Using representativeness indicators to evaluate Understanding Society dataset representativeness given sample attrition across waves and non-response during data collection periods.
This fellowship aimed to improve understanding of representativeness in Understanding Society in several ways. The use of representativeness indicators to evaluate datasets was novel, and enabled their quality to be compared and impacts on representativeness to be quantified. The indicators were used to evaluate how UKHLS datasets are affected by attrition across waves. The project considered representativeness with regard to subject attributes likely correlated with survey responses, and also whether response decisions are impacted on by life processes, such as employment, health changes and family separation. This research informed discussion on UKHLS dataset quality across waves. It was also f use to survey designers, because it identified sources of non-representativeness to target with adaptations to improve quality, and to see whether subject life changes affect responses. In a second sub-project, the fellowship used representativeness indicators to evaluate how dataset quality changes over data collection, i.e. with the number of attempts to contact non-respondents. This considered representativeness with regard to response correlated subject attributes, and also the impacts of life changes. A focus was to identify points after which further responses (and attempts to contact non-respondents) minimally effect quality. As well as identifying targets for adaptations to improve datasets, the project addressed the issue of whether (as many) attempts to contact non-respondents are required to maintain quality.
Outputs
Papers
- Moore, J. C., Durrant, G. B., & Smith, P. W. (2021). Do coefficients of variation of response propensities approximate non‐response biases during survey data collection?. Journal of the Royal Statistical Society: Series A (Statistics in Society), 184(1), 301-323. https://doi.org/10.1111/rssa.12624
Presentations
- Moore, J.C, Durrant, G.B. & Smith, P.W.F. Using representativeness indicators to monitor the impacts of missing data. CEMMAP workshop, London, May 2019.
- Moore, J.C, Durrant, G.B. & Smith, P.W.F. Understanding Society survey dataset representativeness: Across waves and over data collection. ESRA conference, Zagreb, July 2019.
- Moore, J.C, Durrant, G.B. & Smith, P.W.F. Understanding Society survey dataset representativeness: Across waves and over data collection. Understanding Society 2019 conference, Essex June 2019
