Symposium: Advances in Survey Methods for Cohort Studies 2
Insights from Developing an Internet Use Study within A Longitudinal Study of Adolescents (Larissa Pople, UCL Centre for Longitudinal Studies)
Passive data collection via smartphones offers significant potential for improving measurement of behaviours such as internet use, but raises important questions about feasibility, acceptability and data quality in survey contexts. This paper draws on the internet use component of Growing Up in the 2020s, a longitudinal study of secondary-school children in England.
A sub-sample of around 400 participants was invited to install an app on their smartphones to passively track internet use over four weeks in 2025. The app recorded websites and applications used and time spent on them, providing granular, objective measures compared with self-reports.
We assess key feasibility metrics, including consent and installation rates, participant compliance and retention over the tracking period. We also examine practical and ethical considerations, including participant burden, privacy concerns and implications for scaling up app-based data collection in longitudinal surveys.
Web-First or Face-to-Face at the Transition to Adulthood? Experimental Evidence from the Millennium Cohort Study Age 23 Survey (Matt Brown, UCL Centre for Longitudinal Studies)
Longitudinal studies are increasingly adopting mixed mode data collection, especially web-based approaches, in response to declining response rates, rising costs, and widespread internet access. The Millennium Cohort Study (MCS) Age 23 Survey, the first in young adulthood, provided an opportunity to reassess traditional face to face methods. Earlier waves relied entirely on interviewer-led data collection, but at age 23 participants were expected to be more mobile and harder to engage.
This paper reports findings from an experiment embedded in the first phase of fieldwork, with participants randomly assigned to either face to face or a web first mixed mode design. The mixed mode approach achieved higher response rates, faster completion, and reduced interviewer effort. Item non-response was low across both groups, with modest differences overall. Some measurement variation was observed: web respondents reported less socially desirable behaviours, while interviewer-led modes supported more complex responses. Overall, the web first design improved efficiency and participation with limited impacts on data quality.
An Evaluation of the Look-up Tool for Occupation Coding (Helena Koerber, UCL Centre for Longitudinal Studies)
Occupation is a central measure in social survey research. Typically, occupations are coded to standard frameworks such as the Standard Occupational Classification (SOC) in the UK. While post-interview office coding has been regarded as the gold standard, it is both time-consuming and costly. A new approach—the look-up tool—allows participants to code their own occupations during the survey, offering potential savings in cost and time. This presentation will discuss findings from the Next Steps Measurement Lab. The results show that most respondents were able to code themselves successfully and rated their codes as accurate, though agreement with office coding and consistency over time were weaker. These results highlight both the promise and the limitations of the look-up approach for large-scale surveys.
“Thinking back to your childhood”: How accurate is our recall? Evidence from the National Child Development Study Age 62 (Darina Peycheva, UCL Centre for Longitudinal Studies)
Childhood circumstances have a significant impact on adult life, and many studies of older adults collect retrospective information on early life. For example, the English Longitudinal Study of Ageing (ELSA), the U.S. Health and Retirement Study (HRS) and the Survey of Health, Ageing and Retirement in Europe (SHARE) include life history questionnaires. However, little is known about how accurately adults recall childhood experiences, or whether some aspects are recalled more reliably than others.
The National Child Development Study (NCDS), which follows over 17,000 people born in Britain in 1958, included a retrospective childhood questionnaire at age 62 which re-collected data collected prospectively at ages 7, 11 and 16.
This study compares retrospective and prospective reports to assess recall accuracy across domains, and examines whether using retrospective versus prospective data affects estimates of associations with later-life outcomes.
Findings have implications for survey design and analysis of retrospective data.
