Although robodebts are just one aspect of the online compliance intervention (OCI) program implemented by the Commonwealth Department of Human Services (as it then was) in around 2015, this category of debt is arguably at the centre of the program. But what exactly are robodebts?
‘Robodebt’ is a term that has been adopted widely, including by media commentators, to describe a category of overpayment debt that Centrelink imputes to its customers, and which has been calculated by means of a specific method. At the current time, the legal recognition of the invalidity of this class of debt—the class of debt colloquially described as a robodebt—has been established in quite specific terms.
Having said that, it is possible that the kinds of debts currently recognised at law as invalid—and identified by the term ‘robodebt’ outside of legal claims—may expand in the future. For instance, if an affected litigant brings a relevant claim to the Federal Court of Australia (FCA), it could be that a finding of that Court might expand the definition of a robodebt so as to include
debts that have been calculated using the Net to Gross Earnings Calculator and in respect of which residual ATO averaging has been imposed to calculate the total; and
debts that have been calculated from ‘verified’ earnings reports of employers that do not provide information on a fortnightly (or other regular) basis but instead only provide a ‘lump sum’ record of payment, such that averaging is imposed to ‘explain’ the debt content (on a fortnightly basis).
To understand the current definiton of a robodebt, it is neccessary to examine the legal contexts in which these debts have been identified.
The legal contexts
Robodebts have been identified in three legal contexts, although none of these has described a ‘robodebt’ as such. But there is effectively only one kind of debt that has been recognised in these various legal contexts as a debt calculated and charged against Centrelink customers without a sufficient legal basis. It is in respect of this category of debt that the term robodebt has been directed in non-legal contexts, such as in media reporting.
This category of debts was first described in several dozen decisions written by members in the Administrative Appeals Tribunal (AAT). It was then recognised in the landmark decision of Davies J in the Federal Court ot Australia (FCA) in Amato v Commonwealth of Australia (VID611 of 2019). And finally, this category of debt has also been identified and described in the proposed declarations of settlement in the FCA pursuant to the (as yet unapproved) class action in Katherine Prygodicz & Ors v Cth (VID1252/2019).
In this series of posts, I will discuss each of these legal contexts in turn. However, in the first few posts, I will focus only on the AAT.
The AAT
What were to become known as robodebts were first recognised in the Administrative Appeals Tribunal (AAT). Several tribunal decisions make reference to a specific category of debt that was incapable of being validly proved under the Social Security Act 1991 (Cth) (SS Act).
In the Second further Amended Statement of Claim (Second Amended SOC) filed by the applicants in the class action Katherine Prygodicz & Ors v Cth (VID1252/2019), some 76 tribunal decisions are cited in support of a claim that the Commonwealth had knowledge that a specific category of debt had been imputed to centrelink customers that was said in those decisions to be not a debt validly due to the Commonwealth within the meaning of s 1222A(a) of the SS Act.1
However, because these decisions are described as decisions that the Commonwealth ‘elected not to appeal or have reviewed,’ it can be assumed that these decisions were made at the SSCS division (Social Services & Child Support Division) of the AAT, otherwise known as AAT1.2 In the Second Amended SOC, furthermore, these decisions are cited in a form as follows: ‘[Decision of] Member Dr King of 17 February 2017 in proceeding 2016/M102583.’ This is not a citation of a published but of an unpublished decision. And, according to the AAT’s Publication of Decisions Policy, decisions of the AAT1 are not published. Accordingly, since these decisions are not publicly available, it is not almost impossible to identify the way in which each of them, and how different AAT members, characterised any of the debts that they are purported, in the Second Amended SOC, to have identified as robodebts.
Finding the AAT robodebt decisions: an ‘unreasonable diversion of resources’
In the final months of 2020, the unavailability of these AAT1 decisions led one Commonwealth senator to request that some 65 of these decisions be provided to the Senate Standing Committee on Community Affairs. This request was made to Services Australia on 23 September, 2020; and it fell due on 9 October, 2020. The request, which was made during the Senate Standing Committee on Community Affairs’s Inquiry into Centrelink’s Compliance Program, was put in the following terms:
With reference to the Administrative Appeals Tribunal, and noting that Centrelink / Services Australia was a party to each of the relevant proceedings, please provide a copy of each of the following decisions by the Administrative Appeals Tribunal (with any necessary redactions in relation to personal information)…3
The request then went on to list the 65 decisions.
When it was asked that question, Services Australia (SA) it on notice. Later, in response to this request, SA gave the following answer:
To provide a copy of each of the 65 decisions by the Administrative Appeals Tribunal would require each decision to be manually reviewed to identify and then make relevant redactions, along with subsequent quality checking that the correct redactions have been made. This would require an unreasonable diversion of agency resources.4
This response came some two months after the Coalition government agreed to settle the class action.
A year earlier, on 6 December 2019, another senator had asked a similar series of questions in the Senate Standing Committee on Community Affairs. Those questions were directed to the AAT and concerned decisions made by then AAT member Professor Terry Carney. The effect of the questions was to request a decision (or decisions) written by member Carney in relation to robodebts, and to inquire whether those decisions described these robodebts as ‘unlawful’. The AAT took the question on notice and later provided the following answer:
The AAT conducted a search of decisions with written reasons made by Member Professor Terry Carney AO. The AAT has identified five decisions made in the AAT’s Social Services and Child Support Division in relation to six applications for review of decisions made by Centrelink about debts in which Professor Carney considered in detail Centrelink’s online compliance intervention system in determining whether the applicants had been overpaid an allowance.
A review of the attached decisions show that the terms ‘unlawful and or illegal’ are not used in reasoning, findings or decision making of Professor Carney.
Copies of the five decisions are attached. Information that may identify the applicant or any other person has been removed.5
Accordingly, the AAT released these five decisions in , which are decisions of the AAT1 (normally unpublished). These decisions, authored by Terry Carney, give us some insight into the terms in which robodebts were first identified in the AAT1. I will briefly examine the relevant parts of these decisions in turn.
Decision 1: 2016/S1046816
This was a decision relating to the alleged overpayment of a Youth Allowance (YA) debt in the amount of $7,452.76.
In this decision, the member noted that:
This review presents for resolution what, in my experience, is a previously unprecedented issue: namely whether fortnightly earnings and any overpayment based on the difference between such ‘earnings’ and the amount previously paid in that fortnight (based on [the applicant’s] reporting from time to time) can provide a secure legal basis for finding an overpayment or determining its quantum.
The member noted that no ‘prior jurisprudence’ could be found to resolve the issue, save the Full FCA decision of McDonald.7 In McDonald, the Full FCA found, in effect, that the rules of evidence are inapplicable and the evidentiary burden of proof was inapposite. As Woodward J stated:
There is certainly no legal onus of proof arising from the fact that this is an ‘appeals’ tribunal, because the AAT is required, in effect, by s 43 of the AAT Act, to put itself in the position of the administrator in carrying out its review and, in the light of the material before the AAT, (not the material before the administrator, Drake v Minister for Immigration and Ethnic Affairs 1979 24 ALR 577 at 589) make its own decision in place of the administrator's. The AAT itself, in a series of cases beginning with Re Ladybird Children's Wear Pty. Ltd., 1 ALD 1, has taken the view that there is no presumption that the administrator’s decision is correct. This is clearly the right approach to the matter.
While much more could be written on the nature of the onus of proof requirements in administrative law (and in administrative tribunals), it will suffice to note that the member in this decision relied on the reasons of the Full FCA in McDonald to find that ‘there is no ‘onus’ of proof on either party (neither on Centrelink nor an applicant for review).’8
The member further relied on McDonald to identify the way in which, where a tribunal finds that it cannot be satisfied that the requisite standard of proof has been met where materials have been provided pursuant to an application (that standard being the balance of probabilities), then a ‘status quo’ or a ‘‘default’ outcome might result.’9
The member also considered the well-known Briginshaw principle, which was said to state that ‘the strength of the evidence needed to reach [the requisite] level of satisfaction [capable of satisfying the decision-maker that a fact exists on the balance of probabilities] varies according to the nature or the effect of what it is that is to be established.’10
The member’s reliance on what might be called the McDonald onus rule and the Briginshaw principle led him to state that
… if on the present facts… Centrelink is unable to advance sufficiently convincing proofs of a debt or debt amount, then no debt arises in law. It is not a matter of the alleged debtor having advanced further information (though in practice this may be the most efficient way of clarifying it). Rather it is that Centrelink has not established the proposition it was required to establish (here a reliance on a methodology incapable, other than in rare instances of unchanging fortnightly income, of addressing the architecture of the fortnightly rate payments).11
In the above passage, the member effectively states that where material reaching the the requisite degree of proof has not been advanced in respect of a debt or debt amount, then the debt or debt amount cannot arise in law.
In the paragraphs that follow, the member makes a number of statements that are worth quoting in full for the purposes of examining what is probably the first time that a robodebt was ever identified in the AAT1.
In this decision, the reason the overpayment subject of the decision was said to have not been established was because the evidence was not sufficiently strong and because the debt was not mathematically capable of being reliably produced (or reliably reproduced). As the member writes:
The reason it does not establish either an overpayment or its quantum is due both to the lack of sufficient strength of evidence and to simple mathematics.
First, member Carney deals with the evidentiary aspect:
The lack of strength of evidence flows from my characterisation of the overpayment ‘methodology’ (actually an administrative algorithm as I understand) — involving extrapolation of ATO employment income information over a period, divided to produce an average fortnightly [amount], and then applied to YA payment periods to raise a debt — as, at best, raising no more than the sufficient doubt about the accuracy of past payments as to warrant the exercise of powers of enquiry held by Centrelink (or indirectly by this Tribunal: see paragraph 42 of these Reasons above). It is too uncertain, and too slight a basis to satisfy the Briginshaw standard in a fortnightly rate debt matter.
Next, member Carney deals with the mathematical aspect:
The simple mathematics, once my finding of variable and at times episodic weekly earnings is taken into account and feed [sic] into the legislative context of a requirement to determine a fortnightly YA rate based on earnings attributable to that fortnight, as further moderated by any reduction of raw gross earnings figures by reference to the ‘earnings bank’ provisions — is that almost any speculated figure between no debt and the alleged debt amount is capable of being calculated.
Of note in the above paragraph is the effect of the so-called ‘‘earnings bank’ provisions.’ It is worth briefly describing what is meant by that term and its importance in the member’s identification and definition of an invalidly raised debt.
The Income Bank
In respect of the expression ‘earnings bank,’ the member provides a citation (omitted from the above quotation). The text of that citation (footnote 20 in the decision) is as follows: ‘SSA 1991 Module J of section 1067G, points 1067G-J1 [diagram], 1067G-J3 [method statement].’ This a reference to the ‘Rate of Youth Allowance’ provisions under s 1067G of the SS Act. Specifically, the citation refers to Module J, which supplies provisions in respect of the ‘student income bank’ (income bank)— a concept or tool that is used in the calculation of a student’s entitlement to YA.
The student income bank effectively allows a YA recipient to build up credit in this income bank during fortnights in which (1) they do not earn ordinary income, or (2) they do not earn much ordinary income. Having built up this credit, the student, if they earn more income than usual in a fortnight, will be entitled to a higher amount of YA than they would otherwise have been, because the income bank credit they have earned in previous (no- or low-earning) fortnights will be used to offset the additional income they have earnt in the present fortnight. In other words, the income bank is a credit system that operates by ‘saving’ any unused part of the recipient’s fortnightly ‘income free area’ as a credit, which is then available to offset against income/earnings in future fortnights. As a policy, the income bank is probably exists to ensure that those who work spasmodically are not easily disentitled to YA when they happen to earn an unusually relatively high income in a specific period, such as during a university of apprentice holiday period.
The thresholds in relation to how much one can earn are defined in s 1067GH within Module H of the SS Act. These thresholds are organised around terms such as ‘ordinary income’ and ‘ordinary income free area’ (s 1067G-H29) and these subsections make provision for circumstances in which a person’s ordinary income exceeds the ordinary income free area, in which case the person has an ‘ordinary income excess’ (s 1067G-H30). In circumstances where an ‘ordinary income excess’ arises, an ‘ordinary income reduction’ will be imposed on the person’s entitlements (in this case, YA entitlements) under s 1067-H31 of the SS Act. Lower range and upper range reductions are also specified in subsequent sections.
The mechanics of how the income reduction occurs are set out in Module J of the SS Act. Section 1067G-J1 provides the following:
A person's ordinary income under Module H may be reduced under this Module. This diagram sets out how to work out:
(a) whether the person's ordinary income for a particular fortnight, in respect of which youth allowance may be payable to the person, is to be reduced; and
(b) if it is to be reduced, the amount of the reduction.
In the later section 1067G-J3, the method by which a person’s income bank credit is calculated is specified in steps in what is called a ‘Method Statement.’ These steps constitute the ‘moderation’ that member Carney refers to in the above paragraph, in which he identifies the ‘legislative requirement’ to determine a fortnightly YA rate based on earnings attributable to that fortnight’ [and] ‘as further moderated by any reduction of raw gross earnings figures by reference to the ‘earnings bank’ provisions’ (emphasis mine).
Needless to say, the ‘moderation’ of the fortnightly earnings through the income bank provisions is a complicated task. It is for this reason that Centrelink use a variety of software tools to calculate a recipient’s income bank, including the ADEX Debt Explanation and the so-called ‘Student Income Bank Explanation’. There is also some guidance in realtion to the student income bank in the Department’s Guide to Social Policy Law, under 1.1.S.370.12
The steps through which the income bank is to be calculated and applied under the legislation (under s 1067G-J3), however, are as follows:
A person's income bank credit for a particular income bank fortnight of the person is to be worked out as follows:
Method statement
Step 1. Assume that the person’s income bank credit, at the time this Module starts applying to the person, is an opening balance of zero.
Step 2. If, for the person's first income bank fortnight, the person has an income credit under point 1067G-J4, add it to the opening balance.
Step 3. For each subsequent income bank fortnight of the person, up to but not including the fortnight in question, either:
(a) if the person has an income credit for that fortnight under point 1067G-J4 and the person is not a new apprentice—add it to the balance of the person's income bank credit in respect of all the previous fortnights, but not so as to increase the balance beyond $10,000; or
(aa) if the person has an income credit for that fortnight under point 1067G-J4 and the person is a new apprentice--add it to the balance of the person's income bank credit in respect of all the previous fortnights, but not so as to increase the balance beyond $1,000; or
(b) if the person has, in respect of that fortnight, drawn from the person's income bank credit under point 1067G-J5—deduct from that balance the amount drawn, but not so as to reduce the balance below zero.
The result is the person's income bank credit for the fortnight in question.
As may be seen from the above method statement, this kind of prescribed legislative calculation sets up a number of precise guiderails around the allocation of welfare entitlements to the recipient. These calculations may be complex, and the scale at which they must be carried out is very high: on 22 May, 2020, there were 168,095 YA recipients in Australia.13 Thus, the legislative scheme clearly requires that significant administrative systems are developed to compute these complex mathematical problems. Indeed, it might be said that automation is inevitable in the face of these complex methods.
Nevertheless, the finding that the member is making in respect of the calculation of debts is not just that the specific debt is ‘too complex’ to calculate in the present circumstances. Rather, the member’s finding is that the effect of averaging a variable income such that it becomes a consistent (averaged) number over time, would be to ignore and to cast aside the effect of the income bank, in circumstances where the income bank must be calculated in reference to the fortnightly earnings — as a legislative requirement — under s 1067G-J3 of the SS Act. In such circumstances—where the imposition of a legislative requirment in respect of an amount of a statutory benefit has been effectively obviated by an intervention of a government agency—the alleged debt cannot be properly calculated for the purposes of the legislative framework.
Illegal? Unlawful? Legally insufficient?
It may be important, at this point, to emphasise that the member does not describe this problem—the inability to calculate the fortnighlty income by reference to the imposition of the income bank test—as a legal one (that is to say, the alleged debt is not, because the legislative requirement to impose the income bank test has become incapable of being discharged, defective in the sense that it is ‘illegal’). Rather, the inability to impose a legislative test on the debt is described as a ‘mathematical’ problem that gives further force to the invalidity of the specifically alleged debt that is in issue for the purposes of the review. In this regard, the real debt, should there be one, is unknowable, because the information provided, which is an extrapolation of fortnighly amounts derived from a single figure—the annual earnings reported to the ATO by the alleged debtor’s employer—simply does not permit a calculation of the amount of YA that the individual, in this case, was entitled to receive. It is, in this sense, a mathematical—or perhaps even an ‘informational’—problem.
Lumpiness
In subsequent paragraphs, the member elaborates on the problems arising from the averaging process and its ‘moderation’ by the income bank:
An overpayment, if any overpayment exists, depends on how ‘lumpy’ is the true earnings pattern (i.e. compared to the invalid assumption of ‘constant’ earnings in every fortnight, including the 12 weeks when I have found it was impossible for [the applicant] to have performed work for her employers: see paragraph 28 above).
I therefore find no proven overpayment of any quantum. The foundation on which the suggested overpayment rests falls well below even the ‘inexact proofs’ and ‘indirect inferences’ which Sir Owen Dixon found to be an unacceptable form of proof of such matters in Briginshaw (see paragraph 47 above). Within the terms of McDonald, given the structure of overpayment provisions and overpayment debts the failure to establish the overpayment leads to the default of no debt (see paragraph 40 and the preceding paragraph above).
After then considering the waiver and write-off provisions of the SS Act (the member finding they were not applicable), the member then decided to set aside Centrelink’s decision to raise and recover the YA debt, and remitted the matter to be ‘redetermined’ with directions that:
1. No debt or debt component is able to be founded on extrapolations from Australian Tax Office records;
2. The earnings components of any recalculated debts as may be raised must be based on and confined to any fortnightly salary records obtainable in the exercise of statutory powers to do so (if set in train); and
3. Debt amounts (if any) as so varied are recoverable debts (not able to be waived).
Thus, in the above decision, the member effectively sets the boundaries within which debts can be raised by Centrelink, indicating, in (1), that extrapolations from ATO records will not suffice for debt calculation purposes and, in (2), that ‘fortnightly salary records’ (ie, payslips) must form the sole basis (‘confined to’) of the recalculated debt. In (3), the member indicates that debt amounts that are calculated in this way will constitute recoverable debts.
The primary point that might be made about this first decision is that the member does not waive or write off the debt altogether, nor set aside the debt and substitute a decision in which there is no debt (under s 43(1)(c)(i) of the AAT Act). These would be quite significant decisions that would place Centrelink in the invidious position of not being able to proceed against the debtor. Instead, the member appears only to exercise their powers under s 43(1)(c)(ii) of the AAT Act to remit the decision to the original decision-maker with directions.
Given the apparent furore and controversy that these original decisions generated, it is notable that this original decision is a rational and reasonable approach to the problem of insufficient evidence in the administrative setting. It is so because it gives Centrelink and the Department ample scope to recalculate the debt in accordance within stipulated legal boundaries and then to persist in recovering it as against the applicant in accordance with the member’s directions. There is no way in which it can be said that the decision intervenes in a way that impedes Centrelink of the Department from carrying out the OCI program, provided actual salary records are obtained. The decision also clarifies that Centrelink may use its statutory powers to obtain those records.
Although only one decision has been analysed here, it is worth noting the other ‘robodebt’ decisions made by the same member throughout 2016 and 2017. These decisions are as follows: decision 2: 2016/S10439414; decision 3: 2017/M11346915; decision 4: 2017/S11288416; and decision 5: 2017/ M112147 and M112302.17 Each of these decisions includes the same general principles as articulated above, with only minor differences.
Ibid.
Ibid.
See LCC-SBE19-21.pdf.
See AAT1, 2016/S104681, available at LCC-SBE19-21 Attachment A.PDF.
Carmel Elizabeth McDonald v Director-General of Social Security [1984] FCA 59; 1 FCR 354; 6 ALD 6.
AAT1, 2016/S104681, para 37.
Ibid para 41.
Ibid para 46.
Ibid para 52.
See AAT1, 2016/S104394, available at LCC-SBE19-21 Attachment B.PDF.
See AAT1, 2017/M113469, available at LCC-SBE19-21 Attachment C.PDF.
See AAT1, 2017/S112884, available at LCC-SBE19-21 Attachment D.PDF.
See AAT1, 2017/ M112147 and M112302, available at LCC-SBE19-21 Attachment E.PDF.