Our World in Data presents the empirical evidence on global development in entries dedicated to specific topics. This blog post is closely related to a companion article where we discuss the drivers of the gender pay gap. It is also related to our articles exploring female labor force participation trends and determinants.
This blog post benefitted greatly from research contributions provided by Sandra Tzvetkova.
The 'gender pay gap' comes up often in political debates, policy reports, and everyday news. But what is it? What does it tell us? Is it different from country to country? How does it change over time?
In this blog post I try to answer these questions, providing an empirical overview of the gender pay gap across countries and over time. I've structured the analysis around six 'key facts'.
1) The gender pay gap measures inequality but not necessarily discrimination
The gender pay gap (or the gender wage gap) is a metric that tells us the difference in pay (or wages, or income) between women and men. It's a measure of inequality and captures a concept that is broader than the concept of equal pay for equal work.
Differences in pay between men and women capture differences along many possible dimensions, including worker education, experience and occupation. When the gender pay gap is calculated by comparing all male workers to all female workers – irrespective of differences along these additional dimensions – the result is the 'raw' or 'unadjusted' pay gap. On the contrary, when the gap is calculated after accounting for underlying differences in education, experience, etc., then the result is the 'adjusted' pay gap. The key facts below cover unadjusted measures of the gender pay gap. (You can read about the adjusted pay gap in our companion blog post here.)
Discrimination in hiring practices can exist in the absence of pay gaps – for example, if women know they will be treated unfairly and hence choose not to participate in the labor market. Similarly, it is possible to observe large pay gaps in the absences of discrimination in hiring practices – for example, if women get fair treatment but apply for lower-paid jobs.
The implication is that observing differences in pay between men and women is neither necessary nor sufficient to prove discrimination in the workplace. Both discrimination and inequality are important. But they are not one and the same. (You can read about discrimination and 'equal pay for equal work' in our companion blog post here).
2) In most countries there is a substantial gender pay gap
Cross-country data on the gender pay gap is patchy, but the most complete source in terms of coverage is the United Nation's International Labour Organization (ILO). The following visualization presents this data. You can add observations by clicking on the option 'add country' at the bottom of the chart.
The estimates shown below correspond to differences between average hourly earnings of men and women (expressed as a percentage of average hourly earnings of men), and cover all workers irrespective of whether they work full time or part time.1
As we can see: (i) in most countries the gap is positive – women earn less than men; and (ii) there are large differences in the size of this gap across countries. Below we come back to these cross-country differences.
(NB. By this measure the gender wage gap can be positive or negative. If it is negative, it means that, on an hourly basis, men earn on average less than women. This happens in some countries, such as Malaysia.)
3) In most countries the gender pay gap has decreased in the last couple of decades
How is the gender pay gap changing over time? To answer this question, let's consider the following chart, showing available estimates from the OECD. These estimates include OECD member states, as well as some other non-member countries, and they are the longest available series of cross-country data on the gender pay gap that we are aware of.
Here we see that the gap is large in most OECD countries, but it has been going down in the last couple of decades. In some cases the reduction is remarkable. In the UK, for example, the gap went down from almost 50% in 1970 to about 17% in 2016.
These estimates are not directly comparable to those from the ILO, because the pay gap is measured slightly differently here: The OECD estimates refer to percent differences in median earnings (i.e. the gap here captures differences between men and women in the middle of the earnings distribution); and they cover only full-time employees and self-employed workers (i.e. the gap here excludes disparities that arise from differences in hourly wages for part-time and full-time workers).
The conclusion is that in most countries with available data, the gender pay gap has decreased in the last couple of decades.
4) The gender pay gap is larger for older workers
The United States Census Bureau defines the pay gap as the ratio between median wages – that is, they measure the gap by calculating the wages of men and women at the middle of the earnings distribution, and dividing them.
By this measure, the gender wage gap is expressed as a percent (median earnings of women as share of median earnings of men) and it is always positive. Here, values below 100% mean that women earn less than men, while values above 100% mean than women earn more. Values closer to 100% reflect a lower gap.
The next chart shows available estimates of this metric for full-time workers in the US, by age group.
First, we see that the series trends upwards, meaning the gap has been shrinking in the last couple of decades. Secondly, we see that there are important differences by age.
The second point is crucial to understand the gender pay gap: the gap is a statistic that changes during the life of a worker. In most rich countries, it’s small when formal education ends and employment begins, and it increases with age. As we discuss in our analysis of the determinants, the gender pay gap tends to increase when women marry and when/if they have children.
5) Women in rich countries tend to be overrepresented in the bottom of the income distribution – and underrepresented at the top
Another way to look at cross-country gender inequalities in earnings is to focus on the share of women among low-pay earners. This is shown in the chart below. Here, 'low-pay' refers to workers earning less than two-thirds of the median (i.e. the middle) of the earnings distribution.
A share above 50% implies that women are 'overrepresented', in the sense that among those with low wages, there are more women than men.
The fact that women in rich countries are overrepresented in the bottom of the income distribution goes together with the fact that working women in these countries are overrepresented in low-paying occupations. This is also crucial to understand the pay gap: occupational choice is a key contributor to the observed inequalities.
A similar analysis at the opposite tail of the distribution is also insightful. The following chart shows, for the UK, the share of women at different brackets of the income distribution across time.
Here we can see that among those with high incomes, women are a minority in the UK. However, the representation of women at the top of the income distribution has been increasing over time.
Notice also that the rate of increase in the proportion of high-income women in the UK is smaller within the very top income groups. This is usually interpreted as sign of the 'glass ceiling' in female earnings, and is actually a pattern that is observed in other countries too. In the US, for example, the gender pay gap declined much more slowly at the top of the wage distribution than at the middle or bottom.
Percentage of women in top income groups, UK, 1997 to 2001 - Atkinson, Casarico and Voitchovsky (2014)2
6) The gender pay gap is smaller in middle-income countries – which tend to be countries with low labor force participation of women
The following scatter plot shows available ILO estimates on the gender pay gap (vertical axis) vs GDP per capita (on a logarithmic scale along the horizontal axis). As we can see there is a weak positive correlation between GDP per capita and the gender pay gap. However, the chart shows that the relationship is not really linear. Actually, middle-income countries tend to have the smallest pay gap.
The fact that middle-income countries have low gender wage gaps is, to a large extent, the result of selection of women into employment. Olivetti and Petrongolo (2008)3 explain it as follows: “if women who are employed tend to have relatively high‐wage characteristics, low female employment rates may become consistent with low gender wage gaps simply because low‐wage women would not feature in the observed wage distribution.”
Olivetti and Petrongolo (2008) show that this pattern holds in the data: unadjusted gender wage gaps across countries tend to be negatively correlated with gender employment gaps. That is, the gender pay gaps tend to be smaller where relatively fewer women participate in the labor force.
So, rather than reflect greater equality, the lower wage gaps observed in some countries could indicate that only women with certain characteristics – for instance, with no husband or children – are entering the workforce.