Department for Work and Pensions Accounts 2019–20 Contents

2Fraud and error

Pre-COVID-19: fraud and error at record levels

10.Benefit overpayments are at their highest estimated rates and have risen consistently since 2015–16. Excluding State Pension, the estimated rate of overpayments increased again to 4.8% (£4.5 billion) of estimated benefit expenditure of £93.1 billion for 2019–20, from a restated rate of 4.4% (£3.8 billion) in 2018–19. The estimated rate of underpayments, excluding State Pension, decreased to 2.0% (£1.9 billion) in 2019–20, from its restated rate of 2.2% (£1.9 billion) in 2018–19.19

11.The Department does not currently have a target rate of fraud and error for us to use to hold it to account. However, the Department does now “absolutely accept in principle” that it should have a target “given the level of public scrutiny and interest in the question of fraud and error in the DWP and benefits.” It told us that it was going to set a headline target of around 2.3% for 2020–21; this target appears to be for overpayments including State Pension which were at 2.4% (£4.6 billion) in 2019–20.20 However, the Department informed us that it reversed its decision to set a target for 2020–21 because it first needs to establish a “clear baseline” after the effect of COVID-19.21 This Committee has made several recommendations in recent reports for the Department to set fraud and error targets, not only at an overall rate, but for each benefit.22 This would allow for better scrutiny of the impact of the Departments’ initiatives and emerging new risks of fraud and error.

12.Universal Credit has the highest estimated overpayment rate of all measured benefits—9.4% (£1.7 billion) for 2019–20—and it has an estimated underpayment rate of 1.1% (£0.2 billion). This is the highest recorded overpayment rate for any benefit other than Tax Credits (administered by HMRC), which peaked at 9.7% in 2003–04.23 The Department told us that that is it “not happy” with the fraud and error figures for Universal Credit and that it is currently “well off” achieving the fraud and error savings targets in the Universal Credit business case (£1.3 billion a year in steady state).24 Despite its current performance, it reassured us that it is confident it will achieve the business case fraud and error savings in future.25

13.Even before the impact of COVID-19 the Department had not yet delivered the value of savings on fraud and error on which the Business Case for Universal Credit was based.26 As these were intended to be annually recurring savings, every year of delay in achieving them represents a real and significant cost to the public purse.27 The Department is not yet able to tell us how long it will take to achieve the promised level of savings, and it is therefore unable to tell us what the total additional cost will be.28

14.The Department informed us that “many more claims naturally means more fraud and error in the system”.29 The Department reported that the number of people on Universal Credit almost doubled from 2.9 million in February to 5.6 million in August and has continued to grow since.30 Fraud and error in Universal Credit for 2019–20 is £1.9 billion (£1.7 billion overpayments and £0.2 billion underpayments).31 Therefore, as a rough estimate, the effect of a doubling in caseload alone (ignoring the effect of easements to controls) could cause around a £1.9 billion increase in fraud and error in Universal Credit for 2020–21.

Fraud and error impact of COVID-19

15.The Department acknowledges that the easements to controls it has made to respond to the pandemic will increase fraud and error more than would otherwise be expected by the increase in claims. It has produced a range of estimates of the amounts potentially at risk which has been shared with HM Treasury.32

16.The Department informed us that it has taken steps to mitigate the impact of these easements. It said it believes the key thing is to have a real-time data feed that enables it to check that the mitigations it has put in place are being effective. It provided an example of it closely monitoring the referrals that its staff make to its centralised Enhanced Checking Service.33 This is a new function, comprising 600 trained fraud investigators, to whom benefit processing staff can refer any suspicious cases for further investigation and additional verification.34 The Department told us that the portion of claims being referred to Enhanced Checking Service “peaked at one level at about 13%, but it has since fallen back to about 3% on average.” The Department explained that it also gets real time information to monitor the fraud and error impact of control easements from “predictive analytics”.35

17.However, whether these detection activities are successful does not inform us about the impact on the underlying rate of fraud and error. The Department reported that due to the redeployment of its staff to tackle the surge of claims and the difficulties of performing its sampling exercise in lockdown, it will not be able to review cases to produce an estimate of fraud and error in 2020–21 in the usual way.36 The Department told us that its approach is to, as best it can, replicate the sampling that it would be doing in a normal year.37

18.The Department also informed us that as of July, after redeploying staff back into measurement activities, it has been measuring fraud and error on Universal Credit and it is also undertaking measurement work on other benefits such as Pension Credit and Employment Support Allowance. It told us that “it is vital that we do our absolute best” to have an overall estimate of fraud and error in its Annual Report for 2020–21, where it also ‘aims’ to report the fraud and error cost of its easements to controls; the Department said it will look at how much detail it can go into with regards to attributing the fraud and error impact of its control easements.38

Pursuing a cost-effective control environment

19.The National Audit Office’s (NAO’s) work in 2019–20 on the Department’s strategy to tackle fraud and error showed that the Department has a good understanding of the types of fraud and error that occur in the benefit system, but that it needs to do more to understand the cost-effectiveness of individual controls so that it knows that it is doing absolutely everything that it should be doing to counter fraud and error. It also recommended that the Department both risk assess changes to its administrative processes and inform Parliament of those risks.39

20.The Department told us that it wants to get to the point where its accounts are no longer qualified. It acknowledged that it is not where it wants to be, but said it knows what it needs to do.40 It said that it is working with the NAO to understand what it needs to do to demonstrate that it has a cost-effective set of fraud and error controls. The Department added that, although there will always be some fraud and error, having fraud and error at a cost-effective level is consistent with its wider objectives of providing the service that its customers expect. It also stated that there “is always a balance here” (implying between fraud and error and other organisational objectives), but that it needs to get to a point where it has “that balance set in a cost-effective way”, in order to provide confidence that its payments are in line with parliamentary intention.41

21.The Department’s fraud and error strategy relies on modernising its technology and putting more investment into data and data analytics. It told us that “we really do see that putting more investment into data, into data analytics and into that prevention space, is going to get us where we need to go”, with prevention activity not allowing fraud and error into the system in the first place.42

22.The Department launched its Risk and Intelligence Service (RIS) in April 2018 and reported that it was using ‘increasingly sophisticated data and analytical tools’ to tackle fraud and error.43 In response to COVID-19, the Department absorbed the work of RIS and other intelligence teams into its new Integrated Risk and Intelligence Service (IRIS) which the Department reports is ‘significantly increasing’ its prevention capability through the use of new data matching rules. However, despite these claims, the impact of this technology is still largely unproven as the estimated rate of overpayments has continued to rise since this Risk and Intelligence Service was introduced.44

23.The Department told us that it is starting to build a system that is based on ‘transaction risking’; its vision is to be in a place where it can, in real time, or near real time, assess every claim as it is coming through and take a view of how much it trusts the information that is in the claim. For example, having one ‘customer journey’ (quicker and easier) for claimants where the Department trusts the information, and a different ‘customer journey’ with more intervention where the Department does not trust the information.45

24.There are specific risk areas such as capital, living together, self-reported and self-employed earnings where the Department admits it is harder to tackle fraud and error, in part due to the lack of access it has to timely, accurate data.46 In 2019–20, measured capital fraud and error rose by £380 million (73%) to £910 million across all measured benefits; the largest increase in value for any individual risk type.47 The Department told us that it is “looking very carefully, particularly with colleagues in other parts of Government”, to see what data it can use to get better access to capital information.48 Although there are data sets available, there are time lag issues with these data sets. The Department also informed us that it has performed some initial work on tackling capital risk using data from banks which it says gave “really good results”.49

25.The Department told us that the big fraud and error saving that it knew would come from Universal Credit is using real-time information (RTI) on earnings from HMRC in an automated way to calculate the award, and that it ‘knows’ it is “doing well on the RTI part”. However, the Department accepted that there is “more fraud and error in self-reported earnings than had been anticipated” and it has “more to do on the self-employment part”.50 The Department told us that is has other datasets from HMRC, “because everybody has to make returns”, for claimants with self-employment or self -reported income. However, the Department said that there are time lag issues with this data so it needs to supplement it with data from other sources e.g. different agencies, financial institutions or financial companies that would have information on people.51 The third risk area identified by the Department where further progress is required is ‘living together’ (e.g. where an undeclared partner might be living in a household). It informed us that it is looking at using other types of data matching in this area and reported that IRIS has developed data matching rules to help identify cases where an undeclared partner might be living in a household. Alongside looking for data matching opportunities, it also told us that a lot of work is going into making reporting a change of circumstance easier for capital, living together and self-reported and self-employed earnings.52

26.The Department has powers that allow it to ask for the information that it needs when it is doing an individual compliance investigation, but it does not have legal access to the same level of information for the controls it uses to prevent and detect fraud and error. The Department told us it is currently working cross-government with Departments such as Cabinet Office and HMRC to understand whether there are gaps between the powers which it has and the powers which it needs.53


19 DWP ARAC 2019–20, pages 186–188

20 Q 21; DWP ARAC 2019–20, page 70

21 Q 21

22 Committee of Public Accounts, Universal Credit and fraud and error: progress review, Session 2016–17, HC 489, 4 November 2016, recommendations 5 and 8; Committee of Public Accounts, Fraud and Error Stocktake, Session 2015–16, HC 394, 28 October 2015, recommendations 2 and 4.

23 DWP ARAC 2019–20, pages 189, 191, 238

24 Q16; C&AG’s Report, Rolling out Universal Credit, Session 2017–19, HC 1123, 15 June 2018, page 52, Figure 18

25 Q 18

26 Q 17, C&AG’s Report, Universal Credit: getting to first payment, Session 2019–21, HC 376, 10 July 2020, page 41, paragraph 2.24

27 C&AG’s Report, Rolling out Universal Credit, Session 2017–19, HC 1123, 15 June 2018, page 52, Figure 18

28 Q 19, Q 20

29 Q 20

30 Letter from DWP to Committee dated 22 September 2020, page 6; StatXplore (Department for Work & Pensions), People on Universal Credit, https://statxplore.dwp.gov.uk/webapi/jsf/tableView/tableView .xhtml (accessed 15/10/20)

31 DWP ARAC 2019–20, page 238

32 DWP ARAC 2019–20, page 194

33 Q 37

34 DWP ARAC 2019–20, page 76

35 Qq 37, 45

36 DWP ARAC 2019–20, page 185 192

37 Q 28

38 Qq 27, 40

39 DWP ARAC 2019–20, page 185

40 Qq 17, 19

41 Q 19

42 Q 24; DWP ARAC 2019–20, page 17

43 Department for Work & Pensions, Annual Report and Accounts 2018–19, HC 2281, 27 June 2019, page 122

44 DWP ARAC 2019–20, pages 73 and 76, 188 Figure 2

45 Q 24

46 Q 17; ‘Self-reported’ earnings for Universal Credit is income from employment where the Department is not able to receive real time information from HM Revenue and Customs that it can use in its award calculation. Therefore, claimants are required to self-report these earnings.

47 DWP ARAC 2019–20, page 192

48 Q 17

49 Q 30

50 Qq 17, 29

51 Q 29

52 Q17; DWP ARAC 2019–20, page 73

53 Qq 29, 30




Published: 18 November 2020