InTrade and HubDub got lucky that their 2 mistakes (so to speak, in a non-probabilistic way) on Missouri and Indiana (both with 11 electoral votes) canceled themselves perfectly. IT WAS PURE LUCK. If their 2 mistakes had been made in the same direction (say, betting on Obama with the outcome going eventually to McCain), and/or their 2 mistakes had been done on 2 very dissimilar states (say, one with 6 electoral votes and the other one with 27 electoral votes), then we would have had reporters and bloggers bashing the prediction markets for the whole month of November.
Let’-s not confuse media visibility with utility. Aside from the depressed Obama-to-win prices on one exchange, prediction market and polling aggregation results for the 2008 election were essentially the same using squared errors. Despite his insane schematics, Emile Servan-Schreiber has a good point about capturing the interest of the public, something that nerdy academic and libertarian-types aren’-t necessarily good at. An Obama-backing baseball statistician out of Daily Kos nailed that part this year, a year where people were especially skeptical of markets, not to mention unregulated “-offshore”- ones. Likewise, if you put down the lens of considering markets as commission generators, you’-ll see the value of contracts tied to social and cultural outcomes. Of one the biggest assets of prediction exchanges is media goodwill, which should be fostered by distilling information on subjects like global development and art prices.
Other things to keep in mind:
This year happened to have a lot of favorite-longshot States, which turned-out to be favorable to 538’-s error relative to markets.
Prediction markets register information in real time. Since the difference in error is small, this is important.
Markets are more flexible, and useful in situations where you don’-t have a rich data set and obvious statistical analyses. Elections are just one type of question. Even if you have data, it might be less expensive to set up a new contract than to undertake the analysis.
And of course, prediction markets have functions aside from forecasting, and provide incentives for uncovering new information.
Prediction market analyst Lance Fortnow in an e-mail to me:
Right now the electoral college markets are tracking the polls pretty closely. I think we’-ll see some divergence when we get close to the election since the polls can’-t keep up. In past elections the markets were much better than the polls within a few days before the election (though not on election day itself which has too many rumors).
Other thoughts: – There is a long-shot bias —-states which are above 85% (for one candidate or the other) reflect a probability closer to 100%. – The state markets are strongly correlated. There is a small but non-trivial chance that many states will be way off this year. And then people will be reluctant to trust the electoral college markets in the future.
So, I have (at least) one answer to my series of provocative questions: Electoral college prediction markets are more useful than the state polls towards the very end of the presidential campaign (but not on Election Day). Interesting. Thanks.
PS: The discussion about this post goes on in the comment area of another post.