X hits on this document

PDF document

Searching for Google’s Value: - page 23 / 44





23 / 44

IPO return to overcome adverse selection problems. Informed investors will only participate in an IPO if

they know that the IPO is by a “good” company. In this case, uninformed investors receive partial

allocations of shares. However, when the IPO is by a “bad” company, the informed investors do not

participate and uninformed investors receive full allocations. This creates an adverse-selection-based

winner’s curse that must be overcome by underpricing on average to get uninformed investors into the

IPO market. Google’s auction process may have been prone to such a winner’s curse. If so, uninformed

auction participants would need to expect Google to underprice on average to create sufficient returns

(again, on average) to overcome the winner’s curse. Our estimates of the demand curve above suggest

that Google deliberately underpriced, possibly to overcome this problem. In contrast, Reny and Perry

(2003) show that, under the right conditions, double auction markets (like our prediction markets) are not

prone to the winner’s curse and converge to the fully revealing rational expectations equilibrium

(explaining our accurate prices). Finally, differences of opinion (between investors) can also drive the

observed trading in prediction markets (e.g., Harris and Raviv, 1993). This evidence is consistent with

asymmetric information across investors.

However, several pieces of evidence run counter to Rock’s (1986) winner’s curse model. First,

given Google’s stated goals and IPO mechanism, it is unclear whether investors could have reasonably

expected underpricing as an outcome even if Google was a “good” company. Second, according to this

model “good” companies should have a higher than expected actual IPO return ex post and “bad”

companies should have a lower than expected actual IPO return ex post (because the information about

the company is revealed through the IPO process).29 In neither case will the actual IPO return ex post

equal the ex ante expected IPO return. In Google’s case, the ex post return and the ex ante expected

return (derived from IEM prices) were approximately equal. Further, Rock (1986) argues that the ex ante

level of uncertainty will be positively correlated with predicted underpricing. While we cannot estimate a

cross-sectional correlation, we can estimate the correlation for this IPO through time. We estimate the ex


In Rock’s, 1986, model, this would be revealed by the presence of informed investors. Uninformed investors can infer the quality of the issue by seeing whether they were allocated the full number of shares for which they bid.


Document info
Document views167
Page views167
Page last viewedMon Jan 23 01:18:32 UTC 2017