YouGov’s Chief Innovation officer, Andy Morris, has written a couple of articles in the last week showing that YouGov’s online polls have more representative samples than fast turnaround phone polls and explaining why .
These are important pieces getting to the heart of how market research should be conducted. I believe that the only way to protect the reputation of polling is greater information transparency. Therefore I am today calling for all public pollsters to release the composition of their sample on 4 key variables – age, gender, region and education.
Academics and commercial pollsters alike will argue long and hard about methodologies - that, by the way, is not a 'crisis in polling' but the necessary process of an 'art' becoming a 'science'. Representing public opinion in the national debate is an important, vital and still complex issue. In this context, and with a pro-science spirit, there is one thing we can hopefully all agree on: greater transparency. Samples that are supposed to be nationally representative cannot be composed of 45% graduates, as phone samples are (ComRes, ICM and Populus all get their phone samples from the same supplier, Populus Data Services). Nearly half a sample made up of graduates is a huge distortion and significantly skews the polls. Even if one doesn't agree with that, we must be open about the composition and we therefore call on all polling companies from now on to publish 'level of education' as a variable, and thereby allow everyone to draw their own conclusions. We will henceforth be publishing low, medium and high education breaks (with a note on how that equates to NVQ levels in our methodology statements).