uncertainty (change in σˆt ) on days of amendment filings was -0.066. The change on other days averaged
less than 0.001. According to a Mann-Whitney two-sample rank sum statistic, this difference is
significant (z=2.717, p-value=0.0066). This correspondence between the reductions in uncertainty
implied by prices and what one would expect from significant information releases leads further credence
to IEM prices as efficient forecasts.
C. Estimating the Demand Curve for the Google IPO
If we knew the demand curve for Google stock, we could determine whether the IEM predicted
post-IPO market price could have been a feasible market clearing price for the IPO. While Google has
not released information about the bids in its auction, publicly available information combined with
Google’s allocation mechanism, allows us to estimate the demand curve.
It seems likely that the IPO price was set below the market clearing price. Page 40 of the
amended S-1 filing on August 13, 2004 (the day the auction began) states, “If the initial public offering
price is equal to the auction clearing price, all successful bidders will be offered share allocations that are
equal or nearly equal to the number of shares represented by their successful bids” (italics added). If
Google set the price lower than the auction market clearing price, it stated that it would ration shares
using one of two mechanisms (pro rata or maximum share allocation) with a goal of allocating to
successful bidders at least 80% of their quantities bid. While it is possible that there was a highly elastic
demand at or very near the market clearing price that necessitated rationing, it is clear that Google
expected little rationing at the market clearing price and that significant rationing would indicate pricing
below the auction market clearing price.26
One might also argue that the $85 price resulted from tacit or explicit collusion among large investors to lower the price below true value in a Wilson (1979) style share auction equilibrium. However, free entry breaks this equilibrium. Further, supposing it did occur, this outcome would not change the implications for theory developed here because it would not change the evidence on the distribution of information, the lack of discretionary allocations of shares and the inability to pre-commit to underpricing. However, it would mean that underpricing was unavoidable given Google’s auction mechanism.