The Really Interesting Election Result

by F.

Forget the substance for a moment and consider that prediction markets called almost every race. This may be the most interesting result of all: that prediction markets went mainstream. Cass Sunstein gives some background:

In general, such markets have been proving to be remarkably reliable. The Iowa Electronic Markets have been forecasting presidential elections for a while now, and they have usually outperformed polls, often “calling” the outcome within one or two percentage points. The Hollywood Stock Exchange does very well in predicting box office success (and Oscar winners, thus taking a lot of the fun out of Oscar night). Markets that attempt to predict economic trends have also been highly accurate.

Businesses too have been using internal prediction markets. At Microsoft and Google, many such markets have been tried, with substantial success. At Google, the results are stunning, as prices have actually been probabilities. When the market says that an event is 80% likely, it comes true 80% of the time, and so too for 70% forecasts, 60%, and so forth.

Of course such markets have their limits. When investors/gamblers lack a lot of dispersed information, the price won’t be aggregating a lot of knowledge. But for elections, at least, the record of prediction markets has been extremely good. (For those interested in details, and some of the limitations and potential uses of such markets, see the discussion in my new book, Infotopia: How Many Minds Produce Knowledge.)

I was watching the Washington Stock Exchange, which doesn’t use real money. Tradesport, however, does and the latest on the Virginia race is here. More background at Slate.