Understanding Society is a panel survey of UK households with yearly interviews.
The Study began in 2009-10. The overall Study has multiple sample components to enable research of different sub-groups over time and location or geography:
2009-2010: The General Population Sample (GPS): (i) a clustered and stratified, probability sample of approx. 24,000 households living in Great Britain in 2009-10 (ii) a simple random sample of approx. 2,000 households living in Northern Ireland in 2009 (selected with twice the selection probability as the Great Britain part) See Sample design for Understanding Society.
2009-2010: The Ethnic Minority Boost Sample (EMBS): approx. 4,000 households selected from areas of high ethnic minority concentration in 2009-10 where at least one member was from an ethnic minority group. See Screening questions Appendix III and sample design paper – Design of the Understanding Society ethinic minority boost sample.
2010: The British Household Panel Survey sample (BHPS), added in Wave 2: approx. 8,000 households from the BHPS sample.
2022-2023: A General Population Sample boost (GPS2): added in Wave 14: approx., 5,800 households. A clustered in Great Britain but unclustered in Northern Ireland sample (selected with equal selection probability as the Great Britain part) in 2022-2023. See Wave 14 Boost technical report.
What are the different Samples?
This video gives an overview of the sample design for Understanding Society and how to account for it in analysis.
Listen to Peter Lynn Deputy Director of ISER talk about how participants are selected to take part in the Study and the importance of longitudinal data.
Survey sampling:
Tips for analysts:
Analyse all samples that are available together. As the data for all samples are provided in the same data file, as long as you don’t explicitly exclude a sample you will be doing this by default. You can analyse the GPS or the GPS2 by themselves, but it will mean your sample size will be smaller than if you were using all samples and the sample sizes will not be large enough for ethnic minority and immigrant groups. Additionally there are no appropriate weights to use any of the samples by themselves, and so the estimates based on analysing these samples by themselves may be biased.
While the samples are probability samples, not all sections of the population were selected with the same probability. Also, not everyone selected and asked to participate in the interviews did so. To correct for bias due to these two reasons we recommend you use the weights provided. TIP: This is explored in the weighting and sampling section.
All samples other than the GPS-NI and GPS2-NI part had a clustered and stratified design. As statistical softwares assume that the data is simple random sample, to estimate standard errors correctly you will need to explicitly inform the software about the clustering variable w_psu (primary sampling unit) and the stratification variable w_strata. For guidance on this visit the Clustering and stratification page.
The Economic and Social Research Council is the primary funder of the study
The Study is led by a team at the Institute for Social and Economic Research (ISER) at the University of Essex.
