Back in March, our Social Statistics section discussed how we might engage with the Economic and Social Research Council (ESRC) review of the future on investments in large longitudinal studies. A number of people got in touch to say the RSS should make some response. Here, we explain what progress has been made so far and how fellows and interested parties might engage further.
A longitudinal study is a research design that repeatedly observes a group of people (or cohort) over long periods of time. A famous example includes the Up television series, which has followed a group of same-aged people since 1964 when they were seven. They are often used by social scientists to study how life is changing over time.
The ESRC review is covering future need, methodology, potential for data linkage and how its current longitudinal studies portfolio fits with the UK’s broader data infrastructure. It has identified themes led by individual members of the team: these include content, harmonisation, linkage, access and capacity.
We have a concern that methodology for social statistics is on the fringes of the ESRC remit and funding responsibility has the possibility to get lost between research councils. Another issue is that while the UK led in birth cohorts, many countries now have them and issues of international comparability of design and analysis risk being neglected. However, we are pleased that the UK remains in a position of strength and hope that the new UK Research and Innovation body (UKRI) will look to build on this.
While there is a real concern that something will be lost if there is no further cohort investment, we are also interested in how administrative data can enhance our understanding of people's lives. A meeting at the British Academy organised by CLOSER and others heard proposals for a new study based on a survey admin spine/wave approach. It is clear that admin data can tell us new things, but also important to say that certain social relations and behaviours are not collected in this way.
One of the great contentions of the whole discussion about a possible future investment in a cohort study has been representativeness. For social researchers random sampling is now foundational but multivariate longitudinal inference does not require it. Epidemiologists have a different view and enrolment approaches have also been for antenatal recruitment. Furthemore we should remember that studies up to 1970 used a one week census of births as the cohort, with the advantage of being able to add immigrants to the sample. Our event (detailed below) will discuss this issue.
Longitudinal studies are more than just the cohort all sampled at birth to understand a generation. These studies and the potential for a new investment have dominated discussions and member feedback. However, studies of ageing are more powerful sampled later in life and other life stages are of serious interest such as the adulthood transitions in Next Steps. Becoming more important is the idea of linked lives in families and households and how social relations affect life chances broadly conceived.
RSS fellows actually using open data sometimes find the level of approvals required at each stage onerous, although access to data seems to be improving. We are also keen to promote greater capability and extended training materials to support more sophisticated analysis for substantive research. It is sometimes frustrating that the glorious riches of our longitudinal studies are used for simple regressions and cross-sectional analyses. Substantive advances demand capability in sophisticated analysis which for our data has largely untapped potential.
A more difficult issue is about where the academic interest and funding responsibility for this sort of study fits. Proposals for data enhancements go beyond educational data to health records, geocoding and biological data. And the use of these data for understanding the life course is becoming more significant. Thus it is timely to say that UKRI might take an interest in coordinating design as well as methodology.
In June, the Wellcome Trust published its new strategy on longitudinal population studies. This includes commitments to fund longer term, ie five years plus, and methodology in data linkage and analysis. While Wellcome funds studies bearing on human health, social determinants and inequalities are rising up the health agenda. It has also been funding studies in developing nations for some time and so has experience internationally relevant to the Sustainable Development Goals.
In August, representatives of the Social Statistics section met with the ESRC review team, having already submitted a summary of the views from responses and discussion with members. This discussion focused on access to data sharing/linkage as well as capacity building and methodology investments for social statistics. We made the same distinction between capacity and capability seen in the ESRC review of skills, now seeking evidence.
Although it was felt important that the Social Statistics section should hold an event on longitudinal studies, the timetable of the review made this difficult. Therefore we agreed to hold an event on the methodology separate from the review, and this event takes place at the RSS on Thursday 26 October. As well as presentations on aspects of new design and methodology for large longitudinal studies of population, we will have a panel discussion. We want to establish an RSS perspective on methodological research priorities for analytical approaches and enhancements to interdisciplinary substantive study designs.
The event will be recorded if you are interested but unable to attend in person; you can also follow the hashtag #RSSLongPop on Twitter.