DWP Annual Report & Accounts 2022-23

DWP expects the activities set out in its counter-fraud plan to generate £9.4 billion of additional savings over the next five years. In May 2022 DWP set out its high- level plan to tackle fraud and error following the pandemic in Fighting Fraud in the Welfare System. This includes £895 million of additional investment over the three years to March 2025 in counter-fraud staffing, advanced data analytics, and a project to review millions of Universal Credit claims. DWP estimates that this investment will lead to £9.4 billion of savings over five years by reducing fraud and error. Following a recommendation from this Committee, DWP has published a detailed estimate of the amounts saved through its counter-fraud work, which it estimates was £1.1 billion for 2023–23. It has also set a target to save £1.3 billion through counter-fraud work in 2023–24. This will need to increase each year to achieve its forecast reduction in overpayments if DWP’s other assumptions are correct. Taken together, the forecast, new savings estimate and target improve accountability by providing greater clarity on the cost-effectiveness of DWP’s counter-fraud activities. However, DWP has acknowledged that the savings estimate is experimental and requires refinement. Recommendation 2: DWP should report annually on its savings from detecting and preventing overpayments and its forecast of future overpayment levels, to at least the level of detail set out in Figure 9 of the Comptroller and Auditor General’s report, while continuing to refine and improve the underlying methodology. 6 The Department for Work & Pensions Annual Report and Accounts 2022–23

The success of DWP’s plan to reduce fraud and error in Universal Credit is dependent on its ability to review 8 million live claims by 2027–28. The biggest element of DWP’s counter-fraud plan is a project to cleanse the benefit system of incorrect payments by reviewing millions of Universal Credit claims – which it calls Targeted Case Reviews (TCR). DWP is investing £443 million in TCR over the current Spending Review period up to March 2025 and expects TCR to produce £6.4 billion of savings by 2027–28. To achieve this, it expects that staff will need to review around 8 million live claims, which will involve interviewing and chasing evidence from millions of people. TCR appears to be working as expected at a small scale (with 25,000 claims in 2022–23), but DWP faces a significant challenge in scaling up the project. It plans to increase the number of its TCR staff from 2,000 to around 6,000 and to more than triple the average number of claims each agent reviews each day. DWP acknowledged these challenges but says it is on track. However, it plans to outsource around 40% of TCR reviews to private contractors, which brings with it further risks to maintaining effectiveness, quality and customer service in reviews that are outsourced. Recommendation 3: As part of its Treasury Minute response to this report, DWP should set out how it will report on the efficacy, quality and customer service of Targeted Case Reviews, including separate disclosure where this has been outsourced to contractors. This is to provide public confidence that the review of Universal Credit cases is working, is not overly burdensome, and is not leading to legitimate claims being disrupted.

DWP and HMRC face a significant challenge in making back payments to people who have been underpaid State Pension due to missing Home Responsibilities Protection. In 2021–22, DWP identified underpayments of State Pension due to gaps in the National Insurance records of people who were historically entitled to a benefit called Home Responsibilities Protection (HRP). DWP now estimates that 210,000 people may have been underpaid some £1.3 billion, going back decades. This is in addition to the underpayment of £1.2 billion affecting 165,000 pensioners due to historical errors by DWP that we reported on last year. We are very concerned that DWP has found another systemic underpayment, potentially leaving hundreds of thousands of pensioners out of pocket by an average of £5,000. HM Revenue & Customs (HMRC) administers National Insurance records and told us it will be very difficult to identify people who have been impacted because it no longer holds the relevant records. It plans to contact people it thinks may be affected and invite them to make a claim for HRP. It will then correct the National Insurance record so DWP can pay back any missing State Pension. HMRC confirmed that any back payments may be subject to a tax charge, but it has not decided how it will deal with this. Neither DWP nor HMRC were able to tell us when this issue will be fully corrected. Recommendation 4: a) DWP should work with HMRC within the next six months to set out a clear plan and timetable for correcting underpayments of State Pension relating to Home Responsibilities Protection and provide clarity on how any tax issues will be dealt with. The Department for Work & Pensions Annual Report and Accounts 2022–23 7 b) DWP and HMRC should regularly publish updates on their progress correcting this issue. This should include key numbers such as the volume of cases identified as at risk, the number of people asking for a review of their case, and the value of payments made.

DWP is not doing enough to assure itself or Parliament that it can rely on National Insurance records to pay State Pension accurately and that it will not find further historic underpayments. The £1.3 billion underpayment of State Pension relating to missing HRP is one of three ongoing historical issues reported by DWP. It is in addition to the previous underpayment of £1.2 billion affecting some 165,000 pensioners. DWP now also reports that people who claimed Universal Credit over the period 2017–18 to 2022–23 are missing National Insurance credits due to an IT issue, and that it is working to update records. These underpayments raise serious doubts about the accuracy and completeness of the NI records. There is a risk that similar errors may occur with other benefits, as DWP does not routinely check that claimants are receiving the National Insurance credits they are entitled to. DWP and HMRC told us that their internal audit teams are collaborating on a joint review to provide some assurance over the integrity of the National Insurance records. But is concerning that these issues were able to build up over many years before DWP was alerted to them. DWP also told us it was working towards responding to our previous recommendation that it does more to detect systemic underpayments early before they can have a serious impact on pensioners. Recommendation 5: a) DWP should work with HMRC to provide assurance to the Committee within the next twelve months over the integrity of the National Insurance records and how they interact with DWP’s benefit system. b) DWP should report as part of the Treasury Minute what is done to set up an ‘early warning system’ to detect issues before they grow into significant underpayments. This could involve performing more frequent and in-depth analysis of underpayments identified by frontline staff.

DWP has not yet done enough to understand the impact of machine learning on customers and provide them with confidence that it will not result in unfair treatment. DWP is expanding its use of advanced data analytics to tackle fraud. This includes machine learning algorithms to flag potentially fraudulent benefit claims, so the system learns and adapts without following explicit instructions. DWP says it is in an early stage of implementing these tools, but has already piloted them to tackle fraud in Universal Credit advances. There are legitimate concerns about the level of transparency around DWP’s use of these tools and the potential impact on claimants who are vulnerable or from protected groups. DWP has not made it clear to the public how many of the millions of Universal Credit advances claims have been subject to review by an algorithm. Nor has it yet made any assessment of the impact of data analytics on protected groups and vulnerable claimants; though we acknowledge it has recently committed to provide such an assessment in next year’s annual report. Although DWP has internal governance arrangements over its use of machine learning and performs some ongoing analysis of bias, the results so far have been largely inconclusive. 8 The Department for Work & Pensions Annual Report and Accounts 2022–23 Recommendation 6: DWP should, as part of the assessment in its annual report, consider explicitly the impact of data analytics and machine learning on legitimate claims being delayed or reduced, the number of people affected, and whether this is affecting specific groups of people. The Department for Work & Pensions Annual Report and Accounts 2022–23 9 1 The scale of fraud and error in the benefit system

On the basis of a Report by the Comptroller & Auditor General (C&AG), we took evidence from the Department for Work & Pensions (DWP) on its 2022–23 Annual Report & Accounts and the level of fraud and error in the benefits it administers.2 We also took evidence from HM Revenue & Customs (HMRC) due to its role in administering National Insurance records.

DWP estimates that it overpaid 12.8% (£5.5 billion) of all Universal Credit payments in 2022–23, which is much higher than any other benefit.10 We challenged DWP to explain why the fall in fraud and error promised in the Universal Credit business case has failed to materialise. DWP told us that where legacy benefit claimants have migrated to Universal Credit, specific types of fraud—including earnings from employment and childcare— have fallen significantly.11 But it acknowledged that there has been a large build-up in what it calls the ‘stock’ of overpaid Universal Credit claims, which it says relate mostly to the pandemic. It added that it expects to address this primarily through Targeted Case Reviews.12 DWP estimates that the overpayment rate is particularly high for claims that started at the beginning of the pandemic - March 2020 to June 2020 - which during 2022– 23 were overpaid by 21.0%.13

We asked DWP to what extent the fact that 1 in 3 Universal Credit claims is incorrect is a result of the complexity of the system. DWP told us it is trying to make it easier for claimants to declare changes of circumstances through continuous improvements of the Universal Credit system. It gave as an example where it has made changes to simplify the process for claimants to declare their level of savings. DWP added that Universal Credit underpayments are relatively low, which it believes is a result of bringing together legacy benefits.14

We challenged DWP to explain whether it still expects Universal Credit overpayments to fall to 6.5% as it had previously committed to. DWP explained that 6.5% was the level implied in the business case as a result of the expected reduction in fraud and error from merging legacy benefits into Universal Credit. It added that there is no reason to think that it cannot still achieve the expected reduction, but that this would now result in a higher rate than 6.5% because the baseline level of fraud and error has increased. It concluded 7 Qq 13–14 8 Q 15 9 Q 13 10 DWP ARA 2022–23, page 272 11 Q 16 12 Q 16 13 DWP ARA 2022–23, page 111 14 Q 17 The Department for Work & Pensions Annual Report and Accounts 2022–23 11 that it might be that DWP is doing everything it possibly can but still does not achieve 6.5%, and that the key is doing that all it reasonably can and clearly demonstrating that its control activities are cost-effective.15 Forecasting future levels of overpayment

We questioned DWP on whether it expects the impact of the pandemic on fraud and error to gradually fade or to remain elevated for some time. DWP told us that although it expects the impact of the pandemic to fall away over time, it now believes that there is a general increase in the propensity to commit fraud in society, and as a result it may be “difficult to get back to where we were pre-pandemic”.16 This assumption is built into DWP’s forecast, which suggests that benefit overpayments will not return to pre-pandemic levels until 2027–28.17 However, DWP cautioned that there is “considerable uncertainty” about the baseline level of fraud and error.18

We asked DWP to clarify why it assumes in its forecast that there is an ongoing increase in the general propensity to commit fraud in society. DWP explained that this is a judgement based on estimates produced by a range of other bodies. These include Cifas which has reported an 11% increase in fraud against organisations, the Public Sector Fraud Authority which identifies a 7% increase in fraud outside tax and welfare, and the Office for National Statistics that reports in its crime survey that 41% of all crimes are related to fraud. DWP told us it believes that, on the basis of these comparators, it is appropriate to assume that there will be a general increase in fraud of 5% per year. It also observed that the Office for Budget Responsibility was comfortable with building this assumption into its spring 2023 outlook.19

We also asked HMRC if it is also seeing an increase to commit fraud among taxpayers as a whole. HMRC told us it does not forecast a propensity to fraud but that the tax gap, which might be considered an equivalent figure, has been holding steady.20 In written evidence submitted after our session HMRC clarified that the tax gap has fallen over time from 7.5% in 2005–06 to 4.8% in 2021–22, the same level as the preceding year. It also stated that HMRC does not make a separate estimate of the amount of the tax gap due to fraud, but set out that the tax gap due to ‘criminal attacks’ fell from 15% in 2020–21 to 11% in 2021–22.21

We have previously found that the DWP lacks the ability to demonstrate that its counter-fraud activities are having the intended impact and are cost-effective.22 Alongside its forecast that benefit overpayments will not return to pre-pandemic levels until 2027– 28, DWP has set a target to achieve £1.3 billion of fraud and error savings in 2023–24. It has also published a new estimate of the amounts saved by its counter-fraud activities23. 15 Q 100 16 Q 14 17 DWP ARA 2022–23, page 300 18 Q 13 19 Q 98 20 Q 99 21 Correspondence from HMRC dated 28 September 2023 22 Committee of Public Accounts, The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefit system, Twenty-Sixth Report of Session 2022–23, HC 44, 9 November 2022 23 DWP ARA 2022–23, pages 302, 303 12 The Department for Work & Pensions Annual Report and Accounts 2022–23 The NAO has reported that, taken together, DWP’s forecast, target and savings estimate should improve accountability by providing transparency on its performance in tackling fraud and error.24 24 DWP ARA 2022–23, page 274 The Department for Work & Pensions Annual Report and Accounts 2022–23 13 2 Systemic underpayments of State Pension Progress correcting underpayments relating to historical error by DWP staff

We have reported previously on the historical underpayment of State Pension due to errors by DWP affecting some pensioners who are married, widowed or over-80, most recently when we examined DWP’s 2021–22 Annual Report & Accounts. During that inquiry DWP told us it estimated that 237,000 pensioners had been underpaid around £1.46 billion due to human error going back decades.25 DWP’s best estimate is now that 165,000 people have been underpaid a total of £1.2 billion.26 This remans in line with the range of uncertainty that DWP previously set out.27

In January 2021 DWP launched an exercise to identify affected pensioners and make any back payments. We asked DWP to provide an update on its progress in correcting these underpayments. DWP told us that it is on track to complete the married and over-80 groups by the end of 2023 as it previously committed to. DWP has reported that it cleared 108,000 cases in Q1 of 2023, compared with around 47,000 in the previous quarter.28 Underpayments relating to Home Responsibilities Protection

When we examined DWP’s 2021–22 accounts, it told us about another category of historical State Pension underpayment caused by gaps in the National Insurance records of women who had previously claimed Child benefit.29 DWP and HMRC explained to us that between 1978 and 2000, people claiming Child Benefit should have automatically received Home Responsibilities Protection (HRP), which lowered the National Insurance contributions needed to receive a full State Pension.30 Up until 2000, the body responsible for Child Benefit did not record a National Insurance number as part of a claim. This led to missing periods of HRP on the National Insurance records of some women and thus an underpayment of State Pension, which is calculated based on National Insurance contributions.31 This is not the first time that DWP has encountered an issue with missing periods of HRP. In 2010–11 DWP worked with HMRC to pay £84 million in arrears to pensioners for the same issue.32 DWP admitted that it was “assumed that the situation had been addressed and solved”.33

We asked HMRC what it is doing to correct the National Insurance record so that DWP can in turn make any back payments of State Pension. HMRC explained that it no longer holds the relevant records and that identifying all of the people who may be 25 Committee of Public Accounts, The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefit system, Twenty-Sixth Report of Session 2022–23, HC 44, 9 November 2022 26 DWP ARA 2022–23, page 273 27 Department for Work & Pensions, Annual Report & Accounts 2021–22, HC193, 7 July 2022, pages 234, 291 28 Qq 36–37; DWP ARA 2022–23, pages 122–124 29 Committee of Public Accounts, The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefit system, Twenty-Sixth Report of Session 2022–23, HC 44, 9 November 2022 30 Q 50 31 DWP ARA 2022–23, page 292 32 DWP ARA 2022–23, page 292 33 Q 22 14 The Department for Work & Pensions Annual Report and Accounts 2022–23 impacted is “a big problem, and quite a challenging one”. It told us it is performing various scans of its IT systems to find a potential cohort of affected people.34 HMRC intends to write to this group, which numbers in the hundreds of thousands, saying it believes they may have been eligible for HRP and invite them to make a claim.35 Where HMRC believes there is a legitimate claim, it will correct the NI record and notify DWP so any back payments of State Pension can be made.36

We asked HMRC how it can be sure that it will reach the right people and also avoid paying out any illegitimate claims. It told us that it would apply due diligence to all claims. HMRC also told us that alongside targeted letters it is planning a communication campaign to encourage people who think they may have missed out to apply. HMRC added that in August 2023 it launched an online checker that customers can use to understand if they have enough National Insurance contributions for a full State Pension – where this is already full, HRP would not make any difference.37 We challenged DWP to justify its assumption, set out in the 2022–23 accounts, that 75% of the next-of-kin of deceased customers will come forward to make a claim for HRP. DWP explained that this is the based on the take-up rates it observes for the ongoing State Pension underpayment exercise. It told us it has set up a next-of-kin portal that people can register with if they think a deceased relative may have been underpaid.38 However, HMRC acknowledged that it is “incredibly difficult” to find and pay back the right next-of-kin. HMRC added that where there are competing claims from multiple next-of-kin it will have to make a judgement, but it has not yet decided how to handle this.39

We asked DWP how much people could receive in back payments of State Pension. DWP told us it assumes an average back payment of £5,000 for people above State Pension age and £3,000 for the next-of-kin of those who are deceased. DWP’s accounts report that overall 210,000 pensioners may have been underpaid some £1.3 billion. Adjusting for take-up, DWP currently expects to pay £1.0 billion to around 187,000 people. DWP warned us that there was a significant amount of uncertainty around these estimates. The possible range of underpayment that DWP will need to pay out was estimated by DWP in its 2022–23 accounts is between £310 million and £1.5 billion.40 We asked DWP what proportion of the affected pensioners will be women and what level of overlap there might be between the HRP issue and previous State Pension underpayments. DWP admitted it did not know because this is an “incredibly complicated” exercise but that it will have to take that into account, along with any interactions with other benefits.41

We asked HMRC and DWP when they expect to have a firmer idea of the scale of the issue and the timeframe for making back payments. HMRC told us it intends to send out the letters to potentially affected people over the next 18 months. It added it expects to turn around 80% of responses within 15 working days, correcting the National Insurance record and notifying DWP.42 DWP told us it does not have a clear, formal delivery plan for completing back payments. The provision in DWP’s 2022–23 accounts 34 Qq 51,61 35 Q 52 36 Q 56 37 Qq 53, 54, 62 38 Qq 59–60; DWP ARA 2022–23, pages 293, 376 39 Qq 61, 63–64 40 Q 58; DWP ARA 22–23, pages 274, 293, 374 41 Qq 68–71 42 Qq 55, 65–67 The Department for Work & Pensions Annual Report and Accounts 2022–23 15 assumes a completion date in 2027–28, but DWP told us it hopes to do it quicker than this.43 It suggested that caseworkers with experience remediating previous State Pension underpayments have the right skillset to roll forward onto HRP, which may help speed the process.44

We asked HMRC to confirm whether pensioners receiving lump sum back payments would be liable for a tax charge and whether it would take a sympathetic approach to handling any issues. HMRC confirmed that the tax charge could apply. It told us that “Generally, we will be as generous as we can be”, but that discussions were still ongoing within HMRC to understand the level of discretion that can be applied.45 Integrity of the National Insurance records

DWP now reports multiple underpayments relating to issues in the National Insurance record. In addition to the HRP underpayment, DWP has also identified that the National Insurance records for 10 million people claiming Universal Credit have not been updated properly. DWP estimates that 137,000 of these people have already reached State Pension age, and a smaller proportion may have been underpaid.46 We challenged DWP to explain what it is doing to make sure that underpayments relating to inaccurate or incomplete benefit records do not happen again. DWP explained that both the National Insurance and Child Benefit records are the responsibility of HMRC, but that DWP has an interest in making sure it has an accurate National Insurance record to pay the correct State Pension.47 It told us that it has commissioned a joint internal audit review with HMRC to look at how the National Insurance record is produced and whether there are any issues in terms of quality and assurance. It expects this to be complete by the end of 2023, and told us it will say more on this work in its next Annual Report & Accounts.48

DWP noted that the government has recently announced a new National Insurance credit for people subject to the High Income Child Benefit Charge.49 It explained that the charge applies when one parent in a household earns more than £50,000, and means that they must complete a self-assessment tax return to pay back some or all of any Child Benefit they have claimed. We observed that some parents will choose to opt out of Child Benefit entirely rather than complete a tax return, which means they miss out on the associated National Insurance credits.50 DWP told us the newly announced credit is designed to fill this gap. It also acknowledged there is a risk that people miss out on this credit and stressed that it wants to avoid a problem similar to HRP occurring again in future due to incomplete Child Benefit or National Insurance records. It told us it is working with HMRC to decide how to hold onto the relevant information long-term, because may need it over 30 or 40 years as people reach State Pension age.51 We asked HMRC whether this new credit would be awarded automatically as a way to avoid incomplete or inaccurate 43 Q 67; DWP ARA 2022–23, page 376 44 Q 37 45 Qq 73–76 46 DWP ARA 2022–23, page 294 47 Qq 3, 48–49 48 Qq 5, 93 49 Q 12 50 Qq 38–39, 47 51 Qq 12, 38–41 16 The Department for Work & Pensions Annual Report and Accounts 2022–23 records in future. HMRC told us it is working through the legislation and aims to make it as simple and easy as possible for customers to claim the new credit, but could not commit to the process being automatic.52 Detecting and preventing further systemic underpayments

During our examination of DWP’s 2021–22 Accounts, we were unconvinced that its control processes were adequate to detect underpayments before they build up into major issues.53 DWP has previously acknowledged “an inability to pick up patterns of underpayment, which had been going on for many years”.54 We asked DWP to explain its current approach to detecting systemic underpayments. DWP explained that it relies predominantly on manual sampling, including the annual exercise to produce its fraud and error statistics. Where issues are identified they are passed on to its wider counter- fraud teams who will investigate the underlying causes. It further explained that 2021–22 was the first year it had used an enhanced sampling methodology that involved contacting pensioners directly to discuss their claim, which is what enabled it to detect the HRP issue.55 We asked DWP what it is doing to make sure it is able to detect issues early in future. DWP acknowledged that while its manual sampling has been helpful, there is more that it can do to avoid systemic underpayments. It explained that, in response to recommendations by the NAO, it is working on bringing together a wider range of intelligence about underpayments. It added there is a lot of work to do in this space and that it would report back to us on this point.56 52 Qq 42–47 53 Committee of Public Accounts, The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefit system, Twenty-Sixth Report of Session 2022–23, HC 44, 9 November 2022 54 Committee of Public Accounts, Underpayments of the State Pension, Thirty-Third Report of Session 2021–22, HC 654, 12 January 2022 55 Qq 3–5, 7–10 56 Qq 3, 72 The Department for Work & Pensions Annual Report and Accounts 2022–23 17 3 Specific activities to reduce fraud and error

DWP set out its plan to tackle benefit fraud following the pandemic in May 2022 in Fighting Fraud in the Welfare System.57 This includes £895 million of additional investment in counter-fraud activities over the Spending Review period covering the three years to March 2025. Details of the plan now published by the NAO show that DWP expects this investment to generate savings of £9.4 billion by 2027–28 by reducing benefit fraud and error.58 We focused our questioning on two elements of DWP’s counter-fraud plan: • A project to review millions of Universal Credit claims; and • Using machine learning to detect fraudulent benefit claims. Reviewing millions of live Universal Credit claims

The most significant element of DWP’s counter-fraud plan is a project to cleanse the benefit system of incorrect payments by reviewing some 8 million live Universal Credit claims over five years. DWP expects this project—which it calls ‘Targeted Case Reviews’ (TCR)—to generate £6.4 billion of savings by 2027–28. DWP is investing £443 million in TCR over the spending review period to March 2025.59 DWP told us that it considers TCR to be the best way to bring fraud and error back down to pre-pandemic levels, describing it as a “huge, huge intervention”.60

The NAO has reported that in 2027–28 alone DWP expects 2.5 million Universal Credit cases to be reviewed by some 5,900 staff. For comparison, in 2022–23 DWP reviewed around 3,600 Universal Credit claims to produce its fraud and error statistics.61 We challenged DWP to explain whether it has enough people to deliver the project and how it will achieve this ambitious scaling up. DWP told us that it has recruited around 1,700 people since launching TCR in January 2022 and is currently running a “massive” recruitment process. It stressed that it is on track and that it knows how long it takes to recruit the people it needs. It also told us that instead of delivering TCR entirely through civil servants as initially planned, it instead aims to recruit around 3,600 people from the public sector and to outsource the remainder, some 40% of the 5,900 people it needs, to private sector contractors.62

DWP told us it will take two years for TCR to have a measurable effect.63 We observed that TCR is an expensive intervention and questioned DWP on whether it has been under pressure to deliver the project faster in order to demonstrate a credible return on investment.64 DWP acknowledged that it needs to increase productivity significantly to achieve the expected £6.4 billion of savings by 2027–28, and there is therefore some 57 DWP ARA 2022–23, page 96; Department for Work & Pensions, Fighting Fraud in the Welfare System, CP 679, May 2022 58 DWP ARA 2022–23, pages 296, 297, 304 59 Q 82, DWP ARA 2022–23, pages 107, 305–308 60 Qq 16, 96 61 DWP ARA 2022–23, pages 308, 394 62 Qq 82–83, 91 63 Q 20 64 Qq 88, 92, 97 18 The Department for Work & Pensions Annual Report and Accounts 2022–23 pressure to speed up the pace of reviews. It explained that it does not expect the hit rate of incorrect claims to rise much higher than 30%, and so it expects savings will need to be driven by the volume of cases cleared.65 The NAO has reported that DWP expects newly trained TCR agents to review 0.5 cases per day, ramping up to 2.0 cases per day across all agents by 2027–28.66 DWP noted that it intends to ensure that the quality of reviews is not sacrificed to maximise productivity, so that that learning can be captured from the reviews and fed back into continually improving Universal Credit.67 Using machine learning to flag potentially fraudulent benefit claims

DWP is investing some £70 million to March 2025 in expanding its use of advanced analytics to tackle fraud. This includes using machine learning algorithms to flag potentially fraudulent benefit claims. DWP has already piloted an algorithm to detect fraudulent Universal Credit advances claims.68 The NAO reports that DWP is now actively developing similar tools for the four main risk areas of Universal Credit. We have reported previously that DWP could be more transparent in its use of machine learning in order to support public trust in the fairness of the benefit system.69

We received written evidence from the Child Poverty Action Group and from the Public Law Project expressing concern about the potential unfairness of machine learning, particularly with regard to vulnerable claimants and people with protected characteristics.70 We asked DWP whether it understood the concerns of people who have warned of unintentional bias in its use of machine learning. DWP assured us it shared these concerns and that is why a human always makes the final decision on whether to make a benefit payment.71

We challenged DWP to explain how it would address the risk that legitimate benefit claims are unfairly delayed or reduced as a result of an algorithms targeting innocent behaviour, such as frequent changes of circumstances. DWP acknowledged that some level of algorithmic bias is to be expected because of how benefit payments work, for example Universal Credit payments are higher for people aged over 25, so older claimants are more likely to be flagged because fraudsters will tend to claim to be older. It asserted that while there “clearly is a hypothetical risk” of unfair impacts on claimants, that there is no evidence of that risk manifesting now. It explained that it is performing analysis regularly to identify bias in the outputs of its algorithms.72 But the NAO has reported that so far this analysis has been largely inconclusive because of limitations in the available data about claimants.73 DWP told us it did not want to provide any further detail on how it will prevent unfair impacts to avoid tipping off potential fraudsters.74

DWP also told us it did not want to reveal when it planned to go live with machine learning on a large scale to avoid informing potential fraudsters, but added it was 65 Qq 84–90 66 Q 90; DWP ARA 2022–23, page 308 67 Qq 84–85 68 DWP ARA 2022–23, pages 102, 308 69 Committee of Public Accounts, The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefit system, Twenty-Sixth Report of Session 2022–23, HC 44, 9 November 2022 70 DWP0007; DWP0008 71 Q 101 72 Qq 101–103 73 DWP ARA 2022–23, page 309 74 Q 103 The Department for Work & Pensions Annual Report and Accounts 2022–23 19 working closely with the relevant authorities and that Ministers would be aware of its plans.75 However, DWP claimed that it is “taking it very slowly” with regards to rolling out machine learning. It explained that its pilot algorithm to detect fraud in Universal Credit advances did not work very well at first and needed to be tested and iterated using a small number of cases before being released for wider use. It added that it intends to follow this approach going forward and will not roll out new algorithms more widely until they have reached a level of accuracy that avoids unnecessarily holding up legitimate payments.76

In our November 2022 report on DWP’s 2021–22 accounts we recommended that DWP should report annually to Parliament on its assessment of the impact of data analytics on protected groups and vulnerable claimants.77 DWP told us it thought the right way to do this would be to report annually in its annual report and accounts.78 In correspondence after our evidence session it confirmed that would be the case, and that its first such assessment would be included in its 2023–24 report and accounts. DWP stated that its first assessment would provide a view on any bias detected, and whether this is in line with its expectation. DWP also stated that the assessment would provide indications of the type of mitigations put in place to reduce the risk of unfairness within the overall system, or actions taken to address issues.79 75 Qq 105–106 76 Q 102 77 Committee of Public Accounts, The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefit system, Twenty-Sixth Report of Session 2022–23, HC 44, 9 November 2022 78 Q 104 79 Correspondence from DWP dated 24 October 2023 20 The Department for Work & Pensions Annual Report and Accounts 2022–23