“Put your money where your mouth is.”

The old taunt reflects a deep economic principle: talk is cheap, but if someone is willing to risk money, it means they’re serious. Put the principle into action and you realise that electoral forecasters should pay as much attention to the betting odds as to the opinion polls. That is because when money is on the line, informed people, perhaps including insiders, have an incentive to turn their knowledge into cash by making big bets. In the process they make the odds more accurate. And of course, there are several reasons to lie to pollsters but no reasons to make a money-losing bet.

Or are there? Imagine you’re a candidate in the US presidential primaries. One of the big selling points is electability: “If you choose me, I’ll beat the other party’s candidate.” It might be worth placing a few money-losing bets if they made you look electable.

This may be more than a hypothetical scenario. The economists Justin Wolfers and Eric Zitzewitz have pointed out that Hillary Clinton’s chances of becoming president, as predicted by one betting market, InTrade, started to climb dramatically mid-May, topping 40 per cent after months of fluctuating between 20 and 30 per cent. Her odds of winning the Democratic nomination stayed around 50 per cent, implying that if nominated, her chance of then winning the presidency would be about 80 per cent. You can’t get much more electable than that.

So is someone in Hillary’s camp trying to boost her chances by manipulating the market into a self-fulfilling prophecy? Or is it a rival candidate trying to make her look like a manipulator?

Or neither? It’s almost impossible to prove that manipulation is taking place, although if it’s clumsy, people tend to notice. In October 2004 one betting market, TradeSports, saw George Bush’s probability of re-election slump from 54 per cent to 10 per cent in three minutes when bets were placed in the small hours of the morning. Either someone was drunk, or a political hack made a crass attempt to change the odds.

But it’s hard to manipulate markets for long. In the 2004 case, Bush’s re-election probability climbed back to 54 within minutes as profit-seeking traders spotted the free money. “When I heard about it, I started checking the betting markets in the early hours a lot more closely,” says Justin Wolfers. Robin Hanson and Ryan Oprea, economists at George Mason University, argue that manipulators can actually increase the accuracy of prediction in markets: by making big, unprofitable bets they are effectively subsidising the market and paying other traders to pay attention.

Economic historian Paul Rhode and economist Koleman Strumpf point out that the speculative attack on Bush - if that is what it was - required an investment of about $20,000, small change by the standards of a modern presidential election campaign.

But if history is any guide, manipulation will not be so cheap in future. Rhode and Strumpf report that political betting markets were once far larger. In 1916, over $160m in today’s money was wagered on the contest between Woodrow Wilson and Charles Evans Hughes. Now that political betting is back in vogue, volumes are rapidly growing. Even if future hacks do see some advantage in trying to skew the odds of future elections, it may cost them dearly to try.

Tim Harford’s book ‘The Undercover Economist’ (Little, Brown) is out in paperback