Australia's tax office cut inflated claims by writing to the accountant

Most tax-compliance nudges target the taxpayer directly — reminder letters, simplified forms, 'most people pay on time' social-proof messages. Australia's tax office had reason to think that approach was hitting a ceiling for work-related-expense deductions specifically, since most individual taxpayers don't actually calculate these claims themselves: a tax agent does, often applying the same template across dozens of similar clients.

BETA, the Australian Government's behavioural-insights unit, and the ATO targeted the professional filling in the number instead of the person whose name was on the return.

In a trial starting March 2017, the ATO sent letters not to taxpayers but to roughly 1,150 of their tax agents, informing each agent that their clients' work-related-expense claims collectively ranked among the riskiest percentage band compared with other agents' client books, and noting that the ATO's data-analytics systems flag larger-than-expected claims. No taxpayer received anything; the accountant preparing the numbers did.

Sent letters to tax agents, not taxpayers, benchmarking each agent's client claims against peers, targeting the person who actually set the numbers.

Tax agents, not taxpayers, calculated the claims, so letters to agents reached the actual decision-maker.

Benchmarking an agent's entire client book against peers made the risk a reputational fact about their practice.

One letter to an agent affected dozens of returns at once, multiplying the impact per letter sent.

When an intermediary shapes the end behavior, target the intermediary with peer benchmarking, not the end customer.

Agents who received the letter saw their clients' claims fall by $2.2 million and tax paid rise by $0.9 million relative to a similarly sized control group of agents. The ATO projected that scaling the letter to its full at-risk agent population would cut claims by about $4.4 million and lift tax paid by about $1.7 million.