YouGov’s poll in the Sun today has topline figures of CON 41%, LAB 31%, LDEM 18%, so little change in the Conservative lead as we head towards the election.
This poll has a significant methodological change, albeit one which has made hardly any difference to the topline figures. Unlike nearly all of the other pollsters YouGov do not normally take into account likelihood to vote in their topline figures – based I believe on the theory that away from elections polls are snapshots, rather than predictions.
In 2005 in YouGov’s final pre-election poll they did factor in likelihood to vote, which decreased the Labour lead slightly and made YouGov’s final prediction more accurate. At this election YouGov have decided to factor it in from the start of the campaign, so on the assumption that Brown will call the election tomorrow, they are now in election mode and the figures are weighted by likelihood to vote in the same way that Populus do.
This actually makes very little difference to the figures. Without weighting by likelihood to vote the figures would have been CON 40%(+1), LAB 31%(+2), LDEM 18%(-2), so all the likelihood weighting did was push up the Conservatives by one point (which, I should add is pretty typical of the testing we’ve done over the last month. It’s never made more than 1 point difference).
Methodological discussion aside, YouGov’s poll shows Labour rising at the expense of the Liberal Democrats, though obviously the lead has a rather greater contrast! Given the normal doubts about polls over bank holidays, I think the chances are that YouGov or ICM have got a strange bank holiday sample (Opinium don’t have any political weighting, so are likely to be a bit erratic anyway).
With recent polls coming together somewhat around a Conservative lead of around 10 points or just below, it’s probably better to be sceptical of the ICM poll until and unless other pollsters show a similar narrowing of the lead. I expect we’ll have no shortage of polling over the next few days to enlighten us.