h1

Introducing the “Golden Rule” seat calculations

May 5th, 2008

ec-pred-060508.JPG

ukp-pred-060507.JPG

One thing that Friday’s London mayoral result showed was that I term “The Golden Rule of British Polling” still applied. For based on the results of the last four general elections and all three London Mayoral races the most accurate poll has always been the one showing Labour (Ken in 2000) in the least favourable position in relation to the Tories.

The rule was the key factor that determined my betting in last week’s election and in the end I had considerably more than a month’s after-tax income at risk. But I never had any real doubts that it would come good. If we had taken an average of the polls we would not have been right.

So taking the latest output at any one time of the five major firms carrying out general election voting intention polls – ComRes, ICM, Ipsos-MORI, Populus and YouGov – the numbers from the poll that satisfies the rule will be put into the two major online seat predictors and be published here.

The calculators, from Martin Baxter’s Electoral Calculus site and Anthony Wells’s UKPollingReport have a different mathematical approach. There has been lots of discussion about the way these two work and others have also developed their own models which we might include in the future. They are, of course, just a guide and are less robust than the Golden Rule. I think that they both understate the Lib Dem position.

The central thing here is the input data which will be restricted to the poll showing Labour in the least favourable position.

This is controversial but the Golden Rule has been tested against real results over a long time period and until it is proved wrong it seems the best approach to take.

Mike Smithson






Comments are closed.