Betting markets are pricing in only a 24 per cent chance of Donald Trump winning next week's presidential election. Anyone taking comfort from this, however, has a problem: betting markets can be horribly wrong.
We needn't look far for evidence of this of course. In the run-up to 23 June, markets told us there was only around a one-in-four chance of us voting to leave the EU - spookily close to the odds now on Mr Trump winning. We know how that turned out.
How can markets be so wrong? Daniella Acker at the University of Bristol has one answer. The problem, she says, is that gamblers don't differ only in their beliefs. Instead differences in wealth and attitudes to risk can distort prices.
For example, during the EU referendum campaign the average bet placed on Remain at Ladbrokes was £450 - six times as big as the average bet placed on Leave. Although most gamblers backed leave, the weight of money skewed the odds towards Remain.
In this sense, betting markets are very different from the experiment that inspired the notion of the wisdom of crowds. Back in 1906 the statistician Francis Galton visited a fat stock show in Plymouth at which people paid 6d to guess the weight of an ox. Galton found that the median guess (of 1,207lb) was within 1 per cent of the actual weight (of 1,198lb): the mean guess was even closer. This, he said, was "more creditable to the trustworthiness of a democratic judgment than might have been expected". But what if the size of bets had varied, with rich but irrational or ill-informed people betting more? Then the average guess, adjusted for the size of bet, might have been wildly wrong.
This isn't so improbable. Ms Acker points out that two-thirds of those who bet on the 2012 presidential election with Intrade actually made money, just as most people who bet on the UK referendum result did so. In both cases, this was because a minority of big gamblers lost a lot. Even in a slightly negative-sum game, most people won.
These two facts are consistent with Galton's finding - that the median bet is a good one. Market prices, however, don't tell us the median bet, but can instead be distorted by weight of money.
This does not mean that betting markets are always or even often wrong; the weight of money often splits evenly so doesn't distort prices. In one study Erik Snowberg, Justin Wolfers and Eric Zitzewitz concluded that prediction markets "are largely efficient": they did, after all, correctly back Barack Obama in 2008 and 2012.
'Largely', however, is not the same as 'always' - a fact that of course applies to other financial markets.
There are few if any law-like generalisations in the social sciences. Claims such as 'markets are efficient' are true only under particular conditions. And these conditions can be absent.
MORE FROM CHRIS DILLOW...
Chris blogs at http://stumblingandmumbling.typepad.com