Big data and welfare

The idea of big data in the welfare system has come to be associated with the reconciliation of ATO pay data with social security income, and the resulting teething problems of #notmydebt. However, relatively little consideration has been given to how big data might shape government’s ability to better target the welfare system.

One of the challenges in welfare policy is the fact that every recipient’s circumstances — and consequently, needs — are different. Welfare has always been required to trade-off between simplicity (fewer payments) and better targeting. As the system becomes more streamlined, it becomes easier to administer but increases the risk that certain recipients will be poorly served.

As tempting as it may be for believers in small government to simply dismiss these concerns on the basis that keeping welfare lean and mean is the best approach, this would be a mistake. Simplifying the system by broadening the number of recipients of a payment means that even small increases in the base rate can have serious budgetary impacts.

Complexity may be a second-best option if the alternative is that the size of the welfare state increases.

For example, if the parliament accepts the argument that long term recipients of Newstart are struggling to live on the payment, and decides to increase the rate, would it be better to restrict the payments to those who have been unemployed for more than 12 months or to give it to everyone?

Could we go further and adjust the rate of Newstart on the basis of private savings? Could it reflect cost of living differences in different parts of the country? Could government use skill shortage information to direct training obligations?

Individual case workers could do this for a handful of recipients. However the cost of paying people to manage the caseload created by complexity is prohibitive. Even monitoring current compliance obligations currently is hit and miss.

However, if big data can replicate this role it may fundamentally change how we see the welfare trade-off. As we become more sophisticated in processing large amounts of data, we could learn how to target welfare on almost an individual level at a cost we could afford.