Because traders reveal their information in a prediction market, the prediction market alone
provides evidence on IPO underpricing theories that rely on assumptions about the distribution of
information before an IPO. Our evidence suggests that traders were able to estimate accurately the post
IPO value of Google. Further, they revealed this information for very little payment. In addition, we
show that the correlation between ex ante forecasts of underpricing and the implied degree of uncertainty
in traders’ forecasts runs counter to models based on asymmetric information across traders. Because this
evidence does not depend on Google’s unique IPO auction mechanism in any way, we argue that this is
general evidence against three types of asymmetric information based theories: (1) theories that rely on
outsiders being relatively uninformed, (2) theories that rely on outsiders being relatively informed and
revealing that information only in exchange for large payments and (3) theories that rely on significant
winner’s curse problems.
Google’s IPO provides a particularly constructive setting for prediction markets because of its
unique features. Google’s specific and clearly stated goal was to avoid IPO underpricing. To achieve this
goal, they used an auction mechanism for gathering information, setting prices and allocating shares.
Thus, the IPO price provides a natural benchmark for evaluating the accuracy of prediction markets and
their potential usefulness in aggregating information in an IPO setting. But, more importantly, the
auction mechanism provides unique evidence on theory. First, evidence from the auction outcome allows
us to make inferences about Google’s pre-IPO information. The evidence suggests that Google also knew
the degree of underpricing that would result from the price they set. Combined with the prediction market
evidence, we think it makes a compelling case against asymmetric information based theories of IPO
underpricing in the specific case of Google. Second, Google’s auction mechanism gives two additional
pieces of evidence relevant to theory: (1) the auction severely restricted the investment bankers’
discretion in issuing shares and (2) the auction did not allow Google to pre-commit to underpricing.
Because of this design, the underpricing that occurred in Google’s case cannot be explained by models
that rely on either of these features.