This is a repost of a blog by Alissa Goodman, Director of CLS, originally published on the ESRC website.
The ESRC last week published its Longitudinal Studies Strategic Review, a report by an international panel, which was commissioned by the ESRC to review its investment in longitudinal studies.
The panel recognised that, thanks to the ESRC, the UK has a strong and unique mix of cohort and panel studies, which will serve social science and beyond in the decades to come.
What does this mix consist of?
Understanding Society is the largest household panel of its kind in the world, following a nationally representative group of UK households drawn from across the whole population. It covers all members of a household, following them up every year, allowing researchers to study dynamics both within and between households. Its large sample size makes analysis of sub-groups of the population possible, including in different geographic areas, and an ethnic minority boost also allows the study of dynamics within ethnic minority families.
Birth cohort studies, such as those at the Centre for Longitudinal Studies, where I am Director, are quite different, and complementary in design. We run three of the four ‘famous’ UK national birth cohorts, of the Millennium, 1970, and 1958. These studies follow large and nationally representative groups of individuals born in a given year, from ‘cradle to grave’ – i.e. from birth and across the whole of their lives. We also run Next Steps, a cohort born in 1989, originated in adolescence.
The CLS cohorts provide unique evidence on how people develop from infancy, throughout childhood and adult life. They explain the roots of many sources of inequality, showing how early (and earlier-) life circumstances are important for later life outcomes; how advantage and disadvantage are transmitted across generations; and how people manage key life transitions. In combination they are a powerful tool for charting and understanding how changes across generations are unfolding, and why.
Both cohort and panel studies combine biological and social data – making them rich resources for interdisciplinary research.
They each answer very different types of research questions. Their combination makes the UK a world-leader in longitudinal research.
As the review showed, the use of the ESRC’s longitudinal data has grown over time, and extends across a wide range of disciplines, both nationally and internationally. They have supported the development of many government policies, including on breastfeeding, childhood obesity, and welfare-to-work, and have also underpinned medical advances.
As the review made clear, many scientific and policy questions simply could not be answered without the richness and depth of information that longitudinal studies provide, even with the increasing availability of administrative data. The use of administrative data – including the recommended new ‘spine’ – can enhance these surveys in important ways, but the review rightly found that it does not replace them.
The review made a number of recommendations – including that a new birth cohort study should be commissioned. This will be vital to ensure that policy is equipped with the best evidence both to understand, and to address, the needs of new generations of babies and children now, and in the decades to come.
As well as the policy needs of the very young, another policy priority for the future is population ageing, to which the review gave much less mention. As pointed out by the Resolution Foundation in its report published last week, population ageing is as pressing a concern as the challenges faced by ‘millennials’, and indeed the two are linked.
Again, the UK’s longitudinal mix of panel studies and cohorts places us in a position of strength for the study of ageing. Both Understanding Society, and the English Longitudinal Study of Ageing (ELSA), are panel studies representative of the older population, priming us to better understand the dynamics of ageing.
At the same time, the UK’s birth cohort studies are now, after decades of foresighted investment, in the enviable position to provide vital information gathered from across the whole of life about older age groups. The oldest of these (including UCL’s 1946 cohort as well as CLS’s 1958 and 1970 cohorts) are indeed the only national studies in the world to link circumstances from birth all the way through to middle, and older age. Their strong integration of the social, economic and biological now provide a powerful series to inform on some of the major issues facing our ageing societies.
For example, the Department for Work and Pensions is co-funding the age 61 survey of our 1958 cohort, for the rich insights it will provide on the retirement plans and expectations of this generation, and – for those who have them – their pension savings. It will inform policymakers in DWP on the lifelong factors – especially their health – which will impact on their plans for extended working life.
More generally, the older cohorts will enable researchers to address pressing and diverse societal concerns such as cognitive decline and dementia; the factors that can mitigate against social isolation and loneliness in older age; and how patterns of socio-economic inequalities are perpetuated across generations, including through powerful mechanisms such as inheritances and bequests.
The review raised the question of national representation among the older studies, and here we are very much on the front foot. While some attrition is inevitable, continued participation from our study members is in fact very strong, with retention rates of around 60 per cent over six decades in the 1958 cohort, which is as high as in other longitudinal and panel studies which are a fraction of their duration. Well-known statistical techniques, such as multiple imputation or weights, using the very rich individual-level data available from before study drop-out, retain the representativeness of the samples, both to their original target populations, and to contemporary population totals.
The review report encourages innovation, including in the use of new technologies for data collection, and recommends that the ESRC commissions an ‘innovation panel’ in the CLS cohorts which would support methodological development. This could rigorously inform the choice of key developmental measures in large scale longitudinal cohort studies, building on some of the innovations we’ve already introduced.
For example, just last week we released new data from the Millennium Cohort Study teenagers’ time use diaries, which they recorded using either a smartphone app or online, which gives unique information on how they are spending their time, including their social media, gaming and homework habits. Alongside this come data from wearable accelerometers which capture objective measures of physical activity, sedentary behaviour, and sleep. All of these can now be used to answer important questions about how young people’s activities are affecting other aspects of their lives, such as their cognitive development and education, mental health, and wellbeing.
There are also great opportunities that await in applying new statistical methodologies to the data we collect. For example we are now applying machine learning to analyse at scale the language used in the essays that our 1958 cohort members wrote at age 11 in 1969, in order to provide insights into population ageing. We are also using machine learning to construct a ‘Policy and events bank’ in the Millennium Cohort Study to aid with causal identification and policy evaluation in the cohorts.
The review threw down a challenge to the ESRC and its constituents, that evidence of the impact of its longitudinal studies should be better recorded and understood. There’s no doubt that many areas of government social and economic policy, including on education, welfare, health, and family policies, and on social mobility and inequalities of all kinds, have been informed by longitudinal evidence. Fostering more use of longitudinal data and the findings from these data, and creating a stronger chain of evidence from research to policy is an important priority for the longitudinal studies themselves, and the research community who uses them.