Pre-IPO Markets∗

Francesca Cornelli David Goldreich

(London Business School and CEPR) (London Business School and CEPR)
Alexander Ljungqvist
(New York University and CEPR)

This Version: March, 2004

Abstract. We take advantage of the grey market, a when-issued market for shares to
be issued in an IPO, to look at how behavioral biases affect prices in the post-IPO market.
We develop a model in which bookbuilding and a grey market take place simultaneously.
While bookbuilding extracts information from large institutions about the fundamental
value of the securities, the grey market reflects the opinion of (possibly biased) retail and
other small investors. When the publicly observable grey market price is high relative to
fundamental value, bookbuilding investors resell shares to these smaller investors in the
aftermarket, so the offer price and the aftermarket price will be close to the grey market
price, but will revert to the fundamental value in the long run. In contrast, when the grey
market price is low, the offer and aftermarket prices will be close to fundamental value.
Empirical results confirm this asymmetry in the issue price and aftermarket prices.

∗
We thank Alex Stomper for providing some of the data. Thanks for helpful comments go to Pierluigi
Balduzzi, Nick Barberis, Wayne Ferson, Francisco Gomes, Harold Mulherin, and seminar participants at Arizona
State University, Boston College, Brunel University, Columbia Business School, Norwegian School of Management
in Oslo, Norwegian School of Economics and Business Administration in Bergen, Tilburg University, University
of Amsterdam, University of Zurich, the NYSE/Stanford Conference on Entrepreneurial Finance and IPOs, and
the LSE-LBS Corporate Finance Workshop. Dimitry Makarov provided excellent research assistance.
 1. Introduction

The performance of initial public offerings, both in the short and long run, has long shown
numerous anomalies. Some studies have argued that high prices in the short run followed by
abnormally low returns in the long run are due to overenthusiasm among retail investors who may
be less than perfectly rational.1 One difficulty in testing this hypothesis is that it is difficult to
disentangle the expectations of different types of investors when we only observe a single price in
the aftermarket. Thus, more evidence is needed to determine what type of investors are driving
prices in the aftermarket and whether these prices are consistent with rational behavior.
In this paper we take advantage of the existence in many countries of a pre-IPO “grey market”
to distinguish between the valuations of different types of investors. The grey market is a when-
issued market (which takes place at the same time as bookbuilding) in which smaller investors
speculate on the post-IPO share price.2 The publicly observable grey market price reflects the
valuation of these smaller investors. Thus, the grey market provides a unique opportunity to
observe the opinion of a subset of investors before the shares start trading. Using the grey market
price we can study the relation between the valuation of smaller investors and the aftermarket
share price, and to assess its consistency with rationality.
In order to conduct this analysis, we first build a model that gives us empirical implications
that we then take to data. In the model, small investors trade in the grey market, while the
underwriter collects bids during bookbuilding from large institutional investors before pricing
the issue.3 Thus, there are two separate markets with two separate sets of investors.
Bookbuilding investors are endowed with information about the fundamental value of the
shares. Grey market investors also have information which may or may not be relevant to the
fundamental value. In addition, the grey market investors may overweight the relevance of their
information.
To the extent that grey market investors are representative of smaller investors in general,
their valuation is indicative of the reservation price for smaller investors in the aftermarket. If that
reservation price is higher than the fundamental value of the shares, bookbuilding investors know
that they can resell their shares in the immediate aftermarket at a price above the fundamental

1
See Ritter and Welch (2002) for a discussion.
2
Section 3 describes the grey market in more detail.
3
For a description of bookbuilding see Cornelli and Goldreich (2001) and Ljungqvist and Wilhelm (2002).

1
value. If instead the reservation price of smaller investors is low, bookbuilding investors will
hold on to the shares that they are allocated, valuing them at the fundamental value. Thus,
the bookbuilding investors value the shares at the maximum of the fundamental value and the
reservation price of the smaller investors. Since the underwriter can also observe the grey market
price, when the grey market price is high he will anticipate that the shares will be sold in the
aftermarket at a profit and will set a higher offer price to extract this additional surplus. When
the grey market price is low, the underwriter will disregard the valuation of smaller investors.
Empirically, we test the predictions of the model by looking at the relation between the grey
market price and both the issue price and the aftermarket price. The influence of the grey market
price should depend on the degree to which it contains information about the fundamental value.
First, consider the extreme case in which the grey market information is completely unrelated
to the fundamental value. If grey market investors mistakenly use this information in their
reservation price, it will still affect the aftermarket price when these (or similar) investors purchase
the shares in the aftermarket. Thus, we should expect a relation between the grey market price
and the aftermarket price, but only when the reservation price is above the fundamental value.
There should be no relation when the reservation price is below the fundamental value because
then the bookbuilding investors will not resell the shares to the grey market investors. To the
extent that a high reservation price is anticipated by the underwriter, there will be a similar
relation between the grey market price and the issue price.
On the other hand, if the grey market price does contain fundamental information, then there
will be a positive relation between the grey market price and the aftermarket price (and also the
issue price) even when the grey market price is low. However, to the extent that grey market
investors overweight their information, there will still be an asymmetry, i.e., a stronger relation
when the grey market price is high.
Note that the model predicts asymmetry only when the grey market investors show conservatism.4
In other words, although they receive new information about the fundamental value when the
issue price is set, they overweight their own information when updating their reservation price.
Whether or not grey market investors overweight their information is thus an empirical question
which can be answered in the context of the model.
We test the predictions of the model using grey market price data for a large set of European

4
See Barberis, Shleifer, and Vishny (1998).

2
IPOs completed between 1995 and 2002. We find that the grey market price is more correlated
with the issue price and the aftermarket price when the grey market price is high, although
there is a positive correlation even when the grey market price is low. This suggests that the
grey market contains information about the fundamental value, but also that the grey market
investors overweight their information in a way that is reflected in the aftermarket price and
exploited by the underwriter when setting the issue price. We find higher levels of aftermarket
volume when the grey market price is high, consistent with bookbuilding investors reselling their
shares to grey market investors only when the grey market investors have higher valuations. We
use the bid-ask spread in the grey market to proxy for a lack of depth which should be reflected
in a lower aftermarket price. The results are consistent.
The model also has implications for long-run returns. When the grey market price is high,
demand from smaller investors will cause the shares to trade in the short run at a high price
relative to fundamentals. However, in the long run, as the true value is revealed through time,
prices will revert to the fundamental value and we expect negative returns. On the other hand,
when the grey market price is low, the aftermarket price is based on fundamentals and we do
not expect a reversal pattern. Our empirical results are consistent with this prediction.
We should point out one thing this paper does not do. Although the optimal mechanism in
our model has underpricing to facilitate information extraction in bookbuilding (as in Benveniste
and Spindt (1989)), it is not our goal to explain the magnitude of observed underpricing. The
point of this paper is to show how high valuations among grey market investors can lead to a high
aftermarket price, but not why the issue price is often set well below this. However, a number of
explanations given in the literature (for example, agency conflicts) could be added to our model
to explain underpricing.
A notable feature of our model is that prices are sometimes biased upwards, but never down-
wards, even though the potentially irrational investors could be both excessively optimistic or
pessimistic. In the presence of short sale constraints, excessively pessimistic investors are priced
out of the market.
We stress that our results are also relevant for countries that do not have a grey market (such
as the United States). As long as the underwriter and the major institutional investors have
some sense of what smaller investors are willing to pay, the same phenomena would occur. The
significance of the grey market is that it allows us to observe smaller investors’ valuation easily
and directly, enabling us to test for these effects.

3
Related literature
Several papers on IPOs have documented empirical patterns that motivate our study. Ritter
(1991) presents evidence that abnormally high prices immediately after the IPO are followed by
abnormally low returns in the long-run and Ritter and Welch (2002) show that this is particularly
strong in certain period of times. Purnanandam and Swaminathan (2003) compare the offer
price in several IPOs with their “fair value” computed using several price multiples of non-IPO
industry peers. They find that issues which are overpriced relative to their fair value also have
higher returns on the first day of trading but lower returns in the long run. These patterns are
consistent with our results. If some investors are excessively enthusiastic about an issue, the
underwriter sets the offer price above the fundamental value, expecting these investors to buy
the shares in the aftermarket. This leads to a high short run price, and, as the price converges
to the fundamental value, long returns will be low.
A large literature, both theoretical and empirical, has explained IPO patterns with the pres-
ence of small, unsophisticated investors. Lee, Shleifer, and Thaler (1991) show that the annual
number of IPOs is negatively related to the closed-end fund discount which they argue is a mea-
sure of retail investor sentiment. Rajan and Servaes (2003) model two different types of irrational
agents, feedback traders and sentiment investors. The latter are similar to our grey market in-
vestors. They proxy for investor sentiment using the market-to-book ratio and find that it is
positively correlated to first-day returns and negatively correlated to long term returns. In con-
trast, we use the grey market price to measure the valuation of sentiment investors. Ljungqvist,
Nanda, and Singh (2003) argue that an initial price run-up may be due to the existence of “ex-
uberant” investors and may lead to long-term underperformance. Their explanation is similar
to ours, but their model addresses the issue of underpricing, which is needed to compensate
regular investors for losses in case the hot market ends abruptly. Our model is mainly aimed at
introducing the grey market, in order to generate testable implications and does not address the
issue of underpricing. It is interesting to note that the grey market price constitutes an ex-ante
signal that the issue will be hot. Therefore, when the grey market price is high the investors
know that that there is a low probability that the hot market will end. This constitutes an addi-
tional reason why the grey market would ultimately benefit the issuer. Aggarwal, Krigman and
Womack (2002) also relate the price path in the aftermarket to momentum traders, focussing on
the role of research analyst and media in creating this momentum. We do not investigate why
the grey market investors are particularly enthusiastic about a certain issue, and simply look at

4
what happens when they are enthusiastic.
Testing these theories means investigating the role of retail or small investors. While we use
the grey market price as an indication of smaller investors’ opinion, other papers have looked
at who owns the shares in the aftermarket. Ofek and Richardson (2003) show that high initial
returns occur when institutions sell IPO shares to retail investors on the first day. Similarly, Ben
Dor (2003) looks at the level of institutional ownership shortly after the IPO and find that high
institutional ownership forecasts higher returns in hot markets. These findings are consistent
with our paper, since we predict that bookbuilding investors sell the shares to smaller investors
when they are overvalued. In an empirical study that is complementary to our findings, Dorn
(2003) finds that the volume of trade by retail investors in the German grey market is correlated
with high initial returns and low long-run returns. This can be viewed as further evidence that
participation of smaller investors in the grey market can be interpreted as sentiment.
Besides Dorn (2003), other papers study the grey market in Germany. Löffler, Panther and
Theissen (2002) document that pre-issue prices are unbiased estimates of the first-day prices.
Aussenegg, Pichler, and Stomper (2003) also find that IPO offer prices are related to prices in
the grey market but they show that the coefficient is smaller than one, which they argue is
consistent with bookbuilding being used for gathering information, even in the presence of a
when-issued market. Finally, Pichler and Stomper (2004) also model the interaction between
bookbuilding and the grey market. However, in their model, grey market investors have similar
information to bookbuilding investors, and they ask whether the existence of the grey market
hinders or helps the aggregation of the information in bookbuilding. In contrast, we introduce a
class of investors who have different information from bookbuilding investors, in order to explain
certain IPO phenomena and to show how these (possibly biased) investors affect prices in an
asymmetric way.

The paper proceeds as follows. In Section 2 we present the model, the optimal mechanism,
and the empirical implications. Section 3 describes the data. Section 4 presents the empirical
results. In Section 5 we extend the model to allow bookbuilding investors to trade in the grey
market. Section 6 concludes.

5
 2. The model

An issuer wants to sell S shares in an IPO. Each share has a fundamental value v ∈ [0, v̄]. Since
the underwriter does not know v before setting the issue price PI , he conducts bookbuilding
to collect information from institutional investors about v. Simultaneously with bookbuilding,
a publicly observable grey market takes place in which a different group of investors trade the
shares on a when-issued basis.5
The expected fundamental value of a share is a weighted average of the information arriving
from bookbuilding sB and the information from the grey market sG :

E(v | sB , sG ) = αsG + (1 − α)sB , (1)

where 0 ≤ α < 1. When α = 0, grey market investors’ information is irrelevant. We assume that
bookbuilding investors’ information is always relevant.6
The timing is as follows. First the underwriter sets an initial indicative price range, based
on his prior beliefs. Then both bookbuilding and grey market trading begin. At the end of
bookbuilding, the underwriter observes the bids in the book as well as the grey market price
and sets the issue price. When the issue price is set, information about bookbuilding is revealed.
Afterwards, aftermarket trading begins.

2.1. Bookbuilding. We model the bookbuilding process as in Biais and Faugeron-Crouzet

(2002). Three investors take part in the bookbuilding process: two are informed and one is
uninformed. The uninformed investor can buy at most S(1 − k) shares, with 0 < k < 1. The in-
formation from bookbuilding, sB , is the aggregate of the signals si observed by the two informed
investors, i = 1, 2. si is i.i.d. and equals H with probability π and L with probability (1 − π).
The distribution of the signals is such that if both investors observe a signal equal to H then
sB = H; if both observe L then sB = L; but if one observes L and the other H then sB = M ,
where 0 < L < M < H < v̄. The ex-ante expected value of the bookbuilding information is
E(sB ) = π 2 H + 2π(1 − π)M + (1 − π)2 L. For simplicity, we assume that E(sB ) = M .
From the point of view of an informed investor who observes si (as well as the grey market

5
In Section 5, we consider the case in which there is an overlap between the two groups of investors.
6
Cornelli and Goldreich (2003) show that bookbuilding aggregates information that is relevant for both the
issue price and the long-run aftermarket price.

6
information sG , which, as explained below, can be inferred from the grey market price) the
expected fundamental value is

E(v | sG , si ) = αsG + (1 − α)E(sB | si ) (2)

Each investor submits a bid for shares, after observing his own signal, but not knowing the
signal of the other informed investor. The underwriter designs a mechanism (described in Section
2.4) in which he sets the issue price and allocates the shares as a function of the bids. We assume
that the underwriter acts in the interest of the issuer, i.e., he maximizes IPO proceeds.
The bookbuilding set-up described so far is similar to the one in Benveniste and Spindt
(1989). The main difference here is the existence of a grey market whose price is observed before
the issue price is set.

2.2. Grey Market. At the same time as bookbuilding there is a grey market in which investors
trade the shares on a when-issued basis. For now, we assume that bookbuilding investors are
not allowed to trade in the grey market.
Since in reality the grey market price is continuously and publicly observable, while book-
building is a confidential process controlled by the underwriter, we assume that grey market
investors do not observe sB . Instead, they only observe a signal about the value of the shares,
sG ∈ [0, v̄].7
Grey market investors know that the fundamental value is a weighted average of their signal
and the bookbuilding information, but we allow for the possibility that they overweight the im-
portance of their signal. In other words, after observing sG , their expectation of the fundamental
value of the shares is

EG (v | sG ) = α̂sG + (1 − α̂)E(sB ) (3)

where α̂ ≥ α and EG refers to the expectation from the perspective of grey market investors.
The difference (α̂ − α) represents the extent to which grey market investors overweight their

7
The assumption that the signal of the grey market investors is continuous, while the signal of the bookbuilding
investors is discrete, is made only in order to simplify the analysis.

7
signal. When α̂ − α = 0 they are fully rational. Note that only the expectation of sB appears in
equation (3), since grey market investors do not observe the bookbuilding information.
Trading in the grey market results in a price PGM , reflecting investors’ beliefs about the
fundamental value of the shares. Thus, PGM = EG (v | sG ). After observing PGM the underwriter
and the bookbuilding investors, knowing α̂, can perfectly infer sG using the following relation:

PGM − (1 − α̂)M
sG = (4)
α̂

After the underwriter aggregates the bookbuilding information into the issue price (and before
the start of aftermarket trading), the bookbuilding information sB is revealed.8 Grey market
investors then update their valuation, starting from their prior valuation, to

P̂GM ≡ P̂GM (sG , sB ) = α̂sG + (1 − α̂)sB = PGM + (1 − α̂)(sB − M ). (5)

P̂GM differs from PGM because it incorporates the observed sB rather than its expectation. It
differs from the fundamental value if α̂ 6= α. The difference between α̂ and α is motivated
by experimental evidence that individuals are slow to change their beliefs in the face of new
evidence: they update their priors too little relative to Bayesian updating (see Barberis, Shleifer,
and Vishny, 1998).9

2.3. Aftermarket. After the shares are allocated to the bookbuilding investors, trading in the
aftermarket begins. At this point, both bookbuilding and grey market investors have observed
both sG and sB . Grey market investors value the shares at P̂GM , while bookbuilding investors
value the shares at the expected fundamental value, given in equation (1). Again, these two
valuations differ if α̂ > α.
We assume that aftermarket participants include investors who have the same valuation as
the grey market investors. They may be the grey investors themselves or other (perhaps retail)
investors who did not trade in the grey market. As a result, the grey market price is representative

8
A more realistic assumption might be that grey market investors infer the information from the offer price.
However, we assume sB is revealed to avoid modelling situations in which the offer price is manipulated by the
underwriter to hide information.
9
An alternative explanation could be the one studied in Harris and Raviv (1993), where individuals receiving
common information differ in the way they interpret this information.

8
of the valuation of a larger set of investors. For simplicity, we continue to refer to this set of
investors as grey market investors.10
Let us define PAM as the aftermarket price in the short-run, say on the first day of trading.
If the fundamental value of the shares exceeds the price P̂GM grey market investors are willing
to pay in the aftermarket, then the bookbuilding investors will not sell their shares to them.
Thus, there will be no trading involving grey market investors and the aftermarket price will not
depend on their valuation. The expected aftermarket price PAM will then equal the expected
fundamental value. If instead the price that grey market investors are willing to pay exceeds the
fundamental value, the bookbuilding investors can sell their shares to the grey market investors
at this higher price.
The price at which bookbuilding investors can resell their shares depends upon the depth in
the market. If there are many investors willing to buy shares at the price P̂GM , bookbuilding
investors will have all the market power and will set the price equal to P̂GM , extracting all the
surplus from trading. However, if there are not enough investors willing to buy all S shares
at P̂GM , for example if the demand for these shares is downward sloping, then bookbuilding
investors will have to sell some of their shares at a lower price. Assuming a linear demand
curve, bookbuilding investors expect to sell their shares in the aftermarket at P̂GM − λS, where
λS captures the discount necessary to sell all S in the aftermarket. Although more complex
functional forms are possible, this simple linear form suffices to capture the idea that the market
may not be very deep. If the market is deep enough to sell all the shares at P̂GM , then λ = 0.
In the long run, all uncertainty is resolved and the long-run price equals the fundamental
value.

2.4. Optimal Mechanism. We now characterize the optimal bookbuilding mechanism. This
mechanism specifies the issue price and the number of shares to be allocated to the various
bidders, as a function of their bids and the grey market price PGM .11
To find the optimal mechanism, by the Revelation Principle we can restrict attention, without

10
Dorn (2003) finds a strong positive correlation between the volume of retail trade in the grey market and
retail volume in the first-day of aftermarket trade. This supports our assumption that the opinion of grey market
investors is indicative of the valuation of small investors in the aftermarket.
11
By law, the underwriter cannot charge different prices to different investors, so we do not allow the mechanism
to price discriminate among investors. Bennouri and Falconieri (2003) show that, if there is no limit to the
quantities investors can be allocated, price discrimination is never optimal.

9
loss of generality, to a direct revelation mechanism where the bookbuilding investors simultane-
ously announce their signals to the underwriter. The underwriter uses the announced signals
(s̃1 , s̃2 ) (which aggregate to s̃B ) and the grey market information sG to set the issue price and
to allocate the shares. A direct revelation mechanism is described by the outcome functions
(PI , q, qu ), where PI (sG , s̃B ) is the issue price; q(sG , s̃i , s̃j ) is the allocation to an informed in-
vestor who announces signal s̃i when the other informed investor announces s̃j ; and qu (sG , s̃i , s̃j )
is the allocation to the uninformed investor when one informed investor announces s̃i and the
other announces s̃j . We look for an equilibrium of this mechanism in which buyers truthfully
reveal their signals.
In order to derive the optimal mechanism, we must first determine the reservation price of
the bookbuilding investors. After the issue price has been set, the bookbuilding investors observe
both sB and sG . They thus know the expected fundamental value of the shares (equation (1)),
which is the value they will obtain if they hold the shares in the long run. Additionally, since
they observe PGM , they can also compute the price that grey market investors would be willing
to pay in the aftermarket (P̂GM − λS).
When P̂GM − λS exceeds the fundamental value, bookbuilding investors will be able to resell
their shares in the aftermarket to grey market investors. Thus, the relevant value of the shares,
from their point of view, is the short-run aftermarket price. When P̂GM − λS is lower than the
fundamental value, bookbuilding investors will not resell the shares in the aftermarket. Therefore,
after the end of bookbuilding, when a bookbuilding investor observes sB , his valuation will be
the maximum of the two possible valuations, i.e., M ax{E(v | sG , sB ), P̂GM − λS}.
During bookbuilding, each informed investor will have observed his own signal si but not that
of the other informed investor. Therefore, an investor with a signal si = H will value a share at
π ∗ M ax{E(v | sG , H), P̂GM − λS} + (1 − π) ∗ M ax{E(v | sG , M ), P̂GM − λS}. An investor with
a signal L will have a valuation of π ∗ M ax{E(v | sG , M ), P̂GM − λS} + (1 − π) ∗ M ax{E(v |
sG , L), P̂GM − λS}.
The underwriter, who also observes the grey market price, knows that if PGM is high, he
can increase the issue price above the fundamental value. That way, he can extract the surplus
informed investors expect to gain from trading with the grey market investors in the aftermarket.
Yet if PGM is low, the underwriter does not need to lower the issue price, since he knows that
the bookbuilding investors are still willing to buy and hold the shares at a price close to the
fundamental value. In other words, the issue price (and the subsequent aftermarket price) will

10
reflect the grey market price when it is high, but will reflect the fundamental value when the
grey market price is low.
The following proposition presents the optimal mechanism.
1−α
Proposition 1: Assume that α̂ ≥ 2
and that H and L are equidistant from M .12 In the
optimal mechanism the quantities allocated to the bookbuilding investors are:

q(H, H) = S/2, qu (H, H) = 0,

q(H, L) = S, q(L, H) = 0, qu (H, L) = 0, (6)
q(L, L) = Sk/2, qu (L, L) = S(1 − k)

where we have suppressed the argument sG since the quantities allocated do not depend on the
grey market information.
The issue price depends on the grey market and the bookbuilding information as follows:

1−π
PI (sG , H) = VH (H) − π
k [VM (L) − VL (L)]
PI (sG , M ) = VM (M ) (7)
PI (sG , L) = VL (L)

where

VsB (s̃B ) ≡ M ax{E(v | sG , sB ), P̂GM (sG , s̃B ) − λS}. (8)

Proof: See Appendix 1.

VsB (s̃B ) represents the value of a share to an informed investor where sB is the aggregate
information from the two signals of the informed investors and s̃B is the aggregate information
revealed through bookbuilding (and thus conveyed to grey market investors). When sB 6= s̃B ,
at least one investor misreported his signal. VsB (s̃B ) is the maximum of the fundamental value
and the resale value. Note that our notation suppresses the argument PGM in the function VsB
as the maximization is conducted for a given PGM .

12
These assumptions are sufficient but by no means necessary for the proposition to hold.

11
 Although the quantities allocated do not depend on the grey market price, the issue price
PI depends on PGM as well as on the announced signals. To highlight how the issue price, as
expressed in (7), depends on the grey market price and the information in the book, we can
divide the possible values of PGM into intervals and present the issue price for each interval.
The boundaries of the intervals and the actual expression for the issue price in each interval are
derived in Appendix 1 and represented in Figure 1.
The vertical axis represents the possible values that PGM can take. The issue price expression
is different in different intervals. Moreover, within each interval, the issue price depends on the
bookbuilding information, sB . Figure 1 visualizes the asymmetry in the optimal mechanism. Let
us start by looking at the two extreme intervals. When PGM is very high (i.e., above v1 , where
v1 is defined in Appendix 1) the issue price is close to the grey market price (more precisely, the
updated valuation of the grey market investors) and does not depend at all on the bookbuilding
information. This follows because the grey market price is so high that bookbuilding investors
are certain they can resell their shares in the aftermarket at a price higher than the fundamental
value, regardless of what the fundamental value is. The underwriter takes advantage of this by
setting PI above the fundamental value. When PGM is very low (below v4 ), the issue price is
based solely on the fundamentals. This is because the grey market valuation is so low that grey
market investors are unwilling to buy shares in the aftermarket at a price above the fundamental
value. In this case, the grey market is irrelevant and the mechanism is the standard bookbuilding
mechanism. In the middle are cases in which the issue price is set equal to the fundamental
value (minus an informational rent to be left to the bookbuilding investors for revealing their
information truthfully) when the fundamental value is high, and equal to the updated grey market
valuation otherwise. So, on average, the issue price is close to the grey market price when the
grey market price is high, but not when the grey market price is low, creating an asymmetry in
the issue price relative to the grey market price.
The extent of the asymmetry depends on α and α̂. When (α̂ − α) is small, grey market
investors overweight their signal only by a small amount. In this case, their valuation and the
fundamental value will be similar and there will be less asymmetry. When (α̂ − α) is large, grey
market investors are much more overconfident in the relevance of their signal and their valuation
can be much higher than the fundamental value. In this case, the issuer can take advantage of
their overconfidence and set the issue price much higher than the fundamental value.
Recall that α indicates the relevance of the grey market signal for the fundamental value. In

12
the extreme case of α = 0, the grey market price is not relevant for the fundamental value, which
is completely determined by the bookbuilding information. In this case, when PGM is low and
the issue price is set equal to the fundamental value, there will be no relation between PGM and
the issue price. However, when the grey market price is high, it will be closely related to the
issue price even though it contains no fundamental information. When α > 0 the grey market
price includes fundamental information. So, even when the grey market price is low there will
be a positive relation between PGM and the issue price. However, to the extent that α̂ > α, this
relation will be weaker than when PGM is high, so the asymmetry remains.
The issue price is based on the valuation of bookbuilding investors, which in turn depends
on their expectation about aftermarket prices. Thus the asymmetry in the issue price is driven
by the asymmetry in the aftermarket price PAM , and there will be a stronger relation between
PGM and PAM when PGM is high.
Although we allow for shares to be sold in the aftermarket at a discount λS to capture a
potential lack of depth, the results also hold when λ = 0. A larger λ implies that shares will be
sold to grey market investors in the aftermarket less often but the basic asymmetry remains.
As explained above, the bookbuilding investors’ expectations about the aftermarket are cen-
tral to determining their willingness to pay and the choice of issue price. This aspect is similar in
spirit to Busaba and Chang’s (2002) model where investors have an incentive to misreport their
information to the underwriter in order to fool uninformed investors and take advantage of them
in the aftermarket. However, there are two main differences here. First, in our model the under-
writer designs the optimal mechanism to account for bookbuilding investors’ incentives to resell
their shares. As a result, bookbuilding investors will not misreport their information. Second,
since the grey market price signals the potential resale value in the aftermarket, the bookbuilding
investors and the underwriter can take it into account when determining their actions.
This is why the existence of the grey market is beneficial to the issuer, even when it does
not contain any information about the fundamental value of the shares (α = 0). The valuation
of the grey market investors affects the bookbuilding investors’ valuation (because it affects the
short-run aftermarket price) and thus provides a lower bound on their willingness to pay. If grey
market investors are willing to pay a high price, the surplus that can be appropriated increases.
Moreover, since this part of the valuation is publicly observable, the issuer can extract a larger
part of the trading surplus from the bookbuilding investors.

13
2.5. Empirical Implications. The model allows us to make predictions about the relation be-
tween the grey market price PGM , the issue price PI , and the aftermarket prices in the short and
long run, as well as other variables. Here we list the main empirical predictions. A more detailed
analysis is conducted in Section 4, where we present the results.
Hypothesis 1: PI is positively correlated with PGM . Moreover, α̂ > α implies that this corre-
lation is larger when PGM is high. α > 0 implies that this correlation is positive even when PGM
is low.
Hypothesis 2: The short-run aftermarket price PAM is positively correlated with PGM . More-
over, α̂ > α implies that this correlation is larger when PGM is high. α > 0 implies that this
correlation is positive even when PGM is low.
Hypothesis 3: When the reliability of the grey market signal increases, the correlations be-
tween PGM and the issue price and between PGM and the aftermarket price increase.
A more reliable grey market signal means that an investor, who is weighting this signal with
the bookbuilding signal, should give additional weight to sG . In other words, α should be higher.
Thus PGM will be more closely related to the fundamental value. While we do not explicitly
model the grey market investors’ bias, presumably α̂ will also be higher in this case.
Hypothesis 4: When PGM is high, PI and PAM are negatively correlated with the issue size
(S) and positively correlated with the depth of the grey market (−λ).
Hypothesis 5: Aftermarket trading volume is lower when PGM is low and higher for high
values of PGM (when bookbuilding investors resell their shares to grey market investors in the
aftermarket).
Since the share price converges in the long run to the fundamental value, αsB + (1 − α)sG , we
can compare the long-run return to the grey market price and the aftermarket price. To do this
recall that when PGM is high the aftermarket price is PAM = P̂GM − λS = α̂sB + (1 − α̂)sG − λS.
Thus the difference between the long-run price and the short-run aftermarket price (i.e., the
long-run dollar return) is (α̂ − α)(sB − sG ) + λS. To the extent that α̂ > α, this return is
positively related to the bookbuilding signal and negatively related to the grey market signal.
The difference between PAM and PGM (for high PGM ) is (1 − α̂)(sB − M ), which is positively
related to the bookbuilding information. Thus, there is a positive relation between the long-run
return and (PAM − PGM ) if α̂ > α. If α̂ = α the two are unrelated.

14
 Finally, PGM equals α̂(sG ) + (1 − α̂)E(sB )) and so is related to the grey market information
sG . Thus, the long-run return should be negatively related to PGM as long as α̂ > α. Hypothesis
6 summarizes this discussion.
Hypothesis 6: When PGM is high, the long-run return (relative to PAM ) is negatively correlated
with PGM and positively correlated with the difference between PAM and PGM (to the extent
that grey market investors do not fully update for sB , i.e., if α̂ > α).
Intuitively, when PGM exceeds the fundamental value, the immediate aftermarket price (PAM )
is closely related to the grey market investors’ willingness to pay (P̂GM ), which differs from the
fundamental value if grey market investors overweight their own signal, i.e., if α̂ > α. In this
case, we should expect reversal of the share price towards the fundamental value in the long run.
In contrast, the difference between PAM and PGM captures grey market investors updating their
valuation when they learn the bookbuilding information sB . To the extent that they underweight
the bookbuilding information, the share price movement from PGM to PAM is only a partial
movement in the right direction and should continue in the same direction as the fundamental
value is revealed over time.

3. Sample and Data

The dataset consists of 486 European-listed IPOs completed between November 1995 and De-
cember 2002 for which we have grey market prices. Grey market prices are not available sys-
tematically, so our dataset is a subset of the universe of firms going public in Europe over the
sample period. Sample companies come from the following 20 countries: Austria (13), Belgium
(1), Canada (1), Denmark (1), Finland (3), France (13), Germany (321), Greece (2), Ireland (2),
Israel (7), Italy (61), Lithuania (1), Luxembourg (1), Netherlands (11), Norway (2), Spain (5),
Sweden (2), Switzerland (11), the United Kingdom (24), and the United States (4).
Grey markets are usually organized not by an exchange but by independent brokers who make
forward markets in IPO shares on a when-issued basis. Thus the structure of grey markets differs
across countries and even within countries depending on the broker. Brokers quote spreads and
investors can take a long or short position depending on their expectations. Usually, grey market
prices are public information: not only are they available from the broker, but the financial news
media often report grey market prices.
Grey market prices were obtained from two large brokers, based in Germany and the United

15
Kingdom, and supplemented with a news search. Information on the IPOs is derived from an
updated version of the dataset compiled by Ljungqvist and Wilhelm (2002), based on Dealogic’s
Equityware, Thomson Financial’s SDC, information from national exchanges, and a compre-
hensive news [Link] trading prices and trading volumes are from Datastream. We
convert monetary values – such as gross proceeds – into U.S. dollars using exchange rates on the
first day of trading.
Table 1 shows descriptive statistics for the sample as a whole as well as broken down by the
twelve countries on whose exchanges sample companies list. Most sample firms (75 percent) list
in Germany, 54 companies list in more than one country (usually the home country plus Frankfurt
or London), and 43 companies do not list in their home country at all (including non-European
issuers from Israel, the U.S., or Eastern Europe).
Although the sample IPOs span the period from November 1995 to December 2002, the range
of dates varies from market to market, depending on which IPOs we have grey market prices for.
In the UK, for instance, we have grey market prices for firms going public between June 1997
and July 2002. To allow the reader to assess how comprehensive our sample is, Table 1 reports
the number of IPOs in each market during the entire period, as well as during the sub-periods
for which we have IPOs with grey market prices for each country.
Over our sample period, Germany and Italy have the most active grey markets, while London-
based brokers frequently make grey markets in IPOs taking place in other countries. Except in
Germany, grey market trading is more common in larger IPOs. Reflecting the fact that many
of our sample IPOs were completed in the late 1990s, the initial return (PAM /PI − 1) averages
36.3%. Bid-ask spreads in the grey market are quite wide, with quoted half spreads averaging
4.7%. A majority of IPOs (54.1%) are priced at the high end of the initial indicative price range.
On average, the last grey market price before the issue price is finalized exceeds the price range
midpoint by 40.4%.

4. Empirical Results

In this section we discuss the empirical results in light of our predictions.

4.1. The Offer Price. Hypothesis 1 discusses the relation between the issue price and the grey
market price. We normalize these prices by the midpoint of the initial indicative price range in
order to reduce the impact of differences in scale and of heteroskedasticity. The grey market price

16
that we use is the last reported transaction price before the issue price is set (or the midpoint
of the bid-ask spread when transaction prices are unavailable). This corresponds to PGM in the
model.
It is well-documented that issue prices in Europe are rarely set outside the initial indicative
price range; frequently they are set at the endpoints, especially at the top of the range (see
Ljungqvist, Jenkinson, and Wilhelm (2003)). Consequently, the observed distribution of issue
prices in our sample is censored at the range endpoints. To correct for this, we estimate censored
regressions (Amemiya (1973)). These are similar to Tobit models, except that the point of
censoring is observation-specific. Note that 54% of our observations are right-censored.
Our model distinguishes between the cases where the grey market price is higher or lower
than the fundamental value. Because fundamental value is unobservable to the econometrician,
empirical studies usually take the midpoint of the initial indicative price range as a proxy for the
underwriter’s ex ante prior of the fundamental value. Thus if the grey market price is above the
midpoint it is more likely to be above the fundamentals. To capture the predicted asymmetry,
we interact the grey market price with an indicator function that equals one if the grey market
price exceeds the range midpoint , and zero otherwise.13
Table 2 reports the results. Regression 1 examines the relation between the issue price and
the grey market price. Overall, the fit of the model is very good in view of the highly significant
likelihood ratio test. As expected, we find a very significant relation between the issue price and
the grey market price, and an even stronger relation when the grey market price is above the
range midpoint. This result is consistent with Hypothesis 1: the grey market price is positively
correlated with the issue price. The fact that the correlation is positive even when the grey
market price is low suggests that α > 0: the grey market price contains information about the
fundamental value. The stronger relation for high levels of the grey market price reflects an
asymmetry in the relation between the grey market price and the issue price. This is consistent
with the asymmetry in the model when α̂ > α: the underwriter bases the issue price on the grey
market price when it is high, but when the grey market price is low, the underwriter uses it only
to the extent that it contains (partial) information about the fundamental value.

13
The large proportion of right-censored observations is the reason why to capture the asymmetry we introduce
the indicator function rather than splitting the sample between high and low levels of the grey market price, as
we do in later tables. If we were to estimate the censored regression model for the subsample where the grey
market price exceeds the range midpoint, we would have little explanatory power since for most observations the
issue price would equal the top of the range.

17
 In Regression 2 of Table 2 we also include the bid-ask spread quoted by grey market brokers
shortly before IPO pricing and (the logarithm of) the gross proceeds from the issue (S). A wider
bid-ask spread may indicate a lack of depth in the grey market. This may be due to a scarcity of
traders in the grey market or a diversity of opinion among investors. Either way, bookbuilding
investors may not be able to resell all their shares in the aftermarket at the (updated) grey
market price P̂GM , causing the underwriter to price the issue more conservatively. Similarly,
when the issue size is large, the issue price should reflect the greater difficulty of reselling the
shares in the aftermarket. In the model, this is captured by the discount λS and gives rise to
Hypothesis 4, which suggests the issue price should be negatively correlated with the issue size
and positively correlated with the depth of the market (−λ).
Consistent with Hypothesis 4, in Regression 2 we find negative and statistically significant
relations between the issue price PI and the bid-ask spread, and between PI and log proceeds S.
In Regression 3 we interact the bid-ask spread with log proceeds, since in the model the
discount λS is the product of the two terms. We find a negative and statistically significant
relation. Moreover, the relations in Hypothesis 4 refer to the case when the grey market price
is high, since only then do the bookbuilding investors resell their shares to the grey market
investors. To capture this asymmetry, Regression 4 includes the product of the bid-ask spread
and the proceeds times an indicator function that equals one when the grey market price is above
the range midpoint. Both the resulting coefficients are negative, but not statistically significant.
Finally, a wide bid-ask spread reflects greater divergence of opinion among grey market in-
vestors, possibly indicating a less reliable grey market signal (or a smaller α in the model).
Hypothesis 3 predicts that the correlation between the issue price and the grey market price is
weaker when the grey market signal is less reliable. In Regression 5, we attempt to capture this
by interacting the bid-ask spread with the grey market price. We find that the coefficient of
the interaction term is indeed negative and statistically significant, suggesting that the positive
effect of the grey market price on IPO pricing is attenuated when the bid-ask spread is wider.

4.2. The Short-Run Aftermarket Price. The model suggests that when the grey market price
exceeds the fundamental value, the first-day closing price reflects the price grey market investors
are willing to pay for the shares (i.e., the grey market price adjusted for the information in the
book). In Table 3, Regression 1, we see that the aftermarket price (at the end of the first day of
trading) is indeed highly correlated with the grey market price. Since Table 2 demonstrated a

18
positive relation between the grey market price and the issue price, it is possible that Regression
1 simply captures the well-documented positive relation between issue prices and after-market
prices (Hanley (1993)): issue prices contain bookbuilding information and so affect aftermarket
prices. To investigate this further, Regression 2 relates the aftermarket price to the issue price.
As expected, we find that the aftermarket price is higher, the higher is the issue price. The
adjusted R2 , however, is much lower than in Regression 1. In Regression 3, we include both
the grey market price and the issue price. The coefficient estimated for the grey market price
remains highly significant but the issue price loses most of its explanatory power. In sum, high
grey market prices predict aftermarket prices independent of the level of the issue price.14
So far, our results might be interpreted as evidence that PGM is a good predictor of the
aftermarket price in general. However, our model implies that the grey market price should be
a better predictor when it is high. Hypothesis 2 suggests an asymmetry in the relation between
the grey market price and the first-day closing price. When the grey market price is high, the
aftermarket price reflects the (updated) valuation of grey market investors and we expect a close
correlation between the grey market price and the first-day closing price. In contrast, when
the grey market price is low, the aftermarket price depends on the fundamental value, so the
grey market price should only be related to the aftermarket price to the extent that it contains
information about the fundamental value.
At the same time, we also expect an asymmetry in the effect of the issue price on the
aftermarket price. Specifically, issue prices should have a relatively stronger effect when the
grey market price is low. When the grey market price is high, the bookbuilding information that
is incorporated in the issue price affects the aftermarket price only through the updating of the
grey market investors’ valuation. When the grey market price is low, information in the book
about the fundamental value is directly incorporated into the aftermarket price.
In Regressions 4 and 5 we capture the asymmetry by splitting the sample into two subsets.
We find a stronger relation between the aftermarket price and the grey market price when PGM
is high (coefficient of 0.95) than when it is low (0.57), consistent with Hypothesis 2. The fact
that the coefficient is larger when PGM is high implies that α̂ > α. The fact that the coefficient
is positive and significant even when PGM is low is consistent with α > 0. As for the issue price,

14
Note that while the issue price depends on the grey market price, and the aftermarket price depends on the
issue price and the grey market price, the system described by these two equations is triangular. Thus it can be
consistently estimated using recursive, equation-by-equation estimation. See Greene (2002), p. 383.

19
the coefficient is the same in both cases but it is only statistically significant (at 5%) when PGM
is low, as expected.
According to Hypothesis 4, when PGM is high, a wide bid-ask spread or a large issue size
reduces bookbuilding investors’ ability to sell their shares in the aftermarket at the grey market
price. The negative coefficients for these variables in Regression 4 support this hypothesis (though
they are statistically significant only for issue size). When the grey market price is lower than
the range midpoint, on the other hand, the bid-ask spread and log proceeds have no significant
effect on the aftermarket price (Regression 5).

4.3. Aftermarket Volume. Table 4 examines the relation between the grey market price and
aftermarket trading volume (as a percentage of the shares sold in the IPO). Hypothesis 5 suggests
that the relation should be a step function. When PGM is high, we expect high turnover because
bookbuilding investors sell their shares to the grey market investors whose valuation exceeds the
fundamental value. When PGM is low, bookbuilding investors have no reason to sell their shares
in our model and so trading volume will be lower.
We measure aftermarket trading volume both on the first day and over the first week following
the IPO. To capture the step function, we use an indicator function that equals one when PGM
is above the initial price range midpoint, and zero otherwise. We find a positive and statistically
significant relation between volume and the indicator function, both for first-day volume (Re-
gression 1) and first-week volume (Regression 4). This suggests that when the grey market price
is high, bookbuilding investors are more likely to resell their shares in the aftermarket, consistent
with the model.
A high grey market price could simply indicate that either the IPO or the market is “hot,”
which may lead to high volume for reasons outside our model. In Regressions 2 and 5 we introduce
the market index return (measured from the pricing day until the end of the first day of trading)
to capture a hot market. In Regressions 3 and 6 we also include the (normalized) first-day closing
market price to capture whether the IPO is “hot.” Even after controlling for these variables,
the coefficient on the indicator function remains positive and significant. This implies that the
positive relation between volume and the indicator function is not simply due to a high level of
trading in “hot” IPOs or in active markets.

20
4.4. Long-Run Returns. We now consider how the grey market price and the results of book-
building are related to aftermarket returns in the long run.
Since the issue is priced above the fundamental value when the grey market price is high, we
expect the price to revert towards the fundamental value in the long run, i.e., we expect negative
long-run returns (Hypothesis 6). Moreover, Hypothesis 6 predicts that the information in the
aftermarket that comes from bookbuilding should show continuance in the aftermarket.
The long-run returns are studied in Table 5. The dependent variable is the return measured
from the end of the first aftermarket trading day until two, three, six or twelve months later,
less the return on the domestic stock market index. The independent variables are the difference
between the grey market price and the midpoint of the range (PAM − Pmid ), and the difference
between the first-day aftermarket price and the grey market price (PAM − PGM ). The dependent
and independent variables are all measured as percentages relative to the midpoint of the range
Pmid .
The intuition behind our specification is as follows. Using the range midpoint as the prior of
the value of the shares, the difference between the publicly observed grey market price and the
range midpoint reflects any information revealed through grey market trading. In the interval
between the final pre-IPO grey market price and the start of the aftermarket, bookbuilding is
concluded and the issue price is set. Therefore, the difference between the immediate first-day
aftermarket price and the final pre-IPO grey market price represents the change in valuation due
to the revelation of the bookbuilding information (assuming that no other information arrives in
the meantime). These two types of information add up to the total information contained in the
first-day aftermarket price.
Hypothesis 6 suggests that the long-run returns relate differently to the grey market infor-
mation and the bookbuilding information.
When the grey market price exceeds the fundamental value, the first-day aftermarket price
will be close to the grey market investors’ reservation price. To the extent that their reservation
price is unrelated to the fundamental value, the price should eventually revert to the fundamental
value, and we should observe a negative relation between long-run returns and the difference
between the grey market price and the range midpoint.
Bookbuilding information, by contrast, is assumed to be about fundamental value. If so, the
difference between the aftermarket price and the grey market price should not be reversed in the

21
long run, that is, it should not be negatively correlated with long-run returns.
The correlation between long-run returns and the difference between the aftermarket price and
the grey market price also sheds light on how grey market investors update using the bookbuilding
information. If they fully update after observing the information (α̂ = α), there should be
no correlation. If they instead exhibit conservatism (α̂ > α), then the movement from the
grey market price to the first-day aftermarket price is only a partial movement towards the
fundamental value. As they update further over time, the price continues to move towards the
fundamental value. Thus, we should observe a positive correlation between long-run returns and
the difference between the aftermarket price and the grey market price.
In Table 5, in addition to analyzing the full sample, we partition the sample based on whether
the grey market price is above or below the range midpoint. The relation between long-run
returns and the grey market information should hold only when the grey market price is high,
since only then does the first-day aftermarket price relate to the grey market investors’ reservation
value.
In the full sample, for all horizons, we find a statistically significant negative relation between
the grey market price and the long-run return. Furthermore, when we partition the sample into
subsamples we find that the negative relation only holds when the grey market price is high.
α
This suggests that α̂ > α, because the coefficient corresponds to α̂
− 1 when the grey market
price is high. Thus, non-fundamental information is transmitted from the grey market to the
aftermarket and is reversed in the long run. Moreover, depending on the horizon, the coefficients
range from -0.24 to -0.77 (which are all significantly greater than -1), indicating that only part
of the price difference between the grey market and range midpoint is reversed. This can be
interpreted as evidence that the grey market contains some information about the fundamental
value that is not reversed (i.e., α > 0).
In contrast, the percentage difference between the first-day aftermarket price and the grey
market price is not negatively related to the subsequent long-run return. As argued above, this
is consistent with the hypothesis that the information in the book pertains to the fundamental
value and is not reversed in the long run. The coefficient is positive when the grey market
price is high, consistent with investors updating only gradually over time, but it is mostly not
statistically significant.

22
 5. Allowing Bookbuilding Investors to Trade in the Grey Market

In Section 2 we assumed that bookbuilding investors are not allowed to trade in the grey market.
In reality, bookbuilding investors are able to participate in the grey market, but the underwriter
actively discourages it (for example, by threatening to exclude them from future IPOs). Since the
underwriter cannot directly observe whether a bookbuilding investor trades in the grey market,
it is unclear how effective a prohibition would be. In this section we explicitly consider the
possibility that bookbuilding investors trade in the grey market, and we show that a bookbuilding
investor with a position in the grey market may have additional incentives not to report his signal
truthfully. To avoid this problem, the underwriter might prohibit the bookbuilding investors
from trading in the grey market. If he cannot enforce such a prohibition, he has to modify the
mechanism in a revenue-decreasing way.
For simplicity, we consider the case in which bookbuilding investors can buy or sell only a
limited number of shares, γ, in the grey market, so that they are price-takers. That is, their
trades do not affect the grey market price.
The details of the argument are given in Appendix 2. Here we sketch the main steps and
give the intuition of the results. We proceed as follows. We start by assuming the underwriter
is using the mechanism in Proposition 1 and look at whether allowing bookbuilding investors
to trade in the grey market affects the incentive compatibility constraints in the maximization
problem.
We solve the model backward: assuming that a bookbuilding investor has taken a position
(long or short) in the grey market, we look at how his incentive to truthfully report his signal
changes. Then, we look at the investor’s decision to buy or short-sell shares in the grey market,
given that he can anticipate the signal he will report during bookbuilding.
A bookbuilding investor who observed a signal H and has gone long in the grey market
will always reveal his signal truthfully since this can only have the effect of raising the resale
price in the aftermarket. However, if he has a short position in the grey market he may have
an incentive to lie and report a signal L. If the grey market price is sufficiently high (PGM >
αSG + (1 − α)M + λS), he will have an incentive to report a signal L in order to manipulate the
price downward and profit from his initial short position. But if PGM < αSG + (1 − α)M + λS,
the aftermarket price does not depend on the updated valuation of the grey market investors, so
he has no incentive to misreport his signal. An investor who observed a signal L will never be

23
induced to lie by a small position in the grey market.
Given these incentives, we look at whether a bookbuilding investor who observed H prefers
to take a long position in the grey market (and truthfully reveal his signal) or whether he prefers
to take a short position (and, when PGM is high, misreport his signal). We find that the choice
depends on the parameter λ.
If there is no lack of depth in the aftermarket (λ = 0), i.e., the bookbuilding investor knows
that there are sufficient investors to whom he can resell shares at P̂GM in the aftermarket, then
an investor who has observed a signal H will find it more profitable to buy shares than to short
sell them in the grey market. In such case, the investor will truthfully report his signal, so the
underwriter need not be concerned if bookbuilding investors trade in the grey market.
The only situation in which trading in the grey market by bookbuilding investors could
hurt the issuer arises when λ > 0, i.e., there is insufficient depth in the aftermarket. In this
case, for sufficiently high PGM , the bookbuilding investors take advantage of the discount in the
aftermarket (λS) by short selling in the grey market and covering this position at a lower price.
Therefore, although the existence of the grey market is beneficial for the issuer in that it
increases the expected IPO proceeds (as shown in Section 2), participation by bookbuilding
investors in the grey market may not be. This result rationalizes investment bankers’ efforts to
discourage bookbuilding investors from participating in the grey market, particularly when PGM
is high. But the underwriter should only be concerned about bookbuilding investors participating
in the grey market when they short sell and the aftermarket is expected to be thin.

6. Conclusions

We have taken advantage of the existence of a grey market for IPO shares to test whether
conservatism on behalf of a set of investors can explain anomalies in post-IPO prices. We argue
that when small investors are excessively optimistic, they will be willing to pay a price above the
fundamental value, resulting in a high aftermarket price. When they are pessimistic, and value
the shares below the fundamental value, they will be priced out of the aftermarket, in which case
there will be no bias in the aftermarket price.
We use the grey market price as an indication of these investors’ opinion, and predict an
asymmetric relation between the grey market price and the aftermarket price. Using grey market
price data for a set of European IPOs, we test this prediction and find that there is indeed an

24
asymmetric relation. We also find higher aftermarket trading volume when the grey market price
is high, consistent with the model.
Looking at the long run, we find that when the grey market price is high (suggesting that
smaller investors are optimistic), long-run returns are negative, while this pattern does not arise
when the grey market price is low. Moreover, there is evidence that the movement from the
grey market price to the aftermarket price shows continuance in the long run when the grey
market price is high, suggesting that smaller investors update only partially, i.e., they exhibit
conservatism.

25
 Appendix 1

Proof of Proposition 1
Step 1. We focus on symmetric equilibria, i.e., informed investors are treated symmetrically. For
a given sG define Ps̃B ≡ PI (sG , s̃B ) as the issue price set when the aggregate information revealed
through bookbuilding is s̃B . The underwriter chooses the mechanism that maximizes expected
proceeds:
h i
max S π 2 PH + 2π(1 − π)PM + (1 − π)2 PL (9)
Ps̃B ,q

subject to the individual rationality constraints

VsB (sB ) ≥ Ps̃B , (10)

the incentive compatibility constraint for informed investor i who observes a signal si = H,

π S−qu2(H,H) [VH (H) − PH ] + (1 − π) [S − qu (H, L) − q(L, H)] [VM (M ) − PM ]

(11)
≥ πq(L, H) [VH (M ) − PM ] + (1 − π) S−qu2(L,L) [VM (L) − PL ]

and the incentive compatibility constraint for informed investor i who observes a signal si = L,

πq(L, H) [VM (M ) − PM ] + (1 − π) S−qu2(L,L) [VL (L) − PL ]

≥ π S−qu2(H,H) [VM (H) − PH ] + (1 − π) [S − qu (H, L) − q(L, H)] [VL (M ) − PM ]
(12)

where the quantities are written so that the sum of the shares allocated to the two informed
investors and the uninformed investor equals S.
We proceed as follows: we first ignore the second incentive compatibility constraint (12)
and find the optimal solution. We then check that this constraint is in fact non-binding at the
optimum.
Step 2. Since the number of shares sold always equals S and the underwriter charges all investors
the same issue price, the quantity allocated to each investor does not directly affect proceeds.
Thus, we choose the quantities to relax the incentive compatibility constraint (11) as much as
possible since a slacker constraint will allow an optimum with a higher price.
This is achieved by setting qu (H, H) = qu (H, L) = 0; qu (L, L) = S(1 − k); q(H, H) =

26
S/2; q(H, L) = S; q(L, H) = 0; and q(L, L) = Sk/2. Substituting and dividing both sides by
S, the incentive compatibility constraint for type H (equation (11)) becomes

π k
[VH (H) − PH ] + (1 − π) [VM (M ) − PM ] ≥ (1 − π) [VM (L) − PL ] (13)
2 2

When this incentive compatibility constraint is binding, we can write it as

2(1 − π) 1−π
PH = VH (H) + [VM (M ) − PM ] − k [VM (L) − PL ] (14)
π π

If we substitute constraint (14) into the objective function, we obtain

h i
maxS π 2 VH (H) + 2π(1 − π)VM (M ) − π(1 − π)kVM (L) + π(1 − π)kPL + (1 − π)2 PL
Pi
(15)

subject to the individual rationality constraint (10).

The maximization function is increasing in PL , so it is optimal to set PL as high as possible
subject to the individual rationality constraint, i.e., PL = VL (L). The choice of PH and PM is
indeterminate, so we consider the example when PM = VM (M ). The incentive compatibility
1−π
constraint thus implies PH = VH (H) − π
k [VM (L) − VL (L)] .
Step 3. Since we derived the optimal mechanism ignoring constraint (12), we have to check
that this constraint is not violated at the optimum. First, note that if we substitute in (12) the
optimal quantities q and qu and issue price PI , the constraint can be rewritten as:

π
k [VM (L) − VL (L)] ≤ [VH (H) − VM (H)] + 2 [VM (M ) − VL (M )] (16)
1−π
π
where all terms in brackets are non-negative. From the assumption of equidistance k < 1−π
=1
it is enough to show that [VH (H) − VM (H)] ≥ [VM (L) − VL (L)] for the constraint to be satisfied.
To check whether this constraint is satisfied, we have to substitute the values of the functions
VsB (·). Depending on the value of PGM these functions take different values. In the next step
we determine the different values of these functions for different ranges of PGM . We can then
proceed to check that constraint (16) is satisfied for all these areas.
Step 4. The value of VsB (·) depends on whether it is equal to the fundamental or the resale value.
We first determine when VsB (·) is equal to the fundamental value.

27
 For any sB and s̃B , VsB (s̃B ) = αsG + (1 − α)sB (the fundamental value) if and only if

α̂
PGM ≤ (1 − α̂)M + [(1 − α)sB − (1 − α̂)s̃B + λS] ≡ B(sB , s̃B ) (17)
α̂ − α

Proof: Since PGM = EG (v | sG ) = α̂sG + (1 − α̂)E(sB ) we can derive that sG = α̂1 [PGM − (1 −
α̂)M ].From the definition of VsB (s̃B ) in (8), VsB (s̃B ) = αsG + (1 − α)sB if and only if

αsG + (1 − α)sB ≥ α̂sG + (1 − α̂)s̃B − λS

Substituting for sG and rearranging the terms gives (17).

Step 5. Let us distinguish three different areas:

1) PGM ≤ B(M, H).

2) PGM > B(M, L).

3) B(M, H) ≤ PGM < B(M, L).

where B(M, H) < B(M, L) since, from (17), B(sB , s̃B ) decreases in s̃B .
In the first area, from equation (17), VM (L) = VM (H) = FM and VH (H) = FH , where
FsB ≡ αsG + (1 − α)sB (i.e., the fundamental value). Thus the constraint is satisfied if (FM −
VL (L)) ≤ (FH − FM ). Since VL (L) ≥ FL it is sufficient to show that (FM − FL ) ≤ (FH − FM ).
Substituting for FsB we obtain

(1 − α)(M − L) ≤ (1 − α)(H − M )

Since M − L = H − M, it is satisfied with an equality.

In the second area, VM (L) = VL (L) = P̂GM (L) − λS. Thus the left-hand side equals 0. Since
the right-hand side is always non-negative, the constraint is satisfied.
In the third area, VM (L) = FM , VH (H) = FH , VL (L) = P̂GM (L)−λS and VM (H) = P̂GM (H)−
λS. Thus the constraint is satisfied if

[FM − P̂GM (L) + λS] ≤ [FH − P̂GM (H) + λS].

28
 1+α
Substituting for FM , P̂GM (L), and P̂GM (H), and using the assumption α̂ ≥ 2
, the constraint
is satisfied.

The optimal mechanism as a function of the grey market price

The expressions in Proposition 1 are written in terms of the functions VsB (s̃B ), which depend
on PGM . Therefore the mechanism looks different over different ranges of PGM . To see how the
mechanism looks for different ranges of values of PGM it is enough to substitute the actual values
of the functions VsB (s̃B ).
Since for a given PGM we can always compute both the fundamental value and the resale
value using equations (4) and (5), we can redefine the fundamental value and the resale value as
functions of PGM . The fundamental value is defined as FsB (PGM ) ≡ E(v | sG , sB ) and the resale
value is defined as P̂GM (PGM , s̃B ) − λS ≡ P̂GM (sG , s̃B ) − λS.
We distinguish five different intervals of PGM . The boundaries of these intervals are the values
v̄ > v1 > v2 > v3 > v4 > 0, where

• v1 is defined so that P̂GM (v1 , H) − λS = FH (v1 ). In other words, if the grey market price
is PGM = v1 , and the bookbuilding information is H, the fundamental value exactly equals
the resale value of the shares;

• v2 is defined so that P̂GM (v2 , L) − λS = FM (v2 );

• v3 is defined so that P̂GM (v3 , M ) − λS = FM (v3 ); and

• v4 is defined so that P̂GM (v4 , L) − λS = FL (v4 ).

The issue price in each interval is as follows:

1. If PGM ∈ [v1 , v̄] then




 if s̃B = H, PI = P̂GM (PGM , H) − λS,


if s̃ = M , PI = P̂GM (PGM , M ) − λS,
B

 if s̃ = L, PI = P̂GM (PGM , L) − λS.

B

29
2. If PGM ∈ [v2 , v1 ] then



 if s̃B = H, PI = FH (PGM ),


if s̃ = M , PI = P̂GM (PGM , M ) − λS,
B

 if s̃ = L, PI = P̂GM (PGM , L) − λS.

B

3. If PGM ∈ [v3 , v2 ] then


1−π


 if s̃B = H, PI = FH (PGM ) − π
k[FM (PGM ) − P̂GM (PGM , L) + λS],


if s̃ = M , PI = P̂GM (PGM , M ) − λS,
B

 if s̃ = L, PI = P̂GM (PGM , L) − λS.

B

4. If PGM ∈ [v4 , v3 ] then


1−π


 if s̃B = H, PI = FH (PGM ) − π
k[FM (PGM ) − P̂GM (PGM , L) + λS],


if s̃ = M , PI = FM (PGM ),
B

 if s̃ = L, PI = P̂GM (PGM , L) − λS.

B

5. If PGM ∈ [0, v4 ] then


1−π


 if s̃B = H, PI = FH (PGM ) − π
k[FM (PGM ) − FL (PGM )],


if s̃ = M , PI = FM (PGM ),
B

 if s̃ = L,

PI = FL (PGM ).
B

30
 Appendix 2: Allowing bookbuilding investors to trade in the grey market

In this Appendix we show when trading by bookbuilding investors in the grey market can affect
the optimal mechanism and reduce expected IPO proceeds.
We proceed as follows. We assume that the underwriter is using the mechanism of Proposition
1 and check whether the incentive compatibility constraints are still satisfied.
We solve the model backward: assuming that a bookbuilding investor has taken a position in
the grey market, we look at how his incentive to truthfully report his signal changes. Then, we
look at the investor’s decision to buy or short-sell shares in the grey market, given that he can
anticipate the signal he will report during bookbuilding.
Consider an informed investor who has observed a signal si = H. We only need to look at
the incentive compatibility constraint of this investor, since we showed in Appendix 1 that only
this constraint is binding at the optimum.
If an investor who has observed a signal si = H has no position at all in the grey market, his
incentive compatibility constraint would be as in (12). However, if he has a position in the grey
market, the incentive compatibility constraint would change.
If the investor has bought γ shares in the grey market, the constraint becomes

π S2 [VH (H) − PH ] + (1 − π)S [VM (M ) − PM ] + γ [πVH (H) + (1 − π)VM (M ) − PGM ]

≥ (1 − π) Sk
2
[VM (L) − PL ] + γ [πVH (M ) + (1 − π)VM (L) − PGM ]
(18)

The last term on the left-hand side represents the expected profit from buying γ shares in the
grey market if he truthfully declares H. The last term on the right-hand side represents the
expected profit from the grey market position if he falsely declares L.
Since the incentive compatibility constraint (12) was satisfied with equality in the mechanism
of Proposition 1, constraint (18) is satisfied if and only if

γ [πVH (H) + (1 − π)VM (M ) − PGM ] ≥ γ [πVH (M ) + (1 − π)VM (L) − PGM ] .

Since VH (H) ≥ VH (M ) and VM (M ) ≥ VM (L) the inequality is always satisfied. Thus, an

investor who observed a signal H and bought γ shares in the grey market will still tell the truth
and declare H.

31
 If instead the bookbuilding investor has a short position of γ shares in the grey market, his
incentive compatibility constraint is satisfied if and only if

γ [PGM − πVH (H) − (1 − π)VM (M )] ≥ γ [PGM − πVH (M ) − (1 − π)VM (L)] ,

where the left-hand side is the expected profit from the short position if he truthfully announces
his signal H, and the right-hand side is the profit if he misreports his signal as L. In both cases,
the profits are given by the difference between the grey market price and the expected value to
the investor of the shares he must deliver to cover the position. Since VH (H) ≥ VH (M ) and
VM (M ) ≥ VM (L) this inequality is always weakly violated. If PGM ≤ FM + λS (corresponding
to PGM ≤ v3 , where v3 is defined in Section 2.4), it is satisfied with equality, so the investor does
not lie. However, if PGM > FM + λS the inequality is violated and he misreports his signal.
To summarize, when a bookbuilding investor with signal H is long in the grey market, he
will have greater incentive to truthfully reveal his signal in order to raise the aftermarket price.
If he is short in the grey market and the grey market price is sufficiently high, he will have an
incentive to report a signal L in order to manipulate the price downward.
We now look at whether this bookbuilding investor prefers to take a long position in the
grey market (and truthfully reveal his signal) or whether he prefers to take a short position
(and misreport his signal when PGM > FM + λS). Since the original constraint (12) is satisfied
with equality, we only have to compare the additional profits from the respective grey market
positions. The investor will prefer to buy shares in the grey market (and report truthfully) if
and only if

γ [πVH (H) + (1 − π)VM (M ) − PGM ] ≥ γ [PGM − πVH (M ) − (1 − π)VM (L)] (19)

In order to see when this inequality is satisfied, start by considering the special case of λ = 0.
By substituting for the functions VsB (s̃B ) one can verify that the inequality is always satisfied.
Thus, an investor who has observed a signal H will buy shares in the grey market and declare
the truth.
However, when λ > 0 the inequality may be violated. In particular, when PGM is very high
(i.e. PGM ≥ FH + λS), any λ > 0 results in a violation of the inequality. For intermediate values
of PGM (FM + λS < PGM < FH + λS), the inequality will be violated for sufficiently large λ. For

32
lower levels of PGM (PGM < FM + λS) the inequality is irrelevant since, as shown above, even
with a short position the investor has no incentive to misreport his signal.
Thus, when PGM is high and λ > 0, allowing bookbuilding investors to trade in the grey
market results in lower IPO proceeds.

33
 References
Aggarwal, R., L. Krigman, and K. Womack, 2002, “Strategic IPO underpricing, information
momentum, and lockup expiration selling,” Journal of Financial Economics 66, 105-137.
Amemiya, T., 1973, “Regression analysis when the dependent variable is truncated Normal,”
Econometrica 4, 997-1016.
Aussenegg, W., P. Pichler, and A. Stomper, 2002, “Pricing with bookbuilding and a when-issued
market,” mimeo, Boston College.
Barberis, N., A. Shleifer, and R. Vishny, 1998, “A model of investor sentiment,” Journal of
Financial Economics 49, 307-343.
Ben Dor, A., 2003, “The performance of initial public offerings and the cross section of institu-
tional ownership,” mimeo, Northwestern University.
Bennouri, M., and S. Falconieri, 2003, “Price versus quantity discrimination in optimal IPOs,”
mimeo, Tilburg University.
Benveniste, L.M., and P.A. Spindt, 1989, “How investment bankers determine the offer price and
allocation of new issues,” Journal of Financial Economics 24, 343-61.
Benveniste, L.M., and W.J. Wilhelm, 1990, “A comparative analysis of IPO proceeds under
alternative regulatory environments,” Journal of Financial Economics 28, 173-207.
Biais, B., and A.M. Faugeron-Crouzet, 2002, “IPO auctions: English, Dutch, ...French, and
Internet,” Journal of Financial Intermediation 11, 9-36.
Busaba, W., and C. Chang, 2002, “Bookbuilding vs. fixed-price revisited: The effect of after-
market trading,” mimeo, University of Western Ontario.
Cornelli, F., and D. Goldreich, 2001, “Bookbuilding and strategic allocation,” Journal of Finance
56, 2337-70.
Cornelli, F., and D. Goldreich, 2003, “Bookbuilding: How informative is the order book?” Jour-
nal of Finance 58, 1415-44.
Dorn, D., 2002, “Does sentiment drive the retail demand for IPOs?” mimeo, Columbia University.
Greene, W.H. , 2002, Econometric Analysis, Prentice Hall.
Hanley, Kathleen W., 1993, “The underpricing of initial public offerings and the partial adjust-
ment phenomenon,” Journal of Financial Economics 34, 213-250.

34
Harris, M., and A. Raviv, 1993, “Differences of opinion make a horse race,” Review of Financial
Studies 6, 473-506.
Ljungqvist, A., T.J. Jenkinson, and W.J. Wilhelm, 2003, “Global integration in primary equity
markets: The role of U.S. banks and U.S. investors,” Review of Financial Studies 16, 63-99.
Ljungqvist, A., and W.J. Wilhelm, 2002, “IPO allocations: Discriminatory or discretionary?”
Journal of Financial Economics 65, 167-201.
Ljungqvist, A., V. Nanda, and R. Singh, 2002, “Hot markets, investor sentiment, and IPO
pricing,” mimeo, New York University.
Löffler, G., P. Panther, and E. Theissen, 2002, “Who knows what when? The information content
of pre-IPO market prices,” mimeo, University of Bonn.
Purnanandam, A., and B. Swaminathan, 2003, “Are IPOs really underpriced?” Review of Fi-
nancial Studies, forthcoming.
Ofek, E., and M. Richardson, 2003, “Dot-com mania: The rise and fall of internet stock prices,”
Journal of Finance 58, 1113-1137.
Rajan, R. and H. Servaes, 2003, “The effect of market conditions on initial public offerings,” in J.
McCahery and L. Renneboog, eds., Venture Capital Contracting and the Valuation of High-tech
Firms, Oxford University Press.
Ritter, J., and I. Welch, 2002, “A review of IPO activity, pricing, and allocation,” Journal of
Finance 57, 1795-1828.
Sherman, A.E., 2000, “IPOs and long-term relationships: An advantage of book building,”
Review of Financial Studies 13, 697-714.

35
 V

PI = PˆGM ( PGM , sB ) − λ S

FH ≡ αsG + (1–α)H

v1

if sB = H PI = FH
otherwise
PI = PˆGM ( PGM , sB ) − λ S

v2

if sB = H PI = FH – underpricing
otherwise PI = PˆGM ( PGM , sB ) − λ S
PGM

v3
FM ≡ αsG + (1–α)M

if sB = H PI = FH – underpricing
if sB = M PI = FM
if sB = L PI = PˆGM ( PGM , L ) − λ S

v4

if sB = H PI = FH – underpricing
if sB = M PI = FM
if sB = L PI = FL
FL ≡ αsG + (1–α)L

0
Figure 1. The issue price in the optimal mechanism. The figure illustrates the issue price (for
sB = H, M, L), as derived in Appendix 1, over the possible ranges of the grey market price. The
boundaries v2, v2, v3, v4 are defined in Appendix 1. FsB is the fundamental value when the
bookbuilding signal is sB (given PGM). The figure is drawn for small positive values of λS. (When
λ = 0, FM = v3.)
 Table 1: Descriptive Statistics of IPOs with Available Grey Market Prices
We have grey market prices for 486 (mostly European) IPOs completed between November 1995 and December 2002. Sample companies are incorporated in the
following 20 countries: Austria (13), Belgium (1), Canada (1), Denmark (1), Finland (3), France (13), Germany (321), Greece (2), Ireland (2), Israel (7), Italy
(61), Lithuania (1), Luxembourg (1), Netherlands (11), Norway (2), Spain (5), Sweden (2), Switzerland (11), the United Kingdom (24), and United States (4).
Note that there is no grey market in the U.S.; the four American companies are in the sample because they go public in Europe. Most companies go public in
their home country, but some do not. Where a company goes public on more than one exchange, we take the listing country to be its home country or (if it
doesn’t list on a home-country exchange) the country in which most of the shares are placed. The table shows descriptive statistics for the sample as a whole as
well as broken down by the twelve countries on whose exchanges sample companies list. We also show, for each listing country, the first and last date for which
we have an IPO with grey market prices. This sample window varies from country to country. The sample for which we have grey market prices is a subsample
of the 2,723 IPOs completed in the twelve listing countries shown between November 1995 and December 2002. Gross proceeds are shares sold (including the
overallotment option if exercised) times the issue price, converted into U.S. dollars using exchange rates on the first trading day. Initial returns are computed
using the closing price on the first trading day. Quoted half spread refers to the quoted bid-ask spread in the grey market, just before the IPO issue price is set. It
is computed as half the difference between the bid and the ask divided by the midpoint of the spread.

Grey market
Sample window No. of IPOs… Gross proceeds ($m) Initial return (%) Fraction price rel. to
w/ grey during in Nov Quoted priced at midpoint of
market sample ‘95-Dec half spread high end price range,
first date last date prices window 2002 mean median mean [Link]. (%), mean of range mean (%)

Total Nov 1995 Dec 2002 486 1,755 2,723 343.7 53.0 36.3 65.6 4.7 54.1 40.4

By country of listing
Austria Nov 1997 Nov 2000 2 18 25 654.5 654.5 -2.5 6.6 6.9 0.0 5.1
Finland Nov 1998 Dec 1999 3 22 58 686.7 531.0 74.6 91.9 1.6 33.3 53.0
France Oct 1997 Dec 2001 14 409 544 1715.3 650.0 6.5 12.9 4.2 42.9 24.6
Germany Nov 1995 Jul 2002 363 489 504 169.9 42.2 41.5 67.8 5.1 63.4 46.6
Greece Oct 2000 Dec 2001 2 23 180 423.4 423.4 -4.6 0.6 0.0 -7.3
Italy Nov 1995 Dec 2002 61 132 133 599.0 106.1 20.2 63.4 4.3 27.9 13.1
Netherlands Mar 2000 Mar 2000 1 1 56 2829.0 2829.0 0.5 100.0 117.9
Norway Mar 2000 May 2000 2 5 107 139.7 139.7 29.5 44.5 0.6 0.0 85.7
Spain Jun 1999 May 2001 5 12 38 1374.0 915.7 10.5 13.2 5.1 20.0 11.8
Sweden Jun 2000 Jun 2001 2 22 196 4405.4 4405.4 9.9 8.2 0.9 50.0 19.2
Switzerland Dec 1996 Dec 2001 8 59 67 1097.4 153.8 50.1 99.8 1.7 12.5 36.0
United Kingdom Jun 1997 Jul 2002 23 563 815 566.8 265.3 21.5 35.0 1.5 21.7 32.9
 Table 2: Determinants of the Issue Price
The dependent variable in these regressions is the IPO issue price normalized by the midpoint of the initial
price range. The explanatory variables are the last grey market price before the issue price was set (also
normalized by the midpoint of the initial price range), the last bid-ask spread in the grey market (divided by
its midpoint), and the logarithm of the IPO proceeds. To capture the predicted asymmetry, we define an
indicator function set to one when the grey market price is above the midpoint of the initial indicative price
range. Grey market prices are available for 486 IPOs. Nine of these are fixed-price offerings, so we lack
information on their initial price ranges. This reduces the number of observations in model (1) to 477.
Models (2) through (5) include the bid-ask spread, which is available for 442 IPOs. We use censored
regressions because European IPOs are rarely priced outside the initial price range. t-statistics are reported
in parentheses. Three and two asterisks indicate significance at the 1% and 5% level, respectively.

Dependent variable: Issue price / midpoint of initial price range

(1) (2) (3) (4) (5)

Grey Market Price 0.30*** 0.32*** 0.32*** 0.32*** 0.38***

(6.22) (5.77) (5.81) (5.60) (6.50)

Grey Market Price 0.15*** 0.14*** 0.15*** 0.17*** 0.14***

x Indicator(GM Price > midpoint) (7.41) (6.64) (6.66) (5.65) (6.45)

Grey Market Bid-Ask Spread -0.39***

(-3.45)

Grey Market Bid-Ask Spread -0.40***

x Grey Market Price (-3.76)

Log gross proceeds -0.01** -0.01**

(-2.01) (-2.08)

GM Bid-Ask Spread x Proceeds -0.08*** -0.05

(-3.24) (-1.31)

GM Bid-Ask Spread x Proceeds -0.06

x Indicator(GM Price > midpoint) (-1.13)

LR test: all coeff. = 0 (χ2) 483.1*** 446.8*** 444.8*** 446.1*** 449.0***

No. of observations 477 442 442 442 442
No. of left-censored observations 51 50 50 50 50
No. of right-censored observations 263 246 246 246 246
 Table 3: Determinants of the First-Day Aftermarket Price
The dependent variable in these regressions is the stock price at the end of the first day of aftermarket
trading (normalized by the midpoint of the range) adjusted for the market index return from the pricing date
to the end of the first day of aftermarket trading: Pam/Pmid – (1 + market index return). The explanatory
variables are the last grey market price before the issue price was set (also normalized by the midpoint of
the initial price range), the (normalized) issue price, the last bid-ask spread in the grey market (divided by
its midpoint), and the logarithm of the IPO proceeds. White heteroskedasticity consistent t-statistics are
given in parentheses. Three, two, and one asterisks indicate significance at the 1%, 5%, and 10% level,
respectively.

Grey market > Grey market ≤

range midpoint range midpoint
(1) (2) (3) (4) (5)

Grey Market Price 0.99*** 0.95*** 0.95*** 0.57***

(15.93) (13.55) (11.89) (3.76)
Issue Price 3.21*** 0.44** 0.51* 0.51***
(13.50) (2.44) (1.66) (3.07)
Grey Market Bid-ask Spread -0.62 0.65 -0.46 -0.73 -0.06
(-1.52) (1.25) (-1.09) (-1.12) (-0.43)
Log proceeds -0.04*** -0.05*** -0.04*** -0.05*** -0.005
(-3.07) (-3.17) (-3.12) (-3.03) (-1.18)

Adjusted R2 75.4 % 20.8 % 75.7 % 70.7 % 67.8%

F-test: all coeff. = 0 98.0*** 63.7*** 205.4*** 79.8*** 69.2***
No. of observations 442 442 442 330 112
 Table 4: Aftermarket Volume
The dependent variable in these regressions is the natural logarithm of first-day volume (as a percentage of
the shares sold in the IPO), measured over the first day and first week of aftermarket trading. The main
explanatory variable is an indicator function set to one when the last grey market price before the issue
price was set exceeded the midpoint of the initial price range. The controls in models (2)-(3) and (5)-(6) are
the domestic market returns over the bookbuilding period and the (normalized) first-day after-market price.
White heteroskedasticity consistent t-statistics are given in parentheses. Three, two, and one asterisks
indicate significance at the 1%, 5%, and 10% level, respectively.

Log first-day volume Log first-week volume

(1) (2) (3) (4) (5) (6)

Indicator(GM Price > midpoint) 1.08*** 1.07*** 0.78*** 1.00*** 0.99*** 0.71***
(7.64) (7.50) (5.06) (8.72) (8.60) (5.67)

Market Returns (pricing to first-day) 7.54*** 5.86* 7.36*** 5.73**

(2.11) (1.78) (2.50) (2.15)

First-day After-market Price 0.40*** 0.38***

(6.35) (6.79)

Adjusted R2 12.0 % 12.9 % 16.7 % 14.2 % 15.4 % 20.3 %

F-test: all coeff. = 0 58.3*** 31.3*** 40.8*** 76.1*** 40.8*** 50.7***
No. of observations 443 443 443 443 443 443
 Table 5. From Initial Price Range to Long-Run Prices
The dependent variables are market-adjusted long-run returns, measured relative to the first day of trading as (PLR – PAM)/Pmid less the contemporaneous return on
the domestic market index. PLR is the price after the first two, three, six and 12 months of trading. White heteroskedasticity consistent t-statistics are given in
parentheses. Three and two asterisks indicate significance at the 1% and 5% level, respectively.

Full sample Grey market > range midpoint Grey market ≤ range midpoint
Horizon: 42 days 63 days 126 days 252 days 42 days 63 days 126 days 252 days 42 days 63 days 126 days 252 days
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 912)

GM Price – Range Midpoint -0.20*** -0.18** -0.38*** -0.66*** -0.25*** -0.24** -0.47*** -0.77*** 0.42 1.03** 1.92** 1.02**
Range Midpoint (-3.22) (-1.97) (-3.37) (7.27) (-3.40) (-2.17) (-3.51) (-6.95) (1.41) (1.97) (2.16) (2.01)

Aftermarket Price – GM Price 0.13 0.76** 0.84 0.05 0.10 0.73** 0.81 0.03 1.25 2.20 1.32 -0.80
Range Midpoint (1.04) (2.41) (1.51) (0.15) (0.79) (2.21) (1.41) (0.09) (1.30) (1.58) (0.89) (-0.99)

Adjusted R2 2.3 % 6.6 % 5.9 % 4.6 % 2.9 % 7.1 % 7.1 % 5.3 % 6.0 % 8.1 % 2.5 % 0.8 %
F-test: all coeff. = 0 7.3*** 9.9*** 16.9*** 45.3*** 7.7*** 10.4*** 18.5*** 41.0*** 1.1 2.0 2.3 3.1
No. of observations 446 446 446 446 330 330 330 330 116 116 116 116