The Sociology of Opinion Polls

While colleague Mark Gillard is delayed reflecting on the merits of kippers, let me entertain you will some thoughts about opinion polls. Coming to the end of a general election campaign when, as usual, we have been regaled with the views of numerous experts and pundits, its worth pausing to reflect on how much we can trust the opinion polls. Accurately done, polls can be very useful tools. But the trouble is, it is hard to do them well - one of my pet points - students always think research is easy - just dash out a questionnaire or cobble a few interviews together. Wrong. And here is C4’s Gary Gibbon, explaining why the polls may not be quite as valid (technical sense) as some claim.

From Gary Gibbon’s blog:

” Interesting look at some of the instant post-debate reaction polls. Some are weighted to match what the pollsters think is the profile of people who watch the debates.

YouGov have been very open, saying that means they tend to include more prosperous voters, more broadsheet readers, older voters, to be slightly more Conservative and sometimes more male-dominated than the voting population as a whole. ComRes doesn’t weight to viewers but to the voting population profile.

Look at how the ComRes poll sample for last night worked. They polled people who expressed their voting preference as 35 per cent Con, 24 per cent Lab and 36 per cent Lib Dem. And here’s how those same people voted on the instant poll on who performed best in the debate: 35 per cent Cameron, 26 per cent Brown and 33 per cent Clegg. As Michael Howard’s posters used to say in 2005, “Are you thinking what I’m thinking?”

ComRes point out that if you look at their sample for previous debates the numbers do jump about more. People with stated allegiances sometimes cross over and say they think a rival leader did better. But you see the point.”

Indeed we do Gary.