Big data shows the graduate pay premium is bigger for women



A new study using big data has confirmed that those who complete university can expect to earn, on average, a decent premium for their degree. This alone won’t come as a shock – previous studies have shown that attending university pays off for most graduates. But my colleagues (Neil Shephard of Harvard University and Jack Britton of the Institute for Fiscal Studies) and I confirmed that this “graduate premium” is much higher for women than for men.

Using government administrative data, we looked at the median earnings of English men and women ten years after their graduation. We found that women graduates earn three times as much as women without a degree, while male graduates earn around twice as much as male non-graduates.

While women benefit more from their degrees than men, the gender gap in graduate earnings remains stark.

Using big data

Part of what makes this work novel is our use of big data – specifically, government administrative data in the form of tax and student loan records for over 260,000 graduates, collected up to ten years after their graduation. This is the first time a “big data” approach has been used to look at graduate earnings in England.

Previous work has estimated the premium earned by graduates using survey data. But the administrative data give a much more accurate picture than existing surveys, which tend to be based on much smaller samples of people who self-report their earnings, which makes the information subject to biases.

Using this anonymised tax data and student loan data, we looked at cohorts of graduates who started university in the period from 1998 to 2011, and observed their earnings (or lack thereof) in the tax year 2011/12 – though the results hold for graduates in other tax years, too.

The work, funded by the Nuffield Foundation, suggests that survey data has previously underestimated the earnings of graduates, particularly higher earning graduates. These new data indicate that ten years after graduation, 10% of male graduates were earning more than £55,000 per annum, 5% were earning more than £73,000 and 1% were earning more than £148,000.

Ten years after graduation, 10% of female graduates were earning more than £43,000 per annum, 5% were earning more than £54,000 and 1% were earning more than £89,000.


Our analysis is not causal, since we are simply comparing the earnings of graduates and non-graduates. Still, the results add to the evidence that, on average, graduates fare much better in the labour market than non-graduates.

Hit by recession

The study also produced a number of other interesting findings. Crucially, we found that the recession had a large impact on the earnings of people in their 20s and early 30s. This is particularly true for women, who experienced much lower earnings than previous cohorts. Over a four year period, men’s earnings were cumulatively 14% lower than expected, based on previous cohorts. For women over the same period, they were 20% lower than expected.

We’ll have to wait and see what this means in the long term, but there is no doubt that the recession has taken a toll on graduates in these cohorts in the short term.

Yet the research also indicates that graduates fared better than non-graduates through the recession – that is, they saw proportionally smaller drops in their earnings. So it also appears that higher education provided some protection from the economic downturn.

As with any study, there are limitations to the analysis. We used administrative data from both the Student Loan Company (SLC) and Her Majesty’s Revenue and Customs (HMRC) to observe how the earnings of students who take out a loan from the SLC change through the years as they mature in the labour market. This means we can only identify graduates who have borrowed money from the Student Loan Company, which is around 85% of English graduates in the period we looked at.

This means that there are some graduates whose earnings we cannot identify. Even so, we have reason to believe that they are likely to be higher earning graduates, on average if we assume that those who don’t take a loan are likely to come from more advantaged backgrounds and hence on average have higher levels of prior achievement and attend higher status institutions. If anything, the data we used is likely to underestimate average graduate earnings.

Despite these caveats, there is no doubt that this type of big data analysis allows us to better understand how earnings evolve during a graduate’s career. This is important if we are to understand not only the extent of the graduate wage premium, but also how earnings vary across different types of graduates.

Author Bio: Anna Vignoles is Professor of Education, Jesus College, University of Cambridge


Leave your comment