AFR have a detailed article about it's establishment and the activities of the usual suspects. It was a project that was used to model potential gains and initial work suggested a 13% increase in debts recovered. Rapidly implemented in the real world despite concerns from many that it was not accurate or legal. The title of the article is wrong, it was not a spreadsheet error that lead to Robo-debt, raher it was human stupidity and cruelty.
www.afr.com
'The trouble was, the 13 per cent figure didn’t stack up. The analysis had been done on the high-risk cohorts that the historical manual reviews had targeted. By definition, they generated better returns for the departmental debt collectors than other risk groups.
The sample itself was a mere 0.1 per cent of the total client base and was reliant on applying the same 13.06 per cent uplift figure to everyone whose annual income did not match their reported income, with no evidence how closely that aligned with reality.
Nor was there any real data to test how effective the online portal would be (it ultimately turned out to be a train wreck), meaning a central assumption that a majority of online debtors would pay up was simply speculative.
When Collins and her colleagues returned after Christmas they were alarmed the spreadsheet analysis had somehow morphed into a fully blown draft cabinet submission with promised integrity savings that totalled $1.7 billion, the largest Human Services savings proposal in more than 30 years.
How a spreadsheet error spawned the $4.7b robo-debt monster
An analysis created by junior officials looked at just 418 welfare cases and contained a fundamental error. Their bosses seized on it anyway.
'The trouble was, the 13 per cent figure didn’t stack up. The analysis had been done on the high-risk cohorts that the historical manual reviews had targeted. By definition, they generated better returns for the departmental debt collectors than other risk groups.
The sample itself was a mere 0.1 per cent of the total client base and was reliant on applying the same 13.06 per cent uplift figure to everyone whose annual income did not match their reported income, with no evidence how closely that aligned with reality.
Nor was there any real data to test how effective the online portal would be (it ultimately turned out to be a train wreck), meaning a central assumption that a majority of online debtors would pay up was simply speculative.
When Collins and her colleagues returned after Christmas they were alarmed the spreadsheet analysis had somehow morphed into a fully blown draft cabinet submission with promised integrity savings that totalled $1.7 billion, the largest Human Services savings proposal in more than 30 years.
