price on the first day, sold the same number of shares and paid the same percentage spread to investment
bankers, Google would have raised $1,379.2 million (or $210.9 million more) for itself and Google’s
existing shareholders would have received $532.8 million (or $81.5 million more), without the exercise of
the over-allotment option.32 Adding the difference in investment bank proceeds brings the total difference
to $300.7 million that was clearly “left on the table” (see calculations in Table II, column 4). Had Google
set its IPO price at the IEM forecast and managed to sell the same number of shares (which seems likely,
according to the estimated demand curve as discussed above), the total foregone proceeds increases to
$379.19 million (calculations in Table II, column 5).33
The overall evidence is consistent with known, equilibrium underpricing, suggesting that Google
may not have wanted to price at the full post-IPO market level. In such cases, prediction markets can still
serve a valuable role. They can serve as low cost mechanisms for forecasting post-IPO market prices.
These forecasts can be used to set IPO prices to achieve the desired levels of underpricing. In such cases,
we would not argue that prediction markets should replace road shows, book building and other means of
gathering information. Instead, we argue that prediction markets can supplement other mechanisms. This
mirrors observations from political markets. Election prediction markets do not replace polls. Instead,
they provide an additional information aggregation mechanism. Given the stakes involved, any
mechanism that provides additional information about IPO valuations would be extremely valuable.
5. Conclusions and Discussion
Underpricing of IPOs is of great theoretical and practical interest. The distinctive features of the
Google IPO and the IEM prediction markets run in advance of the IPO provide unique evidence on
underpricing theories. From a practical point of view, given the initial underpricing of IPOs, companies
The entire over-allotment option was sold by existing shareholders. Had they sold the full over-allotment at the IEM predicted net price (assuming the same spread) instead of the actual $82.6161, existing shareholders would have made $158.0 million more than they actually did. We have already discussed how the excess demand information can be used to judge the likelihood that the same number of shares could have been sold at the IEM predicted price. In addition, Google closed above the IEM forecasted price on the second day of trading and has risen above this level even after the exercise of the over- allotment option had been made public.