392 A. Ljungqvist
prediction,James and Wier (1990)find that companies that have secured lines of credit
before their IPOs (and thus have a more credible threat to withdraw) experience lower
underpricing.
In the Benveniste and Spindt framework, investors incur no cost in becoming in-
formed. If information production is costly, underwriters need to decide how much
information production to induce.Sherman and Titman (2002)explore this question
in a setting where more information increases theaccuracyof price discovery, resulting
in a trade-off between the (issuer-specific) benefit of greater pricing accuracy and the
cost of more information production.
The idea of costly information production is further investigated byBenveniste,
Busaba, and Wilhelm (2002)andBenveniste et al. (2003)who link the underwriter’s
capacity to ‘bundle’ IPOs over time to the empirical observation that IPOs tend to occur
in waves. The central idea is that valuation uncertainty is composed of a firm-specific
and an industry component. Obtaining information about the industry component allows
investors to evaluate other offerings in that industry more cheaply. Such economies of
scale could result in too few firms going public, because the first firm to do so must com-
pensate investors for their whole valuation effort, while later firms can ‘free-ride’ on the
information production.^9 By establishing networks of regular investors, underwriters
may be able to reduce this negative externality. To do so, they compensate investors
for their information costs across asequenceof offerings. This is consistent with the
observation that investment banks tend to specialize in particular industries, and that
companies tend to go public in industry-specific ‘waves’.
3.2.2. Testable implications and evidence
The most direct tests of bookbuilding theories of IPO underpricing areCornelli and Gol-
dreich (2001, 2003)andJenkinson and Jones (2004). These studies exploit proprietary
datasets from two different European investment banks. The datasets contain informa-
tion on the bids institutional investors submitted into the book, as well as the allocations
they received. Such data are usually kept confidential, so their availability provides a
rare opportunity to test information revelation theories of underpricing. Two potential
drawbacks are that the sample sizes are relatively small, and that the results are bank-
specific and so may not generalize to other banks. Indeed, the fact that Jenkinson and
Jones’ results are at odds with those of Cornelli and Goldreich, as we will see, may in
large part be due to differences in the sophistication with which the two banks carry out
bookbuilding.
(^9) The idea that information spillovers can cause IPO clustering is explored in three papers that are not
based on the Benveniste–Spindt information-acquisition framework.Booth and Chua (1996)point out that
when many companies come to market, the marginal cost of information production is lower, so average
underpricing falls.Mauer and Senbet (1992)argue that IPO companies that start trading in the secondary
market may reduce the valuation uncertainty surrounding companies with similar technologies which are in
the process of going public.Stoughton, Wong, and Zechner (2001)develop a model in which one firm’s IPO
provides information about industry prospects, thus causing many similar companies to go public soon after.