The ghost of Robodebt – Federal Court rules billions of dollars in welfare debts must be recalculated

A recent landmark court decision could have significant ramifications for several million social security recipients.

The ruling means the federal government will need to recalculate more than A$4 billion in debts owed to the Department of Social Services, which administers Centrelink.

Some of the debts – which occurred due to overpayment of benefits – stretch back decades.

Reminiscent of Robodebt, the problem occurred because an unlawful method – income apportionment^[1] – was used to calculate the money Centrelink claimed it was owed.

The judgement

From the early 1990s until 2020, more than 5.3 million welfare debts were calculated using income apportionment.

In the test case Chaplin v Secretary, Department of Social Services^[2], the full Federal Court approved a method proposed by the government to recalculate the debts.

The court was not asked whether the debts were unlawful – a point the department had already conceded – but whether its remedy was legally sound. In a two-judge majority, the court ruled it generally was.

Following the judgement, the department swiftly resumed^[3] debt recovery, which had been paused in 2023, pending the legal decision. It said in a statement:

now there is certainty to the legal position, assessments will recommence in line with the court’s decision.

The scale of the problem

The unlawful debts are worth $4.31 billion^[4] in total, and affect almost three million Australians. About 91% of these debts – $3.93 billion – has already been repaid to Centrelink.

Another 170,000 debts^[5] – totalling $347 million – remain outstanding.

All the debts – either repaid or still owing – must be recalculated using the revised method approved by the court.

According to the government, the median debt^[6] is $330 and has been owed for 19 years, on average.

But the judgement does not compel the government to actually recover the money. Some media^[7] reports^[8] suggest a waiver^[9] is being considered.

For its part, the government^[10] says it will “evaluate” the court decision and develop a “suitable response”.

What is income apportionment?

An internal anti-fraud policy^[11] meant Centrelink was obliged to calculate a person’s income when it was “earned” rather than “received”.

This led to the use of income apportionment^[12] – essentially an educated guess about a person’s fortnightly earnings when their pay cycle didn’t align with their income reporting period.

Income apportionment was used by Centrelink when payslips did not match up with fortnightly reporting periods, leading to many overpayments. Joel Carrett/AAP^[13]

This process, which typically produced overpayments to recipients, spread income outside an instalment period, which was contrary to the applicable law^[14]. It also attributed earnings to a person for days and fortnights they hadn’t worked.

Income apportionment was discontinued in 2020. Three years later, the Commonwealth ombudsman found the method was unlawful^[15].

Is this different to Robodebt?

While Social Services has sought to distinguish income apportionment from Robodebt^[16], the two methods of calculating debt are comparable.

Both attributed a person’s daily income^[17] beyond the timeframe permitted by law^[18].

But there are differences in source and scale.

Where apportionment was personalised by using individual customer payslips, Robodebt used Australian Tax Office records to raise debts en masse.

Significantly, while the ombudsman said the department’s understanding of the law relating to apportionment was “incorrect”, it was also “genuinely held”.

On the other hand, the infamous Robodebt scheme was designed to ramp up debt clawbacks. Claims of misfeasance in public office continue to be litigated^[20].

Other troubling overlaps remain.

Many individuals affected by apportionment debts raised after 2015 will be the same people served with Robodebt notices.

Evidentiary burden

A troubling aspect of the test case was the suggestion by the majority judges – citing High^[21] Court^[22] precedent^[23] – that the evidentiary burden could shift to the welfare recipient when overpayments are believed to occur through “wrongdoing”.

This could force an individual to disprove their alleged debt if a decision-maker concluded the recipient had accidentally under-reported – as occurred in the test case – and a lack of evidence made it difficult for the government to prove its allegation.

The finding arguably runs counter to the Robodebt Royal Commission’s observation that most welfare recipients lack the power to disprove a debt because their historical records are unavailable.

The dissenting judge in the case rejected the government’s proposed recalculation method, finding it “not proper” for recovery action to be taken without probative evidence^[24].

He said the majority decision meant Centrelink could reassess debts in the future after evidence had been lost, and recipients would be powerless to disprove them.

Expensive fix

The administrative burden of reassessing these unlawful debts is immense.

Late last year, a team of 150 public servants, each costing $117,400 per annum, was assigned^[25] to rectify income apportionment.

Their internal sampling^[26] revealed 64% of people issued debt bills were overcharged, 29% were undercharged, while 4% are owed a total refund.

The remediation process has been chaotic.

In the year following the ombudsman’s report, recipients lodged 531 appeals and made 530 complaints^[28], highlighting the human impact of income apportionment.

But in a five-month period, a mere 83 cases^[29] have been finalised.

Controversially^[30], Social Services offered to process debts on request^[31], contrary to a provisional finding of the Administrative Appeals Tribunal^[32], which dismissed the method being used by the department.

Political choice

While the Federal Court has seemingly given the government a legal victory, the ultimate outcome will be costly – especially if the debts are waived.

The court ruling requires recipients be afforded “procedural fairness”, meaning resource-intensive investigations will need to be undertaken into the millions of cases yet to be reviewed.

The final price tag is yet unknown. In the 2025–26 budget^[33], income apportionment was recorded as a “contingent liability – unquantifiable”.

Almost all of the outstanding debts would have already been resolved if the government had implemented the Robodebt Royal Commission recommendation^[34] that welfare overpayments should not be pursued if they are more than six years old.

The court’s decision also fails to address the 159 Australians believed to have been criminally prosecuted^[35] over unlawful debts since 2018. These people – and likely many more before that year – may have been convicted on defective evidence.

The response to these issues will be a test for the government.

Has it learned the lessons of previous egregious mistakes, or will it allow the ghost of Robodebt to continue to haunt our welfare system?

References

1. ^{^} income apportionment (www.servicesaustralia.gov.au)
2. ^{^} Chaplin v Secretary, Department of Social Services (www.judgments.fedcourt.gov.au)
3. ^{^} swiftly resumed (www.dss.gov.au)
4. ^{^} $4.31 billion (www.righttoknow.org.au)
5. ^{^} 170,000 debts (www.righttoknow.org.au)
6. ^{^} median debt (ministers.dss.gov.au)
7. ^{^} media (www.afr.com)
8. ^{^} reports (www.theaustralian.com.au)
9. ^{^} a waiver (www.afr.com)
10. ^{^} the government (ministers.dss.gov.au)
11. ^{^} internal anti-fraud policy (web.archive.org)
12. ^{^} income apportionment (www.ombudsman.gov.au)
13. ^{^} Joel Carrett/AAP (photos.aap.com.au)
14. ^{^} applicable law (www.rudge.tv)
15. ^{^} was unlawful (www.ombudsman.gov.au)
16. ^{^} sought to distinguish income apportionment from Robodebt (www.aph.gov.au)
17. ^{^} daily income (www.legalaid.vic.gov.au)
18. ^{^} timeframe permitted by law (www.rudge.tv)
19. ^{^} Lukas Coch/AAP (photos.aap.com.au)
20. ^{^} continue to be litigated (gordonlegal.com.au)
21. ^{^} High (jade.io)
22. ^{^} Court (jade.io)
23. ^{^} precedent (jade.io)
24. ^{^} probative evidence (www.alrc.gov.au)
25. ^{^} assigned (www.theguardian.com)
26. ^{^} internal sampling (www.righttoknow.org.au)
27. ^{^} Lukas Coch/AAP (photos.aap.com.au)
28. ^{^} recipients lodged 531 appeals and made 530 complaints (www.aph.gov.au)
29. ^{^} 83 cases (parlinfo.aph.gov.au)
30. ^{^} Controversially (www.art.gov.au)
31. ^{^} offered to process debts on request (www.dss.gov.au)
32. ^{^} Administrative Appeals Tribunal (www.austlii.edu.au)
33. ^{^} 2025–26 budget (budget.gov.au)
34. ^{^} recommendation (robodebt.royalcommission.gov.au)
35. ^{^} criminally prosecuted (www.aph.gov.au)