Why new Treasury rules on estimating fraud and error could help drive savings

New Treasury guidance requires that departments report on the total level of loss across their programmes. Here's why it matters and how the NAO can help
Photo: Adobe Stock

By Joshua Reddaway

07 Feb 2025

The NAO has released a new good practice guide on reporting fraud and error. This will help departments to implement new HM Treasury guidance that requires them to report on the total level of loss across their programmes. Here's why it isn’t just accountants who should care.

Fraud and error are a big problem for government

We estimate that fraud and error cost the taxpayer between £55bn and £81bn in 2023-24. Two things stand out in our estimate.

One – the numbers are huge, and it would only take a small percentage saving to give a big bonus to the exchequer.

The other is that it’s a really big range. That’s because, for around £560bn of spend and income, there is no proper fraud and error measurement. In other words – over half a trillion pounds where government simply doesn’t know how much it loses.

Why estimates matter

Government needs to know where the risks are, measure and report how much it loses, and state what it’s doing so it can plan how to bring down these losses.

This measurement must go beyond just tallying up how much it has found - it needs to include the fraud and error losses that government hasn’t yet detected. This is important not only for public accountability, but also to making the business case for doing something about it.

The UK has already made strides in that direction. Since 2021, in the midst of the Covid-19 pandemic, HM Treasury has required all departments to provide an “evidence based” assessment of the level of loss from fraud and error “in all material distribution schemes”. As a result, government recorded £10.5bn of fraud and error in the Covid-19 related schemes in its accounts. This reporting brought much needed focus to the issue of tackling fraud.

But it wasn’t always clear what was required or how to measure something you cannot see.

HM Treasury has now sent new guidance to departments to clarify the requirements around fraud and error for 2024-25 annual reports and accounts. It’s based around two new tests that we consulted departments on when we were producing our new good practice guide. The two tests are:

  • Whether the likely level of fraud and error in an activity is significant
  • Whether that activity is significant to the body as a whole.

If the answer to both these tests is yes, then departments must now report an estimate about the total amount of fraud and error in those activities. This goes in the performance report section of their annual report and accounts.  

How our guide can help

Departments still have a lot of discretion when deciding what exactly significant means, but our guide gives the tools to help make decisions about what is required and how to start managing fraud and error down.

We hope that the new requirements, together with our good practice guide, will inspire much better reporting and a renewed impetus to tackle taxpayer losses to fraud and error.

So how do you estimate a hidden crime?

Essentially there are three ways government can go about estimating its loss from fraud and error:

1. Compare its activities to benchmarks. This is the least informative approach, but at least gives you an indication of the likely loss.

2. It can compare what it is paying or receiving against a model of what it should be. For example, the BBC compares its licence fee income to its model of the number of premises that should have one. To do this well, you need to have good assumptions.

3. Gather more information on a sample of transactions to see if it can detect an overpayment, looking at more information than it normally would before making a payment. For example, the Department for Work and Pensions asks some benefits claimants to provide further justification of their eligibility, and it does inspection visits for some grants. This depends on its ability to see fraud and error when it looks directly at the sample of items – some fraud such as people paid cash in hand – is very hard to detect.

The problem is that no route is easy, and all have drawbacks. One of the key things we urge is that departments shouldn’t let perfection be the enemy of the good, and that they view this as a longer-term operation to build the capability to do it well.

Our guide also lists risky design features in government activities. For example, where there are unfamiliar suppliers, where payments are made in advance, or where government finds it difficult to verify whether it has received what it is paying for. Such features are one of our ‘early warning signs’ for government bodies to consider producing an estimate – along with things like unusual levels of detected fraud and error, and clues from internal or external audit.

There are some good examples of estimating and reporting in government already, including estimates when public bodies don’t have perfect information, and when the activity in question is commercially or legally sensitive. We bring these out in our guide, showing, for example, how different parts of government use "benchmarking" of similar activities to help build more sophisticated estimates, and discussing the Department of Health and Social Care’s disclosures about fraud and error in PPE during the Covid-19 pandemic.

The NAO stands ready to help any government body thinking about its estimates and what it should report about fraud and error.

Take a look and let us know what you think.

Joshua Reddaway is director of fraud and propriety at the NAO. Read our good practice guide on estimating and reporting fraud and error

Read the most recent articles written by Joshua Reddaway - Why we should all be more proactive in declaring potential conflicts of interest

Share this page