The robo-debt pile-on

Services Australia makes about $180 billion in (Centrelink, Medicare and Child Support) payments annually. Anyone connected with the administration of welfare knows that overpayments are a huge problem costing the government billions over time. Besides undeclared income, other areas of Centrelink fraud include failure to disclose a partner, and often exaggerated claims of incapacity.

Data from the Australian Tax Office (ATO), ultimately sourced from employer-generated Group Certificates, have for many years (even under Labor) been matched with income directly reported to Centrelink by welfare recipients. The employer data are accurate, and have been used to help identify those, who had under-declared their income to Centrelink.

There have, however, been major limitations to the matching process:

• The first is that (until recently) ATO income data had been annual and available only from the end of a tax year. Centrelink payments are made fortnightly and are subject to a (current) fortnightly income declaration/testing process.
• [A person's entitlement to social security payments (inter alia) is based on actual fortnightly income. Thus using yearly income information to create averaged fortnightly earnings is not a lawful or accurate method of calculating a debt, especially where clients have been on benefits for only part of a year or had variable incomes. Minister Robert, in recently stating that the robo-debt scheme was "legally insufficient", admitted this.]
• The second problem was that the debt notice issuing process, which historically had been subject to manual supervision and investigation, was automated about five years ago. Vastly more (unsupervised) debt notices were subsequently issued.
• Many debt notices were issued a long time after benefits had been paid, making it difficult for alleged debtors to contest their debts.

The term "robo-debt" refers to the automated debt collection process, following which it was not uncommon for alleged debtors to have their tax refunds seized to repay assessed overpayments. The robo-debt process ended-up being fatally damaged by its inadequate and untimely data, and by insufficient human oversight and follow-up in its administration.

Worse still, robo-debt impacted a vulnerable group, who in many cases were chased up by debt collectors and were often told they had to prove they did not owe the flagged debt (reversing the normal burden of proof). The scheme saw hundreds of thousands of people issued with computer-generated debt notices, some of which imposed penalties. Some of these demands even went to people who did not owe the Government any money, though such cases were very much in the minority. With most welfare recipients living from one payday to the next, many were alarmed, when presented with a debt, especially if it was a large one.

The Government has justifiably been subjected to a pile-on from the media and Opposition over what was a debacle, and it has now agreed to return $721 million, which it had unlawfully recouped from welfare recipients. Services Australia said in a statement that 470,000 debts would be waived, with refunds to be rolled out from July.

Some of the criticisms of robo-debt, however, were over-the-top.

Bill Shorten accused the Government of acting like a "legalised mafia" in respect of the scheme, while some newspaper columnists have characterised robo-debt as "an algorithmic weapon of calculated political cruelty" and "a terror campaign against class mobility...based on flawed maths behind false debts".

The bureaucratic and political stubbornness and laziness that epitomised the robo-debt process continued for several years, despite knowledge of its problems. Emails between senior bureaucrats, released to a Senate inquiry, show the Australian Taxation Office had advised that the scheme was "unlawful". Additionally, in 2017 a senior member of the Administrative Appeals Tribunal ruled five times that debts raised under the scheme were legally unsound.

Instead of just relying on averaged income, when undisclosed income was suspected, bureaucrats should have contacted past employers to obtain accurate information on the relevant recipient's income stream for every relevant fortnight. The bureaucracy persistently failed to do this, thus creating a mess that continued to build up.

Once the (Victorian Legal Aid sponsored) High Court case, Amato v Commonwealth of Australia was under way, the Commonwealth promptly conceded the data averaging method was wholly unlawful. The Government also agreed to pay Amato $92 in interest on the amount that was unlawfully taken from her tax refund.