Martin Baxter introduces his new calculator

The former Cambridge and now city financial mathematician, Martin Baxter, has produced a new version of his famous predictor which allows the user to input suggested party shares and at the click of a mouse button produce a projected new House of Commons.

For years this has been a great tool for all who like speculating about what is going to happen and it is great that Martin has got round to producing the new version in such a quick time. His site also contains all the 2005 results and what would happen in each seat with different vote shares.

Just playing arround it’s clear that the extraordinary way that Labour gets so many more seats for a given vote share is still there. On LAB 31: CON 31: LD 31 the calculators suggests that on a uniform national swing it would be LAB 309: CON 209: LD 96 seats

The new boundaries that should be in place by the next election will ameliorate some of this but you can understand why Labour has no desire at all to change the voting system.

Looking at how his calculator performed on May 5th Martin notes that polls had an average Labour overstatement of 2.7%, there was tactical unwind which caused 8 Labour seats to go Tory, and what he terms “Lib Dem re-alignment helped the party to pick up at least six seats not predicted on the national swings.

In summary Martin notes:- Pollster error of about 3% in Labour’s favour caused a prediction error of 32 seats in the majority. Tactical voting and model approximation error caused another 17 seats to be mis-predicted, giving a total error in the predicted majority of 66 seats. When the model is given the correct national support levels, and reasonable estimates of tactical voting parameters, the error drops to a handful of seats. So the model was more accurate than its inputs.

Mike Smithson