Should you bet on the poll that’s worst for Ken?

Should you bet on the poll that’s worst for Ken?


    Will the “golden rule” apply again next week?

So far four polling firms have produced surveys on the 2008 London Mayoral race and it looks as though each will be doing at least one further survey for publication in the next week.

What’s great about this contest is that it’s about aggregate vote totals across London and we’ll be able to compare directly the performance of each polling firm with the actual result. In this election there is no hiding for YouGov, Ipsos-MORI, ICM and MRUK and their respective performances will play a key part in shaping attitudes to each of them ahead of the next general election. This is one they will be doing everything to get right.

People keep on emailing me to ask which pollster I will “be betting one”? The answer is that I’ll probably go with the one that most fits the polling “form-book”.

For based on performances at the last four general elections and in the mayoral races of 2000 and 2004 the golden rule is to choose the survey that is showing Ken in the least favourable position. For in all of those battles the pollster that was most right about Labour/Ken versus the Tories was the one where the former was predicted to be doing worse.

Just look at the record. In 2005 NOP had the lowest lead for Labour and came out best. In 2001 it was the same as in 1997 and 1992.

Click on the year links to UKPollingReport to see what happened. The same occurred in the last two mayoral races and at the last Euro elections in 2004. The pollster showing Labour/Ken in the most unfavourable position against the Tories turned out to be the least inaccurate.

It is this record of consistent Labour/Ken overstatement that reinforces my opposition to poll averaging. In each case the figures least favourable to Labour/Ken were a much better guide than the average.

The mayoral betting has continued to be stable with Boris at about 0.6/1 and Ken on 2.6/1 1.6/1.

Mike Smithson

Comments are closed.