This is a version of a column published in the Irish Examiner 3 November 2012.
Over the last number of months the opinion polls have swung this way and that in the US presidential election, from an initial Obama high to the present situation where, by most reasonable polls, a virtual tie in the popular vote is expected. The peculiarities of the US electoral system can and occasionally do mean that even winning the popular vote is not enough to win the election, and it is also generally seen that Obama retains an edge there. As of Thursday afternoon the RealClearPolitics poll of polls has the race tied.
Opinion polls are fickle creatures – they are always interpretable with a margin of error that can range from 2-5%; they are dependent on the sample being a representative sample of the likely population, and in an electorate which is as fractured as the US that is not easily guaranteed; they rely on people saying what is truly in their mind and not what they think the pollster wants to hear; and they are dependent on statistical adjustments to make the poll as close to the ideals above as the particular statistician in charge deems appropriate. They are therefore, and accepted to be, at best indicative. What is nice however is that we can utilize a combination of the wisdom of crowds and people putting their money where their mouth is to get a clearer picture. And that picture, if it is as accurate has it has been historically, tells us that Obama will be returned as president.
Keynes had it right when he noted that the market was in some ways akin to a beauty contest in a Sunday newspaper – the object is not to pick the face one finds the most beautiful, but the face one thinks other people will on average find most beautiful. Using this approach we can look at real money markets, both betting and futures markets, on the us election. Lets look at two, one based in the swing state of Iowa and the other in Lucan…Both have startling accuracy, historically. The Iowa Electronic Market (IEM) deals in political futures, while Intrade is more akin to a traditional betting exchange. We should note that money or reward based formally structured prediction markets have been and are being used in large corporates for internal r&d and marketing decisions as well as other uses. One of the more ambitious was to use these markets for economic forecasting, the Economic Derivatives market set up by Goldman in 2002 which closed in early 2007. Its closure was perhaps ironic as these markets thrive on uncertainty and what was to be the great uncertainty was just around the corner…
There are good reasons why such markets would be expected to have at least a decent chance of giving as good a prediction to an event as a poll. Markets exist not just to allocate resources but also to aggregate information. They do this well where a number of conditions can be met: a definite action to be predicted; a clear reward or cost for accuracy or otherwise: liquidity; and a fair trading mechanism. Like al markets they can be subject to attempts at manipulation, but such efforts merely add liquidity. Moreover, polls not only have a significant margin of error – the margin of error itself is not a good indicator of how accurate the poll may be. The problem lies in that an opinion poll is a static snapshot of voter’s intentions at that time. As such it tells little about how the dynamics will change over time. By contrast a futures or prediction market explicitly forces traders to consider this very issue, in a dispassionate manner.
Technically, the IEM is a futures market. Modern finance theory has achieved a deep understanding about the nature of prices in a futures market. One remarkable finding is that the price of a commodity on a futures market is the best predictor of the actual eventual price. The prediction of the market, especially as the date of the election looms, is in fact substantially closer to the actual outcome than opinion polls and historically lies within 1-2per cent of the actual outcome. In fact over the 5 US elections since 1988 the IEM outperformed the polls, in terms of being closer to the eventual result, 75% of the time. These findings have been confirmed in other similar markets including Taiwan.
A further example of prediction markets is betting exchanges. In terms of political markets, one of if not the largest and most influential is the Irish based Intrade. Other markets such as Betfair and Paddypower offer odds on elections but Intrade is perhaps the largest. These are liquid markets: as of Thursday there are over 2m trades on intrade and over £12m bet on Betfair. As these are essentially the same asset, the wnner of the election, the beauty of the market is that the prices, odds of winning, cannot diverge too far from each other. And the results are consistent: Obama to win. On Intrade his odds (Thursday afternoon) are 65% and have not been below 50% all year; on Betfair the odds are closer to 70%, while on paddypower.com they are similar. In fact, the historical evidence is that real money markets, such as bookies odds in the pre 1936 us elections (gallup polling began that year) indicates that they were astonishing accurate.
However, the remarkable thing about these is that they appears to have good predictive power not just immediately prior to the poll, but in the months prior. Recent research has indicated that the predictive power of these markets is greater than opinion polls held three-four months in advance. And these suggest, as does the data now, that Obama will win. Plus, the statistical models across the ideological divide agree. So, thats my prediction. Obama to win….