Student loans are about to look a lot more expensive for the taxpayer

But something else is going on that is likely to mean further pressure on the “value” that governments are getting from their notional subsidies in the four (devolved but not devolved) schemes.

Student loans in UK public accounts are treated as a financial asset. When a government issues a loan, it records an asset because it has a contractual right to receive repayments in the future.

The accounting problem is that repayments depend on graduates’ earnings and policy rules over decades, so the asset can’t be valued just by adding up what was lent out.

Instead, the four sets of accounts estimate the present value of future repayments – forecast what will be repaid each year, then discount those future cash flows back to a value “today” using an HM Treasury discount rate.

Changes to the forecast of repayments, or changes to the discount rate, can shift the reported value by very large sums without any cash moving. And they can make each of the four schemes look more or less sustainable as a result.

That is why the modelling matters. In England the Department for Education (DfE) runs an earnings and repayments forecasting pipeline that is used both for policy and financial planning and to value student loans in its annual accounts.

If the model’s assumptions are too optimistic (for example, it expects higher earnings growth or higher repayment rates than reality), then future repayments are overstated and the loan book looks more valuable than it really is. If the assumptions are revised downward, the loan asset value falls – and the annual “charge” for loans not expected to be repaid rises. That makes the system look a lot more expensive.

There can’t be many people looking at the Welsh and Scottish accounts and audit reports – fewer still the sections on student loans – but both sets of docs flag a gap between (a) an older approach used in devolved accounts and (b) a newer DfE-run (or DfE-aligned) model that is “in the system” but not yet fully adopted in those devolved valuations.

The modelling matters

In Wales, Audit Wales says the Auditor General qualified the 2024–25 accounts because they could not obtain sufficient appropriate audit evidence to conclude that the student loan asset valuation was reasonable, and it links this directly to continued use of an “ageing” model rather than the newer approach.

Audit Wales also says that actual repayments are around a whopping 50 per cent of those forecast under the existing model, and that adoption of the new model is therefore likely to produce a substantial downward revision of the loan asset value in 2025–26 – as documented in the Welsh Government consolidated annual accounts. Ouch.

Meanwhile in Scotland, the Audit Scotland annual audit report says the Scottish Government’s student loan balance is valued using a valuation model run by DfE, notes DfE’s view that methodologies within the model may over-forecast repayments (which would understate impairment and write-offs), and then says that DfE is in the process of providing a new model expected to be used for the 2025–26 accounts.

The practical meaning is the same as in Wales – if the newer model embeds lower expected earnings and repayments, the accounting valuation of the loan book drops, and the write-off/impairment charge rises.

The reason this shows up as a “squeeze” on devolved governments is that these valuation movements feed into budget control totals even though they are non-cash. If a government suddenly has to recognise that a larger share of its loan outlay will never be repaid (or has to revalue the asset down because the model changes), the accounting charge increases in that year and can breach budget limits.

As a result, a model update can translate into a large in-year budget problem, even if student finance policy has not changed and cash spending is unchanged.

If you follow this stuff, you might be thinking “who cares?” The Treasury runs the four schemes, and even though this stuff has to be recorded in the devolved accounts, the whole point of “annually managed expenditure” is that it takes the hit on the ups and downs generated by changes to inflation, or the discount rate, or whatever.

You might be right. But there’s three significant potential implications.

Knock on impacts

Under HM Treasury’s devolution settlement, Scotland, Wales and Northern Ireland can run their own student loan systems, but only on an “equivalence” basis.

That means the loan scheme in each nation must be broadly similar in overall cost to the English system, judged on the structure of the loans themselves rather than the wider student support package.

Individual parameters do not have to match England exactly, but the combined effect of repayment thresholds, interest rates, repayment periods and write-off rules must not be more generous on balance.

If a devolved government chooses to go beyond that, the additional cost has to be met from its own budget rather than being funded through UK Government AME. HM Treasury allows time for devolved governments to move back towards equivalence if changes are required, and will generally provide cover while that happens.

In budget terms, loan outlay is AME, while the expected loss on new lending (the RAB charge) and revaluations of the existing loan book affect resource budgets, with Treasury saying it will cover revaluation movements within a reasonable range – but expecting devolved governments to manage the consequences if costs move materially beyond that.

I’ve noted the lack of transparency over any assessment that HMT makes over devolved nations’ equivalence before – but we do know that Wales was already very close to, or already has, maxed out its credit card via its more generous maintenance offer.

In other words if England’s system gets meaner, or if DfE revises its student loan forecasts modelling such that graduates are paying less back than previously thought, changes are pretty much required to recover more money from graduates.

You can see why the devolved governments (and their auditors) might want some additional assurance. Scotland’s annual audit report says that for 2025/26, it intends to obtain assurance from the National Audit Office over the statistical model.

More generally, once all of this feeds through to the RAB (ie subsidy) charge and/or the amount that is declared as being written off, it generally makes the system look much more expensive to the taxpayer.

In all four nations, that places pressure on education budgets in general and intensifies discussion about the appropriate cost share between graduates and the state – which you can only really control by making terms more hostile, or by having fewer students take part.

We are where we are

But the third implication then gets even more interesting. If graduate repayments aren’t close to the old modelling, that could be because of things like the discount rate, inflation, and other economic factors that aren’t really about them per se.

But it could also be about the subject mix, or the jobs they are getting, or the skills they have, or their prior social background. The point is that the alarmingly low rates of graduate employment among students on private franchised provision, for example, isn’t just a quality signal – at some point it feeds through into the sustainability of the system itself. And then someone has to pay.

The Telegraph version of that would be “graduate student loan terms are worse because too many people are going to university”, but the grain of truth is that by giving itself the power to vary the repayment threshold, governments really can compensate for crap outcomes in one bit of the system by making other graduates pay more.

You might argue that from a student point of view and from a devolved nation point of view, the assessment of the split that’s made between state and graduate is the one that ought to apply when the loan is made. You might also argue that being able to retrospectively punish nations or graduates when things don’t pan out as intended is morally wrong. You might then argue that not correcting for forecasts that undershoot where inflation ends up (impacting how much students can borrow to live on), while very much correcting for forecasts that undershoot how much they’ll pay back (by doing things like freezing repayment thresholds) is quite the cheek.

But we are where we are. And where we are is that the “cost” of the system is about to look a lot higher than previously.