In longitudinal surveys representativeness is important to enable population inferences and to ensure that the sample is diverse enough to allow subpopulations to be accurately analysed.
Understanding Society was designed as a large, equal-probability sample of the general population of the UK. To ensure that meaningful analysis can be done on different population groups, particularly minority ethnic groups, Understanding Society was designed to oversample ethnic minority and immigrant households.
At Wave 1 (2009-10) Understanding Society was very close to the population estimates of the 2011 UK Census. As a longitudinal study, our sample evolves over time with children turning into adult participants, new households forming, individuals moving out of scope of the Study, or passing away. This constant movement means that we employ several strategies to ensure that Study data remains representative.
Comparing response rates
Lower overall response rates create a risk that the sample will become less representative and biased introduced that affect analysis. For longitudinal surveys it is important that high response is maintained wave-on-wave, as large sample sizes are needed to support longitudinal analysis.
In Understanding Society response rates for individuals and households are monitored throughout fieldwork and reported annually in our user guide. Understanding Society has achieved high response since its inception, with response rates of between 80% and 90% for previous wave responders.
We also compare our response rates with comparable surveys in the UK and internationally, which show that Understanding Society performs as well, or better, on response.
Research into the best methods for recruitment and retention has focused on the role of interviewers, communication with participants, survey design and the mode of interview. You can read more about response and attrition in this section. [Link to response page]
Alongside comparison with other surveys, Understanding Society also benchmarks external data on the UK population. Our analysis shows that at Wave 6, compared to ONS 2015 mid-year population estimates, when weighted the Study was still broadly similar to the UK population, especially for region and sex. At this stage in the Study there was a slight underestimation of young adults and people of South Asian and Black ethnic groups. At Wave 8, comparison with ONS data showed that weighted estimates of the key characteristics of the whole sample were comparable, with a slight underestimate for younger age groups, those living in London and some ethnic minority groups.
Members of the original sample of Understanding Society are followed indefinitely, and over time new people join the Study. They could be born into an Understanding Society household, form a partnership with an original sample member, or move into a household in the Study. Depending on the relationship with the original sample member, some of these joiners are followed if they leave the original household (for example, children born into the Study), but many stop being part of Understanding Society when they leave the household.
Research has been done to see who temporary sample members are and the role they have in maintaining the representativeness of the Study. Analysis shows that temporary sample members tend to be drawn from more mobile groups, like students or people who regularly work away from home. Including them in the Study helps to maintain the cross-sectional representativeness of Understanding Society, particularly for migrant groups. A significant number of temporary sample members in Understanding Society are European migrants.
Even with high year-on-year response rates, cumulative attrition reduces the sample size and new immigrant populations who arrive in the country after the last sample were recruited are not represented. To address these issues, Understanding Society periodically ‘boosts’ the sample by inviting new participants to join. Boosts of the general population sample are implemented every 10 years, as is a separate boost specifically for immigrants and ethnic minorities.
It’s important to note that to be representative Understanding Society does not need to be a simple random sample, or a miniature version of the UK population. We do require the Study to have complete population coverage, but unbiased estimates can be obtained from biased samples by using weights. Weights adjust for unequal selection probabilities, survey nonresponse and sampling errors. Unweighted analysis of Understanding Society does not correctly reflect the population structure, as some groups may be over-represented by design.
Understanding Society produces weights to go with each dataset. You can also create tailored weights if your analysis is not covered by the weights provide.
Need help?
Visit our new user pathway
to explore the data and online resources or contact the User Support forum if you have a question for the Study team.
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.
