Top civil servant reveals why the country is in such a mess
Ah, it all becomes clear now.
Gus O’Donnell used to be one of Britain’s top civil servants, running the civil service in 2005-11 and right at the heart of the big decisions made by 10 Downing Street.
In an interview with the Daily Telegraph he reveals an, er…, interesting approach to evidence. He thinks broadcasters shouldn’t pay attention to properly conducted opinion polls carried out in line with the rules of the British Polling Council and instead… should use betting odds to make decisions. He said:
The best form of objective evidence is probably the betting odds.
Not exactly a reassuring approach to making use of evidence for decision-making… [Insert your own joke about betting odds, gambling with the economy and casino banking here.]
UPDATE: Two pieces of background evidence to pick up on points readers have raised –
- The betting market for the London Mayor contest is small, which means that the weight of money matters more than the predicted changes of events happening.
- The UK evidence on the accuracy of betting to predict election results is that it is a poor predictor. (Even in the US the evidence is pretty poor, though one of the first – and so most famous – examples was of it being a good predictor.)
UPDATE: For more on the virtues of betting markets versus opinion polls, see my book Polling UnPacked.
Bookies are very bad at it!
And his view on this has absolutely nothing to do with his friend who's running to be London Mayor who is not really registering in the polls?
How many opinion polls predicted Galloway's win in Bradford? The betting market did.
The keys are the "proper conduct" of opinion research and the amount of money laid.
Frank H Little Given the lack of polls in Bradford West, saying they failed to predict the result is true only in the sense that something that doesn't exist can't really make a prediction 🙂 If you look at the broader evidence of how betting odds stack up against election results, the record is poor – see the research I've linked to in the post.