single IPO prediction market. Third, we argue that Google’s unique goals and mechanism make it an
interesting IPO to study in its own right. Google provides a natural and conservative benchmark for
evaluating the efficiency of IPO prediction markets because of their stated intentions of avoiding large
post-IPO price changes and their auction mechanism.
Some issues cause difficulties for prediction markets. For example, expected value pricing in
such markets depends on traders not using the markets for hedging purposes. Significant hedging
demand could drive prices away from the fundamental values that markets are trying to forecast. But this
simply means that a model of hedging demand needs to be grafted onto market prices to reveal true
probabilities. Other mechanisms also have the potential for improving IPO prices. But, these
mechanisms have limitations, too.34 Given the importance of IPOs, the stakes involved in pricing them
and the lack of agreement on the theoretical reasons for underpricing, we suggest that all potential means
of forecasting IPO values, setting optimal IPO prices and evaluating theory, including prediction markets,
are worthy of further study.
As discussed above “when-issued” markets for IPO stock are run in other countries including Germany. However, as currently structured, they cannot be used in setting the initial price range and, because the initial price range generally creates binding restrictions, they are of little practical use in setting IPO prices. Another alternative would be an actual direct auction to the public. Appropriately designed, auctions can be incentive compatible and truth revealing. However, auctions are often afflicted by the winners curse, as Rock (1986) points out. And, when participation is endogenous, auction mechanisms may lead to increased risk (because of uncertainty about the number of bidders) and a sub-optimal level of information production (Sherman, 2004).