But who is right in the battle of the seat predictors?

The latest prediction from Martin Baxter – the ex-Cambridge and now City mathematician who runs the Electoral Calculus site – suggests that the next General Election would produce a Commons with CON 305 (+107): LAB 285 (-71): LD 24 (-38) seats. This would leave the Tories on the current boundaries 19 MPs short of an overall majority.

Martin produces his prediction from a “poll of polls” that he maintains and applies a uniform national swing (UNS) based on the changed party shares in each seat. The prominent Lib Dem MPs featured above, Mark Oaten, Ed Davey, Lynne Featherstone and Chris Huhne would all, on this calculation, be ousted.

There is a debate going on about how you actually calculate the UNS and Anthony Wells of UK Polling Report has just produced a downloadable seat calculator which is not so devastating to the Lib Dems.

Under the Wells formula the same vote shares that Martin uses – C 38.67: LAB 32.46: LD 17.67 – produce a Commons with CON 293: LAB 284: LD 42 seats.

Given that seat projections from given vote shares play a major part for those who play the spreadbetting markets it’s important to know how the two approaches differ. A £100 a seat buy spread bet on Lib Dem would see an £1,800 difference in the profit/loss .

What Anthony does is take the change in party vote shares and apply the same level in each seat. So here the Lib Dems are down five points on May 2005 so Anthony’s calculator is based on him knocking just under five points off the share in each and every seat. So Ed Davey’s 51.05% in Kingston & Surbiton a year ago is cut to 46.07%. Result Lib Dem Hold.

I stand to be corrected here but my understanding is that the Martin Baxter approach is to take the national proportional drop that the five point fall represents and then calculate a new total. With Kingston the LD share becomes 38.67%. Result Lib Dem loss.

From a mathematical perspective Baxter must be right. The proportional cut is what matters not the crude minus 4.98% in every seat.

The main argument against the UNS is that seats perform differently with incumbents, certainly Lib Dem ones, having a good record of holding on against a national trend.

Where I differ from Martin in his latest forecast is the weighting that he gives to each pollster in his “poll of polls” calculation. He does this by taking the number of people in the survey which, in this case, means that YouGov and Mori are weighted at double the level of ICM.

So the Lib Dems being reduced to 24 seats has a lot to do with the heavy emphasis that Mori’s 10% Tory lead has on Martin’s calculation.

The latest Spreadfair Commons seat spreads Commons seats spreads are: CON 282-287: LAB 277-291: LD 54-56.

Mike Smithson