Free Leo! Why Starmer should release apprentice data

Are higher apprentices better off than graduates? The government has the data to tell us — but won’t.

YouTube videos discuss merits of apprenticeships vs university. But the government is oddly reluctant to share the information it holds about salary outcomes.

If you’re 18 and weighing up university versus a degree apprenticeship, you might reasonably expect the Department for Education to help you compare. After all, we’re talking about the two main pathways for ambitious school-leavers: one that leaves you with £50,000 of debt, one that pays you while you train. And Keir Starmer has this week been bigging up apprenticeships.

If he was serious, he could transform the debate by releasing detailed salary data held by HMG. It would, I suspect, show apprenticeships offering far better long-term prospects than a lot of degrees. I say ‘suspect’ because while I managed to code up a graduate salary database reader over the summer - the so-called LEO database - apprenticeship data is not included in that redacted release. If we had the full picture, then debate would be transformed. I said so on Twitter yesterday and was (to my surprise!) asked on to Radio 4 to discuss this nerdy point about LEO database access. So someone, somewhere must recognise what a potentially revolutionary change could be made by pushing for open calculations.

An inequality of information

Leo is a classic example of something that sounds dull and technical but has the power to change futures, prospects and debates. The technology and data exists for a kid predicted, say, BCC at A-Level to plug the grades into a computer and be told in an instant…

• Uni courses that last year accepted students with that level or more

• The completion rates of that course

• The salary of graduates of that course three, five and ten years after

• How this compares to a range of other options, like apprenticeships

• And how all this adjusts for your age, area and social background.

But while this info exists, it is not released. For reasons I’ve explained elsewhere, government’s instinct is to withhold information. There was a while when I wanted to take a year off trying to develop this. I noticed that the kids at my children’s (private) school were given stellar information: careers advisers, university advisers, parental networks: the works. But the pupils I meet in the speaking that I do for deprived state schools had none of this coaching. This is a real inequality of information, and it will of course affect the decisions kids make.

The Leo revolution - and how it was thwarted

The architects of LEO knew they were working on something revolutionary. Marry the tax data with education data and you have a rich, complex database - the world’s first - that shows exactly how education and training correlates to salary. Once completed, every child would know. No longer would those steered towards student debt be buying a pig in a poke, or fooled with misleading HMG claims that “the average graduate earns £100k more over a lifetime”.

But a great many people want this information kept quiet: mainly for what it says about the prospects of certain degree (like mine: history). The university lobby is powerful - you can argue that HMG is the university lobby - so only a redacted version of Leo ever became available. And, even then, as an unreadable data dump. The Leo revolution was thwarted to the dismay of those who worked on creating this now-buried treasure trove. One was LSE’s Tim Leunig, now chief economist at Nesta who says:-

‘When we created the database, I always assumed and intended that it would be available to sixth form students, along with equivalent data for other routes Free Leo!

The case for freeing Leo sounds obscure. But what might it show? Taking what fragments of information we have, I’ve had an estimate. It can be no more than that. But LEO would tell us exactly. Here is my maths:-

How apprenticeships could pay £240k more

HMG likes to use cumulative earnings and say a degree pays you £100k more over a lifetime. I disagree with this method: no one thinks in terms of cumulative earnings. But if you’d deploy this method to apprenticeships vs uni, then what? By my rough calculations, a Level 5 apprentice could be £200k better off than a typical graduate by age 40, once you factor in earnings during training. Even £240k, if you include debt. But we can’t be certain, because the government publishes the data in a way that makes direct comparison almost impossible. So the below is an illustrative example of the kind of chart that the government could publish for real.

Given how important the question is - how many young people’s futures may hang on it - has anyone tried to answer? Or do a better job than my rough stab? Here’s what we know, what’s been tried, and what it would take to get an actual answer.

What LEO and FEO tell us…

The Department for Education tracks both graduates and apprentices into employment using linked tax records. The underlying system is called LEO (Longitudinal Education Outcomes) and it’s genuinely impressive: school records linked to university records linked to HMRC earnings data linked to DWP benefits data. You can follow someone from their GCSEs to their payslip fifteen years later. It’s amazing that the UK has (at great expense) put this dataset together. And infuriating that it won’t open it up.

The outputs are published as two entirely separate releases:

For graduates: “LEO Graduate and Postgraduate Outcomes” is a redacted version of the file which gives median earnings at 1, 3, 5 and 10 years after graduation, broken down by subject group (using 35 CAH subject codes), institution, prior attainment, gender, ethnicity, and region. The most recent data covers 2016/17 graduates measured in tax year 2022/23. The Times has published a LEO reader here.

For apprentices: “Further Education Outcomes” (FEO) gives median earnings at 1 and 5 years after completion, broken down by level (L2-L5+), sector subject area (using 15 categories), age, and region. The most recent data covers 2020/21 achievers measured in tax year 2021/22.

Spot the problems? The measurement years in the redacted, publicly-released figures don’t match (8pc inflation between them). The subject classifications don’t align. The cohort years differ. And crucially, there’s no control for prior attainment — a Level 6 degree apprentice at Rolls-Royce is being compared raw against an English Literature graduate from a mid-ranking university, with no adjustment for the fact that they probably had very different A-level grades. LEO can adjust for grades. This is the killer: how do kids with similar school attainment fare in apprenticeships vs uni? Is the debt even remotely worth it?

You can query even the redacted LEO database to look at salary by school attainment for all graduates (below). It makes you wonder if school attainment is a more powerful predictor of later earnings than university.

The story so far…

The best academic work on apprenticeship returns comes from the Centre for Vocational Education Research (CVER) at the LSE, particularly Chiara Cavaglia, Sandra McNally and Guglielmo Ventura. Their 2020 paper “Do Apprenticeships Pay?” used the full linked ILR-HMRC dataset to track the 2003 GCSE cohort to age 28.

They found positive returns to apprenticeships but their comparison group was people with the same qualification level who didn’t do an apprenticeship, not graduates. So a Level 3 apprentice was compared to someone with Level 3 qualifications from college. This tells us apprenticeships beat the alternative vocational route, not whether they beat uni.

Crucially, the paper notes that sample sizes for Level 4+ apprentices “were not large enough to extend the analysis.” Higher and degree apprenticeships have only scaled recently (starts tripled between 2015 and 2020) so there simply wasn’t enough data to study them properly.

The IFS, meanwhile, are the LEO masters. They have done extensive work estimates of lifetime earnings by subject and institution. Their headline finding is that 80pc of graduates gain financially from attending university, while 20pc would have been better off not going. But the counterfactual is “not going to university” — which includes both apprenticeships and doing nothing. They haven’t published a direct graduate-vs-apprentice comparison.

So we have detailed research on whether apprenticeships beat other vocational routes (yes), and detailed research on whether degrees beat not having a degree (usually). What we don’t have is research on whether a degree apprenticeship beats a traditional degree for similar students.

The data we need

The data to answer this question already exists. It sits in the ONS Secure Research Service, accessible to accredited researchers (journalists tend to be refused such permission). The components are:

1. ILR (Individualised Learner Record): covers all apprenticeships

2. HESA: covers all higher education students

3. NPD (National Pupil Database) includes GCSE and A-level results

4. HMRC earnings data: P45/RTI employment records

5. DWP benefits data

A researcher with SRS access could, in principle:

• Take two cohorts: (a) students who started degree apprenticeships in 2017/18, (b) students who started traditional degrees in 2017/18

• Match on prior attainment (A-levels), gender, ethnicity, region, family background (using FSM as a proxy)

• Measure earnings in the same tax year (say, 2022/23, five years after start)

• Control for sector/subject as far as the different classification systems allow

• Produce genuine like-for-like estimates of the earnings premium (or penalty) from choosing the apprenticeship route

This is exactly the methodology the IFS used to estimate returns to university versus not attending. The same approach could compare university versus apprenticeship.

So why hasn’t it been done?

Three possible explanations

1. It’s a new question. Higher apprenticeships only became a serious alternative to university around 2017. The first cohorts with degree apprenticeships are only now reaching the 5-year measurement window. Maybe the research is coming.

2. Nobody’s commissioned it. The IFS graduate work was funded by DfE. It was intended and used by the department as pro-university advertising - even though it actually found one in five courses was uneconomical. The CVER apprenticeship work was funded by DfE via ESRC. If DfE wanted this comparison, they could commission it tomorrow. They haven’t.

3. The answer might be embarrassing. If higher apprentices genuinely do earn more than graduates in comparable fields, it raises awkward questions about the value of the £50,000 student loan, the expansion of low-return degrees, and the entire structure of post-18 funding. Sometimes questions don’t get asked because people don’t want to know the answer. And if it leads to fewer students taking ropey uni courses, those unis would face a funding crisis and go asking for bailouts. Ministers don’t want this headache.

Also, the idea that more education is always better is a political creed that ministers often don’t like to challenge with now-available evidence. In the late 1990s, the government ended up paying teenagers to stay in sixthform vis the Educational Maintenance Allowance (EMA). It cost a bomb. The LEO data is now so good that it can trace those kids through the tax systems. It shows the EMA lot ended up ended up actually paid less than those who dropped out of school age 16 and entered the workforce early. It turns out that getting a couple of Ds in A-Level doesn’t do much for your earnings. Who’d have thunk? This finding - that EMA cost billions and made the kids poorer - was reported by no one when it came out earlier this year. Not even in Scotland and Wales, where this policy still exists. It’s odd: ministers talk about parity of esteem with graduates vs vocational training, but the state system is one big university sales machine.

What Starmer can do (in one email)

If Keir Starmer wants parity of information for apprenticeships - and to give all pupils of all backgrounds the best information to decide their next steps, he can…

Commission the research. Ask the IFS or CVER to run a proper controlled comparison using the existing linked data. This would cost perhaps £100,000-200,000 and take 6-12 months. Given that we spend £20 billion annually on student loans, this seems worth doing.

Publish unified statistics. Align the LEO Graduate and FE Outcomes publications so they use the same measurement year, comparable cohorts, and harmonised subject classifications. This wouldn’t give us causal estimates, but it would at least let people make informed comparisons without manually adjusting for inflation and guessing at sector mappings.

Or just publish the full LEO dataset. Rather than license it to special people for special circumstances, publish the full LEO dataset - anonymised, of course - so organisations like The Times can get to work and give people the comparability they need. We have the resources and the professionalism. We just need the data.

Until Starmer moves on transparency, we’re left with a situation where the most consequential financial decision an 18-year-old makes — university or apprenticeship — cannot be properly evaluated using official data. The information exists. It’s just not being published in a useful form.

I’ve submitted an FOI to DfE asking whether they’ve conducted any internal analysis comparing graduate and apprentice earnings using the linked LEO dataset. I’ll report back on what they say.

In the meantime, if you’re choosing between a degree apprenticeship at a good employer and a middling university degree, the incomplete evidence suggests the apprenticeship is probably the better bet financially. But I’d feel more confident saying that if the government would actually crunch the numbers. If Starmer would release the full dataset, other would do the work for him. Sending such an email would cost him nothing.

Employers are already paying for apprenticeship training via the levy. The government is spending billions on student loans. Yet ministers won’t publish the data that might help young people choose the option that costs the state less. It makes no sense. And yes, some students go to university for non-economic reasons, never expecting to be able to repay the loans. Let them continue to do so: but with their eyes open.

If I were a backbencher wanting to fight for the pupils in my constituency - their right to know about earnings outcomes of the options being sold to them - I’d press the Prime Minister to release the files. There are some amazing prospects out there via vocational paid-to-learn course where you earn from your first hour of your first day and they lead to well-paid, fulfilling and well-regarded non-graduate jobs. This should not be a dirty secret. A Labour government, surely, should do all it can to share information about the value of labour.