HMRC error and fraud estimates

1Summary

The United Kingdom pays research and development (R&D) relief to companies through Corporation Tax, and much of it is paid in cash. Because of that, HMRC reports the reliefs as expenditure in its accounts and must estimate how much of that expenditure was paid in error or as a result of fraud.1 It has published an estimate every year since 2019 to 2020.2

The estimates have changed a great deal. The first ones, based on compliance cases selected for risk, put error and fraud at between 3.6 and 4.9 percent of the cost of the reliefs.21 In 2023 HMRC began using a random enquiry programme, which examines a random sample of claims in full. Its revised estimate for 2020 to 2021 was 16.7 percent, and for 2021 to 2022 it was 17.6 percent, with the SME scheme at 25.8 percent.34 HMRC's July 2026 annual report gives a final estimate of 6.4 percent for 2023 to 2024 and illustrative estimates of 5.3 percent for the two following years.5

This paper explains how the estimates are made, why they were revised, what the random enquiries found about claims and agents, and the measures that followed. It relies only on HMRC and National Audit Office publications. It describes the UK record and makes no recommendation for the United Kingdom.

2Why HMRC estimates error and fraud in the reliefs

HMRC operates the R&D reliefs as part of Corporation Tax. Its Resource Accounts report spending on the schemes as expenditure because they often result in cash payments, unlike most other tax reliefs.1 Where error or fraud leads HMRC to overpay a claimant, the National Audit Office treats the payment as not conforming with Parliament's intention, and so as irregular.1

HMRC has included an estimate of error and fraud in the reliefs in its accounts since 2019 to 2020, "in response to the increasing take up of these reliefs over recent years".2 The Comptroller and Auditor General has qualified his opinion on the regularity of HMRC's accounts on this ground every year since then.6 When it first asked HMRC to produce an estimate, the National Audit Office pointed to HMRC's own "Strategic Picture of Risk", the growth in spending on the schemes, and "an increase in public reporting of instances of abuse of these schemes, including through unregulated agents".1

The term "error and fraud" is wide. HMRC uses it to cover "mistakes and failure to take reasonable care through to deliberate non-compliance".3 To classify a case as fraud, a caseworker must have found evidence that the claimant "deliberately set out to misrepresent their circumstances to get money to which they were not entitled".3 A common example of error given by HMRC is claiming for expenditure that was not incurred on qualifying R&D activity.7

3The first method: risk based compliance reviews

3.1How the early estimates were made

For 2019 to 2020, 2020 to 2021 and 2021 to 2022, HMRC estimated error and fraud by combining results from claims it had reviewed through its compliance work with an estimate for the claims it had not reviewed, using comparable error rates from its tax gap work.2 Its 2021 to 2022 annual report listed the uncertainties: the rate was extrapolated from historic compliance activity; it assumed that error in claims not reviewed matched general error rates in the tax gap; and the cost of the reliefs was itself an estimate. HMRC said that, given these uncertainties, it could not produce "a meaningful range of outcomes" for the estimate.2

Year	SME scheme	RDEC	Overall
2019 to 2020	5.6% (£271m)	1.0% (£40m)	3.6% (£311m)
2020 to 2021	5.5% (£303m)	0.9% (£33m)	3.6% (£336m)
2021 to 2022	7.3% (£430m)	1.1% (£39m)	4.9% (£469m)

Figure 1. HMRC's estimates of error and fraud under its earlier method, as reported by HMRC and the National Audit Office.21

HMRC said the rise in 2021 to 2022 "reflects improvements in HMRC's risk identification process".2

3.2What the National Audit Office said about them

The National Audit Office later explained that these estimates rested on compliance cases where HMRC had assessed the claims as at risk, together with assumptions including "that unreviewed cases have a lower rate of error and fraud". It said they were enough to show that error and fraud was material, but were limited by the data on unreviewed cases, so that the actual level "could be significantly different". It added: "This is borne out in the significantly higher levels of error and fraud reported in HMRC's current estimates."1

3.3The 2022 attack on the payable credit

In April 2022 HMRC identified a pattern of irregular claims, which it later described as "a criminal attack on the R&D payable tax credit regime".23 It introduced a threat risk assessment of all R&D claims, more payment identification and verification controls, and an accelerated R&D Anti Abuse Unit.2 By July 2023 it reported that it had refused £85 million of suspected fraudulent claims, issued over 2,500 letters challenging suspect claims, and arrested 9 people suspected of conspiring to submit £16 million of fraudulent claims.3

4The random enquiry programmes

4.1What a random enquiry programme is

A mandatory random enquiry programme (MREP) takes a random sample of claims and subjects each to a full enquiry. HMRC describes the approach as representing "international best practice", and the National Audit Office describes it as "good practice in the measurement of error and fraud".81 The programmes cover claims made by small and medium-sized enterprises (SMEs), under the SME scheme and under the R&D expenditure credit (RDEC). Large businesses claiming RDEC were not included, because HMRC considered that its customer compliance manager model gave it sufficient insight into those claims.8

Random enquiries introduce a lag. Companies claim up to two years after incurring the expenditure, and enquiries take time to close, so each programme measures claims made some years before the estimate is published.1

4.2The four programmes so far

Programme	Sample	Claims examined	Share with some non-compliance
MREP 1	Random sample of 500 SME claims filed in 2020 to 2021	484 closed when the estimate was made	244 (50%)
MREP 2	Stratified random sample of 404 claims filed in 2021 to 2022	387 closed	200 (52%)
MREP 3	Stratified random sample of 350 claims filed from 1 November 2023 to 29 February 2024	334 closed by July 2026	109 (33%)
MREP 4	Not stated in the reports reviewed	61% of enquiries closed by July 2026	Not yet used

Figure 2. HMRC's random enquiry programmes, as described by the National Audit Office.146

The sampling changed over time. For MREP 2 HMRC moved from a wholly random sample to a stratified one, by type of relief and size of claim. For MREP 3 it sampled proportionately more high value claims, which earlier programmes had found to yield more.49 In each programme HMRC applied a "non-detection multiplier" to allow for error or fraud that caseworkers may have missed, and estimated the likely yield from enquiries that were still open.146

MREP 3 is the first programme drawn wholly from claims made after the additional information form became mandatory on 8 August 2023.10 When HMRC produced its 2024 to 2025 accounts, it had closed 296 of the 350 enquiries, and 88 of them (30 percent) had found some non-compliance.9 By the 2025 to 2026 accounts it had closed 334, and 109 (33 percent) had found some non-compliance.6 HMRC did not use MREP 4, because only 61 percent of its enquiries had closed when the latest estimate was produced.6

4.3The National Audit Office's observations

The National Audit Office reviewed samples of the random enquiry cases each year. Its observations included the following.

In 2022 to 2023, more than half (32) of the 60 cases it sampled related to claims for software and information technology R&D. Referral of technically complex cases to HMRC's internal technology experts was at the caseworker's discretion; in one case a recommended referral was not made because of capacity.1
HMRC had set no minimum value below which error would not be pursued, so caseworkers decided individually whether to pursue small errors. The National Audit Office said it "would expect HMRC to include all error or fraud in its estimate, irrespective of value".1
HMRC applied no non-detection adjustment to large business RDEC claims, which the National Audit Office said could mean that error and fraud in that scheme was understated.14
In its sample for 2023 to 2024, HMRC had initially used the wrong yield figure in 6 of 56 cases, compared with 3 of 60 the year before.4
In one case, HMRC received no contact from either the taxpayer or the agent; it referred the case to its Fraud Investigation Service, "and the agent is currently being investigated". HMRC told the National Audit Office that the mandatory additional information form, which includes information on agents, had increased its ability to identify suspicious claims at source.4

For 2023 to 2024, HMRC replaced its assumption based method for large business RDEC claims with desk reviews of a stratified random sample of 108 cases, 106 of which were completed. The National Audit Office said this produced a more accurate estimate, reflected in a higher error rate for large businesses.6

5The estimates and their revisions

5.1From first estimate to latest estimate

Each year HMRC publishes an illustrative estimate for the current year, projected from the most recent random enquiry results, and later replaces it with a final estimate once a programme covering that year has closed.45

Year	Earlier estimate	Latest estimate	Range for the latest estimate
2020 to 2021	3.6% (£336m), earlier method	16.7% (£1.13bn), July 2023	13.0% to 23.2%
2021 to 2022	4.9% (£469m), earlier method, July 2022	17.6% (£1.34bn), July 2024	13.8% to 23.2%
2022 to 2023	13.3% (£1.05bn), illustrative, July 2023	9.9% (£759m), July 2025	6.7% to 14.1%
2023 to 2024	7.8% (£601m), illustrative, July 2024	6.4% (£493m), July 2026	3.8% to 11.1%
2024 to 2025	5.9% (£481m), illustrative, July 2025	5.3% (£442m), July 2026 (illustrative)	None published
2025 to 2026	None	5.3% (£475m), July 2026 (illustrative)	None published

Figure 3. HMRC's overall estimates of error and fraud in the R&D reliefs, as a share of the estimated cost of the reliefs.2341056 The ranges are those HMRC published with each final estimate.34105 HMRC notes that later years are not directly comparable because of the merger of the schemes.5

For the SME scheme alone, the final estimates were 24.4 percent for 2020 to 2021, 25.8 percent for 2021 to 2022, 14.7 percent for 2022 to 2023 and 11.1 percent for 2023 to 2024. The illustrative estimates for the SME scheme and enhanced R&D intensive support (ERIS) are 11.0 percent for 2024 to 2025 and 2025 to 2026.41056 In July 2025 HMRC published an intermediate estimate of 6.5 percent for 2023 to 2024, before revising it to 6.4 percent in July 2026.105

5.2Why the numbers moved

Several separate things explain the revisions.

The method changed. The jump in the 2020 to 2021 estimate, from 3.6 to 16.7 percent, came from replacing assumptions about unreviewed claims with random enquiry results. HMRC said the programme showed that "the assumptions for the SME estimate were previously underestimating the true rate and value of error and fraud".3
Projections built in assumed effects. The illustrative estimate for 2022 to 2023 assumed that compliance interventions would reduce error and fraud by £250 million.18 The illustrative estimate for 2023 to 2024 started from a projection of £1,045 million and deducted £444 million for "changes in the compliance environment".4 The National Audit Office noted each time that HMRC would not know the actual effect until it completed random enquiries for that year.14
The basis changed to accruals. From the 2024 to 2025 accounts, HMRC estimates error and fraud by the year in which expenditure was incurred, rather than the year in which claims were received.10
Claims before and after the new form were weighted. The final estimate for 2022 to 2023 weights the MREP 2 rate, for claims received before 8 August 2023, and the MREP 3 rate, for claims received after it.10
The expenditure base was revised. HMRC revises its estimate of the cost of the reliefs over five years. The cost for 2022 to 2023 fell from £10.19 billion as first reported to £7.69 billion, and for 2023 to 2024 from £9.44 billion to £6.94 billion, which affects the monetary value of the estimates.105
Large businesses were measured differently. The final estimate for 2023 to 2024 is the first to use the new desk review method for large business claims.56

5.3Statistical caveats

HMRC itself qualifies the trend. In its 2024 to 2025 accounts it said that the SME rate for 2022 to 2023 "is not statistically different from the rate for 2021 to 2022 of 25.8%", and that the overall rate for 2022 to 2023 was not statistically different from 2021 to 2022.10 In its 2025 to 2026 accounts it said that the SME rate for 2023 to 2024 "is not statistically different from the rate for 2022 to 2023 and reflects natural variability in the results based on a sample of cases".5 The illustrative estimates carry no confidence intervals.96

The wide ranges reflect sample sizes. For 2023 to 2024, HMRC's range for the SME scheme runs from 7.4 to 15.5 percent, and for the RDEC scheme from 1.4 to 8.2 percent.5

6What the random enquiries found about claims

6.1The shape of non-compliance

HMRC's analysis of the first random enquiry programme found that around half of claims contained at least some non-compliance.8 In 25 percent of claims no qualifying R&D activity took place, in 19 percent R&D took place but the company overclaimed, and in 2 percent there was a technical misinterpretation of the legislation.8 Fraud indicators were found in fewer than 10 percent of claims, accounting for less than 5 percent of the value claimed, which HMRC said "indicates that the majority of non-compliance is down to other behaviours".8 The second programme found similar overall results: 26 percent of value non-compliant, 49 percent of claims compliant, 30 percent wholly non-compliant and 21 percent overclaimed.7

The first programme also showed patterns by claim.

Size. In claims with expenditure below £10,000, 78 percent of the value claimed was non-compliant; in claims over £1 million, 13 percent was, and around 75 percent of those claims were fully compliant.8
First claims. 38 percent of the value of first time claims was non-compliant, against 22 percent for companies that had claimed before.8
Sector. Non-compliance by value ranged from 9 percent in professional, scientific and technical activities to 87 percent in education.8

6.2Agents

HMRC states that more than 90 percent of R&D claimants are represented by an agent.87 For the random enquiry samples it assessed whether the company used a specialist R&D agent.

Programme	Share of claims with a specialist R&D agent	Non-compliance by value, with such an agent	Without an identified agent of that kind
MREP 1 (2020 to 2021)	51%	25%	24%
MREP 2 (2021 to 2022)	54%	28%	21%

Figure 4. Non-compliance by value of claims, by whether the company used a specialist R&D agent.87

HMRC described non-compliance in the second programme as "slightly higher in those claims that were submitted with the support of a specialist R&D agent".7 In the first programme the largest claims, over £500,000, showed non-compliance of 15 percent with a specialist agent and 18 percent without, while claims under £10,000 showed 90 percent and around 75 percent.8

A separate analysis of the first programme, published with HMRC's 2024 consultation on the tax advice market, divided claims by whether the agent belonged to a professional body.11

Representation	Claims in sample	Claims with non-compliance	Cost of relief non-compliant
Agent affiliated to a professional body	303	49%	20%
Agent not affiliated	118	53%	53%
Affiliation unknown	41	56%	42%
No agent	38	66%	12%

Figure 5. HMRC analysis of the 500 claims in the first random enquiry programme, by agent affiliation.11

HMRC cautioned that some groups are small, that claim size is correlated with non-compliance, and that companies with no agent tend to make higher value claims and are likely to be larger businesses able to manage their own claims.11

None of these breakdowns divides claims by how the agent was paid. HMRC's published analyses distinguish the presence of an agent and the agent's affiliation, not the basis of the agent's fee.8711

7The policy response

7.1Parliamentary scrutiny

The Committee of Public Accounts made recommendations in February 2022 on HMRC's approach to abuse of the reliefs. HMRC accepted them, and its formal response in April 2022 committed it to a cross cutting team of around 50 staff focused on abuse, digital claims with more detail endorsed by a named senior officer, customer education, analysis of the growing cost of the reliefs, and continued random enquiries.1 In 2023 the Comptroller and Auditor General recommended that HMRC consider whether the scale and focus of its compliance work were adequate, and whether to inspect more claims from before the period covered by its estimates.1

7.2The measures

Measure	Applies from
Cap linked to PAYE on the size of SME claims	Accounting periods from April 20217
R&D Anti Abuse Unit created	July 20228
SME additional deduction cut from 130% to 86%, SME payable credit from 14.5% to 10%, RDEC raised from 13% to 20%	April 20234
Advance notification of claims	Accounting periods from April 20237
Digital claims with additional information, a named company officer and the identity of any advisers	Claims from 8 August 202347
No new assignments of R&D tax credits; payment to the company rather than to nominees	Assignments from November 2023; claims from April 20247
Merged RDEC and enhanced R&D intensive support	April 20244
Online disclosure service for companies that claimed too much and can no longer amend their return	Published 31 December 202412

Figure 6. Measures listed by HMRC and the National Audit Office.

HMRC also increased its R&D compliance staff from around 100 in 2020 to 2021 to more than 500, and opened compliance checks on 17 percent of claims in 2023 to 2024.7 The National Audit Office recorded that the Finance Act 2023 rate changes were "in part, a response to the estimated level of error and fraud for the SME scheme being considerably higher than for the RDEC scheme".4

The disclosure service lets a company tell HMRC that it claimed too much where the time limit for amending its return has passed. It may be used by a director or company secretary, or by a tax adviser on the company's behalf, but not where the company's behavior was deliberate.12 HMRC warns that a company "may be charged extra interest and penalties if they wait for HMRC to contact them".12

7.3How HMRC connects the estimates to the measures

HMRC's 2025 to 2026 annual report says that the 2023 to 2024 rate "is lower than estimates since 2020 to 2021, reflecting legislative and operational changes to tackle error and fraud, notably the mandation of digital claims and additional information requirements introduced in 2023".5 It attributes the lower illustrative rates for later years to the additional information form and the new ERIS and merged RDEC schemes.5 HMRC told the National Audit Office that requiring companies to provide more information "has helped reduce the number of non-compliant claims".6

The National Audit Office said that HMRC "is beginning to see the impact of compliance and legislative changes made in recent years", that there are "currently no further changes planned", and that HMRC is reviewing recent random enquiry results "to consider whether the measures taken have gone far enough".6 It also noted that claims by SMEs typically have a higher rate of error and fraud than claims by large businesses, and that HMRC will not know the full impact of the changes until it has completed enquiries on 2025 to 2026 returns, which it expects will most likely be reported in HMRC's 2028 to 2029 annual report.6

8What the estimates do not show

The estimates are the most detailed public measure of claim quality in the UK reliefs, but they have limits that HMRC and the National Audit Office acknowledge, and others that follow from what is measured.

Lag. The most recent final estimate is for 2023 to 2024. Later figures are projections.56
Attribution. Several measures took effect within a short period: rate changes, notification, the additional information form, digital claims, the payment restriction and the merged scheme. HMRC has not published a separate estimate of the effect of each.105
Behavior. The headline figure combines mistakes, carelessness and fraud. HMRC's thematic analysis indicates that most non-compliance is not fraud.87
Agents. The analysis records whether a specialist R&D agent was used and whether an agent belonged to a professional body, but not how any agent was paid.8711 HMRC states that it "does not regulate tax agents including those who advise on R&D claims".7
Responsibility. HMRC states that a customer who uses a tax agent remains responsible for its own tax affairs.7

9What the record shows

HMRC's measurement of error and fraud in the R&D reliefs changed when it moved from risk based reviews to random enquiries. The earlier method assumed that unreviewed claims were cleaner than reviewed ones; the random enquiries showed that around half of SME claims in 2020 to 2021 and 2021 to 2022 contained some non-compliance, and that error and fraud was 16.7 and 17.6 percent of the cost of the reliefs in those years.34 The National Audit Office described the SME rate as among the highest reported across all government spending programmes.14

The latest final estimate, for 2023 to 2024, is 6.4 percent, and the illustrative estimates for the next two years are 5.3 percent.5 HMRC and the National Audit Office both connect the fall mainly to the 2023 requirements for digital claims and additional information.56 HMRC also states that some of the year on year differences in the SME rate are not statistically significant.105

The record identifies who prepared a claim, whether that person was a specialist R&D agent and, in one analysis, whether the agent belonged to a professional body. It does not record the basis on which agents were paid, and no published UK estimate relates claim quality to fee arrangements.8711

References

National Audit Office, Report by the Comptroller and Auditor General: HM Revenue & Customs 2022-23 Accounts, Summary paragraphs 9 to 11, 17 and 18, and Part Two, paragraphs 2.1 to 2.21 and Figure 11. https://assets.publishing.service.gov.uk/media/64da2f04c8dee4000d7f1c01/Report_by_the_Comptroller_and_Auditor_General.pdf (accessed 11 October 2026). Back
HM Revenue and Customs, Annual Report and Accounts 2021 to 2022 (print version, July 2022), "Research and development tax relief error and fraud" (page 116) and Resource Accounts note 5.1.5, "Corporation tax reliefs: R&D error and fraud" (pages 241 to 242). https://assets.publishing.service.gov.uk/media/639c4a30d3bf7f7f8cebfe8f/HMRC_Annual_Report_and_Accounts_2021_to_2022_Print.pdf (accessed 11 October 2026). Back
HM Revenue and Customs, Annual Report and Accounts 2022 to 2023 (July 2023), "Research and development tax reliefs" (page 23), Principal Accounting Officer's report, "Research and development tax relief error and fraud" (pages 115 to 116), and Resource Accounts note 5.1.5 (pages 256 to 257). https://assets.publishing.service.gov.uk/media/64e34f1c3309b700121c9baa/HMRC_annual_report_and_accounts_2022_to_2023.pdf (accessed 11 October 2026). Back
National Audit Office, Report by the Comptroller and Auditor General: HM Revenue & Customs 2023-24 Accounts, Summary paragraphs 10 and 11, and Part Two, paragraphs 2.4 to 2.21 and Figure 7. https://assets.publishing.service.gov.uk/media/6889f286048fff613a4d5b72/Report_by_the_Comptroller_and_Auditor_General.pdf (accessed 11 October 2026). Back
HM Revenue and Customs, Annual Report and Accounts 2025 to 2026 (July 2026), "Our accountability", Control challenges, "Corporation Tax research and development tax relief error and fraud" (page 107), and Resource Accounts note 3.1.4, "Corporation tax reliefs: R&D error and fraud" (pages 251 to 253). https://assets.publishing.service.gov.uk/media/6a4f66c406e7256c331df8a2/HMRC_annual_report_and_accounts_2025_to_2026.pdf (accessed 11 October 2026). Back
National Audit Office, Report by the Comptroller and Auditor General on HM Revenue and Customs 2025-26 Accounts, published within HMRC's Annual Report and Accounts 2025 to 2026, Summary paragraph 11, and Part Two, paragraphs 2.8 to 2.15 and Figure 11. https://assets.publishing.service.gov.uk/media/6a4f66c406e7256c331df8a2/HMRC_annual_report_and_accounts_2025_to_2026.pdf (accessed 11 October 2026). Back
HM Revenue and Customs, Approach to Research and Development tax reliefs 2023 to 2024 (published 30 October 2024), Introduction, sections 1.2, 2.1 (Table 2), 2.4 and 2.5, Annex A and Annex B (Themes 1 and 5, Tables B.4 and B.5). https://www.gov.uk/government/publications/hmrcs-approach-to-research-and-development-tax-reliefs-2023-to-2024/approach-to-research-and-development-tax-reliefs-2023-to-2024 (accessed 11 October 2026). Back
HM Revenue and Customs, HMRC's approach to Research and Development tax reliefs (corporate report, published 17 July 2023), sections 1.2, 2.1, 2.3, 3.1 (Table 1) and 3.2 (Themes 1 to 5 and Tables 2 to 6). https://www.gov.uk/government/publications/compliance-approach-to-research-and-development-tax-reliefs/hmrcs-approach-to-research-and-development-tax-reliefs (accessed 11 October 2026). Back
National Audit Office, Report by the Comptroller and Auditor General on HM Revenue & Customs 2024-25 Accounts, published within HMRC's Annual Report and Accounts 2024 to 2025, Summary paragraphs 9 and 10, and Part Two, paragraphs 2.5 to 2.17, Figure 8 and paragraph 2.41. https://assets.publishing.service.gov.uk/media/687e5d318adf4250705c96d8/HMRC_annual_report_and_accounts_2024_to_2025.pdf (accessed 11 October 2026). Back
HM Revenue and Customs, Annual Report and Accounts 2024 to 2025 (July 2025), "Research and Development tax reliefs" (pages 52 to 53), "Corporation Tax research and development tax relief error and fraud" (page 105), and Resource Accounts note 4.1.5, "Corporation tax reliefs: R&D error and fraud" (pages 252 to 254). https://assets.publishing.service.gov.uk/media/687e5d318adf4250705c96d8/HMRC_annual_report_and_accounts_2024_to_2025.pdf (accessed 11 October 2026). Back
HM Revenue and Customs, Raising standards in the tax advice market: strengthening the regulatory framework and improving registration (consultation, 6 March 2024), Annex C, "Analysis: non-compliance according to tax practitioner membership of professional bodies", section "Research and development tax credit". https://www.gov.uk/government/consultations/raising-standards-in-the-tax-advice-market-strengthening-the-regulatory-framework-and-improving-registration/raising-standards-in-the-tax-advice-market-strengthening-the-regulatory-framework-and-improving-registration (accessed 11 October 2026). Back
HM Revenue and Customs, Tell HMRC if you've claimed too much Research and Development (R&D) tax relief (published 31 December 2024, last updated 14 August 2026), sections "Who can use the R&D disclosure service", "When you should not use it" and "How to work out the penalties due". https://www.gov.uk/guidance/tell-hmrc-if-youve-claimed-too-much-research-and-development-rd-tax-relief (accessed 11 October 2026). Back

Cite this paper

Summary

1Summary

2Why HMRC estimates error and fraud in the reliefs

3The first method: risk based compliance reviews

3.1How the early estimates were made

3.2What the National Audit Office said about them

3.3The 2022 attack on the payable credit

4The random enquiry programmes

4.1What a random enquiry programme is

4.2The four programmes so far

4.3The National Audit Office's observations

5The estimates and their revisions

5.1From first estimate to latest estimate

5.2Why the numbers moved

5.3Statistical caveats

6What the random enquiries found about claims

6.1The shape of non-compliance

6.2Agents

7The policy response

7.1Parliamentary scrutiny

7.2The measures

7.3How HMRC connects the estimates to the measures

8What the estimates do not show

9What the record shows

References

About this paper

Cite this paper

Summary

1Summary

2Why HMRC estimates error and fraud in the reliefs

3The first method: risk based compliance reviews

3.1How the early estimates were made

3.2What the National Audit Office said about them

3.3The 2022 attack on the payable credit

4The random enquiry programmes

4.1What a random enquiry programme is

4.2The four programmes so far

4.3The National Audit Office's observations

5The estimates and their revisions

5.1From first estimate to latest estimate

5.2Why the numbers moved

5.3Statistical caveats

6What the random enquiries found about claims

6.1The shape of non-compliance

6.2Agents

7The policy response

7.1Parliamentary scrutiny

7.2The measures

7.3How HMRC connects the estimates to the measures

8What the estimates do not show

9What the record shows

References

About this paper

Related papers