Survey Representativeness

When compared to the 2011 Census and the 2010 general election results, the survey sample is representative in terms of sex, region and party vote but less so in terms of age (in part because YouGov panellists must be aged 18 or over), ethnicity, education, and non-voting. Those in their late teens, 20s, 40s, and 80s or older are underrepresented, whilst those in their 30s, 50s, 60s, and 70s are overrepresented. In terms of ethnicity, the sample overrepresents White British respondents whilst underrepresenting those in other ethnic groups (On this topic, see Ford, Janta-Lipinski and Sobolewska 2015). Further, the sample overrepresents those with higher level qualifications (A-level and above) and concomitantly underrepresents those with lower level (GCSE or below) or no formal qualifications. Finally, although the sample is largely representative in terms of choice in the 2010 general election amongst those who turned out, it underrepresents those who did not vote. The application of the weights (Lynn 1996) provided as standard by YouGov improves the representativeness of the sample to a limited extent for some of these demographic and political measures. The comparisons between the weighted sample, unweighted sample, census data, and 2010 general election results are presented below in figures 1-6 and Table 1.

The low number of non-voters indicates that the sample is likely to contain a higher proportion of people who undertake political activities than is the case in the population. However, this is beneficial in the sense that rare political activities such as direct action are better represented, whilst response quality also tends to be higher in internet surveys than alternative modes (Chang and Krosnick 2009). Indeed, it is this mode of fielding the survey that allowed space to ask about the topics of interest in detail and thus enabled the research to be conducted. Nevertheless, the fact that the sample was not representative of the British population in some key respects has implications for the generalisability of the descriptive analysis presented on this website. In other words, some of the estimates of how many people undertake various activities and hold different views may not accurately represent the prevalence of those activities and views in the population at the time.

Beyond descriptive analysis, the focus of this research project is on relationships between variables and, although there is some evidence to the contrary (Malhotra and Krosnick 2007), the descriptive representativeness of the sample is less important for this purpose (Ansolabehere and Rivers 2013; Pasek 2016; Sanders et al. 2005). There is also notable variation in answers to the variables such that there are large numbers of respondents who undertake little or no political activity, a wide range of levels of economics, social and cultural capital, and a range of perceptions of social structure and explanations for it. Ultimately, although the data is not from a probability sample, there is sufficient reason to think that the relationships between the variables will not be significantly different in the sample from in the population at the time.

Figure 1. See also Table A1.
Figure 2. See also Table A2.
Figure 4. See also Table A4.
Figure 3. See also Table A3.
Table 1.
Figure 5. See also Table A5.
Figure 6. See also Table A6.

Published by joegreenwoodhau

Joe Greenwood-Hau is a British Academy Postdoctoral Fellow in the School of Government & Public Policy at the University of Strathclyde, where he works on the Capital, Privilege and Political Participation in Britain and Beyond project.

One thought on “Survey Representativeness

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: