Are Prediction Markets Accurate?

Galton and the Ox

 

In 1906, Francis Galton was researching how breeding affects intelligence. He observed a competition at a cattle show where the crowd predicted the weight of the meat from a slaughtered and dressed ox, the animal was alive at the time. The 800 participant's average guess was 1197lbs, the actual weight was 1198lbs. The crowd's judgement was essentially perfect.

 

Evidence from Public Prediction Markets

 

Public prediction markets to forecast political, sports and commercial outcomes have been around for decades. The Iowa Electronic Market (IEM) was founded in 1988 and enables members of the public to forecast political outcomes. IEM has an average of 23% less error versus more traditional polls when used for forecasting political events.

The Hollywood Stock Exchange (HSX) was started in 1996 and enables the public to forecast Oscar winners and opening box office receipts. The results prove the validity of the model. The 700,000 active traders on the HSX have predicted 92% of the Oscar winners in the last 3 years.

Xpree's academic advisors - Justin Wolfers (Wharton) and Eric Zitzewitz (Dartmouth) - have shown that probabilities revealed in prediction markets accurately represent probabilities of the events with the prices of binary prediction markets being closely related to actual frequencies of event in the real world.

Evidence from Corporate Prediction Markets

Corporate prediction markets have evolved over the past 10 years, and today are used by leading companies such as HP, Intel, Microsoft, Google and Electronic Arts. HP was a pioneer in the field. Working with academics from Caltech to forecast sales of printers in 1996, HP found that the crowd forecast was more accurate than the management forecast in 75% of the trials. Similarly, Intel is using prediction markets to forecast chip demand and has found that the crowd based forecast has 20% less error than the official management forecast. Google is perhaps the furthest in adoption and allows all employees to forecast product launches, office openings and market shares. Electronic Arts is using Xpree prediction markets to forecast sales units, ship dates and quality of games. They have found that forecast error is 50% lower using prediction markets compared to more traditional forms of forecasting.