Benefits · Fraud, error and viral claims · Great Britain · 2025–26 (financial year ending 2026)

Fraud, error and viral claims: how much benefit is paid wrongly?

DWP estimates 3.2% of benefit spending was overpaid in 2025–26 (£9.9bn): 2.2% through fraud, 0.6% through claimants’ mistakes and 0.4% through DWP’s own. A further 0.4% was underpaid. The rate is below its pandemic peak.

Evidence last reviewed 3 October 2026

Three kinds of overpayment, and underpayments

Fraud, 2025–26: 2.2%; Inaccurate or missing information with evidence or suspicion of fraudulent intent.
Claimant error, 2025–26: 0.6%; Inaccurate or missing information without fraudulent intent.
Official error, 2025–26: 0.4%; Mistakes or delays by DWP, councils or HMRC that lead to overpayment.
Underpayments, 2025–26: 0.4%; People paid less than they were due through official error.

The three overpayment types add up to 3.2% of spending (£9.9bn of £308.6bn). After recoveries, the net loss was 2.8% (£8.6bn). Shares are of money, not of claims or claimants.

Over time, and by benefit

Is benefit fraud rising or falling?

Overpayments 3.2% of spending in 2025–26, down from a 4.0% peak in 2021–22

Estimated share of DWP benefit expenditure overpaid (all causes, and fraud alone) and underpaid through official error, Great Britain, financial years 2005–06 to 2025–26.

Great Britain · 2005–06 to 2025–26

1996–2026 displayed window · observations only. Shading marks years outside the displayed record.

Overpayments rose sharply in 2020–21 and 2021–22, driven by fraud in Universal Credit as claims surged during the pandemic, and have fallen since: the 2025–26 rate of 3.2% is not statistically different from the year before, and remains above the roughly 2% seen before 2020. Fraud is the largest part. Underpayments — people paid too little through DWP’s own mistakes — have stayed around 0.3–0.5%.

Great Britain · 2005–06 to 2025–26 · Accredited official statistics; sample-based (2025–26 total overpayments 95% interval 3–3.5%, fraud 2–2.4%)

Sources, definitions and data

Geography:: Great Britain
Measure:: Per cent of benefit expenditure
Period:: 2005–06 to 2025–26
Status:: Accredited official statistics; sample-based (2025–26 total overpayments 95% interval 3–3.5%, fraud 2–2.4%)
Calculation:: As published (central estimates, Tables 2 and 7). The latest edition’s revised figure is used where a year was restated.
Compare with:: 2.1% overpaid in 2018–19, before the pandemic
Not included / breaks:: Claimant-error underpayments moved to a separate “unfulfilled eligibility” publication from 2022–23, and State Pension was fully reviewed again only from 2021–22, so the line is not perfectly comparable over time. Changes in the benefit mix (Universal Credit replacing older benefits) move the all-benefit rate even when individual rates do not change.

Source: DWP, Fraud and error in the benefit system, financial year ending 2026 estimates, Reference tables 2 and 7 (released 14 May 2026). Chart BEN-05.

View the data as a table

Financial year	All overpayments	Fraud	Claimant error	Official error	Underpayments
2005–06	2.1%	0.6%	0.8%	0.7%	0.4%
2006–07	2.2%	0.6%	0.8%	0.8%	0.3%
2007–08	2.0%	0.6%	0.7%	0.7%	0.4%
2008–09	2.2%	0.8%	0.8%	0.6%	0.4%
2009–10	2.2%	0.8%	0.8%	0.6%	0.3%
2010–11	2.1%	0.8%	0.8%	0.5%	0.3%
2011–12	2.1%	0.7%	0.9%	0.5%	0.3%
2012–13	2.1%	0.7%	1.0%	0.5%	0.3%
2013–14	2.1%	0.7%	0.9%	0.4%	0.3%
2014–15	1.8%	0.8%	0.6%	0.4%	0.3%
2015–16	1.9%	1.1%	0.5%	0.3%	0.4%
2016–17	2.0%	1.2%	0.5%	0.4%	0.3%
2017–18	2.2%	1.2%	0.6%	0.4%	0.4%
2018–19	2.1%	1.1%	0.6%	0.4%	0.4%
2019–20	2.4%	1.4%	0.6%	0.4%	0.4%
2020–21	3.9%	2.9%	0.6%	0.4%	0.4%
2021–22	4.0%	3.0%	0.7%	0.3%	0.5%
2022–23	3.6%	2.7%	0.6%	0.3%	0.5%
2023–24	3.6%	2.7%	0.6%	0.3%	0.4%
2024–25	3.3%	2.2%	0.7%	0.4%	0.4%
2025–26	3.2%	2.2%	0.6%	0.4%	0.4%

Scroll the table sideways to see every column.

Which benefits are most often paid wrongly?

Pension Credit and Universal Credit have the highest overpayment rates; the State Pension the lowest

Estimated overpayment rate (all causes) by benefit, % of that benefit’s expenditure, Great Britain, 2025–26 (financial year ending 2026). Money overpaid and the underpayment rate are printed under each name.

Great Britain · 2025–26 (financial year ending 2026)

Pension Credit£620m overpaid · 1.3% underpaid10.0%
Universal Credit£6.72bn overpaid · 0.4% underpaid8.5%
Housing Benefit£800m overpaid · 0.3% underpaid6.2%
Personal Independence Payment£660m overpaid · 0.2% underpaid2.3%
Disability Living Allowance£70m overpaid · 2.3% underpaid0.9%
State Pension£230m overpaid · 0.3% underpaid0.2%

Means-tested benefits, which depend on reported income, savings and household, are overpaid far more often than the State Pension, which depends on a contribution record. Universal Credit accounts for most of the money overpaid; in Pension Credit, a third of claims were overpaid, though most by under £10 a week. DLA is the only benefit here with a higher underpayment than overpayment rate.

Great Britain · 2025–26 (financial year ending 2026) · Accredited official statistics; Housing Benefit working-age and DLA adult rates were last measured in earlier years and rolled forward

Sources, definitions and data

Geography:: Great Britain
Measure:: Per cent of each benefit’s expenditure
Period:: 2025–26 (financial year ending 2026)
Status:: Accredited official statistics; Housing Benefit working-age and DLA adult rates were last measured in earlier years and rolled forward
Calculation:: As published in the bulletin (central estimates).
Compare with:: All benefits: 3.2%
Not included / breaks:: Rates are shares of each benefit’s own spending, so a high rate on a small benefit can be little money: Universal Credit is most of the £ total. PIP figures exclude £1.03bn of cases where claimants could not reasonably have known to report a gradual improvement.

Source: DWP, Fraud and error in the benefit system, financial year ending 2026 estimates, Reference tables 2 and 7 (released 14 May 2026). Chart BEN-06.

View the data as a table

Benefit	Overpaid (%)	Overpaid (£m)	of which fraud (%)	Underpaid (%)
Pension Credit	10.0	620	3.5	1.3
Universal Credit	8.5	6,720	6.8	0.4
Housing Benefit	6.2	800	3.6	0.3
Personal Independence Payment	2.3	660	1.4	0.2
Disability Living Allowance	0.9	70	0.2	2.3
State Pension	0.2	230	0.0	0.3

Scroll the table sideways to see every column.

Claim checks

“8.4 million people are claiming unemployment benefit”

Incorrect

What is known: 8.4 million people were on Universal Credit in Great Britain in May 2026. 1.6 million were in the group required to look for work; 4.3 million had no work requirements, and 3.1 million were in paid work (April 2026).

What would change this verdict: A published count showing all 8.4 million were unemployed and required to look for work.

“Most benefit spending goes to unemployed people”

Incorrect

What is known: Pensioners received 54% of welfare spending in Great Britain in 2024–25. Jobseeker’s Allowance cost £0.3bn in 2024–25 (DWP Table 1a); job-seekers on Universal Credit are 1.6 million of 8.4 million people on it.

What is not known: DWP does not publish Universal Credit spending split by conditionality group, so the exact cost of support for job-seekers is not published.

“Benefit fraud costs £10 billion a year”

Misleading

What is known: £9.9bn was overpaid in 2025–26, but that combines fraud (2.2% of spending), claimant error (0.6%) and DWP’s own mistakes (0.4%). Fraud alone was about two-thirds of it.

What is not known: These are estimates from a sample of claims, with confidence intervals; they cannot be attached to individual claimants.

“Most PIP claims are fraudulent”

Incorrect

What is known: DWP estimates 1.4% of PIP spending was overpaid through fraud in 2025–26 — up from 0.4%, a real rise — and 2.3% overpaid in total. 4.1 million people received PIP in England and Wales.

What is not known: Fraud rates are not conviction counts, and a low fraud rate does not mean every other payment was right: claimant error, official error and underpayment are measured separately.

“Tribunals prove two-thirds of PIP decisions are wrong”

Misleading

What is known: 67% of PIP appeals decided at a tribunal hearing in April to June 2026 went the claimant’s way. Across all initial decisions after an assessment (2021–26), 3% were overturned at a tribunal.

What is not known: How many decisions that were never challenged were wrong.

What would change this verdict: A published overturn rate measured against all initial decisions showing two-thirds were wrong.

Viral “family gets £X a week” stories usually describe a household exempt from the benefit cap, most often because of disability or caring: how the cap works.

What is being done

The Public Authorities (Fraud, Error and Recovery) Act 2025 lets DWP ask banks to flag accounts that break eligibility rules, such as savings above Universal Credit’s limit. DWP estimates around £500m a year in savings once fully rolled out; the measure is being introduced gradually. Its trade-offs — privacy, false positives and the hardship of recovering debts — are set out with the other options in What would move the needle?

Sources and methods (3)

DWP, Fraud and error in the benefit system, financial year ending 2026 estimates, Reference tables 2 and 7 (released 14 May 2026) — overpayment and underpayment rates over time and by benefit
DWP, Unfulfilled eligibility in the benefit system, financial year ending 2026 (released 14 May 2026) — unfulfilled eligibility
GOV.UK, Benefit cap — benefit cap — the key to viral “£X a week” claims

Method notes

Estimates come from a specialist review of a sample of claims and are applied to each benefit’s spending; the financial year ending 2026 sample ran from September 2024 to October 2025.
Some benefits are not reviewed every year; their last measured rates are rolled forward and applied to current spending.
Cases where a review could not be completed (£1.1bn in the latest year) and PIP/DLA cases where claimants could not reasonably have known to report an improvement are excluded from the headline rates.