Friday, April 16, 2010

Sorry, the new ComRes poll is still worthless

I don’t mean the limited poll of debate viewers, which recorded an implausibly massive swing to the Lib Dems. Obviously that’s worthless outside its own very narrow context. I mean the national figures that they’ve produced this lunchtime: Con 35% (unchanged from Wednesday), Lab 28% (down 1), Lid Dem 24% (up 3).

These numbers look a lot more credible – you’d expect a Lib Dem boost after Clegg’s performance – but they’re still not valid.

The methodology (via John Rentoul) was not that of a standard poll. ComRes did not survey non-viewers of the debate. Rather, they took their poll of 4000 pre-recruited debate viewers and extrapolated it to all 10 million viewers, and then combined this with the data from their normal poll from Wednesday to cover people who didn’t watch the debate.

There are two problems here: it doesn’t allow for the possibility of non-viewers being influenced by coverage of the debate, or indeed by anything else since the previous poll; and it assumes that all viewers of the debate will have responded in the same way as the small minority who watched it knowing that they were due to receive a call from a pollster afterwards.

ComRes appear to have been completely honest and open about their methods here, but what this means is that we can’t compare these numbers with other polls. We’ll have to wait a little longer to find out the impact of the debate.

No comments: