Do higher level vocational qualifications lead to higher earnings than degrees?

The Prime Minister's comments on an earnings boost for higher technical courses need to be caveated. David Kernohan explains.

We can’t be as sure about the earnings benefits to higher technical study as the Prime Minister might think.

His argument comes from a National Institute of Economic and Social Research (NIESR) study published back in 2019, which  finds that in general there is an early advantage in holding sub-degree technical qualifications as a highest qualification, but that as time passes a degree is of more value. In a very small number of cases overwhelmingly male students who study STEM vocational courses earn more than overwhelmingly female students who study care, education, and business. It would be simply incorrect to claim that the average higher technical qualification holder earns more than the average graduate. But even the specific cases need further explanation.

Enter the lion

However there is also one really big problem with the findings. Or rather, a small one.

If you saw “graduate earnings” and thought “LEO”, you are dead on. The data underpinning the report is derived from the Longitudinal Educational Outcomes (LEO) data set, looking at 2002-03 school leavers. The usual LEO effects apply, namely:

  • As LEO does not understand the concept of part-time work, salary figures will understate the full-time equivalent earnings of part-time workers – which, overwhelmingly at these ages, tend to be women
  • There is also a gendered effect – women (still) earn less than men for the same work
  • There is a prior attainment affect – good pre-tertiary results link to good career outcomes
  • There is a socio-economic background effect, closely linked to the prior attainment effect.
  • There is a regional effect – some regions (indeed, some localities) where young people may live are linked to lower earnings.

We know all this from LEO, and it all applies here. The econometric study – which is competently done via a machine learning technique that rejoices in the name of LASSO – controls for gender, work experience, ethnicity, free school meals eligibility, region (of domicile, not current residence), IMD at a very low area size, GCSE results, subject area of qualification, and type of school . That’s a long and impressive list of variables, and – indeed – the kind of thing I’m usually quite keen on.

With a small “n”

But you can tell I’m hedging here so I may as well bring the central problem with this study into the open – there are a very low number of people with a level 4 or 5 (tertiary, sub-degree, vocational) qualification as their highest qualification. The overall sample size is around 622,000 – but the sub-group we are interested in contains only about 9,000 people (1.5 per cent of the total). This is being compared with 25.35 per cent of the sample – 157,700 people with an undergraduate degree.

The problems of comparing information about a large group of people with information about a small group are legion, but the main problem is one of outliers when using median and modal averages. For example, a small number of very high earners is likely to be found in both groups – earnings, frankly, that may or may not have anything to do with their study choices at 16 or 18. In a larger group the rarity of these cases will mean that there is little impact on statistics about the whole group – but the same number of cases will have a significant effect on a smaller group.

There might be other strangenesses in the smaller group compared to the main sample – for example, those who have a level 4 or 5 vocational qualification as their highest qualification are more likely to be White than a graduate chosen at random, more likely to be male, more likely to be older, and less likely to have free school meals eligibility than a non-Russell group graduate. These are controlled for, but as controlling means disregarding some of the data this exacerbates our small numbers problem.

So there is some evidence that there might be an earnings premium for higher technical qualifications in some limited cases, but we don’t have enough data to know for sure. I know it feels odd to doubt the words of Boris Johnson, but here we are (again).

What about higher apprenticeships?

There’s surprisingly little data on salaries expected by apprentices at any level. I found this experimental release from 2016, covering five years of data, which suggests someone studying on a higher apprenticeship (level 4) between 2010 and 2013 could expect to earn £23,600 four years later – less than most graduates five years after graduation.

Leave a Reply