrated in the better-paid families and Filipino men in the worse-paid ones, so the all-employee average carries the sorting as well as the gap, and the sorting is the larger of the two. Across the whole set the median economy’s job-family gap runs 5.8 points above its all-employee gap.
| Economy | Year | All employees | Median job family | Families paying men more, of nine | Rank of 80 |
|---|---|---|---|---|---|
| Philippines | 2024 | -15.3% | 6.3% | 7 | 5 |
| Pakistan | 2021 | -5.1% | 12.0% | 7 | 11 |
| Colombia | 2025 | -4.6% | 2.4% | 5 | 13 |
| Thailand | 2025 | -4.2% | 9.5% | 7 | 14 |
| Italy | 2020 | -2.5% | 4.2% | 6 | 16 |
| Mexico | 2025 | 0.1% | 5.7% | 8 | 25 |
| Greece | 2020 | 3.5% | 11.3% | 8 | 30 |
| Türkiye | 2025 | 5.0% | 11.1% | 9 | 34 |
| France | 2024 | 5.4% | 4.9% | 8 | 36 |
| Portugal | 2025 | 7.7% | 11.7% | 9 | 43 |
| Vietnam | 2024 | 8.7% | 13.8% | 9 | 48 |
| Brazil | 2025 | 10.0% | 15.8% | 8 | 52 |
| United Kingdom | 2025 | 13.4% | 13.0% | 8 | 54 |
| Indonesia | 2023 | 13.5% | 23.8% | 9 | 55 |
| United States | 2025 | 19.2% | 23.6% | 9 | 68 |
| India | 2025 | 19.5% | 28.5% | 8 | 70 |
| Uganda | 2021 | 48.3% | 46.4% | 9 | 80 |
Twenty-two economies, the same reversal
Counted over all employees, the gap favours men in 56 of the 80 economies, and the median figure is 6.8%. Counted inside the median job family, it favours men in 78 of the 80, and the median figure is 12.5%. Only Afghanistan and Belize pay women more in their middle job family.
Twenty-two economies report an overall gap that is level or in women’s favour while their median job family pays men more. Italy is one, at minus 2.5% overall and 4.2% inside the median family. Thailand is another, minus 4.2% and 9.5%. Pakistan, Jordan, Tunisia, Malaysia and most of Central America sit in the same group, and the reversal is not a property of rich countries or of poor ones.
Income orders none of it. The all-employee gap correlates at 0.11 with the log of income per person and the job-family gap at minus 0.11, and neither moves in one direction across the income thirds: overall medians of 11.2%, 1.5% and 7.6% from the poorest third to the richest, job-family medians of 17.9%, 10.4% and 12.0%. Uganda carries the widest gap in the set, 48.3% over all employees, and twelve economies sit above the United States’ 19.2%.
What holding the job constant does not do
It does not make this equal pay for equal work. Nine job families is the coarsest possible control, and the finer sorting survives inside every one of them: which specialism, which employer, which seniority, how many years in the role. A gap that remains after grouping a surgeon with a primary school teacher has not been adjusted for anything much.
Nor does it make the number a hiring signal. The widest family is craft and trades, at a median 18.1% across the 80; the one where men lead in the most economies is professionals, at 74 of the 80; the narrowest is clerical support, at 5.9%. Those are averages over very different jobs, and none of them describes any particular vacancy.
And the gap is the same measure on both sides of the comparison. In 15 of the 80 economies the median job family is NARROWER than the all-employee figure, France among them at 4.9% against 5.4%. Holding the job family constant does not automatically widen a gap. It happens to widen it in 65 of the 80 because of how men’s and women’s work is distributed, which is a fact about the distribution and not about the measure.
- The gap is HOURLY, and it is raw. Hours are already divided out, so none of it is the part-time work that women do more of. But nothing else is held constant: it is the distance between two averages, not an estimate of what a woman is paid for a man’s work. The published series was checked against the ILO’s own hourly earnings, which reproduce it in 5,172 of 5,172 comparable cells, where the monthly series reproduces 47.
- A job family is one of NINE ISCO-08 major groups, not an occupation. "Professionals" holds a surgeon and a primary school teacher; "services and sales" holds a shop assistant and a restaurant manager. Holding the family constant removes the coarsest sorting between men’s and women’s work, and leaves every finer sorting inside it exactly where it was.
- Who enters the average is decided by the labour market, not by the statistician. Only employees with earnings are counted, so where few women hold a paid job the ones who do are unusually qualified and the overall figure can favour women. Afghanistan reads minus 125%. That is why the comparison here is between two counts WITHIN an economy, never between the levels of two economies.
So the Philippines’ minus 15.3% does not say Filipino women are paid more for the same work, and its 6.3% does not say Filipino men are paid 6.3% more for the same work: they say that in 2024 the average employed woman earned 15.3% more an hour than the average employed man, and that of the nine job families she might have been working in, seven paid men more, the middle one by 6.3%.
One hourly gap, computed twice over the same respondents, and a reading that forgets what is held constant says what the source does not.
What these gaps do not say
They do not say women are paid less for doing the same job. The control here is a group of occupations, not a job, and no hours, contract, seniority or firm is held constant. What the two counts together establish is narrower and firmer: the overall average is not a summary of what happens inside the work.
They do not rank the economies. Who enters the average depends on who holds a paid job, so a country where few women are employed can report a gap in women’s favour without anything about its labour market being favourable. Afghanistan’s minus 125% is that, and it is why every comparison here is made within an economy.
And they name no cause. Occupational segregation, hours, tenure, firm, bargaining, discrimination and the composition of who works for a wage all sit behind these two numbers, and none of them is measured here. Each economy appears once, in its own latest year: nothing here says a gap widened or narrowed.
