Billions Spent Making Disabled People, Parents, And Apprentices Unemployable

An employer offers a job in February. The candidate is deaf and needs an interpreter for perhaps six hours a week. There is a government fund for exactly this situation, and she applies the same afternoon. The interpreter is not funded in March. Nor in April. Nor across the summer.

Somewhere around the thirtieth week of waiting, eight months later, the vacancy has either gone to somebody else or been quietly redrawn into a role which does not involve meetings.

The fund is called Access to Work. Preventing what has just happened is its entire purpose.

The money exists, was voted through Parliament, and sits in a departmental budget with her name notionally attached to some of it. The failure occurs in the gap between the promise and the possession, and Whitehall has never taken much interest in what happens there.

British government has become highly skilled at creating entitlements and almost wholly uninterested in whether anybody can practically use one.

106 Working Days Instead Of 25

The Department for Work and Pensions gives itself a target of twenty-five working days to process an Access to Work application. Five working weeks to decide whether a disabled person may have a chair, a screen reader, a taxi account, or a few hours of human assistance.

Average processing hit 109 working days in November 2025 before easing to around 106. Applicants ringing the helpline are told the wait may run to thirty-seven weeks.

Not everybody waits thirty-seven weeks, and urgent starters are prioritised. Thirty-seven weeks is nonetheless the figure the department itself thinks worth warning people about, and it is almost longer than a pregnancy, longer than all probationary periods, and considerably longer than the patience of a small employer who has held a desk empty since spring.

The queue tripled in four years.

The department's own estimate for restoring a functioning service runs to between eighteen months and two years, with September 2027 as the target for clearing the backlog. It amounts to a candid admission the current cohort will have to make other arrangements.

The Public Accounts Committee heard the utterly predictable consequences: people could not start jobs; offers were withdrawn.

Citizens Advice submitted cases of employees risking their health or their employment while waiting for support to arrive (which is a polite way of saying they went to work in pain, or in silence, or in the dark). SignHealth described onboarding disrupted. And employers grew warier of hiring disabled candidates at all.

A programme created to remove the employment disadvantage attached to disability had begun manufacturing a fresh one. The Committee said so in terms, describing a support system operating as a barrier to work.

More Staff, More Money, = Longer Queues

The obvious explanation is understaffing, and it is wrong. DWP had 247 full-time equivalent staff on Access to Work in 2021–22. By 2024–25 the figure was 588, reaching 648 by March 2026. Administrative staff costs rose in real terms from £12.6m to £24.4m across the same stretch.

Performance collapsed anyway.

Another 480 staff were announced after the PAC hearing, which is the traditional Whitehall response to a process problem: apply more people to it and hope arithmetic does the rest.

More than double the staff. Nearly double the administrative spending. Three times the backlog.

PAC noticed the arithmetic did not work and asked whether the real culprits were unsuitable IT, poor administrative process, and an absence of grip.

DWP could not say how many cases a case manager can actually handle in a day, which is a remarkable admission from an organisation whose entire function is processing cases.

A factory which cannot state its output per worker per shift is not a factory with a staffing problem. It is a factory with a management problem, and hiring 480 more people will simply produce a larger version of whatever is currently going wrong. As everyone with any common sense knows.

A Form Which Demands the Abilities It Exists to Compensate For

The application process itself is, in the Committee's assessment, frequently inaccessible. Applicants described complex forms, repeated requests for the same evidence, long email chains, unclear deadlines, unanswered correspondence, and being passed between officials.

Which is to say the scheme requires precisely the administrative stamina some applicants are applying for help in compensating for.

A person seeking support with concentration, communication, correspondence, or executive function must first demonstrate an ability to sustain a nine-month bureaucratic campaign against a department which does not reply.

Anyone designing a system to filter out the people it was built to serve would struggle to improve on it.

The scheme is also built around the notion of a labour market which has been actually dissolving for twenty years.

Evidence to PAC suggests it is constructed around stable full-time employment, and works far less well for freelancers, the self-employed, part-time workers, and anybody with fluctuating earnings.

Whitehall has spent two decades celebrating flexible work while running its disability employment support on the assumption every applicant has one employer, one contract, and one set of predictable hours. Which anyone with a set of eyes or a working brain can see from a look at a single job board website.

Demand Doubled and Nobody Noticed

Applications rose from roughly 76,100 in 2018–19 to about 157,000 in 2024–25. Spending went from £163m to £321m in real terms, with £517m forecast by 2029–30. In 2024–25 the scheme supported 74,200 people, and individual awards can reach £69,260 a year.

The composition changed: recipients with mental health or learning conditions went from 11,200 to 37,900, and now make up more than half the caseload. This, of course, mirrors issues inside the SEND debacle.

Cases often need individual judgement rather than a catalogue number for a piece of equipment, and they are genuinely harder to decide.

DWP concedes it did not immediately recognise the caseload had changed character. Its processes had been built around a different sort of applicant and a different sort of support, and they were left running while the population passing through them turned into something else.

Demand doubled in six years, in full public view, inside a department which publishes the figures itself, and the operating model was not rebuilt to match.

DWP told the Committee earlier efforts to speed up processing had been undertaken for good reasons and produced the wrong outcomes, because the scheme's principles were not being applied consistently. Some people renewing support then received smaller awards despite no change in their circumstances, on the basis the earlier decisions had apparently been wrong.

The department is therefore stuck between deciding quickly and deciding consistently, and is currently managing neither.

A disabled worker whose award is cut on renewal is not receiving a correction. She is receiving the bill for somebody else's process failure.

Even the payment system is backlogged. Outstanding payment requests stood at 31,700 in March 2025, against 6,900 in 2022. Complaints ran to 800 in the first six months of 2025–26 alone, having totalled 234 across the whole of 2022–23.

Approval is not the finish line. Approval is where the second queue begins.

40% Of Claims Make Up For Employer Laziness

The other half of the scandal is less discussed, because it embarrasses everybody. Access to Work is meant to fund support over and above the reasonable adjustments employers are already legally obliged to make.

DWP reviewed 200 cases in 2025 and concluded roughly 40 per cent of the support it had funded should properly have come from the employer, the employee, or the NHS. In around 18 per cent of the sampled cases, employers had failed to make the adjustments the law already required of them.

DWP gave the Committee examples of large employers applying to the taxpayer for ergonomic chairs, items the department says such an employer should simply buy.

It also found cases where funded job aides were effectively doing work an employer would otherwise have hired a second employee to do, which the department itself described as misuse.

A £2,000 adaptation is a rounding error for a bank and a serious blow to a three-person shop, and any sane scheme accounts for the difference. Discretion is not the problem. Discretion without consistent guidance is, and case managers have asked for clearer rules, which suggests they know they are improvising.

The upshot is a scheme which can leave a legitimate applicant waiting nine months for an interpreter while processing, deciding, and paying for a chair a FTSE company should have ordered from a catalogue. Every misdirected award consumes money and staff time which could have gone to the queue.

£321 Million And No Idea If It Works

Access to Work operates a principle whereby funded support must represent value for money. DWP cannot establish whether Access to Work represents value for money.

Its meaningful outcome research dates from 2009 and 2018, and is largely qualitative. It has not built a quantitative account of benefit payments avoided, additional tax collected, employment retained, employment gained, sickness absence reduced, time spent in work, or the comparative usefulness of different types of support.

The departmental defence is genuinely reasonable as far as it goes: a proper controlled trial would mean deliberately denying support to a group of disabled people in order to observe what happens to them, which is not an experiment any decent department should want to run.

It does not begin to explain the rest.

• Administrative data exists.
• Matched cohorts exist.
• Historical comparison, phased rollout, natural experiments, and economic modelling all exist, and government uses every one of them elsewhere when it suits.

DWP has not attempted anything as elementary as modelling the benefit expenditure avoided or the tax receipts generated by keeping somebody in work, which is the exact calculation the Treasury would demand from any private applicant for public money.

The state is spending hundreds of millions a year (heading towards half a billion), on a programme whose aggregate employment and fiscal effects it has never bothered to quantify, while requiring every individual grant within it to pass a test the programme itself would fail for want of evidence.

Ministers meanwhile continue to announce that more disabled people must move into work.

PAC found DWP had not treated the scheme as a priority despite demand rising in front of it. Policy says get a job. Administration says allow up to thirty-seven weeks.

How An Incentive Becomes a Deterrent

A small firm interviews two capable candidates. One can start Monday. The other may need an interpreter, adapted transport, or workplace support routed through a government process of unknown duration.

The problem here is obvious, regardless of the company's good intentions.

The employer must then work out for one candidate:

1. Whether funding will be approved;
2. When it will arrive;
3. How much must be fronted meanwhile;
4. Whether the award might later be revised downwards, and ;
5. Which portion of the cost the firm is legally obliged to carry regardless.

Nobody can answer any of it with confidence. Not the applicant, not the employer, and on current evidence.... not the department.

Faced with an unanswerable set of questions and a second candidate who raises none, a manager under commercial pressure does the cowardly thing quickly and tells himself it came down to experience.

Even the capable disabled candidate understands why.

The law forbids the outcome.

The administration all but arranges it, and the Committee has the evidence of employers hiring fewer disabled people as a result.

The state has not removed the cost of hiring here.

It has converted the cost from money, which a business understands and can budget for, into delay and uncertainty, which it cannot.

What's the next iteration in this cycle?

If you're disabled, don't apply for help as it will hinder you in getting the job.

Pay the Nursery Now, Ask the State Later

Disability support is the sharpest case and by no means a lonely one. Universal Credit childcare exists to help low-income parents meet nursery fees so they can work. The default arrangement requires those parents to pay the bill themselves and reclaim up to 85 per cent afterwards, subject to a monthly cap, and only once the childcare has actually been provided and paid for. Paraphrased:

Government: we will help you pay for childcare so you can take a job. Parent: excellent. Government: settle the nursery invoice yourself first.

The Work and Pensions Committee said in 2018 the arrangement directly conflicts with the aim of moving people into work, since some parents face a choice between borrowing and refusing the job.

It repeated the criticism in 2022.

A remedy does exist, and its history is instructive.

The Flexible Support Fund can cover up to a month of childcare costs up front, paid straight to the provider and never repaid, for a parent starting work or increasing hours. The Committee's own examination found the fund poorly known to claimants and inconsistently promoted by work coaches, with the department unable to say much about how it was being used because the money is managed locally and the spending is not centrally recorded.

Written evidence noted the awkward detail of 12 per cent of the fund going unspent in 2015–16 while parents turned down jobs for want of a deposit. There is no entitlement to it. Payment sits at the discretion of an individual work coach, so the cure for a discretionary gap is a discretionary grant, and the parent must know to ask.

Around 177,000 households were receiving it in November 2024, amounting to 29 per cent of Universal Credit households with pre-school children where every claimant had earnings. A proportion which had barely moved in two years.

Roughly seven in ten eligible working households with young children were not claiming money set aside for them which is genuinely useful.

A parent weighing a job offer is not calculating the theoretical annual value of a benefit. She is calculating whether the nursery wants payment on Friday, whether wages land in four weeks, whether reimbursement follows later still, and whether the bank balance survives the same period of time.

The scheme solves the accounting problem with some elegance and the cash-flow problem only for those who find the right official on the right day.

Businesses Ran From State Handouts

The apprenticeship levy makes the same argument from the corporate end of the market. Government wanted employers to train more people, so it built a levy and a funding apparatus to make training cheaper.

Starts then fell off a cliff, and employers initially spent about 9 per cent of the levy funds available to them on new apprenticeships.

Parliamentary evidence has ground through the same complaints ever since. A Transport Committee report in January 2026 was still listing the levy's complexity and inflexibility among the reasons starts had fallen, still reporting employers frustrated by constraints on how the money may be spent, and still asking for flexibility, nine years after the thing arrived.

The government rejected the recommendation. Because, of course it did.

Ten thousand pounds obtainable through a cumbersome process is not worth ten thousand pounds.

It is worth ten thousand pounds minus the staff hours, the professional advice, the risk of getting the paperwork wrong, and the opportunity cost of the person doing all of it rather than running the business.

For a small company of nine people with no compliance department, the subtraction can consume the entire benefit.

The state measures the headline figure with great care and never once measures the deduction.

Whitehall's answer to employers leaving the money untouched arrived in April 2026, when the levy became the Growth and Skills Levy. Among the changes: funds now expire after twelve months rather than twenty-four, the 10 per cent government top-up has gone, and employer co-investment rises to 25 per cent once the pot runs dry.

Confronted with a subsidy too cumbersome to spend, the department has made it expire twice as fast. If there is any stupider thing than taxing growth and skills to pay for more of them, it will take much discovery.

Legal Standing Before Childcare Places Exist

The expansion of funded childcare belongs to the same family of dysfunction, with a variation. The National Audit Office found the 2023 expansion was announced at such speed the Department for Education could not complete the analysis needed to give full assurance on deliverability, likely cost, or expected benefit, and could not properly consult providers beforehand.

Those providers then warned about workforce capacity and poor information.

Announcement first, capacity later.

A parent holding a legal entitlement to a nursery place which does not exist has been given a document rather than a service.

The policy succeeds administratively and fails at the only counter which matters, namely the reception desk of a nursery with nothing available in the two-year-old group until next September.

Whitehall Creating Perverse Reverse Incentives

A useful discipline for reading any departmental press release: the advertised sum is where the calculation starts.

Take the headline subsidy.

Subtract:

1. The administrative burden of applying;
2. The financing cost of paying up front;
3. The delay before reimbursement;
4. The probability of rejection;
5. The compliance obligations attached to acceptance, and;
6. The uncertainty running through all of it.

What is left is the effective subsidy, and it is the only figure with any influence on behaviour.

In bad cases the effective subsidy is a fraction of the headline.

In the worst cases it goes negative, and the offer of help leaves the intended beneficiary less employable than silence would have.

An employer who works out hiring a particular candidate unlocks government support, but only after months of correspondence and unquantifiable risk, has been handed a reason to hire somebody else.

The support is real. The incentive has reversed.

Always Asking The Wrong Questions

Departments assess themselves against questions of impressive irrelevance. Was the budget spent? Were applications processed? Were awards made? Were entitlements created?

Every one can be answered triumphantly while the policy fails completely.

1. Childcare support exists and seven in ten eligible households do not draw it.
2. Disability support exists and arrives after the job has gone.
3. Apprenticeship funding exists and employers decline to touch it.
4. Funded childcare exists and there are no places.

Four programmes; four respectable sets of departmental statistics; four failures to move anybody in the intended direction.

The only test worth applying is whether the intervention made the desired decision easier for the person taking it. British politicians do not routinely ask, have not built the evidence base to answer, and would not enjoy the results.

A Good Idea With Zero Common Sense

These schemes need fraud control, eligibility decisions, safeguarding, financial accountability, and a clear line between what an employer must fund and what the state will. The 40 per cent of wrongly claimed accommodation is itself evidence of what happens when the line is policed too loosely, not too tightly. Some cases take longer because they are harder, and the growth in complex applications is real rather than an excuse. Abolishing the checks would produce even worse numbers.

Administrative friction is a cost of the policy, borne by the citizen rather than the department, and it belongs in the accounts.

A support programme should be judged not merely by what it pays but by what somebody has to do to receive it. Where the burden of claiming grows heavy enough to deter the behaviour the money was meant to encourage, the department has not delivered a subsidy.

It has delivered an obstacle course with a cheque at the end and no guarantee of finishing.

Britain now runs a whole class of these things. Schemes conceived to remove barriers, administered with sufficient delay, complexity, and uncertainty to become barriers themselves.

The money is voted. The entitlement is drafted. The press notice goes out.

And a woman who needs six hours of interpretation a week sits at home in week thirty-one, formally entitled to help, in receipt of nothing, while a department recruits 480 more people to explain the delay.