Radical analytics power welfare change

In what is claimed as a world first, the lifetime cost of welfare dependency is revealed, analysed and segmented.
Written by Rob O'Neill, Contributor

Since 2007, analytics have been at the heart of welfare reform in New Zealand, minister of social development Paula Bennett told an SAS user conference in Wellington this week.

paula bennett
Paula Bennett

That effort has seen analytics software and techniques used to quantify the lifetime cost of welfare dependency, to segment and analyse that liability – with sometimes startling results – and to increase transparency around the impact of government spending.

But the impact of the programme remains blunted as long as it is restricted to just the Ministry of Social Development (MSD), one of the project’s architects warned.

“MSD is a great start, but it’s not nearly enough,” said James Mansell, director of innovation and strategy at the Ministry. Mansell advocated an all-of-government approach to data analytics and more comprehensive tracking of individuals within the system.

The analysis produced a baseline cost figure of a kind Bennett said had never been generated anywhere else and which is now recalculated annually to track changes and the success or failure of government programmes.

All up, the lifetime cost of welfare in New Zealand was $78 billion at 30 June 2011.

Predictive risk modelling also allows agencies to identify people at risk and to target interventions as well as to guide policymaking, Bennett said. Segment analysis over time helps government to track the success of programmes and invest money where it will make the biggest difference.

Alan Greenfield, of Australian analytics and actuarial consultancy Taylor Fry which worked on the lifetime cost valuation, told the conference that the work had already revealed that far more effort was being expended on managing unemployment beneficiaries than was justified by their contribution to the overall cost.

The lifetime cost of the Unemployment Benefit was $3.3 billion compared to $21.4 billion for the Invalid Benefit and $15.1 billion for the Domestic Purposes Benefit.

Greenfield said another finding was that 75% of the cost came from people who first went on a benefit at the age of 20 or less.

Bennett told the conference 13% of the working age population of New Zealand received welfare of some kind with 161,000 receiving a benefit for at least half of the previous 10 years. 139,000 people had spent more than a decade on a benefit since 1993.

The welfare system was not giving people the support they needed to build their futures, she said. It was passive rather than active.

“Without the use of analytics we were putting our finger in the air and throwing money at the problem,” she said, claiming an early success in registering the lowest number of youth not in employment, education or training since 2008.

Greenfield said he most recent baseline recalculation, at 30 June 2013, was down to $76.5 billion.

“I have Treasury wanting to give us money,” Bennett said. “It’s unheard of.”

As important as the savings, analytics can shift agencies from being ambulances at the bottom of the cliff when dealing with the abused and neglected to giving them the ability to identify and target the people most at risk, the minister said.

“Analytics means we can get into their lives and make a difference,” she said.

Bennett said government was catching up with business in the use of analytics tools.

“I’m excited about the future and where we can take this work,” she said, adding that a team of people were advising on issues of privacy.

Mansell said MSD’s experience had revealed bigger challenges and opportunities, such as extending business analytics across the whole of government and matching performance indicators with outcomes rather than the delivery of processes.

Every day people in government are making life and death decisions, he said, not just in child protection, but in police and on the border and elsewhere. But they don’t have a real view of the people they are dealing with.

“There is no free lunch in the public sector when you make a mistake, but you have to keep making decisions.”

Would knowing someone was incarcerated or abused early in life make a difference to agency decisions, he asked. Would knowing their educational outcomes and benefit history?

The problem is government is broken up into service silos, in education, health, law enforcement and more, that obstruct real progress on social problems because they obscure the long term effects of interventions.

Mansell showed a mock-up timeline for two individuals with full lifetime data as an example of what could be possible with whole of government analytics.

Referring to one of his hypothetical examples, he asked: “The people who dealt with him probably met their KPIs [key performance indicators], but did they get a good outcome?”

 It’s hard to get a coordinated response if you can’t track. You also can’t talk about outcomes unless you have transparency about what happens after a service has been delivered.

For that reason, Mansell said, while most government agencies have diagrams showing the client at the centre of what they do, most of their data is really system or service focused, not customer focused.

It was possible, for instance to identify the 2000 kids most likely to offend in future, but that would set off wails about “pre-crime” and “Minority Report”.

Until there is an integrated view and joined up interaction not  much can be done about these kinds of cases, he said.

There are barriers to that though. For instance, the Inland Revenue Department would not want to share data as if it starts giving it out, businesses will be wary about putting it in.

“A lot of people have a lot of interest in data,” Mansell said.

Greenfield said overall the project took an "investment" approach. What was being built was a framework to measure how best to invest and measure the cost and effectiveness of return-to-work initiatives.

The project followed recommendation of the 2010 Welfare Working Group to reduce long-term dependency among people of working age, he said. It wasn’t about reducing costs up front, but about improving employment outcomes to then reduce costs.

Editorial standards