Longitudinal Data Conference 2016
Written by Paul Koshy and Mike Dockery, NCSEHE
The National Centre for Longitudinal Data (NCLD) hosted the inaugural Longitudinal Data Conference from 25 to 27 October at the National Convention Centre in Canberra. The conference focused not only on on-going longitudinal data collection, but also new developments in linkage to administrative data sets and very important recent initiatives in view of the Government’s Public Data Policy Statement, released in December 2015.
In introducing the Conference, Adam Rowland, Executive Manager of the NCLD, observed that “data is now cool” and that the opportunities that big data initiatives provide for better policy making are seeing a marked increased interest in its use in Australia, be it from government, academics, private citizens or companies.
Two important themes emerged in the first day workshops: the challenge of how to manage longitudinal and administrative data sets in integrated systems and the availability of high level public data for academic and policy inquiry.
In a workshop on managing and sharing longitudinal data, Dr Steven McEachern of the Australian Data Archives at the ANU, outlined the challenges of the former, specifically those relating to the harmonisation in structure and definitions between data sets through the use of reference models and the requirement for a strong commitment to data security, most notably through analysis using the ‘five safes’ model: safe people, safe projects, safe settings, safe outputs and safe data.
Of immediate relevance to policymakers and practitioners of course is the use of such data sets in policy and program analysis. The conference provided no shortage of examples of such use. These included a large number of papers based on the Longitudinal Study of Indigenous Children, looking at outcomes such as physical and mental health, wellbeing and school performance and incorporating explanatory factors as diverse as parental experiences of discrimination, nurse home visits and dietary factors. Analyses of the Household, Income and Labour Dynamics in Australia (HILDA) survey featured prominently, and Deputy Secretary Nigel Ray of the Commonwealth Treasury cited how data from the HILDA wealth module and estimates of labour supply responses to child-care availability were being utilised to shape fiscal policy settings and policy design.
The future for data in Australia and use in such settings is rapidly emerging. ABS Chief Statistician David Kalisch, outlined key initiatives which are of interest to people in the equity space, including: the Australian Census Longitudinal Dataset (ACLD) which links data across census without compromising its commitment to confidentiality over time; the Multi-Agency Data Integration Project (MADIP), which allows government departments such as Health and the Social Security to integrate data sets to get a better picture of socio-economic conditions in Australia; and the continuation of existing collections such as the Longitudinal Study of Australian Children (LSAC).
The emergence of these data sets saw many speakers stress the importance of the ‘social license’ which underpins their release, with Professor John Brynner from University College London and Stephen England-Hall, CEO of Loyalty NZ and executive board member for New Zealand’s Data Futures Partnership (DFP), indicating that these concerns were magnified in the absence of a proper communications strategy and just as important, a translational strategy, where the benefits of such initiatives can be seen.
Professor Mark Western from the University of Queensland echoed these issues as well as warning about the temptation to see data as an end in itself, as opposed to a mechanism for describing and analysing important social issues using research methods, both quantitative and qualitative. This point was emphasised in a keynote address by Tom Calma who discussed research affecting Aboriginal and Torres Strait Islander People. He outlined the extensive community engagement efforts and the involvement of Aboriginal Community Researchers by the CRC for Remote Economic Participation, the Longitudinal Study of indigenous Children and the ‘Tackling Indigenous Smoking’ program.
A highlight of the conference program was a very energetic and entertaining presentation by Professor Mick Couper, an expert in survey research from the University of Michigan, who provided a thought-provoking reality check on the potential for big data to replace carefully constructed research designs. Professor Couper pointed out the many methodological issues associated with making inferences from harvesting on-line data when compared to probability weighted samples, and provided some great examples of premature claims of the ability of big data to cheaply replace existing data collections, such as the ‘Google Flu Index’ in which trends in Google searches for ‘Flu’ or ‘Flu symptoms’ mirrored epidemiological data on outbreaks of the flu in the United States – but, as it turns out, only for a very limited period of time.
The success of the 2016 Longitudinal Data Conference is further proof that longitudinal data and big data continue to emerge as important policy tools after 20 years of development and extension. Higher education policy, and equity policy in particular, will increasingly be discussed in view of the development of an ongoing evidence base.