The question of whether internet polling is better or worse than face-to-face or telephone polling quite often causes sparks (friendly of course) to fly better different pollsters and pundits. So I thought it would be useful to compare YouGov against all the other political pollsters (except for BPIX – because they also use the internet but also aren’t a member of the British Polling Council and so are of questionable reputation*).
Here are the monthly average scores since the general election for the three main parties; the solid line is the average of all the polls with fieldwork dates in that month by firms other than YouGov; the dotted line is the average of the YouGov polls from that month.
The conclusion? YouGov gives Labour and the Conservatives regularly roughly the same ratings as other pollsters – sometimes a little higher, sometimes a little lower. But their Liberal Democrat ratings are consistently lower than other pollsters. Why and who is right? Debate away…
* Rather strangely, the people behind BPIX do seem credible and knowledgable. But they don’t seem very open about how BPIX operates. Their website has been “under construction” for as long as I can remember, they aren’t a member of the British Polling Council and my previous email query to them didn’t solicit a reply – though of course emails do sometimes go astray. If anyone from BPIX is reading this please do use the comments facility to clarify these matters.
3 Comments
Speculation: the way they weight may be messing things up, or they may be over adjusting? They want a representative sample, so use IIRC newspaper read and votes at previous general to determine how much an individual respondent should count for.
Given that they’re polling online users, odds are they’ve got a disproportionate number of younger, male, educated geeks signed up, and then they have to weight them out, and may be doing so too heavily (we know we generally poll better within that demographic, students, etc).
Could be anything, naturally, or it could just be that they do get poor responses from our supporters, but that’d be weird. I suspect their weighting is wrong.
Anthony Wells has a good response to this piece on his website: http://www.ukpollingreport.co.uk/blog/archives/983
Anthony Wells offers various explanations, but to me the most plausible seems to be this one:
“ICM and Populus weight using recalled past vote and YouGov weight using party identification.”
Carrying out a survey earlier this year, a fair number of people (though not a lot) identified themselves as Labour. However, from their comments it is quite clear that last time (or even several times) they have voted Lib Dem.
So ICM and Populus would have classified these as Lib Dem (because of their remembered last vote); while You Gov would have classified them as Labour.
So the weighting would be different for the various pollsters: ie You Gov would have given a lower weighting to current Lib Dem supporters.
It would also explain You Gov’s over-weighting for Others: lots of people identify as Green, for example, but would not vote Green (either in the past or in the future) for lack of a candidate.