DWP’s Targeted Case Review

What is happening?

DWP anti-fraud activities include a targeted case review (TCR) of universal credit (UC) cases. This is described as the most significant component of DWP’s counter fraud plan.1

This work will build on the work already done to ‘retrospectively verify’ the 900,000 awards made while ‘easements’ on checking and verifying claim information were in place during the pandemic and identified as the ‘riskiest’ cases by the Integrated Risk and Intelligence Service (IRIS).2

Intention

TCR agents will review awards for ‘incorrectness’, which may or may not be the result of fraud. A claimant must have been receiving UC for six months to be eligible for review.3

The DWP expects to act on potential incorrectness identified by IRIS. The aim, assuming funding is extended beyond 2024–25, is to conduct around 8 million reviews by the end of the TCR and that this will, by 2027–28, prevent £6.4 billion of losses that would otherwise have been incurred. The TCR budget was £28 million in 2022–23, rising to £129 million in 2023–24, and £286 million in 2024-25. The DWP’s TCR exercise will recruit 5,900 agents by March 2025.4

The DWP states that ‘customers are supported to provide the right evidence, which verifies their circumstances and entitlement to claim universal credit. Our agents work closely with vulnerable customers to support them through the process’. The DWP says that overpayments will be recovered and, where appropriate, prosecution or penalty action instigated. The DWP does not record data on time taken to contact claimants whose case is being reviewed.5

Progress so far

The TCR has started by reviewing the lower risk cases which entered the UC system during the pandemic, and which are still in payment.6

By March 2023, the DWP had recruited 840 staff, conducted 25,000 reviews and claimed to have saved £39 million (of which £14 million related to 2022–23 expenditure).7

The DWP has begun recruiting a large number of TCR staff. Over 40 per cent of staff carrying out the TCR reviews will be outsourced, with training provided by the outsourced provider. It appears that the DWP’s internal recruitment into TCR roles is causing knock-on problems in other areas of the DWP.8

Areas of focus

The DWP will conduct a detailed review of the major drivers of fraud and error, generating intelligence on new and emerging areas of fraud, and will look at more areas of incorrectness as it carries out the TCR. Intelligence gained will inform changes to UC policies, service design and processes.9

The DWP has stated that the TCR will help to identify incorrect awards among the self-employed and people living together as a couple, and will presumably use the TCR to target other high-risk areas, including capital and housing.10

The TCR will partly align with the system of conditionality, so that certain sanctioned claimants who disengage with conditionality will be expected to engage with the TCR.11

Blanket approach

The DWP says that ‘[w]here incorrectness is found, claims will be corrected retrospectively’ and resulting overpayments recovered. Any such blanket approach could result in unlawfulness – for example, it would be unlawful to revise a decision where there is no statutory ground to do so. The DWP took the same blanket approach to the ‘retrospective verification’ of awards made during the pandemic, and CPAG is aware of decisions apparently taken in accordance with that blanket policy which were successfully challenged at the First-tier Tribunal without the DWP seeking to appeal further to the Upper Tribunal.12

Prior to any decision, if the award has been suspended advisers should consider whether the DWP is correctly using its suspension power – if not, then in these cases although there is no right of appeal, the available legal remedy is judicial review. Once an entitlement decision (eg, following suspension) has been made, claimants have a right of appeal, subject to the mandatory reconsideration requirement. Advisers assisting with revisions and appeals should ensure that they consider whether the DWP correctly applied the legislation on suspension and termination, and on revision and the effective date of supersessions.13

Effectiveness

Thirty-two per cent of reviewed awards have so far been found to have contained inaccuracies, according to the DWP – which the National Audit Office (NAO) notes is similar to the level of incorrectness found in the DWP’s random sampling exercises.14

The DWP has identified uncertainties around assumptions on productivity rates of TCR activities and the value of overpayments detected. The DWP estimates a roughly constant rate of incorrectness identified – around 30 per cent – over the lifetime of the TCR, described by the NAO as a ‘conservative assumption’.15

Automation, data matching and machine learning

IRIS uses automation and data matching to identify cases to refer to the TCR. Data matching to investigate benefit fraud has caused problems and hardship in the past. It is worth noting that the TCR will make use of special category data (ie, particularly sensitive data which could include data concerning health, racial origin, etc) or data which reveals or concerns these types of data.16

The DWP intends to improve its analytics to better target the TCR at awards that are more likely to be incorrect. The DWP intends to use machine learning to better target TCRs. But, although the DWP is developing and piloting new models, it seems that machine learning is not currently in use in the TCR.17

The NAO states that there is an inherent risk that machine-learning algorithms are biased towards selecting cases for review from vulnerable people or groups with protected characteristics.18

The DWP says it is mitigating the risk of machine-learning bias by, first, ensuring that a human makes the decision and that false positives are included in the cases presented to decision makers and, second, that small-scale testing is carried out initially with analysis of any resulting bias. The DWP has committed to reporting annually on the impact of data analytics on protected groups and vulnerable claimants.19