False accusations of fraud: lessons from the Carer’s Allowance Overpayments review

Is there anything more self-defeating than a benefits system that falsely accuses people of fraud?

Is there anything more self-defeating than a benefits system that falsely accuses people of fraud? The system works hard to provide people with more security, but then it takes this money away (sometimes with added penalties on top) in a way that is both humiliating and unjust, making people feel even less secure than they did before claiming.

Yet this is something that benefits systems around the world have sometimes done on a significant scale. There’s the Robodebt scandal in Australia; the Dutch childcare benefits scandal – and closer to home in the UK, there’s the Child Benefit travelling abroad scandal, and (my focus here) the Carer’s Allowance Overpayments scandal.

After years of increasing pressure, then-Secretary of State Liz Kendall asked Liz Sayce in late 2024 to conduct a review into Carer’s Allowance Overpayments. The review came out in November 2025, but in the face of everything else going on in the world, I don’t think it got the attention it deserves – hence this post, where I want to talk about some lessons that have largely been missed. (There’s a lot of lessons, so this post sprawled to longer-than-usual…).

Full disclaimer: I was on the advisory group to Liz Sayce’s review, along with a bunch of people that knew more about Carer’s Allowance than me: Jeremy Moore, Seyi Obakin, Imelda Redmond, and Sue Yeandle. The excellent review is very much Liz’s own, and equally, this blog doesn’t reflect the views of Liz nor the other advisor members nor the excellent civil service team supporting the review – but I learnt a huge amount from our discussions, which I’m enormously grateful to everyone involved for.

The biggest problem, and how DWP are fixing it

You can think about what went wrong on two levels: design and implementation. The deeper problem is that Carer’s Allowance (CA) is a badly-designed benefit. It’s paid to people providing 35+ hours/wk of care to someone claiming disability benefits, but it has a cliff-edge condition – you are only entitled to CA if you earn less than a fixed earnings cut-off (currently £196/wk net, equivalent to 16 hrs/wk work on the National Living Wage). Cliff-edges like this are horrible; Liz Sayce spoke about examples of people turning down promotions, cutting hours or even changing to worse-paid jobs because they are so much worse off if they go over the threshold.

But if there’s a stupidly-designed threshold, then from the DWP’s perspective, claimants still have to stick to it – it’s Parliament’s job to change the benefit itself. (As a side note: there’s work underway in DWP to figure out what a better alternative might be, and there needs to be a big push to make this happen).

The other way of thinking about ‘what went wrong’, though, is that claimants couldn’t reasonably comply with this (badly-designed) requirement. In particular, lots of low-paid people have fluctuating earnings,¹ and if your income variously goes above and below the cut-off, it’s difficult to know if you’re still eligible. CA claimants are told that their income may be ‘averaged’ over time, which seems reasonable – but as Liz’s review explains, “current operational guidance within DWP is so restrictive that staff say averaging never or virtually never happens in practice.” There was no way of knowing in advance if you were eligible; there was no way of even checking in real-time; you were left in the dark. As one carer quoted in the report put it, “It’s like playing a game where only the other side knows the rules”.

DWP has put its hands up on this issue, accepting that the DWP’s operational guidance was not consistent with the decisions made by Parliament about how the benefit should work. They are not only changing this going forward, but they’re also going through about 210,000 historic overpayment decisions by hand, and DWP estimate that debts will be cancelled or reduced for 26,000 carers (which will cost DWP £75m, including the costs of figuring out which debts to cancel). This shouldn’t have happened; but given that it did, this is a serious attempt to put right some past injustices.

Where concerns remain

However, this is only part of the picture, and there’s three areas where concerns remain.

Firstly, the averaging guidance relates to only 10 of Liz’s 40 recommendations. The DWP rejected two recommendations outright² and also one part of a crucial recommendation on ‘allowable expenses’: these are things that carers had to spend for work (ranging from uniforms to travel to a replacement carer) that get taken off their income, potentially pushing it under the threshold. Liz argued that this process was unfair and opaque: carers aren’t told which expenses are permitted or not. But DWP said that “we do not agree that the allowable expenses guidance was unclear or ambiguous”, so they’re not compensating people for this. This is just a difference of opinion, and interested readers should look at Chapter 4 of Liz’s review to form their own view.

Secondly, more worryingly, to some people it seems like the DWP is just pretending to agree with the review, while seeking to undermine it behind the scenes. Much of this impression comes from an ill-advised internal blog post by Neil Couling (one of the most senior people in DWP) that said that ‘at the heart’ of the problem is carers’ ‘failure to report changes in circumstances’, rather than the DWP’s mistakes. (The Secretary of State had to publicly say that “what he [Couling] said is not the position of the department”). This in itself has led to some damaging media stories and was discussed at the Work & Pensions Select Committee (both in questions to Liz Sayce and to senior DWP staff).

But this concern also comes from some of the small print in the DWP’s response. The Government said that they have accepted 38 of Liz’s 40 recommendations – but lots of these are ‘partially accepted’, and when you look at what this means, it often looks pretty close to rejecting them. For example:

• One recommendation was to create an online calculator so that claimants with fluctuating earnings can easily see if they’re going over the earnings limit. DWP’s way of ‘partially accepting’ this is to launch a user research project on the best way of giving people information, without committing to create a calculator.

• One recommendation was to publish standards for better treatment for Carer’s Allowance customers. DWP’s way of ‘partially accepting’ this is to point out that they already have a customer charter.

There’s also been an angry letter from Debbie Abrahams (Select Committee chair) to the DWP about how slowly the Department is taking things forward; carers are still being hit by accusations of fraud while we’re waiting for changes to take place.

For what it’s worth, I don’t think it’s fair to say that DWP is only pretending to make major improvements in response to Liz’s review – the DWP appointed a senior responsible owner to coordinate their response (sounds boring, but crucial for making things happen), it’s reassessing 200,000 debts and is going to hand £75m back to claimants (at a time when the public finances are stretched), and it’s got specific plans to improve several processes, which it’s committed to publicly reporting on every six months. This isn’t the hallmark of tokenism.

But still: while things are definitely going to get better, there are ways in which DWP is dragging its heels – as Liz Sayce put it, there are forces for change in DWP, but also ‘forces of resistance’.³

What needs to change in DWP

My final lesson, though, is that we need to pay attention to a related but different issue: the gap between fraud repayments teams and the rest of DWP.

To be clear: we need to have teams that investigate benefits fraud. But it seems like there is no connection between the wider aims of the benefits system and the aims of the fraud prevention team. If we want to provide people with security, in as dignified and supportive and fair a way as possible, then this needs to be integral to how the DWP treats people when it’s investigating fraud too – particularly because the overwhelming majority of people investigated for fraud are found to have done nothing wrong.

The Carer’s Allowance Overpayments scandal is a case in point. It’s not just that the system was so confusing that it was very difficult for overstretched people juggling work and caring to comply – it’s that when this happened, the DWP went after people aggressively, often making them feel humiliated, unfairly treated and wildly insecure in any future interactions with the DWP. As Anna Dixon MP put it, “We see that carers are saying they were pursued as if they were criminals, although they were trying to do the right thing.” To make matters worse, this sometimes echoes people’s feeling of being distrusted and humiliated in other ways, ranging from conditionality to disability assessments.

Fraud investigations are one of the most important and least-understood parts of the benefits system. Lots of the DWP’s recent/future planned savings come from investing in fraud prevention, using a combination of people power, data linkage and AI tools (to be informed about this, I thoroughly recommend reading/listening to this and this). By last April there had been over a million Targeted Case Reviews, with millions more planned. And HMRC is doing the same, making a spectacular mess out of a recent attempt to reduce Child Benefit fraud by removing it from people that it thought was living abroad (without checking that the data on people’s movements was accurate). But there is almost no research on all of this – an admirable exception being this essential reading from the Administrative Fairness Lab.

As far as I can tell, the problem isn’t about individual members of staff – it’s about people doing what they’re told to do. Some DWP staff are told to support claimants, and some do it well. But other people are told to save the taxpayer as much money as possible, and in trying to do that job well, the experience for claimants is far from their minds. If we are to have a benefits system that genuinely provides security, in a dignifying way, and feels supportive, then this needs to inflect fraud prevention much more than it currently does.

Finally, in a blog post I can only scratch the surface of everything that’s happened around Carer’s Allowance, so if you’re interested then do read Liz’s full review.

1. This is an opportunity to flag John Hills, Rachel Smithies & Abigail McKnight’s amazing study of short-term income fluctuations, which doesn’t get as much attention as it deserves.

2. One was a recommendation to ‘Re-categorise error and fraud figures in published statistics so that the ‘fraud’ category does not include cases where no intention of fraud has been found’, the other was to commission a detailed operational audit.

3. Liz’s quote in full: “I was really distressed by that blog [by Neil Couling], as I am sure many people were. What we were hoping for at that point was that senior people would share with colleagues across the Department the seriousness of this, what had been learned and what would be put right, not an attempt to almost minimise it or again to place the responsibility back on to carers, as if it was their fault, which the review clearly found that it was not. …

“When I was doing the review, I found officials at different levels, including at the frontline, who were serious about wanting to improve things, trying to be more empathetic and so on. Since then, I can see that there are some people who want to learn and make change, but there are also forces of resistance, which worry me.”