Appendix C: Comparison with previous survey results

The Migration Observatory/Ipsos MORI survey was broadly consistent with other polling since 2009 on the general question of whether the level of immigration to Britain is too high, as shown in Table 1. Different organisations have asked different versions of this question, and sample from slightly different populations, so the Observatory survey should not be expected to exactly replicate other results. Moreover, there is a degree of random error inherent in polling and survey research, further diminishing any expectations of a precise match. Nonetheless, it is encouraging that the results do not diverge dramatically from other findings, particularly when taking into account differences in survey procedures.

These differences include question wordings. The Migration Observatory survey used the same wording as the Government’s Citizenship Survey, but other organizations ask similar questions in different ways, as Table 1 shows. Sample frames also vary. Ipsos MORI Capibus sample includes Scotland, and also include respondents as young as age 15. These decisions have the advantage of representing a broader population but mark a difference from other surveys that sample only in England, or only in England and Wales, and that only sample from age 16 or age 18 and above. This affects results, because both younger respondents and respondents in Scotland are significantly less likely than others to support reduced immigration.

The difference in sampling explains part of the difference between Migration Observatory/Ipsos MORI results and the Citizenship Survey results from 2009-2010. Excluding Scotland raises the percentage favouring reduced immigration to 70%, four percentage points less than the 74% from the Citizenship Survey as shown in Table 1 (Note that the 78% favouring reduced immigration reported in Migration Observatory briefings and charts comes from analysis excluding “don’t know” responses, while analyses in this report include “don’t know” responses, yielding lower percentage estimates for other responses.)

Table 1