Ashcroft marginals poll looks good for Labour, and possibly not just in the headline

Lord Ashcroft's latest round of polling in 12 English bellwether seats shows a swing of 4.4% from Conservative to Labour, with 9 of the 12 seats turning red. On a perfectly uniform swing of the same size, all 12 would turn red (the swings required range between 2.4% to 3.6%) but variation in swing is to be expected.

In fact variety is quite a theme in this poll with swings of just 1.5% in Kingswood and in Blackpool North & Cleveleys, but 7% in Bury North. Remember though, that even though the sample as a whole is huge (12,000), each constituency on its own is just a sample of a thousand, so all of the swings are within the 3.1 point margin of error of the 4.4% average. Thus some of the seat-to-seat variation could just be sampling.

What is interesting this time is that the 4.4% swing is exactly equal to the swing in Lord Ashcroft’s national polling over the same period. Contrary to popular belief, Labour has typically done worse in the marginals than nationally in the recent past. The reason for this misconception is likely to be that constituency polling tends to have slightly older fieldwork at the time of publication, and since Labour’s lead has steadily been narrowing, when the polls have been published they have tended to show bigger swings than national polling with more recent fieldwork. If you compare them to polls by the same pollster at the same time, the swings in the marginals have mostly been smaller (and none by Ashcroft recently have been larger):

The differential between the two swings is now zero (the same swing), the very low end of its recent range. Over the past few months it has varied between equal swing and a point less swing, and has averaged about half a point less swing (ie one point net). Is the swing to Labour more resilient in the marginals than nationally?

If the it really has gone, that would be very good news for Labour indeed – each point of swing in these sorts of constituencies is typically worth an extra 15 to 20 seats. But it’s too soon to be sure of that, it could just be sampling error or that these 12 seats aren’t typical of the battleground as a whole. We would need more evidence to conclude that.

Source for the Ashcroft data is here and the sources for the ComRes data are here and here.

About The Author

Related Posts