Research and analysis

Error and fraud in the COVID-19 schemes: methodology and approach (an update for 2023)

Updated 17 July 2023

Executive summary

The table below sets out the complete picture of the UK government’s main HMRC administered COVID-19 support schemes: the Coronavirus Job Retention Scheme (CJRS), the Self-Employment Income Support Scheme (SEISS), and Eat Out to Help Out (EOHO). Our up-to-date assessment includes the level of error and fraud; the amount of COVID-19 yield that has been recovered through post payment compliance activity across the lifecycle of the schemes; and an estimate of the error and fraud that remains unrecovered by the end of 2022 to 2023.

These are revised estimates since those published in July 2022 in the HMRC Annual Report and Accounts (ARA) 2021 to 2022 and the original estimates published in November 2021 in the HMRC Annual Report and Accounts (ARA) 2020 to 2021. These revised estimates take account of an improved understanding of error and fraud on the COVID-19 schemes from data that was not previously available.

These will be HMRC’s final estimates of COVID-19 support scheme error and fraud.

Across the full lifecycle of the three financial support schemes (covering 2020 to 2021 and 2021 to 2022), the total value of error and fraud is now estimated to be between £3.3 billion and £7.3 billion, with a most likely estimate of £5.0 billion (an error and fraud rate of between 3.3% and 7.4% with a most likely estimate of 5.1%).

This is an increase from last year’s published estimate from the 2021 to 2022 Annual Report and Accounts (between £3.2 billion and £6.4 billion with a most likely estimate of £4.5 billion, which corresponds to an error and fraud rate between 3.3% and 6.5%, with a most likely rate of 4.6%), but less than was originally estimated in the 2020 to 2021 Annual Report and Accounts (between £4.5 billion and £8.0 billion, with a most likely estimate of £5.8 billion; an error and fraud rate between 5.5% and 9.9%, with a most likely rate of 7.2%). The final most likely estimate of £5.0 billion (5.1%) is within the previously published estimated ranges.

Overview: error and fraud

Error and fraud 2020 to 2021

	Coronavirus Job Retention Scheme	Self Employed Income Support Scheme	Eat Out to Help Out	Total
Claims plus accruals £ million	61,507	19,745	842	82,094
Unprompted repayments and disclosures £ million	(939)	(43)	(2)	(984)
Net cost £ million	60,568	19,702	840	81,110
Error and fraud %	5.1% [2.9% to 8.0%]	3.2% [2.8% to 3.5%]	8.5% [5.1% to 11.8%]	4.7% [2.9% to 6.9%]
Error and fraud amount £ million	3,082 [1,784 to 4,816]	629 [560 to 699]	71 [43 to 99]	3,782 [2,387 to 5,614]

Error and fraud 2021 to 2022

	Coronavirus Job Retention Scheme	Self Employed Income Support Scheme	Eat Out to Help Out	Total
Claims plus accruals £ million	8,727	8,372	–	17,099
Unprompted repayments and disclosures £ million	(71)	(8)	–	(79)
Net cost £ million	8,656	8,364	–	17,020
Error and fraud %	4.9% [3.0% to 7.4%]	10.1% [7.5% to 12.9%]	–	7.5% [5.2% to 10.1%]
Error and fraud amount £ million	427 [258 to 640]	843 [629 to 1,078]	–	1,270 [887 to 1,717]

Cumulative totals: 2020 to 2021 plus 2021 to 2022

	Coronavirus Job Retention Scheme	Self Employed Income Support Scheme	Eat Out to Help Out	Total
Net cost £ million	69,224	28,066	840	98,130
Error and fraud %	5.1% [3.0% to 7.8%]	5.2% [4.2% to 6.3%]	8.5% [5.1% to 11.8%]	5.1% [3.3% to 7.4%]
Error and fraud amount £ million	3,498 [2,045 to 5,418]	1,472 [1,189 to 1,776]	71 [43 to 99]	5,041 [3,277 to 7,293]
COVID yield achieved £ million	–	–	–	1,056
Error and fraud amount minus COVID yield to date £ million	–	–	–	3,985 [2,221 to 6,237]

Notes

1. Claims and accruals as reported in HMRC’s 2021 to 2022 ARA. Accruals are amounts relating to claims made for a financial year that were paid out in a later financial year.

2. Repayments occur when customers entitled to a grant choose to repay it voluntarily, or when customers repay due to error or fraud whether unprompted or following a prompt from HMRC. For the error and fraud calculation, when the repayment is prompted, this is considered as COVID-19 yield so is deducted later in the calculation. The error and fraud estimate then represents the amount available for recovery by HMRC through post payment compliance. See Appendix 2 for more detail.

3. Ranges as presented in square brackets represent the lower and upper limits of the simulated 95% confidence interval presented together with the most likely estimate. Total percentages for combined schemes and financial years have been rounded to the nearest 0.1%.

4. COVID yield is defined as post-payment compliance only.

5. Of the £1,056 million COVID yield achieved to the end of the 2022 to 2023 year, £536 million is from before the Taxpayer Protection Taskforce commenced.

Updated evidence has significantly improved our understanding of the nature of error and fraud in the COVID-19 support schemes.

For the Coronavirus Job Retention Scheme, insight from a second Random Enquiry Programme and newly available operational compliance data has filled some evidence gaps in our statistical modelling and has led us to revise the estimate. The final whole scheme (2020 to 2021 plus 2021 to 2022) error and fraud rate for the Coronavirus Job Retention Scheme lies in the range of 3.0% to 7.8%, with a most likely value of 5.1%. This is similar to last year’s published most likely estimate. Although we now have stronger evidence overall to inform our estimate, the range has widened slightly since last year’s estimate – this is not because we are less certain in our estimate, but is because a wider range of overclaims were found in the second Random Enquiry Programme.

The error and fraud estimate for the different Coronavirus Job Retention Scheme claim periods to which the policy applied are shown in Appendix 1.

For the Self Employment Income Support Scheme, we now have results from a new Random Enquiry Programme as well as complete Self Assessment tax return data covering the whole of 2021 to 2022. The final whole scheme (2020 to 2021 plus 2021 to 2022) error and fraud rate for the Self Employment Income Support Scheme lies in the range of 4.2% to 6.3%, with a most likely value of 5.2%. The increase in this year’s revised error and fraud estimate compared to last year’s estimate is driven by results from the Random Enquiry Programme for the Reasonable Belief Test in grant periods 4 and 5.

The error and fraud estimate for the each of the Self Employed Income Support Scheme grants are shown in Appendix 1.

1. Background

The COVID-19 support schemes have helped millions of people and businesses through the pandemic. The Coronavirus Job Retention Scheme has helped to pay the wages of people in 11.7 million jobs and 1.3 million business. Around 2.9 million self-employed workers have received a Self-Employment Income Support Scheme grant.

From the beginning it was clear the schemes could be targets for fraud and that customers could make mistakes. We designed the schemes to prevent as much error and fraud as possible before any payments were made, while still quickly supporting those who needed it. These design measures gave us the confidence to get the schemes up and running quickly, recognising there would still be error and fraud risks.

Error and fraud estimates

The government has been open and transparent about the risks of error and fraud in the Coronavirus Job Retention Scheme and the Self Employment Income Support Scheme.

We published early provisional estimates in November 2021, given the understandable level of public interest. We were clear at the time that our initial provisional estimates were tentative, often based on only ‘mild’ evidence with at best some elements of ‘adequate’ evidence across the schemes, meaning that we used the best available evidence at the time and that there was a high probability that the estimates would change on receipt of more complete datasets. Further detail on the early provisional estimates, strength of evidence definitions and methodology can be found in our November 2021 technical report.

We revised our provisional estimates, publishing updated estimates in July 2022. These updated estimates incorporated a stronger evidence base including the first Random Enquiry Programme for the Coronavirus Job Retention Scheme which covered claim periods from 1 March 2020 to 31 October 2020, Self Assessment tax return data for 2020 to 2021 for the Self Employment Income Support Scheme and further insight from compliance activity. Further detail on the updated provisional estimates can be found in our July 2022 technical report.

For our final estimates of COVID-19 support scheme error and fraud, we now have further evidence from data that was not previously available, giving us a more comprehensive understanding of the claimant population.

For the Coronavirus Job Retention Scheme, we have completed a second Random Enquiry Programme covering claim periods from 1 March 2020 to 30 September 2021, across a range of different business sizes and sectors. We have also introduced new operational compliance data at a more granular level than in the previous estimates.

For the Self Employment Income Support Scheme, there were particular evidence gaps for grants 4 and 5 (21 April 2021 to 30 September 2021), so for those grants, we have conducted a new Random Enquiry Programme covering the ‘Reasonable Belief Test’, the ‘Financial Impact Declaration’ and the ‘Not Intending to Trade’ compliance risks.

We also now have complete Self Assessment tax return data covering the whole of both years 2020 to 2021 and 2021 to 2022 and more compliance outcome data which informs the estimates across all grants.

The level of error and fraud reflects the prevalence of non-compliance remaining after pre-payment compliance activity. This means that this represents the level of error and fraud before subsequent recoveries through post-payment compliance activity by HMRC compliance caseworkers and latterly the HMRC Taxpayer Protection Taskforce.

2. Headline findings

Tables 1a, 1b and 1c provide the latest error and fraud estimates across all schemes and summarises the differences between the original provisional estimates published in the HMRC ARA 2020 to 2021, the updated provisional estimates published in the HMRC ARA 2021 to 2022 and the final estimates published in the HMRC ARA 2022 to 2023.

In the HMRC ARA 2020 to 2021 we stated that the planning assumption for Coronavirus Job Retention Scheme error and fraud during 2020 to 2021 was between 5% to 10%. Our final most likely estimate for the whole scheme lifecycle is at the lower end of this range.

The final error and fraud estimates for Self Employment Income Support Scheme grants 1 to 3 (2020 to 2021) remain above the planning assumption of below 1% to 2% and the previous provisional estimates as published in HMRC ARA 2020 to 2021 and HMRC ARA 2021 to 2022.

Table 1a: Comparison of error and fraud figures for the Coronavirus Job Retention Scheme between HMRC ARA 2020 to 2021, 2021 to 2022 and 2022 to 2023

	ARA 2020-21 error and fraud estimate % rate	ARA 2021-22 error and fraud estimate % rate	ARA 2022-23 error and fraud estimate % rate	ARA 2020-21 error and fraud estimate range	ARA 2020-21 error and fraud estimate % rate	ARA 2022-23 error and fraud estimate range	Difference in evidence assessment
Coronavirus Job Retention Scheme estimate 2020-21	8.7% [6.7% to 12.0%]	5.3% [3.7% to 7.6%]	5.1% [2.9% to 8.0%]	£5,279 million [£4,065 million to £7,281 million]	£3,218 million [£2,246 million to £4,614 million]	£3,082 million [£1,784 million to £4,816 million]	Improvement: Adequate evidence level with elements of strong evidence
Coronavirus Job Retention Scheme estimate 2021-22	NA	2.8% [2.0% to 3.7%]	4.9% [3.0% to 7.4%]	NA	£241 million [£172 million to £319 million]	£427 million [£258 million to £640 million]	Improvement: Adequate evidence level with elements of strong evidence

Table 1b: Comparison of error and fraud figures for the Self-Employment Income Support Scheme between HMRC ARA 2020 to 2021, 2021 to 2022 and 2022 to 2023

	ARA 2020-21 error and fraud estimate % rate	ARA 2021-22 error and fraud estimate % rate	ARA 2022-23 error and fraud estimate % rate	ARA 2020-21 error and fraud estimate range	ARA 2021-22 error and fraud estimate range	ARA 2022-23 error and fraud estimate range	Difference in evidence assessment
Self-Employment Income Support Scheme estimate 2020-21	2.5% [1.8% to 3.2%]	3.2% [2.4% to 4.1%]	3.2% [2.8% to 3.5%]	£493 million [£355 million to £631 million]	£631 million [£473 million to £808 million]	£629 million [£560 million to £699 million]	Improvement: Adequate evidence level with elements of strong evidence
Self-Employment Income Support Scheme estimate 2021-22	NA	4.5% [3.5% to 6.3%]	10.1% [7.5% to 12.9%]	NA	£376 million [£292 million to £526 million]	£843 million [£629 million to £1,078 million]	Improvement: Adequate evidence level with elements of strong evidence

Table 1c: Comparison of error and fraud figures for Eat Out to Help Out (EOHO) between HMRC ARA 2020 to 2021 and 2021 to 2022

	ARA 2020-21 error and fraud estimate % rate	ARA 2021-22 error and fraud estimate % rate	ARA 2022-23 error and fraud estimate % rate	ARA 2020-21 error and fraud estimate range	ARA 2021-22 error and fraud estimate range	ARA 2022-23 error and fraud estimate range	Difference in evidence assessment
Eat Out to Help Out provisional estimate 2020-21	8.5% [5.1% to 11.8%]	NA	NA	£71 million [£43 million to £99 million]	NA	NA	No further update in 2022 or 2023: Mild evidence

Notes

1. Ranges are modelled as 95% confidence intervals in square brackets for error and fraud % rates and amounts. 2020 to 2021 provisional estimates were presented in Figure 17, R54 (page 357) of HMRC ARA 2020 to 2021.

2. 2021 to 2022 updated provisional estimates were presented in Table 7 (page 30) of HMRC ARA 2021 to 2022.

3. For description of evidence levels please see Section B.3.1 from the 2020 to 2021 published technical document.

3. Coronavirus Job Retention Scheme: Assessment of error and fraud

HMRC’s final estimate for Coronavirus Job Retention Scheme error and fraud now uses results from two Random Enquiry Programmes. The first Random Enquiry Programme covered claim periods from 1 March 2020 to 30 June 2020 and from 1 July 2020 to 31 October 2020. The second Random Enquiry Programme covered claim periods from 1 March 2020 to 30 September 2021.

Random sampling of taxpayers for auditing in a Random Enquiry Programme is an internationally recognised robust approach to estimating error and fraud (see for example International Monetary Fund 2021 publication).

The additional data from the second Random Enquiry Programme, as well as further operational compliance data, gives us a stronger evidence base than our previous approaches to estimating Coronavirus Job Retention Scheme error and fraud. Our first provisional estimate in HMRC ARA 2020 to 2021 relied on assembling evidence from a variety of internal HMRC and external evidence sources (see our archived technical publication from November 2021 for more detail), whilst our updated provisional estimate in HMRC ARA 2021 to 2022 used results from the first Random Enquiry Programme and some operational compliance data (see our archived technical publication from July 2022 for more detail).

Our updated approach still makes some necessary assumptions. We use corroborating evidence from internal HMRC sources and third-party survey data for non-detection of risks, which is an accepted limitation to Random Enquiry Programmes.

Our assessment of the evidence base is that we have ‘adequate’ evidence to support our final estimates, but with elements of ‘strong’ evidence from the additional Random Enquiry Programme data and operational compliance data that have become available since the previous published estimate. These elements of ‘strong’ evidence are an improvement to the strength of evidence compared to our previous estimate.

Coronavirus Job Retention Scheme: Detailed findings on the nature of error and fraud

The scale and nature of Coronavirus Job Retention Scheme error and fraud changed throughout the different periods of the scheme, as scheme design and eligibility rules changed (see Table A3, Appendix 1).

Error and fraud was found to be highest in the claim period of the scheme covering 1 March to 30 June 2020, particularly because of the risk from employers claiming the Coronavirus Job Retention Scheme for employees who were working. This is because, in this period of the scheme, the eligibility criteria excluded furloughed employees who did any work at all, whereas the scheme became more flexible from July 2020 onwards to allow part-furloughed employees to qualify. The same behaviours that could be non-compliant in this period of the Coronavirus Job Retention Scheme could therefore be compliant or only partially non-compliant in later periods of the scheme.

Whilst the risk ‘employers claiming for employees who are working’ contributes the largest proportion of error and fraud across the lifecycle of CJRS, new insight from the second Random Enquiry Programme shows that, in the final phase of the scheme, there is a lower proportion of error and fraud attributed to that risk than the total of other risks. This is likely because of changes to scheme design and eligibility rules, which introduced more scope for error from other risks during the final phase of the scheme than originally estimated.

After incorporating the results of the second Random Enquiry Programme, we now attribute a larger proportion of the overall Coronavirus Job Retention Scheme error and fraud estimate to error than we previously published.

Both Random Enquiry Programmes were designed to identify organised crime as well as other error and fraud risks. There was very little organised crime identified through either Random Enquiry Programme, re-enforcing the assessment that HMRC’s pre-payment controls were effective and organised crime on the scheme was very low.

The most likely error and fraud estimates split by risk types are shown in table A1 and A2 in Appendix 1. The proportional splits are based on additional insight from the Random Enquiry Programme and expert judgement, with the tables illustrating that the proportion of losses due to fraud decrease from 2020 to 2021 to 2021 to 2022. The error and fraud rates for the Coronavirus Job Retention Scheme risks are shown in table A5 in Appendix 1.

4. Self Employment Income Support Scheme: Assessment of error and fraud

HMRC’s updated and final error and fraud estimate for the Self Employment Income Support Scheme now uses insight from a Random Enquiry Programme as well as up to date data from Self Assessment tax returns covering the whole of 2020 to 2021 and 2021 to 2022 and updated compliance outcomes data.

The Random Enquiry Programme covered only grants 4 and 5 because of how different the economic conditions were during these final two grant periods, with lockdown restrictions lifting. For this reason, it was suspected that error and fraud would be higher in grants 4 and 5 compared to earlier grant periods.

These updated data sources improve our estimate, giving us a more comprehensive understanding of the claimant population and allowing us to fill some evidence gaps in our statistical modelling.

Our assessment of the evidence base is that we have ‘adequate’ evidence to support our final estimates overall, but with elements of ‘strong’ evidence from incorporating the full 2021 to 2022 Self Assessment data and new Random Enquiry Programme data that were previously unavailable. These elements of ‘strong’ evidence are an improvement in the strength of evidence compared to our previous estimate. Our updated approach does still make some necessary assumptions, for example that the updated statistical modelling captures the subjectivity within the ‘Reasonable Belief Test’ compliance risk.

Self Employment Income Support Scheme Policy Changes

The Self Employment Income Support Scheme was expanded to include the latest information submitted on 2019 to 2020 Self Assessment tax returns that were filed by 2 March 2021. This ensured that the scheme was based on the most up-to-date complete data held by HMRC and meant that self-employed customers previously ineligible for the Self Employment Income Support Scheme could be eligible to claim the Self Employment Income Support Scheme grants 4 and 5. This included those that were ineligible for support based on their previous returns and those newly self-employed in 2019 to 2020.

To reflect the reopening of the economy, and ensure support was targeted at those most affected by the pandemic, the level of grant 5 was determined by a turnover test, known as a ‘Financial Impact Declaration’ (FID). Whilst increasing targeting to support those who needed it most, there may be error and fraud introduced when claimants were calculating and entering their turnover figures (although eligibility and the grant calculation remained based on information that HMRC already held).

The final estimate for the Self Employment Income Support Scheme grant 5 accordingly reflects the additional risk of error and fraud associated with the FID.

A ‘Reasonable Belief Test’ (RBT) was also introduced for later grant periods to improve targeting. This test required traders to confirm they would suffer a reduction in their trading profits compared to what they would otherwise have expected in the absence of the pandemic. The RBT was devised to act as a ‘nudge’ for claimants intended to induce a behaviour change, and asked the customer to keep evidence to prove their reasonable belief that they would be impacted. Post payment, it is difficult to establish the counterfactual to prove that the customer’s belief they would suffer a reduction in trading profits was not well grounded.

New insight from the Random Enquiry Programme (REP) has allowed us to fill some evidence gaps and has shown that error and fraud from the RBT is larger than previously thought. Some customers were unable to provide evidence to support their claim.

The RBT is now the largest error and fraud risk within the Self Employment Income Support Scheme, despite only being introduced in later grant periods. The RBT drives the increase in the overall scheme error and fraud rate in the final estimate, compared to our previous published estimate. Meanwhile, the FID is a smaller error and fraud risk than previously thought.

Adding additional criteria such as the ‘Financial Impact Declaration’ or ‘Reasonable Belief Test’ increases targeting and reduces overall spending, but holding all other things equal, will increase the scope for error and fraud either because people may misunderstand the more complex eligibility rules or because they try and push at the boundaries of entitlement. The error and fraud estimates for each of the Self Employment Income Support Scheme risks are shown in table A6 in Appendix 1.

Self Employment Income Support Scheme: Findings on the nature of error and fraud

The most likely error and fraud estimates split by the risk types of organised crime, opportunistic fraud and error are shown in table A1 and A2 in Appendix 1.

The proportional splits are based on insight from the Random Enquiry Programme and expert judgement, with the tables illustrating that the proportion of organised crime within the Self Employment Income Support Scheme error and fraud estimate decreases from 2020 to 2021 to 2021 to 2022 despite the overall level of error and fraud increasing. The scheme design and HMRC’s compliance controls had a clear positive impact.

Following the introduction of the ‘Reasonable Belief Test’ and ‘Financial Impact Declaration’, the proportion of error in Self Employment Income Support Scheme grants 4 and 5 increases significantly, suggesting that increasing controls and complexity has the effect of increasing the scope to incur errors.

Appendix 1: Detailed error and fraud estimates

Table A1: Detailed error and fraud assessments for Coronavirus Job Retention Scheme, Self-Employment Income Support Scheme and Eat Out to Help Out during 2020 to 2021 most likely estimates by risk type

Scheme	Most likely error and fraud % rate	Most likely value £million	Proportion organised crime	Proportion opportunistic fraud	Proportion error
Coronavirus Job Retention Scheme	5.1%	£3,082 million	<1%	44%	56%
Self-Employment Income Support Scheme (1 to 3 combined)	3.2%	£629 million	36%	26%	38%
Eat Out to Help Out	8.7%	£71 million	7%	75%	18%

Notes

1. Figures here are largely assumption driven on mild evidence including judgement obtained through a panel of experts, evidence from the Random Enquiry Programmes and further completed post payment compliance activities.

2. For the Coronavirus Job Retention Scheme, the balance between opportunistic fraud and error remains uncertain. For example, where employers claimed Coronavirus Job Retention Scheme for employees who were working, some cases will have been driven by fraud and others by error.

3. For the Self-Employment Income Support Scheme the balance between opportunistic fraud and error is also uncertain in some instances. For example, in risks related to claimants ‘not trading or intending to trade’.

Table A2: Detailed error and fraud assessments for Coronavirus Job Retention Scheme and Self-Employment Income Support Scheme during 2021 to 2022 most likely estimates by risk type

Scheme	Most likely error and fraud % rate	Most likely value £million	Proportion organised crime	Proportion opportunistic fraud	Proportion error
Coronavirus Job Retention Scheme	4.9%	£427 million	<1%	40%	60%
Self-Employment Income Support Scheme (4 and 5 combined)	10.1%	£843 million	4%	29%	67%

Notes

1. Figures here are largely assumption driven on mild evidence including judgement obtained through a panel of experts, evidence from the Random Enquiry Programmes and further completed post payment compliance activities.

2. For the Coronavirus Job Retention Scheme the balance between opportunistic fraud and error is particularly uncertain. For example, where employers claimed Coronavirus Job Retention Scheme for employees who were working, some cases will have been driven by fraud and others by error.

3. For the Self-Employment Income Support Scheme the balance between opportunistic fraud and error is also uncertain in some instances. For example, in risks related to claimants ‘not trading or intending to trade’ or incorrectness related to the FID.

Table A3: Detailed error and fraud estimates for Coronavirus Job Retention Scheme by grant claim period

Claim periods to which the policy applied	1 March 2020 to 30 June 2020	1 July 2020 to 31 October 2020	1 November 2020 to 30 September 2021
Coronavirus Job Retention Scheme rate	5.4% [3.2% to 8.1%]	4.5% [2.2% to 8.0%]	5.0% [3.0% to 7.4%]
Coronavirus Job Retention Scheme value	£1,522 million [£916 million to £2,312 million]	£569 million [£280 million to £1,004 million]	£1,406 million [£850 million to £2,106 million]

Notes

1. Most likely estimates presented together with the simulated 95% confidence interval for each phase of the Coronavirus Job Retention Scheme.

2. The grant claim periods referred to in the table above are the phases during which Coronavirus Job Retention Scheme claims could be made. Employers may have had multiple claim periods during each phase, and the length of claim periods would have been different for different employers.

3. The final claim period of the Coronavirus Job Retention Scheme policy runs runs across the two financial years 2020 to 2021 and 2021 to 2022 as shown in the table above.

Table A4: Detailed error and fraud assessments for Self-Employment Income Support Scheme by grant phase

Grant Phase	Self-Employment Income Support Scheme 1	Self-Employment Income Support Scheme 2	Self-Employment Income Support Scheme 3	Self-Employment Income Support Scheme 4	Self-Employment Income Support Scheme 5
Self-Employment Income Support Scheme rate	3.3% [3.0% to 3.6%]	3.3% [2.9% to 3.6%]	3.0% [2.6% to 3.5%]	8.0% [5.8% to 10.5%]	14.1% [10.9% to 17.6%]
Self-Employment Income Support Scheme value	£246 million [£223 million to £269 million]	£195 million [£174 million to £215 million]	£189 million [£163 million to £214 million]	£442 million [£319 million to £577 million]	£401 million [£311 million to £501 million]

Notes

Most likely estimates presented together with the simulated 95% confidence interval for each grant phase of the SEISS scheme.

For this table the grant phases correspond to claims made for the following periods:

1. Self-Employment Income Support Scheme 1 claims were submitted between 13 May and 13 July 2020
2. Self-Employment Income Support Scheme 2 claims were submitted between 17 August and 19 October 2020
3. Self-Employment Income Support Scheme 3 claims were submitted between 20 November and 29 January 2021
4. Self-Employment Income Support Scheme 4 claims were submitted between 22 April and 1 July 2021
5. Self-Employment Income Support Scheme 5 claims were submitted between 29 July and 30 September 2021

Table A5: Detailed error and fraud assessments for Coronavirus Job Retention Scheme by risk for whole lifecycle (2020 to 2021 and 2021 to 2022)

Coronavirus Job Retention Scheme risks	Error and Fraud Percentage Rate	Error and Fraud Value
Employers claiming for employees who are working	3.3% [1.9% to 5.1%]	£2,270m [£1,327m to £3,516m]
All other risks	1.8% [1.0% to 2.7%]	£1,228m [£718m to £1,902m]

Note

Most likely estimates presented together with the simulated 95% confidence interval for each risk within the CJRS scheme.

Table A6: Detailed error and fraud assessments for Self-Employment Income Support Scheme by risk for whole lifecycle (2020 to 2021 and 2021 to 2022)

Self-Employment Income Support Scheme risks	Error and Fraud Percentage Rate	Error and Fraud Value
Not Intending to Trade (NIT)	1.3% [0.9% to 1.7%]	£373m [£264m to £482m]
Boundary Pushing (BP)	0.5% [0.3% to 0.7%]	£130m [£72m to £189m]
Organised Crime (OC)	0.9% [0.8% to 1.0%]	£260m [£240m to £281m]
Reasonable Belief Test (RBT)	2.2% [1.4% to 3.0%]	£613m [£404m to £842m]
Financial Impact Declaration (FID)	0.3% [0.2% to 0.5%]	£95m [£73m to £142m]

Note

Most likely estimates presented together with the simulated 95% confidence interval for each risk within the SEISS scheme.

Appendix 2: Repayments

The error and fraud estimate is presented on a net cost basis, which we define as total claims plus accruals (as detailed in the ARA) minus the repayments and adjustments which were made by claimants without direct HMRC intervention. This is separate to the net expenditure presented in the ARA.

There are small but methodologically important differences between net expenditure and net cost.

1. Prompted repayments: In 2020 to 2021 and 2021 to 2022, some prompted Coronavirus Job Retention Scheme and Self Employment Income Support Scheme repayments, resulting from HMRC activity, were classed as repayments and are thus deducted from claims to arrive at net expenditure for accounting purposes. These prompted repayments form part of our error and fraud estimate so we do not deduct them upfront when we calculate the net cost.

2. Repayments timing: For accounts purposes, repayments are attributed to a financial year based on when the money is received by HMRC. For estimating error and fraud, we want to deduct voluntarily repaid money from the financial year in which the original claim it relates to was made.

For these reasons there are small differences in the repayments reported for error and fraud purposes and the repayments reported in the ARA. Where HMRC data was not available to precisely determine the amount of repayments on this basis, we have used available internal data to form an estimate.