Bringing robo-debts before the law: why it’s time to right a legal wrong

What is a robo-debt? Robo-debt is the popular name for a technological innovation introduced by Centrelink in mid-2016 to streamline social security payment obligation compliance and debt recovery. This ‘on line compliance’ (‘OCI’) program automated the way tax data from the ATO on total earnings over a period of usually 26 weeks, was averaged per fortnight and then matched against the earnings previously reported by social security clients, to determine the correctness of their past fortnightly rates of payment.

When payment rates based on the ATO average over the period of up to 26 weeks diverged from fortnightly amounts previously paid, clients were issued with a notice of a debt for the difference and invited to supply PAYE slips or other evidence showing they had correctly reported their earnings. These ‘debts’ were raised from as far back as 2010, so few people had kept such records.

The new OCI robo-debt program turned on its head Centrelink’s long-standing practice of treating data match discrepancies as a sufficient reason to conduct its own investigation of a possible debt and if necessary, use its extensive legislative powers to compel employers or banks to supply records when the social security client had not retained them.

Many of the supposed debtors held multiple short-term casual jobs, where earnings fluctuated from fortnight to fortnight. To smooth out some of this fluctuation, the social security means test for working age payments like Newstart (‘NSA’) and Youth Allowance (‘YA’) effectively enable unused free-of-income-test earnings allowances to be ‘banked’ and carried forward to a later fortnight where income has been earned. In combination this means that debts projected from ATO averages are often highly inflated, or in some instances there is no debt at all.

This was revealed by the Commonwealth Ombudsman in early 2017. Worked examples of alleged debts of up to $7000 or morefell to zero or a few dollars when correctly calculated (Commonwealth Ombudsman, ‘Centrelink’s Automated Debt Raising and Recovery System’. Canberra: Commonwealth Ombudsman, April 2017). These inflated or false ‘debts’ are at the heart of the problem.

Why is this an issue for law and lawyers?

The reason that robo-debt is an issue for law and lawyers is a simple one. It is because the debts do not exist in law. The debts do not exist in law because statute prescribes that a debt arises ‘if and only if’ a provision makes it so (Social Security Act 1991 (Cth) s 1222A(a)). Section 1223 does provide that there is a debt in the difference between monies paid to and the true entitlements of the person. But for working age payments like NSA and YA, the Rate Calculators in the Act insist on a fortnight by fortnight rate calculation (see for example the provisions of ‘Benefit Rate Calculator B’ at s 1068). Because an ATO average of earnings never speaks to its constituent fortnightly parts, Centrelink fails to comply with this requirement other than in the (very rare) case of a person who had but a single job with constant fortnightly earnings. Consequently robo-debt operates without evidence regarding the central issue in any debt calculation.

This is no mere debating point. Centrelink must prove any debt it raises. The Full Federal Court long ago held in the McDonald case ([1984] FCA 59; 1 FCR 354) that Centrelink bears the onus of proof of such matters (it cannot be shifted to the alleged debtor). And in the case of debts specifically, that burden of proof is heightened by the Briginshaw principle. This is due to the moral stigma and other consequences of such a finding (Briginshaw v Briginshaw (1938) 60 CLR 336; [1938] HCA 34). This is all detailed in its full legal glory elsewhere (Terry Carney, ‘The New Digital Future for Welfare: Debts without legal proofs or moral authority?’ (2018) (March) UNSW Law Journal Forum 1).

The preceding analysis of invalidity has never been contested by contrary legal argument. Centrelink’s only justification is the elliptical one that debtors have adequate ‘opportunities’ to correct the record by lodging documents. But that ignores that there is no debt at all until Centrelink establishes one in law. Even a Centrelink advice letter, not always sent, that the ‘debt’ is based on an average and inviting fortnightly records, is a flimsy protection. Few people grasp the arithmetic and earnings bank rate calculation dispensations that mean that the two calculations can differ so much that an ATO averaged debt calculation of a few thousand dollars may actually be nil or just a few hundred dollars when correctly calculated.

In any event, the documents Centrelink invites people to supply were not required to be kept for more than six months according to its own website advices during the period leading up to introduction of robo-debt. Moreover, such documents are frequently now unobtainable (employers no longer in business or operating their business ‘on the black’).

How many people and how much money?

Robo-debt was first announced in the December 2015 mid-year Budget revision. Its revenue yield was then dramatically increased to cope with a Budget deterioration, in an announcement on the eve of the 2016 Federal Election (Peter Martin, ‘Extortion is no Way to Fix the Budget’, Sydney Morning Herald, Thursday 12 April 2018).

Estimated revenue from robo-debt over the forward estimates has recently been placed at $3.7 billion, though additional administrative expenditures in the early stages of the roll-out meant that net savings have been comparatively modest to date. Departmental figures in answer to recent questions during Senate Estimates hearings give a better picture of robo-debt operations. But even these figures are rough, because much of the data collected is about numbers or value of debts raised, rather than the amounts actually collected by Centrelink itself or after referral to debt collection agencies.

In the period from the commencement of robo-debt in July 2016 through to October 2018, some 409,572 debts were raised (many dating back to 2010 onwards, from which point reliable data was available for ‘matching’). Of these, 17 per cent were changed in some way — 29,888 debt amounts were reduced, 14,621 were wiped altogether, and 26,104 were waived or written off permanently (under Social Security Act discretionary powers in
s 1237AAD [special circumstances waiver] and s 1236 [write-off]). The average debt amount was $2,184, though some range to $10,000 or more (Luke Henriques-Gomes, ‘Centrelink cancels 40,000 robodebts, new figures reveal’, The Guardian 6 February 2019).

The latest available figures through to March 2019 show a cumulative 500,281 robo-debts, valued at $1.25 billion, of which 57,386 were reduced, 24,788 waived in part, and 31,160 fully waived (Luke Henriques-Gomes, ‘Centrelink still issuing incorrect robodebts to meet targets, staff claim’, The Guardian 29 May 2019). In short, this extra-legal program has a very large impact on a very large number of people.

How has robo-debt survived so long?

One of the many issues for lawyers is how such an arguably straightforward contravention of principles of the rule of law failed to be corrected, for so long, despite the numbers of people and the total amount of money at stake. The answer is both complicated and unedifying.

In the first place, the much-vaunted accountability machinery of ombudsman, national audit office, parliamentary oversight committees, legal aid commissions and civil society institutions appears to have been over-rated. None of the oversight bodies asked or obtained an adequate answer to the ‘law question’ of whether there was a legal foundation for robo-debt (Terry Carney, ‘Robo-debt Illegality: The seven veils of failed guarantees of the rule of law?’ (2019) 44 Alternative Law Journal 4). As already explained, there is no such foundation. Yet even the 2019 follow up report by the Ombudsman continued to overlook this fundamental point, despite it being on the public record (Stephen Easton, ‘Commonwealth Ombudsman “pleased” by robodebt changes, legal challenges remain’, The Mandarin 4 April 2019).

Rather belatedly a Legal Aid Victoria challenge was launched in the Federal Court in February 2019 to the lack of legal foundation of robo-debt; and this remains on foot (Madeleine Masterton v Secretary, Department of Human Services of the Commonwealth VID73/2019, https://www.comcourts.gov.au/file/FEDERAL/P/VID73/2019/order_list). However the Commonwealth subsequently waived the debt at issue in that case, and is reportedly seeking to settle rather than have the legal arguments adjudicated (Cameron Houston and Chris Verdelago, ‘Centrelink wipes “robo-debt” at centre of test case’, Sydney Morning Herald 5 May 2019).

In the meantime many robo-debts were overturned by the Social Services and Child Support Division of the Administrative Appeals Tribunal. But these decisions (and hearings) remain private at that level. Centrelink did not once challenge the invalidation reasoning by appealing to the General Division of the AAT, where the ruling would have become public. Nor did it desist from raising debts on the basis of the invalidated reasoning, or advert to those over-rulings when pursuing debts in the AAT. A clear breach, surely, of the Commonwealth’s ‘model litigant’ obligations.

What is so unedifying about robo-debt?

Robo-debt is a most unedifying saga on several levels.

First, it is chastening to lawyers when law fails to deliver to vulnerable citizens the justice promised by the principle of the rule of law. Even if this is by no means the first time such has occurred in the long history of law, it is dispiriting to the profession and undermines public confidence.

Second, it is unedifying because social security clients are among the most vulnerable of Australian citizens. There is a high incidence of mental health conditions such as anxiety and they are often young and unschooled in how government operates (and thus easily intimidated by the might of the State or prone to assume that ‘if Centrelink says it’s a debt it must be a debt’). Even highly qualified professionals such as lawyers and tax advisers understandably do not grasp the complexities of social security rate calculations because the concepts and principles do not align with other areas of law such as taxation principles. So how is such oppression tolerable here?

Third, and perhaps most worrying of all, it represents a breakdown in standards of government administration. A breakdown in the very design of robo-debt by ignoring crucial ‘law compliance’ items of the Administrative Law Council’s protocol for automation in government (Administrative Review Council, 2004. ‘Automated Assistance in Administrative Decision Making’. Canberra: Administrative Review Council 2004; principles 4 & 7). A breakdown in departmental procedures and standards in failing to properly administer the beefed up version of robo-debt announced days before the 2016 Election to deliver dramatically higher fiscal savings (Terry Carney, ‘Danger! Election 2016 delivered us Robodebt. Promises can have consequences’, The Conversation, 16 May 2019). And a breakdown of ethical administration in failing to adequately adjust the program once its flawed legal foundations were exposed in 2017 and 2018.

Finally, it is concerning for the legal profession and welfare rights advocates that a test case, lobbying or other strategies were unable to be promptly mobilised to redress such a compelling injustice. Can we seriously believe that other citizens, if faced with a ‘debt’ for which no legal foundation is able to be advanced, would have been advised to ‘find the documents that prove this’? Certainly, this was not the case when taxpayers ran into equivalent difficulties with the ATO (where the law does put an onus of disproof on taxpayers).

It surely behoves us all to reflect on why it should be that taxpayers apparently receive a higher quality of justice than do social security clients. For the rule of law supposedly applies equally to all.

Law and lawyers owe special vigilance to ensure that the rule of law is respected, as the Hon Kenneth Hayne warned in 2017 when writing about the application of new technologies, ‘[s]howing that something can be done does not mean that it should be done’ (Kenneth Hayne, ‘Opinion: “Change”’ (2017) 42(2) Alternative Law Journal 87).

Surely it is well past time that robo-debt was brought before the law rather than apparently permitted to operate beyond it?

Terry Carney AO is Emeritus Professor of Law at the University of Sydney and Visiting Research Professor, at UTS. He was a member of the Social Services & Child Support Division of the AAT for nearly 40 years.