At 4 days past the election day, we know who the next President will be and we know the results from every state bar one. Missouri has yet to deliver a final answer with a total of 1,442,673 votes for Obama and 1,436,814 votes for McCain. However, since this is a difference of 5,859 votes in favor of Obama, we will assume the state is a blue state at this point in time.

Unfortunately, by looking at the data from our last post, we can see that not only were the prediction markets correct, but the polls were correct as well. So, while this shows the legitimacy of prediction markets, it does not do much to show if they are superior or not to polls. We can see that this year pollsters were very careful with their results. We were, however, able to extrapolate some very interesting conclusions about the mechanics of prediction markets.

First of all, let’s look at Intrade.com, whose results matched the polling results perfectly.

Secondly, let’s look at Betfair.com, whose results also matched the polls with one exception: Missouri. Missouri had a T, which meant “Too close to call”. This is clearly reflected in the fact that it is the only state still currently counting votes. Therefore, Betfair could be said to show even more precision.

Thirdly, let’s look at the Predictify’s Election Showdown results. This only allowed betting for certain states. However, it also works in a different way. Bettors do not have to place their own money. All of the money is put up by the people asking the question who are ultimately interested in the results. This brings many more predictors, but with less personal stake.

Here we can see that North Carolina and Missouri were predicted incorrectly, while Ohio and Virginia were deemed ‘too close to call.’ These were obvious mistakes, which are interesting to us because they show the difference between having a personal stake and not having one.

James Surowiecki, in The Wisdom of Crowds, stated that a monetary interest does not have to be at stake. From what we have seen here, it seems the opposite is true. Not only will the results be more precise if participants have an interest in obtaining a reward (financial or other), but if they stand to lose something for getting the wrong answer, you will gain even more precision.

Therefore, when people have to bet with their own money, as in with Intrade and Betfair, the results will be more precise than when they stand only to gain and not to lose.