New BPIX poll

As well as the ICM and YouGov polls I reported last night, there is also a BPIX poll in the Mail on Sunday. The topline figures with changes from January are CON 36%(-3), LAB 34%(+4), LDEM 18%(nc). The poll was of 5655 people, so large enough to offer cross breaks of marginal seats. In the 75 Labour held seats where the Conservatives need a swing of less than 5% the shares of the vote are CON 40%, LAB 33%. This represents a swing of 6%, so far larger than the 2.5% swing on the headline figures. When the marginal cross-break is expanded to include the 110 seats where the Conservatives need a swing of up to 7.5%, the totals are CON 37%, LAB 36%, representing a swing of 4%.

The Mail on Sunday's quote from Paul Whiteley about what the swings would mean is very odd indeed. We don't have any tables to examine the data properly, but from the figures in the Mail on Sunday it doesn't work out at all. On a swing of 2.5% as represented by the topline figures in this poll, the Conservatives would gain 33 of these Labour held target seats. If they recieved a 8 point swing in the 75 more marginal seats as suggested by the figures in the article, they would gain all of them, so would gain 42 extra seats rather than the 9 claimed in the article.

UPDATE: Mea culpa, that swing in the closest 75 seats is actually 6 points, not 8. It would still be enough for the Conservatives to will all the seats in that group (a 5 point swing is needed to win a seat with a 10% majority).