Counting the gains from Britain’s scam refunds

Britain’s requirement that payment firms reimburse certain scam victims has produced an encouraging estimate of lower losses, alongside a less settled verdict on its overall benefits. An evaluation published by Frontier Economics on July 1st 2026 estimated that losses from authorised push-payment fraud on Faster Payments fell by around 21% following the reimbursement requirement. That was equivalent to roughly £73 million a year. The Payment Systems Regulator commissioned the study.

Frontier compared a 2023 baseline with the first post-implementation year, adjusting for seasonal patterns and using a consistent six-month reporting window for fraud values. It combined that analysis with evidence about firms’ behaviour. The researchers could not construct a reliable untreated control group because the policy affected firms across the market. Their causal conclusion rests on the combined evidence, rather than an experiment isolating the rule’s effect.

Authorised push-payment fraud exploits the payer’s willingness to send money. The customer gives the instruction, but deception lies behind it. Reimbursement policy therefore has two jobs: reducing the harm after a scam and changing the incentives to prevent the payment beforehand. Success at one does not automatically prove success at the other.

Frontier reported that firms had strengthened fraud controls on both sending and receiving payments. It also estimated annual increases of £44 million to £56 million in payment providers’ costs, including prevention, claims handling, disputes, reporting and compliance. Set against the £73 million fall in losses, those costs implied a short-term net benefit of £17 million to £29 million a year, a figure Frontier described as conservative. Its assessment was that benefits probably exceeded costs in the first year, while longer-term effects on competition, innovation and behaviour remained uncertain.

The wider sensitivity analysis shows how much that judgement depends on assumptions. After adding assumptions about delayed fraud reports and displacement into international scams, Frontier’s expanded calculation ranged from a £4 million annual net cost to an £8 million net benefit. The authors still judged an overall benefit likely, partly because fraud might otherwise have grown. The resulting range crosses zero, leaving room for either a net cost or a net benefit under those assumptions.

Reimbursement is a separate measure

A separate regulatory dashboard helps describe claim handling. Updated on July 30th, it reported £316 million reimbursed over the 18 months from October 7th 2024 to March 31st 2026, representing 88% of the value of claims counted as reimbursable under its reporting measure. It also reported that 82% of claims were closed within five business days. The dashboard concerns reported, closed claims and Faster Payments transactions within its specified scope. It excludes internal transfers where both accounts are at the same firm.

The reimbursement percentage measures a different outcome from the estimated reduction in fraud. The dashboard and the commissioned evaluation also organise their observations differently. The regulator’s account of Frontier’s work notes that the evaluation uses the date of the scam transaction, whereas other industry statistics can use the date a claim closes. Coverage differs too. A change in the population being measured can change a reported rate without any change in how a particular victim was treated.

A useful assessment therefore has to keep several questions separate. Are losses falling in the covered payment channel? Are eligible victims receiving money promptly? Are prevention and administration costs proportionate? And is fraud being displaced into other channels? A single repayment percentage answers only part of that set.

A December consultation on claims

The regulator itself acknowledges uneven implementation. Its published roadmap schedules a December 2026 consultation on policy parameters and more consistent outcomes. Subjects include claims unresolved after 35 business days, the treatment of investment returns, and possible further guidance on customer caution, civil disputes and transfers between a person’s own accounts. These are planned consultation topics, not changes already in force.

The definitions determine which losses qualify and how claims are resolved. A customer who has lost money cares whether the payment qualifies, how the bank interprets the facts and how long the decision takes. Consistent definitions also matter to firms: rules that are applied differently can reward a restrictive interpretation rather than better fraud prevention. Clearer treatment of disputed cases would make it easier to judge whether firms are meeting the same standard.

There is a limit to what the payments industry can fix alone. The PSR’s response to the evaluation calls for technology companies and telecommunications firms to play their part in tackling the channels criminals exploit. Reimbursement reallocates losses inside the payment relationship; it does not, by itself, remove a fraudulent advertisement or stop an impersonation attempt from reaching a potential victim.

The first evaluation gives policymakers evidence of lower covered losses and stronger controls, with important limits on causal attribution and the net-benefit estimate. December’s consultation addresses a separate problem: whether comparable claims receive consistent treatment. A policy can reimburse victims promptly while remaining costly to administer, or reduce losses in one channel while fraud moves elsewhere. The evidence has to follow each outcome.