The Second Workshop on Prediction Markets

Yiling Chen June 30th, 2007

I spent some time in San Diego earlier this month for the The Second Workshop on Prediction Markets in conjunction with EC’07, which is part of FCRC 2007. The workshop was very successful. The participants were as diverse as one always expects, coming from different fields: economics, business, computer science, law, medicine, and statistics. Needless to mention, it also attracted many industrial practitioners.

The more academic part of the workshop was the presentations of 10 high-quality research papers, ranging from theory, experiments, and evidence. 4 papers directly studied the market scoring rule mechanism. (Congratulations, Robin! Your mechanism is very popular.) Since papers are available at the workshop’s website, I’ll save my words for other parts.

The industrial panel brought our attention from research to practice. Russell Anderson shared HedgeStreet’s experience at dealing with legal hurdles of using real money. (Anyone thinking of running real money public prediction markets is strongly recommended to hear his advice.) Matthew Fogarty talked about his experience of running a corporate prediction market for Electronic Arts, a large video game publisher. The prediction market is used to predict ship dates, quality of video games, and etc. Unlike many corporate prediction markets that are often thin, the prediction market at Electronic Arts is blessed to have more than 200 traders. David Perry presented Consensus Point’s experiences with running prediction markets for companies, such as GE and Bestbuy. He also pointed out that Consensus Point is willing to share some data for research use, which is great news for researchers like me. Emile Servan-Shreiber, in stead of talking about prediction markets, talked competitive forecasting, another forecasting platform provided by NewsFutures. In short, competitive forecasting directly elicits forecasts and creates competition among forecasters. Participants are asked to give a range of forecast; correct forecasts with narrow ranges are rewarded; early forecasts are rewarded (introducing competition). Competitive forecasting may be simpler to use, compared with prediction markets. But as Robin Hanson pointed out in a blog post, such forecasting method may have some biases. I look forward to seeing how it works.

In answering the question “what should we study?”, the panel directed several questions for researchers: (1) Understand why people like trading and create psychological profiles for traders; (2) How to make markets as simple to use as possible; (3) Study the organizational behavior related to prediction markets; and of course (4) Design better mechanisms to provide the right incentives.

The workshop also featured a mini panel on political elections. Lance Fortnow and Eric Zitzewitz each shared their work related to prediction markets and political elections. Both of them made the point that we should educate people on what prices in prediction markets mean. Since market prices are probabilities, people should be prepared to see that an event does not happen even when the market gives a high probability.

The field is healthily growing. I can’t wait to attend the Third Workshop on Prediction Markets.