Robo-debts lack any legal foundation

Nearly two years after social security’s popularly-called ‘robo-debts’ began to be raised without any adequate proofs, Australia’s much vaunted justice machinery has failed to expose the emperor’s lack of (legal) clothes.

The Online Compliance Initiative (‘robo-debt’) essentially required social security recipients to prove that they were not overpaid whenever Australian Tax Office (‘ATO’) data-matching calculated a fortnightly average income different from earnings they had reported to calculate each fortnight’s rate of Youth Allowance (‘YA’) or Newstart Allowance (‘NSA’). Previously, just seven percent of ATO data-match discrepancies resulted in raising of debts, and only after precise fortnightly earnings data was verified by Centrelink (either by the person providing pay slips or through invoking Centrelink’s information gathering powers to require employers or banks to provide that information). By purportedly shifting the onus of proof, catching all data-match discrepancies, and automating input of any pay slips, the December 2016 half yearly Budget up-date projected raising $2.1 billion over the forward estimates. Yet in truth, the overwhelming bulk of such ‘debts’ were either false (zero) or greatly inflated, as found by the Commonwealth Ombudsman. And, crucially, government was completely wrong in law in believing that it could avoid its legal responsibility to prove the existence and size of ‘debts’, or raise a debt where a person could not disprove its supposed existence.

It can confidently be said that the government was completely wrong in law. For two years it has assiduously avoided putting its supposed legal argument in favour of its ability to reverse the onus of establishing a debt. Despite bald ‘claims’ of some basis, Centrelink has neither exposed its reasoning publicly (such as to the Ombudsman and Parliamentary Committees), nor mounted contradictory argument before the Administrative Appeals Tribunal (‘AAT’), nor appealed against the (many) legal invalidations of debts by the first tier of the AAT when it has an absolute right to go straight to the second tier if it believes the reasoning is incorrect. Because first tier AAT (‘AAT1′) decisions are not made public (unlike AAT2), might it be that the emperor fears being seen in public?

Whatever the reasons, the fact of the matter is that Centrelink continues to bear the onus of establishing the existence and size of any debt. This is described more fully elsewhere but the bones of it can be very simply explained.

First, s 1222A(a) of the Social Security Act 1991 states that there can only be a debt if another provision creates it. There is no relevant provision ‘automatically’ creating a debt just because data-matching shows a discrepancy, so Centrelink is obliged to establish that there is a difference between the amount paid and the amount to which a person was entitled (s 1223). Also, considering that it’s been estimated that automation is expected to take over up to 46% of Australian jobs by 2030, this could mean more people on some sort of UBI (Universal Basic Income) therefore meaning these discrepancies in the amount due and the amount paid need to fixed rather quickly should they want to stay within the law, as well as out of debt.

Second, rate entitlements for payments such as YA and NSA must legally be determined each fortnight, taking into account any ‘income bank’ or other offsets and adjustments. And, just as rainfall in a given fortnight cannot be calculated by dividing the ‘annual’ figure by 26, nor can an ‘average’ from ATO records of earnings over many months or weeks speak to what was earned in particular fortnights. Indeed, correctly ascertained, supposed debts in the thousands of dollars evaporate entirely, or reduce to a few hundred dollars. That is why raising debts against vulnerable people on an average of their often multiple jobs with casual and fluctuating earnings is so morally bankrupt (or even ‘extortionate‘).

Third, Centrelink bears the onus of marshalling sufficient material to show that the person was overpaid by reference to inaccuracies in the actual earnings in each of the particular fortnights in question. It must ‘enliven’ or set in motion the relevant debt creation section. It is simply impossible to read this part of the legislation any other way. Consequently, as determined by the Full Federal Court in McDonald v Director-General of Social Security [1984] FCA 59; 1 FCR 354, Centrelink is obliged to show some proofs. But unless it is a constant, an average never speaks to its constituent parts. So, aside from the rare case of a person having held just one job (and most robo-debts involve half a dozen or more) the rate of pay for which remained constant, extrapolation of fortnightly averages from ATO data is no material or relevant evidence at all; it is no more helpful than reciting the words of a football club theme song. But like the midnight TV adverts, legally speaking ‘there is still more’: when the issue to be decided is as grave as an allegation of debt, it attracts the High Court’s ‘Briginshaw’ principle. As Sir Owen Dixon put it in that case, ‘In such matters ‘reasonable satisfaction’ should not be produced in inexact proofs, indefinite testimony, or indirect inferences…’. Yet, at best, that is what Centrelink seeks to do.

Fourth, and finally, failure by Centrelink to marshal any or sufficient material results in there being no debt in law. This too was made plain in the McDonald case.

Australia’s much vaunted rule of law machinery fails to engage

Australia has rightly prided itself for the innovative character and range of machinery to protect the rights of citizens in their dealings with government. Its system of merits review of administrative decisions by the AAT is lauded, along with the remedies available in the Federal Court on a point of law under the Administrative Decisions (Judicial Review) Act 1977 and associated avenues of judicial review. Likewise the institutions of the Office of the Commonwealth Ombudsman or Auditor General, and the work of Parliamentary Committees are held in high regard, along with access to Legal Aid Commissions and welfare rights bodies. So how is it that, for nearly two years, this clear breach of the rule of law has gone unnoticed and uncorrected? This is a particularly poignant question given the speed with which an equivalent illegality was struck down by the High Court in its original jurisdiction in Green v Daniels ([1977] HCA 18, 13 ALR 1 before any of this machinery was brought into existence by invoking the very prerogative remedies so rightly criticised as archaic and inadequate and thus the spur for all of the later reforms.[1] The answer is not a pretty one. For it exposes some glaring deficiencies in the institutional edifice. Deficiencies unable to identify and correct an unlawful impost of up to two billion dollars on vulnerable Australians.

First, why did the AAT fail? Of course generally it did not fail those individuals with the stamina to press their case through up to two internal review (original decision-maker review and then authorised review officer) and AAT1 hearings. But the closed AAT1 hearings and non-publication of its decisions (partly a failure of AAT leadership in not publishing any de-identified precedent cases), coupled with what appears to be a deliberate policy by Centrelink not to appeal to AAT2, meant that the AAT failed to exercise any normative influence, and led to justice being administered in secret (and thus its ruling not being open to public debate). This latter is not of small moment. Many alleged debtors were unable to locate or obtain pay slips because debts often went back to 2010 and many employers either had gone out of business or were unwilling to cooperate. Fearful of having ‘done something wrong’ (even though many had not put a foot wrong at all) many recipients of such letters simply accepted that Centrelink must know what it was doing and sought to clear the ‘debt’ as rapidly as possible. Public knowledge of AAT invalidation of such debts undoubtedly would have contributed to questioning of such behaviour. By continuing to press robo-debt cases at AAT1 without adequate legal argument and in the knowledge that they were being overturned for lack of legal foundation also put Centrelink in breach of government-wide ‘model litigant’ protocols.

Second, why did the institutions fail? In the case of the Ombudsman the answer must be that it did not ask, or did not properly press, the ‘law question‘. In an otherwise impressive report into administrative and other aspects, this lack of proper critical analysis is a glaring omission, which strikes the reader immediately. This is an unforgivable mistake. The report by the Senate Community Affairs Reference Committee did a little better, but unfortunately the witnesses who appeared were familiar only with general principles of law outside social security. So they did not guide the Committee towards the social security specific flaws, which leave Centrelink lacking any legal authority for its debts, and presumably its support staff could not be expected to rectify this.

Why Legal Aid commissions, welfare rights, and even pro bono divisions of major law firms all failed to manage to mount a judicial challenge is more complicated. Because of the rule about ‘exhausting’ administrative remedies and there being few AAT1 decisions upholding robo-debts once challenged, a potential case would be harder to find; and, of course, people have many valid reasons for not putting themselves forward as a test case. But even so, the Commissions and pro bono lawyers appear to have blotted their copybook in not being more active. Welfare rights bodies, better placed to have the expertise to identify the lack of a legal foundation for robo-debt, for their part face the invidious moral dilemma of balancing off public campaigns on such issues and jeopardising on-going funding for their casework.


Respect for the rule of law is a core democratic principle. Its importance is not diminished when the breach may be driven by worthy motives like keeping administration at the forefront of technological developments (machine learning) or a failure to tick all the boxes in the rush for a department to deliver a Budget savings measure. To the contrary, respect for legality is a constant irrespective of scale or motive. The failure of administrative, judicial, and institutional machinery to hold robo-debt to account is however, particularly troubling from a moral or ethical standpoint.

This is not the place to explore the lessons from this episode in any depth, but it would be remiss not to propose some remedies. First, merits review cannot be permitted to be open to being gamed by government agencies. Any practice of what US commentators call ‘administrative non-acquiescence’ is unacceptable: accepting adverse rulings in the individual case while leaving unchanged the policy overturned rewards the resourceful few at the expense of the vulnerable many. While model litigant protocols should ensure it doesn’t occur, transparency of justice (published reasons, avoidance of undisclosed ADR settlements) is the best antiseptic. Second, institutions such as the Ombudsman should be obliged always to seek, scrutinise and make public the legal reasoning behind any new program initiative. Parliamentary oversight should include monitoring compliance with this injunction. Third, funding of civil society bodies such as community legal centres, welfare rights organisations and legal aid needs to be provided in ways that secure rather than jeopardise their capacity for fearless and independent advocacy. Diversification of funding through partnerships or funding from the private profession (pro bono lawyering) and other non-government sources should also be vigorously pursued, including by provision of tax concessions or other incentives to encourage such investment in social issues.

A failure of institutional protections of the rule of law of the magnitude of the lack of legal authority for robo-debt surely cannot be permitted to result in ‘business as usual’ for the bodies and parties responsible. To allow it constitutes a form of moral and ethical bankruptcy. Respect for the rule of law calls for far better.

[1] See Peter Hanks, ‘Administrative Law and Welfare Rights: The 40-year story from Green v Daniels to “robot debt recovery”‘ (2017) 89 AIAL Forum 1.

For further reading please see: Terry Carney, ‘The New Digital Future for Welfare: Debts without legal proofs or moral authority? (2018) [March] UNSW Law Journal Forum 1-16.


Terry Carney AO, FAAL is an Emeritus Professor at Sydney Law School and Visiting Research Professor at UTS.

Suggested citation: Terry Carney, ‘Robo-Debt Illegality: A Failure of Rule of Law Protections? on AUSPUBLAW (30 April 2018) >