With the US presidential election race coming to an end tomorrow (we hope!), it is time to take one last one last look at the RealClearPolitics (RCP) poll average and the Iowa Electronic Markets (IEM) winner-take-all contracts and see where things stand.
One thing to point out about the IEM that we mentioned in the initial post – the contracts represent a probability, as does the much-cited FiveThirtyEight model that we’ll look at later in this post. That means that as it stands in the IEM, Obama is the favourite but far from a sure thing.
The RCP poll average appears neck and neck:
Obama closed the slight lead that Romney had opened up, but with four lead changes in October, hard to really favour either candidate. The IEM victory probabilities are nowhere near as close:
With time running out for Romney to make up the necessary ground in the battleground states, the IEM market has completely decoupled from national polls.
Here’s an overlaid look:
The mid-October charge by Romney can be seen in the narrowing of the RCP poll average (which comparatively barely budges) and the shifts in the IEM and FiveThirtyEight victory probabilities.
But as the polls began to level out around mid-October, so did Romney’s deficits in the important swing states. And the with dwindling time to make up that ground, the IEM and FiveThirtyEight both began to steadily climb for Obama.
This speaks to the usefulness of the IEM, Intrade, and other futures markets.
The IEM currently projects about a 60 per cent likelihood of victory for President Barack Obama. If you were to go back and look at every time this market projected a 60/40 favourite and that favourite won every time, the market wouldn’t be very accurate. In fact it would be wrong nearly half the time. For the market to be pricing accurately the 60/40 favorite should win six out of ten times, not ten out of ten. Obama trading at reasonably better odds than a coin flip.
This election has also shown that some of the simplest numbers can be the most misleading. We used the RCP average of national polls because it’s a common, popular indicator, but it is, in fact, not a particularly good number to tell us each candidate’s chances of victory.
The structure of the electoral college ensures that national polls can only be so accurate. Nowhere is that more evident than on FiveThirtyEight, Nate Silver’s blog on the New York Times website – where he currently has Obama projected as an 86.3 per cent favorite thanks to his leads in various swing states. Silver made his name with a model that accurately predicted the 2008 US election. But as has been pointed out many times recently, Silver’s model also faired very poorly in the 2010 UK elections.
With Silver’s model such a stark different to the RCP average, the Iowa market serves as a good hedge between the two – favoring Obama in a race that should come down to the wire.