A key issue in this policy failure was the speed at which a complex policy was pushed through by the bureaucracy to meet the desires of the government of the day. As the Albanese Government seeks to deliver on its election commitments, what lessons should be taken from the commission’s inquiries so far?
The robodebt scheme used inaccurate matching of welfare and tax office data, and replaced the intensive compliance checks that were in place before debt notices were issued, in favour of an automated process. Once the policy direction was set, there appears to have been an intense focus by public servants on implementing the Government’s policy regardless of the impact. The program involved structural changes to the nationwide implementation of welfare services across a range of communities, and yet it was assumed its implementation would be costless to both government and welfare recipients (they would lose money, but they were never “entitled” to it anyway). Greater use of digital solutions is key to unlocking future productivity gains in Australia. However, the robodebt case shows that technology in and of itself is not a quick fix and relies on effective design and implementation to succeed.
The Government’s direction to public servants was clear, best articulated in a speech by former Prime Minister Scott Morrison: “The public service is meant to be an enabler of Government policy, not an obstacle.” In this rush to enable, the new system glossed over the potential impacts on the most important person in welfare service delivery – the individual or family receiving support. In introducing an automated system and removing contact with welfare recipients via phone and letter, the process ignored the complexity of the welfare system, the use of weighting to calculate income and the need to cross-check errors in records.
At the time of the program roll-out, the Federal Government boasted that it would send 1.7 million compliance notices to welfare recipients. In testimony at the commission from compliance officer Colleen Taylor, she spoke of the difficulties of implementing an automated system: “You can't assume that the data match implies that there's a debt there … there are lots of examples where there's nothing in that. There's no discrepancy.”
The policy objective of robodebt was blinkered to focus on revenue from “fraudsters”. It was projected to improve government revenues and return the Federal Budget to surplus. In 2016 this was outlined as “strengthening the integrity of the welfare system by cracking down on fraud and overpayments ” to recover an estimated $4 billion in welfare-benefit overpayments between 2010 and 2018. Policies laser-focused on government revenues ignore the impact on Australians today, in favour of reducing the potential tax burden of future generations. The program had limited policy merit then, and the forecast revenues pale in comparison to the $895 billion gross debt the Australian Government has now accrued during the response to the COVID-19 pandemic.
These policy design and implementation issues were neither novel nor ground-breaking, even in 2016. In fact, as robodebt was being rolled out, the Royal Commission into the Home Insulation Program had just delivered recommendations to government on how to improve public-service delivery when there is intense political pressure to meet fiscal targets and to be seen to be acting. In its final report, Royal Commissioner, Ian Hanger AM QC, said: “Systemic or fundamental shortcomings can be identified which not only are capable of repetition ... but which might be avoided through diligence and the taking of some additional measures… mistakes identified in the report, many of which can be traced to overconfidence and unrealistic optimism.”
Even as the robodebt royal commission continues, there are already lessons for public-service leaders. As a new federal government seeks to make progress on its own commitments – such as emissions reduction targets, adjusting immigration settings and addressing overall cost-of-living pressures – we cannot afford another round of quickly implemented policies that deliver poor-quality services.
Policy is only as good as the implementation. This program was designed to increase revenue and streamline compliance processes – and that was how it was delivered. Implementation glossed over the human impact and there was little-to-no user voice in the design of this new system. This was a program that would have benefited from the voice of welfare recipients and scrutiny from community organisations and academics. While this would have delayed the implementation of the program, given its size and complexity, it would have improved its design.
Technology is often used as a scapegoat for human error or to obscure the human decision-making that will always play a role in service delivery. Regardless of the technology used, ethical guardrails are key to unlocking the benefits of data and avoiding future implementation failures. There are many frameworks available, including those outlined in CEDA’s disrupting disadvantage reports. Pilot programs can help identify the risks when trialling new technology, especially if using evaluation techniques such as randomised controlled trials. A pilot approach would have likely identified that data-matching and averaging were insufficient to be used as the only evidence for an accusation of welfare fraud and beginning a debt-recovery process.
Successive Intergenerational Reports have shown that the need for more innovative policy solutions will grow as demand for services rises and revenue declines. The public service must be empowered to deliver frank advice on both development and implementation for these reforms to be successful.