A new poll by ComRes in Tuesday’s Indy tells a completely different story to ICM and MORI’s recent figures. The topline figures, with changes from ComRes’s last poll, are CON 37%(-6), LAB 36%(+4), LDEM 17%(+5). The poll was conduced between Friday and Sunday.
In recent weeks we’ve become used to polls showing differing results – methodological differences have resulted in some pollsters showing much larger Tory leads than other ones. However, until now polls have been consistent in showing the same trends – until last week the pollsters, despite differences in extent, were all showing the Conservative poll shrinking. The two polls published since the press reaction to the PBR, conducted by ICM and Ipsos MORI, both showed that trend reversing, with the Conservatives again gaining support. This poll is not just out of line with the figures other companies have been presenting, but also it’s the exact opposite of the trend ICM and MORI are showing.
Normally when polls disagree I look at the methodologies and try to explain the difference. There is no obvious explanation here. I’ve looked through the ComRes tables and there is nothing obviously freaky or wrong – the Lib Dems appear to have been weighted to a much higher figure than their previous poll – 12% of the sample said they voted Lib Dem in 2005, as opposed to 8% of the sample in ComRes’ previous poll – which goes a long way towards explaining the jump in the level of Lib Dem support, but not that of the two main parties: the overall shares of recalled past vote are pretty similar to those ICM use to weight their polls.
This is the first poll conducted since Damian Green was arrested, so theoretically it could be possible that the public reaction to it has been the polar opposite to the media’s, but that would be unusual. It could also be that the the increased Conservative lead we saw from ICM and MORI was just a reflection of the bad press coverage the PBR was receiving at the time those two polls were carried out, and the public reaction to the PBR now it has sunk in properly is actually more positive. The alternative possiblity is, of course, that either MORI & ICM, or ComRes are “rogue polls”.
I am cautious about the term “rogue poll”. It tends to get thrown around willy-nilly against any poll people don’t like and sometimes at companies with methodology people see as faulty. What is actually refers to is that the 3% margin of error commonly quoted for polls is at the 95% level of confidence. In layman’s terms it means that 19 out of 20 times the “real” figure will be within 3 points of the figure quoted in the poll. A rogue poll is that 1 in 20 where the figure is more than 3 points out. It is inevitable that these things happen, and happen to all companies – but realistically it is impossible to ever be certain whether a poll that looks out of line is a rogue poll, or the start of a new trend. We won’t know for sure until we see some other polls that confirm or contradict this one, but until then I will urge my normal caution against polls that show large changes in support, or sudden reverses in the trend.