From the polling in the first week of the official campaign it is hard to draw any conclusions. See Wiki list pic.twitter.com/kjoEbkEFjs
— Mike Smithson (@MSmithsonPB) April 7, 2015
Projecting seats numbers: Bottom up or top down?
The Easter holidays are over. Now it gets serious. Will one of the main parties end the stalemate and draw clear across a range of pollsters?
Hard to say. All the post-debate polling has shown LAB on the same 33%. What has differed is the CON share ranging from 31-34%. Everything is, of course, within normal margins of error.
The big question at the moment is how do we convert GB polling numbers into projections on the total of seats at stake and this is no easy task.
There are two main approaches. The traditional top down method where you take the difference between now and the last election and compute swings between the parties. You then seek to apply them to each seat on the basis that the movement is the same across the country.
With the very different political environment that has emerged in post-indyref Scotland seats north of the border are sensibly being treated separately. But once you’ve done that then you can’t apply the vote changes in the all GB polls without making some adjustment. I’m endeavouring with each new poll to compute from the data what’s happening in England and Wales and then produce a swing figure.
My problem is that I do not know how credible, in terms of projecting seats, the E&W sub-sets are.
The other way of computing seat numbers is to take a bottom up using what data that’s available. We have, of course, the Ashcroft seat polling and, in addition, some of the university forecasters are provided with a unique YouGov stream of data from each seat. Unfortunately that’s going to a few select academics and is not generally available.
What we do know is that “bottom-up” analysis is producing better seats totals for the Tories then the traditional top-down mechanism.