A bargain the government is proposing to throw away

In March 2026 the Department for Education proposed to stop funding most categories of specialist assistive software through the Disabled Students’ Allowance, paying for it only “in exceptional circumstances,” on the grounds that free and built-in tools now do the job for most students.^[1] The consultation on that proposal closed in June, and the decision is expected in the autumn. It has not been made yet. This is not an argument about whether disabled students deserve support. That case is well made by people better placed to make it. The point here is narrower and harder to wave away. The government has a binding method for decisions like this, and when that method is applied honestly, the proposal does not stand up. What it would scrap turns out to be one of the better bargains in the budget.

The test the government sets for itself

The Treasury’s Green Book is the rulebook for appraising public spending. It judges a proposal by its cost set against its wider benefit, discounted to the present at a prescribed rate, with the resulting net present value the primary basis for deciding whether government action is justified.^[2] A saving that creates higher costs downstream is not really a saving. The government made that same point itself only a few weeks ago, in its review of young people not in education, employment or training: it spends far more managing disengagement than prevention would cost.^[3] Prevention beats repair, in its own words. The test, then, is the government’s own. Whether this proposal was ever held to it is the only real question.

The saving

Start with the saving. The first surprise is that there is no government figure for it. The consultation talks throughout in the language of value for money, yet nowhere does it say what the change actually saves, per student or in total. I asked, under freedom of information, for the derivation of any per-student saving. The department holds no such information.^[10] The central justification for the whole policy, the money it saves, has never been quantified.

Nor can the agency that pays for it produce the figure. I asked the operator of the scheme, Student Finance England, directly, for the total spent on DSA assistive technology over recent years. It refused under the cost limit in the Act: isolating that spend would mean manually checking tens of thousands of student accounts, because assistive-technology software is not recorded apart from other equipment, and the training is logged under another heading again.^[11] So the spend on the very thing being withdrawn is not held in any readily usable form, by the department proposing the cut or by the body that funds it. Asked separately to substantiate the “value for money” its 2024 reforms were reported to deliver, the same body was clear that those gains came from the wider support model rather than the cost of the technology, and that assistive-technology costs are if anything rising.^[11]

So it has to be estimated, and I will estimate it generously, against my own case. Specialist software is sold at DSA prices, which are not published, usually as a package that includes training. The government’s average spend per student is simply the total DSA budget, £203m, divided by the 88,000 students who use it, or about £2,300 per student.^[1] But that covers much more than assistive technology, so for this analysis I will use £1,000 per student: a deliberately generous figure, at the top of any plausible range. I have erred high on purpose, since a smaller figure only shrinks the break-even and strengthens the case. It is a one-off cost: the licences are bought once, at the start of a course, and the budget renews each year only because a new cohort of students arrives each year.

The benefit

The benefit at risk is what a graduate returns to the Exchequer, and here too the figure is the government’s own. The Institute for Fiscal Studies, commissioned by the department, put the net return on an undergraduate degree at around £110,000 for the average man and £30,000 for the average woman, in present value on the Treasury’s discounting basis.^[4] Rather than pick one, I have weighted the two by the government’s own record of who actually receives DSA, which is 68.4% women at undergraduate level.^[1] That gives a single honest average of about £55,000 per recipient. These are conservative numbers. As averages across everyone who enrols, they already allow for students who drop out or never earn enough to repay.

The break-even

The whole argument fits in one line. The share of affected students whose outcome has to change for the proposed cut to break even is simply the saving divided by the benefit:

£1,000 ÷ £55,000 ≈ one in 55

Below that, the country loses money. Written as a number, then, the proposal rests on a single claim: that funded specialist support, for students already assessed as needing it, changes the eventual outcome for fewer than one in fifty-five of them. You need not even accept my £1,000. Even the £2,300 whole-allowance average pays for itself at one in 24.^[1]

Turn it the other way and the bargain is obvious. The Exchequer return on a single student who stays the course covers the specialist support given to all fifty-five. Not many things the state funds clear the bar by that distance.

Is one in fifty-five plausible?

The case that it might be rests on disabled students now sitting close to parity with their peers on continuation and attainment.^[5] But that parity is unlikely to be a baseline the support just happens to sit alongside. It is far more likely to be what the support produces, and the proposal pulls out one of its inputs without modelling what then happens to the gap. The change also lands on a particular group. Its single biggest element, text-to-speech software, names them: students with dyslexia, ADHD, autism and mental-health conditions.^[1] That is roughly two-thirds of all DSA awards,^[6] and it is the same cohort that the government’s own NEETs review identifies as the fastest-growing driver of young people falling out of education and work.^[3]

Two pieces of evidence identify the real effect, and they point in the same direction. The department’s own 2016 analysis found DSA recipients running two to three points above the adjusted average for attainment and progression, and unsupported disabled students two to three points below it.^[7] That is a gap of four to six points, several times the one in fifty-five the proposal needs.

The closest thing to the model this proposal drifts toward, with support left to institutions and unfunded at the level of the individual student, is the United States. There the gap in degree attainment between disabled and non-disabled adults sits at around twenty points, and has barely moved in years.^[8] That figure is directional rather than decisive. It is an attainment gap rather than a continuation rate: it reflects who enrols as well as who completes, not the effect of support on any given student. The United States also has no clean national measure of completion by disability, which is part of why an attainment gap is the closest comparison available. Even so, to believe the support changes outcomes for fewer than one in fifty-five, you would have to put almost none of that twenty-point gap down to support of this kind. You can see where this road ends, and it is not parity.

Did the department do the work?

As part of the consultation, the Department for Education produced an equality assessment and a pricing mechanism. It did not produce the half that matters. It never weighed the saving against the benefit it puts at risk. I asked, under freedom of information, for any appraisal of the saving against downstream cost, any modelling of the effect on retention or completion, any derivation of the saving, and any estimate of the numbers affected. The department confirmed in writing that it holds none of it.^[10] Asked what lay behind the claim that free tools are comparable, it confirmed it had set no accuracy threshold. It had judged the tools equivalent because they are widely available and used, not because they had been tested.^[9]

The safety net is in the same state. The “exceptional circumstances” provision, meant to catch the students whose free tools fail, is undefined. The consultation actually asks respondents what it should be, and the department has no estimate of how many students would qualify.^[1][9] Nor can it claim it did not know the risk. In 2016 it ringfenced assistive software as too essential to move, and wrote in its own equality analysis that withdrawing such support could lead students to withdraw from their course or not enrol at all.^[7] The 2026 proposal comes for exactly what it protected then.

What this is really about

The pattern is what matters. This is the same government that, only weeks earlier, was capable of months of careful, bottom-up analysis in its NEETs review, yet has applied none of it to a proposed cut affecting a protected group. Almost everything needed to do the job properly is already to hand, and most of it belongs to the government: the appraisal method, the discount rate, the benefit figures, the evidence that the support works. The one missing step is the one the Green Book exists to require, which is putting them together.

There is also the order of it. The defensible sequence is to work out whether a saving exists, and then consult on the cut. Here the question went to public consultation first, through the exam term and the months applicants spend deciding whether to enrol at all, before the appraisal that would show it was worth asking. The alarm it creates is itself a cost, and it is exactly the kind of thing the appraisal exists to weigh before it is run up.

A government is entitled to make a cut. It is not entitled to make one on a group the law protects, without the appraisal its own rules demand, and then call the result a saving. The proposed cut rests on a single unstated bet, that this support changes nothing for at least 54 of every 55 students who get it. The government has not shown the bet is sound. It has confirmed that it never worked it out. The decision is still to come, and on the figures the government has so far, this was never a saving at all. It is one of the cheapest things the state buys, and one of the best, set to be scrapped on a sum nobody has done. It is a cost the appraisal has yet to count.

Disclosure: I co-founded a company in the assistive-technology sector and have since exited it, so I come to this with an interest, which I state plainly. I am trained as an economist; what I have done here is apply the government’s own appraisal framework rather than assert any authority of my own. The argument stands or falls on the figures, all of which are the government’s.