Occupation coding in longitudinal studies presents challenges: we must ensure high precision while maintaining consistency over time. Firstly, jobs and occupational structures evolve alongside labour markets, necessitating periodic classification updates that are not easily integrated into panel data. I begin by discussing these challenges and potential ways of addressing them within the Understanding Society data.
Secondly, major changes in questionnaire design intended to capture these evolutions can also disrupt longitudinal consistency. Here, I focus on the implications of revisions to occupation measurement in Wave 13 of Understanding Society, driven by the introduction of a new multiple-jobs module. Before Wave 13, respondents in continuous employment confirmed whether their previously reported occupation description was still accurate (dependent interviewing). If they indicated a change, they provided a new description. In Wave 13, dependent interviewing was suspended, and respondents were instead asked to provide full descriptions of the tasks and duties for each of their jobs, which were then independently coded. While this approach provided up-to-date information on all jobs, it also led to a sharp increase in recorded occupation changes in the main job variable, with 45 per cent of continuously employed respondents appearing to change occupation.
To address this, a post-hoc adjustment compared prior and current verbatim job descriptions, emulating the pre-Wave 13 validation method based on respondents’ assessments. Ultimately, 70 per cent of these occupation changes were deemed spurious, reducing the overall rate of change to 15 per cent.
This case study highlights how changes in survey design can significantly impact longitudinal data consistency. It underscores the value of dependent interviewing in longitudinal research and demonstrates how retrospective validation techniques can mitigate artificial variability, contributing to best practices in occupational classification and survey methodology.