A Generational Miscalculation

Unpicking Centrelink’s Unlawful Income Apportionment

Ever since the the Commonwealth Ombudsman announced the widespread use of an unlawful practice to calculate welfare payments last August, questions have swirled about the fallout.

This week, a document released under Freedom of Information laws provides fresh insight into the debacle. It shows that an entire generation of welfare debts have been substantially miscalculated and are currently left on the books, unremedied, in an unlawful state.

The Minister for Social Services now faces a range of serious ethical and financial choices. Will the department launch remediation, including by waiving the debts? Does the greater good militate against attempting to recalculate them, given the resources involved could endanger the effort to improve the currently dire processing times at Centrelink?

The Apportionment Flaw

Apportionment debts were born from Centrelink’s tendency to do what was convenient for it, and the sheer impracticability of employment income reporting prior to 2021.

To calculate a customer’s payment rate, the law required Centrelink to determine a ‘daily rate’ of employment income for each payment fortnight in the Centrelink calendar.

• Recipients were generally asked to report the income they had earned in the Centrelink fortnight. Their payslips, however, might not fit that calendar timeline. The dates couldn’t line up. The complexities of this timeline mismatch were such that the Department expressly warned people not to copy their payslips into their income reports.

• Instead, recipients could fill in a worksheet to attempt to figure out their imputed daily earnings figures. But there was no obligation on any employer to include a daily breakdown of hours and pay rates in payslips. The welfare system assumed people in casual and volatile employment relationships could somehow reverse engineer or predict their earnings without access to key payroll data.

Indeed, even when Centrelink took over, things tended not to go too well on the actual daily earnings front. As the Ombudsman pointed out as far back as 2002:

When requesting details of earnings from an employer Centrelink send the employer a form letter that asks for details of days and hours worked and a breakdown of dates worked, dates paid, gross pay and any allowances included. However, depending on the nature of records maintained by the employer, this breakdown of information may not always be provided as requested. Often the employer will report the gross and net pay for each pay period, but may not include details of days worked, hours worked each day or when the payment was actually paid to the person. Centrelink then calculates any entitlement (and overpayment) based on the information provided. This necessarily involves assumptions about those matters on which there is no information and those assumptions will rarely be correct in any particular case.

Those affected were precarious casual workers employed by volatile businesses on the edge of the economy. As notmydebt consistently argued: these people were set up to fail by a system that had little regard for their time, resources or workplace realities.

Apportionment debts were produced when the dates on the payslips went across two Centrelink fortnights. Instead of identifying what actual daily earning was earnt, derived or received in Centrelink fortnight 1 or Centrelink fortnight 2 (as required by law), the agency took an unlawful shortcut. It simply took the total figure on the payslip, divided it by 14 and slapped that figure into each day —regardless of which fortnight it was in.

And, unlike robodebt, this isn’t just about debts. This was the method by which payment rates were calculated. The sample exercise includes an as-yet-unfinished study of arrears cases that include the unlawful technique.

Can Services Australia — in theory — fix these files?

Under the approach currently prevailing at the tribunal level, the Agency can recalculate these debts if it has the dates on which the fortnightly income was received by the person. This could be included on the payslip, or a bank statement might supply it (at least in terms of the net payments) in some cases. The Administrative Appeals Tribunal has supported a right of election — whereby the Department moves away from when income was earned to when it was received in its calculations.

But if dated payslips for each fortnight are not on file, recalculation would not be possible. In a tribunal case this week, we saw acceptance of the fact that a debt cannot be raised if there are no documents evidencing the dates of receipt. If such records aren’t on file, most of these debts lie outside the retention period for bank statements and payrolls records.

There are a number of reasons why documentation might not be there. Perhaps copies or evidence of verification were not retained. We know from pre-2015 AAT cases and page clxxxiii of the Robodebt Royal Commission Report that a small proportion may feature residual, last-resort averaging. While Commissioner Holmes took the approach that such small-scale practices were marginal to her inquiry, the sampling exercise would capture them. Needless to say, their existence provides no exoneration for the creators of robodebt. This new sample again actually proves the fundamental change robodebt represented. But having personally encountered such debts, I would take delight in them being refunded.

How Big Is the Mountain of Unlawful Files?

The FOI released this week allows us to start answering this question. The core of this second sample is an analysis of the pre-2010 determined debts. In theory, I thought this second sample was the cohort of debts that could grade well for the Agency. The first sample was – confusingly – made up of mainly undetermined debts – backlogged debts – that hadn’t been issued yet. As Rick Morton reported: it didn’t go well. Prosecutions were discontinued and debt recovery pauses announced. Halfway through the email exchanges a senior public servant makes comments on the exercise, noting the ‘low strike rate for matters that we have been able recalculate’ (See LEX 76245 Page 24 of 61).

Centrelink staff used to investigate cases in depth prior to 2010 – policy required them to go direct to employers before issuing any debt. Might things look better this time round?

Here’s how I’d communicate the results:

Of 902 employment income debts sampled:

• 484 (54%) were unlawfully income apportioned

• 326 (67%) have sufficient evidence available on the record to conduct a recalculation

• 17.5% of all sampled pre 2010 employment income debts had insufficient information on file to perform a recalculation.

A recalculation was only conducted ‘where there [was] evidence available for the full debt period. Where any information is missing, it is deemed insufficient evidence is available.’ It’s extremely unlikely records prior to 2010 exist — even if you wanted to pay to search for them. Therefore, the 17.5% is likely irretrievably unlawful.

Where unlawful apportionment was applied, it led to substantial miscalculation of the debt. 64% of the recalculated debts were reduced by an average of 36% of the debt value. 30% of the recalculated debt went up, with the average being 43%.

In short, Centrelink employment income debts have been incorrectly calculated, at scale, for a generation.

Recalculation: An Impossible Task?

The fact that 67% of files were capable of being recalculated, might, on first viewing, have you thinking government may choose to reprocess the files.

But there’s a catch.

Staff carrying out the sample have recorded the time taken on each file.

• Time to screen and investigate a file: 1 hour and 11 minutes

• Time to recalculate the debt: 3 hours and 23 minutes

This means that a trained member of staff would be unlikely to re-calculate even two debts in a full working day. It would consume extraordinary workforce resources to find and recalculate files scattered across 18 years at least. The average value of these pre 2010 debts ($511) in proportion to staffing costs also points towards a recalculation approach being uneconomic.

The scale of it all looks pretty substantial even to a grizzled robodebt observer. If the prevalence of apportionment in determined debts in this sample was replicated across all years, then roughly 40% of debts issued by Services Australia over the past 20 years would need to be reworked.

The Unique Silences of Social Security Law

The second sample also analysed AAT decisions to see the prevalence of unlawful calculation in cases that had been through independent merits review. Apportionment occurred at an even higher rate. 78% of files were income apportioned, with 55% having sufficient evidence for recalculation. 72% of the recalculated debts went down by 24% of the debt value.

While I’m a massive fan of the new escalation mechanism in the proposed new tribunal (the Administrative Review Tribunal), this underlines why tribunal reform is no cure-all post robodebt. For nearly 20 years, tribunal members seem to have failed to frame the interpretive issue at the heart of this case.

Ultimately, it was notmydebt case workers and two principled staffers working in Rex Patrick’s office that finally pressed the issue. This shows that what is really needed is funding for advocates, community legal centres and legal aid — funding that would allow them to put their hands on more files as needed.

Despite efforts to sunder them, robodebt and apportionment have a shared root cause. It’s the 98.8% dynamic. The true heart of robodebt was that, even with all the outcry, only 1.2% of cases were ever formally reviewed in the lifetime of the program. For all the academic and political rhetoric about flawed algorithms and AI, robodebt was about power and the stigma that fills the space between people.

Apportionment again speaks to the prevailing reality — the reality in which Services Australia’s practices go unmarked and untested. On display is the same silent power that saw the Department refuse to proactively publish its operational instructions regulating the recalculation process. The same unchallenged power that can create a mountain of unlawful calculations, for 20 years, in the small, quiet places, far from the public eye.