So, one of the most entertaining books I’ve read of late has been James Surowiecki’s The Wisdom of Crowds. Surowiecki examines ways in which people can tap into the collective knowledge of everyone. Some of the cool topics include prediction aggregations, how quickly markets are at getting to the right answer given individuals incomplete information, how hive type interactions occur (siting some fascinating studies on the foot traffic in NYC).
By far though my favorite topic is the emerging field of non-good based futures markets, known as decision markets or prediction markets. The most famous of which is the Iowa Electronic market.
The Iowa Electronic market works like this… Two separate propositions are put up as to whether the next president will be a democrat or a republican. If on the day the election is complete, the democratic candidate has more popular votes than the republican candidate then each share of the democratic proposition held is worth $1 and each share of the republican prop is worth $0, if the republican candidate has more popular votes then each share of the democratic proposition is worth $0 and each share of the republican proposition is worth $1. The chart below is for both of the propositions in the 2008 election (blue is the democrat, and red is the republican).
So what? … Well, it turns out that the election market is better at predicting than polls. This has some very interesting implications. It means that markets are better at incorporating a group of individuals who each have incomplete information, than polling that same group about their intent. Now other markets have sprung up that do pretty much the same thing for winners of sporting events, Hollywood box office, and all kind of other things in the public domain.
One of the really interesting surprises in this field is that play money prediction markets have no appreciable difference with real money prediction markets in terms of the validity of their results. There was a study comparing TradeSports (real money market) and NewsFutures (play money market) for NFL games predictions, and there was no better set of predictions using the TradeSports results than the NewsFutures results.
This has large implications in how feasible and useful these prediction markets are. A company can setup a play money market for predictions of certain events internal or in their market at large, and with the correct incentives for traders to keep liquidity (free iPods for example), and then tap into the results of the market. If one were to put a ticker of not just the stock of the company on the employee portal, but the stock of whether pending legislation will pass which would significantly impact the business, there would be a new wealth of information generated.
One of the next pieces of interest for me who be to compare the market predictive elements with analytics tools. My gut tells me that the market’s invisible hand will far outstrip the analytics model (communism vs. capitalism).