Social welfare swallows up a quarter of New Zealand's gross domestic product, so it is important to identify where that money can be most productively and efficiently used, says Mike O'Neil of the Ministry of Social Development (MSD).
O'Neil is using data analytics software from SAS to assist in the process of welfare reform. Accumulated factual evidence on client behaviour is useful in targeting help where it will do most good and hence helping to ensure better social, fiscal and economic outcomes for the long-term.
This effort accords with the government's major exercise in welfare reform.
In an initiative that has been under way for almost 18 months, MSD has been "identifying cohorts of clients and evaluating their risk of long-term dependency," O'Neil told the SUNZ (SAS Users of NZ) conference last month.
Set against the history of those clients it gives some indication of the likely effectiveness of preventative measures applied earlier in life. "Then we can stop what doesn't work and continue what does," he says.
"We're trying to identify 15-17 year-old school-leavers most at risk of not being in education, employment or training, with an emphasis on likelihood of long-term benefit liability," he says.
To analyse the relationship between historical background and later outcomes, O'Neil took the unusual route of using Campaign Studio, a SAS product usually applied to managing marketing campaigns, along with SAS Data Flux for matching the data from the Child, Youth and Family (CYF) division of MSD and the Ministry of Education (MoE).
As with a marketing campaign, the key long-term objective is to optimise spending on various initiatives to achieve best results.
There are obvious privacy implications in correlating the data from MoE, CYF and Work and Income and some changes had to be made to the law to accommodate this.
There were challenges in the matching of identities between records and the likelihood of a mismatch was evaluated as part of the exercise.
And there were significant ethical questions surrounding the whole exercise -- "should we be doing this at all?" -- O'Neil acknowledges. The exercise was justified on the strength of MSD's responsibility to use taxpayer funds responsibly. The ministry has "a fundamental duty" to optimise allocation of the money it receives, O'Neil says.
NZ Police
Meanwhile, the New Zealand Police have a similar end purpose in view to the Ministry of Social Development -- to analyse the causes and incidence of undesirable behaviour so plans can be made to minimise it and to apply their resources appropriately to this end. While MSD's activities are currently being reformed, Police went through that process in the 1980s and 90s, but are still dealing with its outcome.
The Police decision to use SAS software was indirect; "we looked at what other police forces in the region were doing and Victoria's seemed to be best practice. They used SAS." NZ Police arrived at a similar business model and the technology to use "fell out of that", Knight says. Elements of the system's architecture came from the Western Australian police. The development is centred on an analytical data mart.
Statistical analysis enables the Police to count crime in a consistent way and, like MSD, to extrapolate the current and future situation based on the history of the people involved -- both offenders and victims. Collection of systematic data on victims has been a gap until recently, Knight says.
Management in Police is still made up to a large extent of promoted active officers, whose decisions to date have been heavily dependent on "good judgement, with a heavy element of intuition," Knight says. Objective statistical analysis gives a firmer basis.
Business parameters for the new way of working included low cost and speed to market and this has led to a rejection of complex systems development lifecycle management in favour of faster processes. The coding was outsourced to SAS Institute NZ.
Police have the basic software complete and are currently working on applying the analytics, refining business processes and service-level agreements and formulating a business intelligence strategy.
Likely future steps include predictive analysis and the application of geospatial data.