!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> Streamline Training & Documentation: How much participation does a prediction market need?

Sunday, June 14, 2009

How much participation does a prediction market need?

In April 2008 McKinsey published an edited and abridged transcript (pdf) of a roundtable on prediction markets moderated by Renée Dye, a consultant in McKinsey's Atlanta office. The participants were four experts on prediction markets:The discussion is quite substantive and well worth reading in its entirety (eleven pages). Here I will call attention to a helpful sidebar Dye provides in which she summarizes the decisions you have to make in setting up a prediction market at your own organization.

Dye cites six key decisions:
  • How to define the variable the market will forecast. "Express [the variable] in a precise, intuitive unit (such as '2nd-quarter revenue, in euros, for new product X') to avoid confusion among participants."

  • With whom to share the results. Dye points out that results can be embarrassing to management (e.g., a prediction that a product under development will fail in the market). Results can also raise legal issues (e.g., a prediction that future financial results will show that the company's current stock price is too high).

  • Who should participate. "Markets involving only internal participants are easiest to organize." On the other hand, if you include appropriate external participants, you will generally increase the accuracy of the results.

    Dye also points out, "Front-line employees often are the most active and excited participants."

  • The nature of the market. "Markets with real-time buying and selling of contracts yield rich, continuous results but require large numbers of participants, some of whom may need training."

    "Simple surveys and other single-point forecast mechanisms are easier to administer. Companies getting started may want to proceed gradually through a series of increasingly sophisticated experiments."

  • Incentives. Cash can present legal issues, since it can make your prediction market look suspiciously like a gambling operation. Fake money can be a workable substitute. Modest prizes, such as t-shirts, have worked well at companies like Google (pdf).

    See here for Adam Siegel's suggestion of using access to top management as an incentive. (Siegel is a co-founder of Inkling, a company in Chicago that provides a prediction market platform that anyone can use to set up a prediction market. See this earlier post.)

  • The role of experts. "Departments dedicated to forecasting [e.g., marketing] will see the establishment of a prediction market as a threat."

    Dye argues that shifting the mindset of experts concerning their role is important. Rather than being the people "with all the answers," experts should view their role as formulating the right questions and helping with "analyzing the answers [yielded by the prediction market] in creative ways and using them to guide decision making.".
BTW, if you want to track academic work in the field of prediction markets, one source you can use is the Journal of Prediction Markets, published since 2007 by the University of Buckingham Press.