How Good Are Intrade Contracts As Predictors of Senate Election Results? A look at 2006

The Intrade futures contracts on the outcomes of political events are one of the available metrics that can be used as predictors of elections. Intrade is a real money open market that has been in operation for several years. With respect to the US Senate, they offer contracts on all of the races and on several “derivatives”, such as Senate control after elections, the number of seats that a Political Party will control, and currently, the appointments that will follow recent or projected resignations. The contracts for the races are not based on the individual candidates, but rather on the Political Party that would prevail.

The use of real money futures markets began, as far as I know, with research at the University of Iowa – the Iowa Electronic Markets http://www.biz.uiowa.edu/iem/markets/. Beginning in 1988, their goal was to investigate the relative accuracy of predictions based on such markets as opposed to those based on polling data. Without going into details, there seems to be some evidence that futures markets provide very slightly better predictive ability. The Intrade market is completely independent of the University of Iowa market – it is a private entity.

I have monitored the Intrade contracts for US Senate races off and on since early 2006. How well did these contracts “predict” the outcomes of the US Senate races in the last two elections? This is the first of two posts in which I will examine the outcomes. Here I will examine the 2006 election cycle. Very shortly, I’ll look at last year’s results.

In 2006, the last time I obtained the numbers was on 2 November, five days before Election Day. The contract for Senate control was still firmly pointing to the Republican Party. However, in each of the individual races, the eventual winner was successfully predicted by the Intrade contracts. To refresh your memory, the 2006 election saw the Democrats pick up six Senate seats: Bob Casey, Jr. defeated incumbent Rick Santorum in Pennsylvania, Sherrod Brown defeated incumbent Mike DeWine in Ohio, Sheldon Whitehouse defeated incumbent Lincoln Chafee in Rhode Island, Jon Tester defeated incumbent Conrad Burns in Montana, Jim Webb defeated incumbent George Allen in Virginia, and Claire McCaskill defeated incumbent Jim Talent in Missouri.

I monitored the Intrade contracts two to three times per month starting in February 2006. At that point, these data were already indicating victories in Montana and Pennsylvania. The other four races gradually went in favor of the eventual winners. That took place in mid-April for Ohio, mid-July for Rhode Island, but not until November for Missouri and Virginia.

It may seem odd that all of the individual races were “predicted” correctly by the Intrade contract data, and the Democratic Party did take control, while the contract for “Senate control” still strongly indicated the Republican Party. For those well versed with statistical inference, this is not at all surprising. The best way to verbalize this is that while each of these six events (Democratic pick-ups) were favored, some were only very slightly favored, and the consensus was that it was unlikely that all six would go in favor of the Democratic candidates.