Robodebt: How did it come to this?
A retrospective of one of Australia’s most disastrous policies
No doubt fuelled by A Current Affair’s frequent exposés of “dole cheats”, which began appearing during the 1980s, Australians for a long time haven’t liked welfare fraud. In the mid 1980s, survey data had us claiming that defrauding the social security system of $1000 was (somehow) worse than tax evasion worth five times that. The Australia Card was supposed to crack down on dole cheats until it was killed off by a well-coordinated campaign highlighting its privacy implications. But in 1990 we got the Data-Matching Program (Assistance and Tax) Act, Australia’s first law allowing departments to match limited social security data with that held by the Australian Taxation Office. John Howard campaigned strongly on the need for more crackdowns before the 1996 election, and, by the end of his prime ministership, Howard had created a pilot scheme that matched Centrelink’s “customers” with their Commonwealth Bank accounts. It was an unalloyed success, netting more than double the anticipated savings of $184 million over five years. Kevin Rudd and Julia Gillard both built on those early forays, involving other banks as well as routine ATO matching. One man who thought Labor’s efforts went too far in these years was Brendan Nelson, who as Opposition leader was worried the Rudd government wanted “some sort of big brother”. “Every person in this country”, Nelson said straight-faced, “is entitled to at least a minimum level of privacy.”
Until 2015, “discrepancies” were checked manually by public servants, and the process went something like this: an individual’s annual income as reported to Centrelink was matched against the income assessed by the ATO for the same period; and the Department of Human Services would use its information-gathering powers under the Social Security Act to demand payment summaries and payslips from employers to enable it to check any discrepancies raised. Then came the Income Compliance Program (ICP), ushered in as part of the 2015 Budget’s “Better Management of the Social Welfare System” measure. “Better” meant “cheaper”, at least for the department: rather than have it run around asking people’s old bosses for collateral, the onus was shifted to the individual to explain any alleged discrepancy. Centrelink began sending out threatening letters asking people to prove that they really were entitled to every dollar of Youth Allowance they received in each fortnight up to seven years earlier. If people didn’t reply, or they didn’t reply adequately, a departmental “compliance officer” would – instead of hunting down the missing evidence themselves, as they’d done previously – simply take the annual income as assessed by the ATO, average it across the year, assess it against the Centrelink entitlements for each fortnight and raise a debt accordingly.
There were two obvious problems with this approach from the beginning, both of which flowed from the fact that the ATO’s assessments are annual whereas Centrelink’s assessments are fortnightly. The first is that “averaging” a person’s annual income across 26 fortnights only makes sense if that person’s income didn’t change from fortnight to fortnight. But that would only happen if a person was employed on a permanent basis by the same employer for the entire year. Many people claiming Newstart or Youth Allowance worked casually or intermittently, meaning that “averaging” is a useless exercise: its results must be erroneous, by definition. The second problem is the sheer difficulty of proving that you correctly reported your income for a particular fortnight six years later.
These problems were so obvious to anyone who’s ever had to negotiate Centrelink’s system of income reporting that it was difficult to believe that highly paid bureaucrats and ministers couldn’t see them. Yet in 2016, they compounded these problems by automating the ICP. Now called the Online Compliance Intervention (OCI), it required people to disprove (often bogus) discrepancies by uploading their own seven-year-old payslips. People who didn’t had their income “averaged” and a debt automatically – and illegitimately – raised. The whole thing happened without an assessment by a real person. The number of “compliance interventions” went from 20,000 a year to 20,000 every week. The OCI and its successors – the Employment Income Confirmation (EIC) and the Check and Update Past Information (CUPI) systems – cost $606 million, but they more than paid for themselves. Or so it seemed. Scott Morrison (then treasurer) and Christian Porter (then social services minister) congratulated themselves on a job well done.
The decision to automate, with algorithms, the cross-checking of ATO and Centrelink data was probably inevitable. It wasn’t just the prospect of the savings promised by instant, computerised matching. It was also the lure of surveillance, which bureaucracies tell themselves is really about information integrity. Even in 1840, Alexis de Tocqueville could see that modern democracies depended on bureaucratic documentation. He also observed the irony that this produced: privatised citizens so concerned for their individual welfare that the state institutions created to protect it end up crushing their autonomy. The modern anxiety about privacy is itself a response to the welfare-oriented bureaucracy.
Max Weber witnessed the emergence of the 20th-century bureaucracy, which is forever looking for better means to collect, store, retrieve and use the more and better data it needs to rationalise the tasks it performs. Obviously, this means technological innovation. By the 1950s, the French sociologist Jacques Ellul claimed that technology exhibited its own logic. If something could be done more efficiently, or more exactly, or more certainly, then it would be – regardless of social consequences. Add some misplaced neoliberal ideology into the mix, and at some point our experience of the welfare state feels less like the New Deal and more like a Franz Kafka novel.
And there were social consequences. Thousands of people never received the initial letter asking them to disprove the bogus discrepancies, so the first they knew about the whole thing was when they got a call from a debt collector (to whom Centrelink now routinely outsources debts). Thousands more did try to engage, but encountered online clunkiness and unhelpful staff – when they could even get to them, as hold times for Centrelink calls blew out to multiple hours. Many of the debts were wrong, but people were finding they couldn’t do anything about them. Centrelink was routinely adding a 10 per cent penalty loading to each automated debt. Regular media stories from late 2016 reported distressed and even suicidal social security recipients. OCI became known as “robodebt”.
In February 2017, the Greens and Labor used their numbers in the Senate to refer robodebt to the Community Affairs References Committee. Liberal James McGrath said the government “strongly” believed in the OCI, and promised: “This practice will continue.” The committee reported in June and found multiple problems, echoing a Commonwealth Ombudsman report two months earlier. In a dissenting report, Liberal senators acknowledged teething issues but were satisfied that the department had addressed them. Robodebt continued. In the May 2018 budget, the then treasurer Scott Morrison forecast its extension into 2021.
The following month, Centrelink raised a robodebt of $3725.04 against a Victorian woman, Madeleine Masterton, and added its 10 per cent penalty. The following February, Victoria Legal Aid filed an application for judicial review in the Federal Court, alleging that both the debt and the penalty were unlawful. Centrelink responded by first reducing Madeleine’s debt to $602.87, and then, after removing the “averaging”, to zero. So VLA tried again, this time on behalf of Deanna Amato. Amato had switched from full-time to casual employment in the February of the year during which her robodebt of $3215.38 was raised. She only became aware of the debt when her subsequent tax refund was entirely retained as payment. As soon as she filed her case, Centrelink again recalculated her debt – and reduced it to $1.48, which it generously waived before refunding the amount it had already recovered. These recalculations were now commonplace whenever people challenged their robodebts. Staggeringly, a Freedom of Information request revealed that Amato had in fact been underpaid by nearly $500. Robodebt was clearly farcical.
By now, Labor and the Greens had referred robodebt to a committee for the second time. Then, two weeks before Amato’s final hearing in December 2019, Centrelink stopped its four-year-old “averaging” of ATO incomes across each fortnight, and admitted to the Federal Court that debts produced by this process weren’t valid. Centrelink couldn’t defend its case. It consented to court orders declaring Amato’s debt – and her 10 per cent penalty – to be invalid, and requiring the government to pay Amato’s lost interest and legal costs. The orders were as unequivocal as they were predictable. An inevitable departmental review identified 470,000 robodebts worth $721 million to be similarly illegal, and the government said they would be repaid.
Three months before the Amato decision, Bill Shorten – the shadow minister for government services – stood with lawyer Peter Gordon in Canberra and announced a major class action on behalf of 600,000 social security recipients, who were suing the government for damages and “unjust enrichment” (the civil version of theft). In May this year, Christian Porter – whose capacity for making decent choices is by now highly suspect – conceded its illegality. In June, Prime Minister Scott Morrison apologised for it. And famously this week, the federal government agreed to a whopping $1.2 billion settlement to prevent the class action going to trial.
The point at which the welfare state tipped into a Kafkaesque nightmare was, it turns out, the point at which the state began literally stealing from some of its most disadvantaged citizens. Robodebt, a technique designed to recoup hundreds of millions, has now cost the country close to $2 billion. It’s difficult to think of a more disastrous and inept domestic policy in Australia’s entire political history. How had very senior public servants created robodebt – with its very obvious and fatal design flaws – in the first place? How had ministers prosecuted it? And why did the government use up enormous public resources defending it?
These are questions Labor wants answered by a royal commission, which it began calling for in mid 2020. But royal commissions are themselves extraordinarily expensive. For it to benefit anyone other than lawyers, this one would need to enquire into the public service’s lost ability to tell ministers what they don’t want to hear – a consequence of Howard-era reforms that made departmental heads disposable and therefore highly sensitive to government directives – and the ease with which the machinery of the welfare state was turned against the very people it was built to help. Morrison, busy recasting his public image as an everyman – won’t want an independent inquiry going anywhere near those questions. But not only are robodebt’s architects – Alan Tudge, Porter and Morrison himself – still there. They’re the government’s most senior ministers. In other eras, they wouldn’t have survived a scandal like robodebt.
Russell Marks is a lawyer and an honorary research associate at La Trobe University. He is the author of Crime and Punishment: Offenders and Victims in a Broken Justice System (Black Inc., 2015).
No doubt fuelled by A Current Affair’s frequent exposés of “dole cheats”, which began appearing during the 1980s, Australians for a long time haven’t liked welfare fraud. In the mid 1980s, survey data had us claiming that defrauding the social security system of $1000 was (somehow) worse than tax evasion worth five times that. The Australia Card was supposed to crack down on dole cheats until it was killed off by a well-coordinated campaign highlighting its privacy implications. But in 1990 we got the Data-Matching Program (Assistance and Tax) Act, Australia’s first law allowing departments to match limited social security data with that held by the Australian Taxation Office. John Howard campaigned strongly on the need for more crackdowns before the 1996 election, and, by the end of his prime ministership, Howard had created a pilot scheme that matched Centrelink’s...
Nothing without context. Politics, society, culture.