Last week's Prediction Markets Summit demonstrated how far prediction markets have come, while also highlighting significant hurdles that have yet to be addressed.
Although some 40,000 markets have been established since 2000 and at least one market application provider is predicting 500% year over year growth, the summit still had the intimate, yet cautious feel of a high school reunion. And while prediction or information markets have shown to have very real, practical applications – companies such as Corning, Intel, Siemens, Hewlett Packard, Eli Lilly and Google have embraced the use of internal markets to guide corporate strategy – widespread acceptance and adoption of these markets will elude the industry until major legal, public relations and design/usability battles are won. That the conference was an uneven mix of research symposium, sales pitch and trade association meeting says a lot about the still evolving state of affairs within the prediction market industry.
The highlight of the day was the presentation by James Surowiecki, author of The Wisdom of Crowds and something of a celebrity in the prediction markets community. Surowiecki helped popularize the notion that randomly selected groups of amateurs can provide more accurate insights than rigorously trained panels of “experts” – in fields as wide ranging as medicine, science and policy analysis. Even more provocative, Surowiecki has found that so-called expert judgments are poorly calibrated (e.g. there is little correlation between an expert’s confidence in his or her judgment and the accuracy of that judgment). These ideas provide the theoretical underpinnings of all prediction markets, which aim to aggregate information from diverse sources in a speculative, betting environment to provide more accurate forecasting data than traditional market research.
Some key insights from the summit:
· Prediction markets aren’t new. Although many people believe the Iowa Electronic Markets, launched in 1988, was the first the prediction market, Professor Koleman Stumpf’s explained that prediction markets – and efforts to manipulate the results of those markets – date back to the1800s. Stumpf used New York Times archives to show how attempts to manipulate the presidential election markets in 1880 and 2004 were equally ineffective.
· While the basic premise of prediction markets isn’t new, recent advances in computing and communication technologies have made widespread application much more feasible. The best prediction markets cast a wide net to recruit a large and diverse group of participants to achieve market liquidity – and today’s cheap, ubiquitous internet access makes it easy for virtually anyone to voice their opinion on a given subject, without the traditional constraints of time, space or money.
· Prediction markets also provide a non-combative forum in which make a point without bluster and emotion. Charles Polk, in his discussion of the use markets to predict pandemic flu outbreaks, suggested that prediction markets work best if participant anonymity is maintained, so individuals don’t need to reveal how often they participate and how strong their opinions are.
· However, anti-gambling regulations and other legal hurdles may delay more widespread adoption of both internal and external prediction markets, especially in the US. To side step this challenge some markets are experimenting with the use of fake “money” and stating explicitly that participants won’t win any real money (even though the fake currency can be converted - wink, wink - to real currency at the conclusion of the market).
· And not surprisingly, usability matters – a point stressed by Bo Cowgill at Google (which hosts 55 separate internal markets a quarter, soliciting participation from nearly a third of all firm employees). Usability here includes both web design concerns (Google has experimented with ‘skins’ to make their internal markets more user-friendly) and fundamental market design considerations: simple 0-100 contracts (see Tradesports) are less complicated and intimidating for many people then the continuous double auction bid/ask model. We also heard from a number of speakers that people won’t participate in a prediction market if the information generated by that market isn’t relevant to them; the most successful markets will therefore proactively parse and summarize market information to reward individual participation.
The summit concluded with an interesting paradox: although prediction markets, wikis, and other collective intelligence tools are becoming increasingly commonplace, we nonetheless live in the age of "superstar" CEOs, many of whom are inclined to manage from the gut. If prediction markets are to wield greater influence on resource allocation, marketing, even public policy decisions, then we need to do a better job helping CEOs ask two fundamental questions: is there anyone - customers, vendors, employees - who may know more about my business than I do? And, how do I build a market where they can profit from that knowledge?