He talks to us about the importance of quantitative skills being better integrated into the education system and how the use of admin data and surveys are evolving, especially with the census consultation underway.
It's been a few months now since the Quantitative Skills in the Social Sciences Methods programme was launched. How has the programme developed since it was announced and what kind of interest has there been from universities?
The Nuffield Foundation/HEFCE/ESRC call for Undergraduate Centres of Excellence in Quantitative Methods opened in October 2012 and closed earlier this year. There was a strong response to the call, and following peer review and a Commissioning Panel meeting, 15 Centres have been offered funding. We believe this will produce a step-change in the provision of highly trained social scientists in the years to come, which will be of benefit to academia and the wider workforce.
An important factor in developing quantitative skills is the way pupils get to engage with data and statistics in schools and colleges. What is your view on improving the provision for including statistical and quantitative skills within the GCSE and A level curriculums?
We agree that engagement with data and statistics is a vital component in boosting quantitative methods skills more generally and we think that this needs to be further embedded in the school curriculum. It is not for the ESRC to comment on policy at this level, but we would advocate that students need to be exposed to maths and statistics throughout the curriculum, including in social science subjects. We have been increasingly emphasising, along with other funders, the importance of advanced training at early stages to improve the pipeline of skilled researchers who will go onto postgraduate study and academia, or into the workforce more generally.
In the last few years there has been an increasing debate on the merits of government use of surveys versus admin data. How do you see the future of the traditional survey?
We have a very long history of funding surveys for socio-economic research and we expect them to continue to play a major role in data gathering. Overall, there remains considerable demand for large survey data as the UK Data Service has shown with its 23,000 non-academic users. The impacts that can be generated from these studies can be impressive – for example, Grandparents Plus and Age UK recently lobbied for changes to retirement and state pension provision based on the evidence from Understanding Society that grandparents care for 1.6 million children saving families £7.3billion a year.
At the same time, ESRC is also leading the way in funding better use of existing ‘Big Data’, be they qualitative, quantitative, visual, administrative or born digital. The establishment of the Administrative Data Research Network is an excellent example of this, with the aim to facilitate access to linked government administrative data for socio-economic research across England, Wales, Scotland and Northern Ireland. Together with the National Statistical Authorities, four Administrative Data Research Centres will provide access to such data for accredited researchers.
However, while there are exciting opportunities opening up with these existing data, they are not designed for research and have many gaps that can only be filled by dedicated surveys. Hence we need a dual approach that combines the strengths of both types of data collection. Of course, how user-defined surveys are conducted, given the falling participation rates requires some thought. Hence the ESRC funds an Innovation Panel in Understanding Society which experiments with different survey modes including web-based and mixed approaches.
In order to improve the quality of admin data collection and sharing, what do you think is the best way to address the privacy concerns around this?
In recent years, the research community has provided opportunities to explore the potential use of administration data and the value it can offer to the social science community. Enabling access to potentially sensitive data has been challenging: researchers have had to contend with issues arising from datasets being primarily designed for operational purposes where the data structure is not conducive to easy analysis; and there are issues around the need to protect the confidentiality of individuals - such as legislation and access limitations.
Despite the challenges the ESRC has continued to work closely with Government and we have established important relationships to enable the use of administrative data sets, subject to appropriate safeguards. Approved researcher status, accredited safe havens for access, and appropriate application processes are critical to giving assurances to data owners, stakeholders and the public that robust safe guards are in place. At the heart of any process must be the principle of maintaining the confidentiality of individuals whose data is being used for research purposes. Strong governance frameworks, including the role of Ethics Committees, are central elements of any infrastructure which provides access to such data.
With census consultation underway, what is your view on whether another census is necessary in 2021 or do you think admin data could be used to produce census information?
Our first priority will be to consider the research value of the data to be produced under whatever option is being proposed going forward. The ESRC supported research draws on the full range of census data products and on the results of multiple censuses to address a very wide range of social science questions. Census datasets are used to investigate specific thematic topics such as: social inequality; changing household structure; patterns of migration; neighbourhood segregation and differentiation; impact of education or environment on social status; and health later in life.
Following on from this the question then becomes one of whether or not this data can only be produced by conducting a conventional census? It is important to acknowledge that the census is not without its problems - although the 2011 census was very successful the cost of maintaining response rates is high and rates are declining internationally. Even the conventional census employs a large coverage survey and dual system estimation methods in order to produce the final estimates. Extensive use of online data collection will need to address further complexity in coverage adjustment, so this option would present new challenges compared to the traditional 2011 design. However it does have the potential to deliver the traditional census strengths of social and geographical detail and a range of different outputs.
ONS work on administrative data linkage is currently focused on population estimation by linking major administrative sources to deliver statistics primarily at the local authority level, but with the potential to produce small area estimates. Implementation of this approach would involve legislative change and is much less well developed with regard to production of statistics on socioeconomic characteristics. It is in this area that the greatest research risks appear to be presented. The option being presented for consultation offers very limited socioeconomic detail for the smallest areas and would only be able to produce sub-local authority level data by pooling survey results for multiple years. This option would not produce a data product comparable to what is currently available. For example, no work has been presented on how interaction datasets, which include internal migration flows and journeys to work between small geographical areas, could be generated from administrative data.
In order that a combined administrative and survey option can deliver all the data that are required it is essential that we clarify which data are required for small geographical areas. However, to date insufficient attention has been paid to these issues and the societal importance of the insights from such research has been underestimated. Although it is clear that the first objective to produce timely population estimates at local authority level could be achieved by the methods being evaluated, it is not yet clear that the research data required could be produced more efficiently by these means. Some subject experts are yet to be engaged by the ONS consultation processes, and many researchers risk losing key data sources without having recognised and assessed the implications. We would therefore join with RSS in encouraging the research community to respond to the current ONS consultation.