Journal of Financial Intermediation 11, 9–36 (2002)
doi:10.1006/jfin.2001.0319, available online at http://www.idealibrary.com on
IPO Auctions: English, Dutch, . . . French, and Internet∗
Bruno Biais
Université de Toulouse, IDEI, Place Anatole France, 31200 Toulouse, France
and
Anne Marie Faugeron-Crouzet
Université de Paris Val de Marne, 61, avenue du Général de Gaulle, 94010 Créteil Cedex, France
Received December 6, 1999
Unseasoned shares are sold through the Book Building process in the United States and
the United Kingdom, fixed price offerings in several countries, uniform price auctions in
Israel or the new internet-based Open IPO mechanism, and an auction-like mechanism called
the Mise en Vente in France. We analyze and compare the performance of these various
IPO mechanisms within the context of a unified theoretical model. Fixed price offerings
lead to inefficient pricing and winner’s curse. Dutch auctions can also lead to inefficiencies,
to the extent that they are conducive to tacit collusion by investors. The Book Building and
Mise en Vente can lead to optimal information elicitation and price discovery. We document
empirically the similarity between the Book Building and the Mise en Vente. We discuss
the implications of our analysis for the design of optimal Internet IPO auctions. Journal of
C 2002 Elsevier Science (USA)
Economic Literature Classification Numbers: G24, G3, D82. ∗ Many thanks to Tom Chemmanur and the two referees for helpful comments, and to the Paris Bourse
for financial support, access to data, and many insights, especially Jacky Billard, Martine Charbonnier,
Didier Davydoff, Dominique Leblanc, Bernard Mirat, Alain Morice, and Pascal Samaran. We are also
grateful for helpful discussions with Michael Fishman, Jean–Pierre Florens, Julian Franks, Sylvain
Friedrich, David Goldreich, Richard Green, Michel Habib, Shmuel Kandel, Kjell Nyborg, Pegaret
Pichler, Jean Charles Rochet, Chester Spatt, Bill Wilhelm, Luigi Zingales, and seminar participants at
the University of Chicago, the London Business School, the London School of Economics conference
on trading markets for smaller companies, the Utah Winter Conference, the CEPR Summer Symposium
on Financial Economics, the conference “Raising Capital in Different National Markets” in Frankfurt,
the 2000 Meetings of the American Finance Association and the Journal of Financial Intermediation
Conference on New Technologies, Financial Innovation and Intermediation at Boston College.
9
1042-9573/02 $35.00
c 2002 Elsevier Science (USA)
All rights reserved.
10
BIAIS AND FAUGERON-CROUZET
1. INTRODUCTION
One of the major benefits of public listing on stock exchanges is the rich flow
of information, about the firm’s prospects and the investors’ willingness to hold
its shares, that is reflected in prices and trades in the secondary market. Prior to
the IPO very little such information is available. Consequently, the task of finding
the IPO price is a difficult one. The auction by which the unseasoned shares are
sold can therefore play an important role in eliciting information from the market
participants about their valuation of the stock. Yet IPO auction mechanisms vary
quite significantly across countries, and their ability to elicit information revelation
from investors is likely to also vary (see Loughran et al. (1994) and Sherman
(1999)):
• In the United States and the United Kingdom, in the Book Building method,
the investment banker elicits indications of interest from institutional investors and
uses these indications to set the IPO price and allocate the shares (see Benveniste
and Spindt (1989), Benveniste and Wilhelm (1990 and 1997), Spatt and Srivastava
(1991), Hanley (1993), Hanley and Wilhelm (1993), and Cornelli and Goldreich
(1998)).
• In Singapore, Finland, and the United Kingdom, fixed price offers are used,
whereby investors submit demands at the fixed price, and (possibly random) rationing rules are used to allocate the shares (see Koh and Walter (1989), Levis
(1990), and Keloharju (1993)).
• In Israel, IPOs are conducted according to the standard uniform price, market
clearing, Dutch auction, whereby the price is set to equate supply and demand (see
Kandel et al. (1997)).
• In the Mise en Vente, an auction-like IPO method commonly used in France,
investors submit limit orders and then the auctioneer sets the price as a function of aggregate demand (the name of this auction procedure has been recently
changed to Offre a Prix Minimum, but its workings have not been altered). In
this IPO mechanism the price does not clear the market, and prorata rationing
is used. For descriptions and analyses of this IPO method see Belletante and
Palliard (1993), Derrien and Womack (1999), Dubois (1989), Husson and
Jacquillat (1989), Jacquillat and Mac Donald (1974), Jacquillat et al. (1978),
Leleux (1993), and Mirat (1983, 1984).
The diversity of unseasoned shares selling mechanisms has actually even increased in the recent past as new, internet-based IPO auctions have been recently
proposed (see Wilhelm (1999)):
• Open IPO offers to sell unseasoned shares through a Dutch auction. In its
advertising (for example on its website: www.openipo.com) it emphasizes that
the use of this standard uniform price, market clearing auction ensures that “IPO
offfering prices are set by the market” and reflect “what people are truly willing to
pay for the stock.” It also advises that “investors should make a bid at the maximum
11
IPO AUCTIONS
TABLE 1
IPO Selling Mechanisms in Different Countries
U.K. U.S. Wit capital
Singapore Finland U.K.
Israel Open IPO
France
Dutch (uniform
price) auction
after
Mise en
Vente
after
Institution
Book building
Price set before
of after
demand
Price clears
market
Rationing rule
after
Fixed price
auction
before
no
no
yes
no
discretionary
prorata or
random
—
prorata
price at which they are comfortable owning shares of the issue.” In fact, Open IPO
has sometimes been presented as an alternative for the standard Book Building
process, potentially more efficient than the latter, thanks to the use of the Dutch
auction mechanism.
• Wit Capital follows a somewhat more standard strategy. It seeks to
entice individual investors to bid in IPOs, but does not contribute in a major way
to the price discovery mechanism. The latter is in large part operated by the
lead managers of the IPOs in which Wit Capital participates. These lead
managers, which are major investment banks such as Goldman Sachs, Morgan
Stanley, or Merrill Lynch, determine the IPO price based on the traditional
Book Building process, which is more focused toward large institutional
investors.
The overall diversity of IPO auction formats (summarized in Table 1) raises the
following issues:
• What is the optimal IPO mechanism in terms of expected proceeds, price
discovery, and information elicitation from informed investors? How should IPO
prices be set, in response to investors’ demand? Is the frequently used fixed price
method optimal?
• Or is the a priori appealing Dutch auction the optimal selling mechanism?
• How do different auction-like IPO procedures, such as the Book Building, the
Mise en Vente, and Dutch auctions compare?
• How should Internet-based IPO auctions be designed?
These questions are of interest to finance and economics scholars aiming
to understand price formation and the workings of auctions. They are also very
relevant in practice for investors, shareholders, and investment bankers, whose
profits can be significantly affected by the nature and efficiency of the IPO
mechanisms.
12
BIAIS AND FAUGERON-CROUZET
To address these issues, this paper develops a unified theoretical model to analyze
the workings of the different IPO mechanisms listed above: the Book Building
method, the fixed price auction, the Mise en Vente, and the uniform price market
clearing auction.
The participants in the IPO process are the sellers, financial intermediaries, and
potential investors. In the model presented in Section 2 we assume the sellers
seek to maximize the proceeds from the IPO, and the financial intermediary acts
in their best interest. Some investors are large, strategic, institutional investors,
with private information about the valuation of the stock in the secondary market. For simplicity, we assume that only the investors have private information,
and abstract from the important problems arising when the sellers have private
information (see, e.g., Allen and Faulhaber (1989), Grindblatt and Hwang (1989),
Welch (1989), and Chemmanur (1993)). In addition to the large investors, there
are smaller, and uninformed, but rational, investors. The funds available to these
small investors are limited, hence they cannot absorb the whole issue. In Section 2,
in the line of Benveniste and Spindt (1989), Benveniste and Wilhelm (1990), and
Spatt and Srivastava (1991), we consider the optimal direct mechanism, and its
implementation within the context of the Book Building method.
In Section 3, we show that, in the fixed price auction, since prices cannot adjust to
demand, winner’s curse problems, as in Rock (1986), lead to severe underpricing.
Since the winner’s curse effect reflects the absence of adjustment of prices to
demand it is natural to expect that the market clearing uniform price auction
should perform better. Yet, we show that this auction can be conducive to tacit
collusion between bidders, as in Wilson (1979), leading to large underpricing. This
provides an alternative interpretation for the underpricing empirically evidenced
by Kandel et al. (1997) in IPOs in Israel. In contrast, we show that the Mise en
Vente can be structured to implement the optimal mechanism and to rule out tacit
collusion, in contrast with the uniform price auction. In this auction, investors with
good signals submit large demands, this drives prices up and thus enhances price
discovery.
In Section 4, we present empirical evidence from Mise en Vente auctions consistent with our theoretical result that it can implement the optimal mechanism,
similar to the Book Building method. Hanley (1983) showed that, consistent with
the implications from Benveniste and Spindt (1989) and Benveniste and Wilhelm
(1990), prices somewhat underreact to demand; i.e., the covariance between price
adjustment and underpricing is positive. This is also the case empiricaly in the
Mise en Vente. Cornelli and Goldreich (1999) show empirically that, in the Book
Building method, prices are set deliberatey below the market clearing level and
that, anticipating this, investors inflate their demand. This also arises in the Mise
en Vente.
Conclusions emerging from our analysis are the following:
• Efficient price discovery in IPOs requires some adjustment of prices to demand. Consequently, fixed price selling methods are inefficient. This is consistent
IPO AUCTIONS
13
with the empirical finding by Loughran et al. (1994) that auction-like mechanisms
lead to more efficient pricing than fixed price offers.
• Yet, prices should not adjust to demand too strongly, lest this should spur tacit
collusion. Consequently, the standard market-clearing uniform price (or Dutch)
auction is not the optimal mechanism. In this auction, bidders can tacitly collude
by placing demand functions such that the market clearing price is very low,
and such that, any attempt to bid more aggressively, to gain market share, would
push prices too high to be attractive. In the Mise en Vente, in contrast, since
prices underreact to demand, tacit collusion on low prices can be unravelled.
Indeed, underreaction implies that, if the other bidders were to bid at low prices,
one investor could gain a large market share, without impacting the price too
much.
• In spite of obvious differences in institutional characteristics, two auctions,
the Mise en Vente and the Book Building method, can be designed to implement
the optimal IPO mechanism. From a theoretical perspective, it is not unnatural that
different institutions can be used to implement the same optimal direct mechanism.
From an empirical perspective, it is interesting that, consistent with our theoretical finding that the two institutions are comparable, data generated by the Book
Building method in the United States or the United Kingdom and by the Mise en
Vente in France, exhibit the same observable patterns, in particular underreaction
of prices to demand and oversubscription.
• This analysis has implications for the design of Internet-based IPO auctions.
While the Dutch auction format proposed by Open IPO seems a priori attractive, our theoretical model shows that it can lead to tacit collusion on the part
of bidders, and in that case it can be quite inefficient. In these collusive equilibria, the optimal strategy of the investors is to shade their bids rather than to
“make a bid at the maximum price at which they are comfortable owning shares
of the issue” as advised on Open IPO’s website. In fact, our analysis suggests
that, in contrast with the Dutch auction, it can be optimal to let prices underreact to demand. Note that, in a recent IPO (Andover.net, Inc.), Open IPO actually set the IPO price at a significant discount relative to the market clearing
price, more in line with the rules governing the book building or the Mise en
Vente than with those of the Dutch auction. A challenge facing internet IPO auction designers is to translate into explicit computer algorithms the rather implicit
rules that map demand into prices in the Mise en Vente or the Book Building
method.
2. MODEL
There are S unseasoned shares for sale. The seller faces N large, strategic,
institutional investors with private information about the valuation of the firm by
the market and large bidding capacity, and a fringe of small, retail investors who
are uninformed and cannot absorb the whole issue. All agents are rational and
14
BIAIS AND FAUGERON-CROUZET
risk neutral, and this is a common value auction. The objective of the seller is to
maximize the proceeds from the sale.
Each large investor can buy the whole issue and observes a private signal si , i =
1, . . . N . The signals are identically and independently distributed and can be good
(g), with probability π, or bad (b) with the complementary probability. The value v
of the shares on the secondary market is increasing in the number of good signals n
and its realization is denoted vn .
The retail investors, as a whole, can purchase up to S(1 − k) shares, with k ∈
[0, 1].
The financial intermediary is assumed to act in the best interest of the seller.
She is in contact with the large institutional investors, and she has a distribution
network, collecting the orders from the retail investors.
Consider the following direct mechanism. Each informed investor i sends a
message m i ∈ {g, b}. The mechanism maps these N messages into a price and
into allocations to the informed agents and the retail uninformed agents. The
mechanism is subject to several constraints. First, the price must be the same for
all. This is to reflect the constraint, observed in practice, that IPO auctions involve uniform pricing (it should be noted, however, that Benveniste and Wilhelm
(1990) show that investment bankers acting in the best interest of the firm could
increase expected proceeds by using price discrimination). Second, since we assume the N large traders are ex ante identical, the mechanism is symmetric. Hence
the price is simply a function of the total number of investors who report good
signals n̂, and is correspondingly denoted p(n̂), while the quantity allocated to informed agent i depends only on her message m i and the number of other informed
agents who reported good signals, li . Correspondingly it is denoted q(m i ; li ). Similarly, the quantity allocated to the uninformed agents depends only on n̂, and
it is denoted qu (n̂). Third, the allocation must be such that exactly S shares are
sold.
∀{m i }i=1,....N , n̂ ∈ {0, . . . N },
N
q(m i ; li ) + qu (n̂) = S.
i=1
The mechanism opens the possibility to allocate different quantities to investors
reporting different signals. Indeed, quantity discrimination is crucial to obtain
information revelation from the investors.
The program of the mechanism designer is to maximize expected proceeds
Maxqi (.;.),qu (..), p(.) E(Sp(n̂))
under the incentive compatibility and participation constraints of the investors. The
incentive compatibility constraint of the informed investor i is that she must be better off announcing her true signal than misreporting it, while rationally anticipating
15
IPO AUCTIONS
that the others will truthfully report their own signals. When the investor has observed a good signal this amounts to
N −1
πl q(g; l)(vl+1 − p(l + 1)) ≥
l=0
N −1
πl q(b; l)(vl+1 − p(l)),
l=0
where πl is the probability, from the point of view of the informed investor, that l
out the N − 1 other investors have a good signal.
We impose the rationality condition in its most demanding form. Investors must
still be willing to participate to the mechanism ex post, i.e., after the messages of
all the agents are known. This implies that the price set in the optimal mechanism
must be lower than or equal to the value of the shares: p(n) ≤ vn .
This problem is similar to the problem analyzed by Benveniste and Spindt (1989)
and Benveniste and Wilhelm (1990). However, there are three differences. First
and most importantly, in the present paper, the uninformed agent participates in
the direct mechanism, along with the informed agents; the empirical analysis of
the book-building process by Cornelli and Goldreich (1998) is consistent with
this view. Considering the game in which both informed and uninformed agents
participate enables us to nest in our analysis the study of winner’s curse effects as
in Rock (1986) (see the analysis of the fixed price sale in the next section). Second,
we do not impose the additive form used in Benveniste and Spindt (1989), where
the value of the share is proportional to the number of good signals, which would
imply, in our context that vn = αn, where α is a constant. Third, in contrast with
Benveniste and Spindt (1989), we assume that each large investor could buy the
whole issue. We think this is a reasonable assumption, given the bidding power
of the large financial institutions participating regularly to IPOs, compared to the
relatively small size of most of these operations. In addition this assumption simplifies the analysis. We discuss below how changing this assumption would alter our
results.
The optimal mechanism is given in the next proposition, with the proof of this
proposition, as well as the following propositions, given in the Appendix.
PROPOSITION 1.
tics:
The optimal direct mechanism has the following characteris-
• If all informed investors report a bad signal, the IPO price is equal to the
value of the shares: v0 , the amount allocated to the retail investors is maximized,
by setting qu (0) = S(1 − k), and the shares which cannot be sold to the retail
investors are equally split between the informed investors: q(0; 0) = Sk
.
N
• If there is at least one informed investor reporting a good signal, informed
investors reporting a bad signal receive no shares, i.e., q(0, n̂) = 0.
• When some informed investors announce good signals, there is underpricing,
and correspondingly expected proceeds are lower than the expected value of the
16
BIAIS AND FAUGERON-CROUZET
shares:
E( p(n)) = E(vn ) − kπ (1 − π ) N −1 (v1 − v0 ).
• The optimal mechanism can be implemented with the following price schedule:
p(n) = vn , ∀n < N ,
1−π N
p(N ) = v N − k
(v1 − p(0)).
π
The proposition has a simple intuition. The mechanism designer must ensure
that investors with a good signal announce their information truthfully rather
than pretending they have a bad signal. To makes misrepresentation unattractive, the optimal mechanism minimizes the amount allocated to investors who
announce bad information. This implies setting this amount equal to 0, except
when it is not possible, because all informed investors announce bad signals.
In that situation it is not possible to exclude all the informed agents, because
the retail investors cannot absorb the whole issue. Hence underpricing must
be used, in addition to quantity discrimination, to induce truthful revelation.
This gives rise to informational rents for the large investors. The corresponding wedge between the expected proceeds and the expected value of the shares
is proportional to the amount which the uninformed retail investors cannot buy
(k S). Indeed, if the uninformed retail investors could buy all the shares, allocating them the whole issue would be a simple way to avoid adverse selection and sell all the shares at their a priori expected value, thus eliminating
underpricing.
Note that, as in Rock (1986), a winner’s curse problem, arises for the uninformed investors in the optimal mechanism, as the amount they are allocated is
maximı́zed when the the value of the stock is lowest. Yet, they are still willing
to participate to the auction, because the mechanism is designed to satisfy their
individual rationality constraint.
Next we discuss the extent to which our results are robust to the informational
requirements of our approach:
• In contrast with Allen and Faulhaber (1989), Welch (1989) or Chemmanur
(1993), we do not assume that corporate insiders selling their shares in the IPO have
private information about the market value of the stock. While clearly important
and interesting, the case where there is two-way information asymmetry, i.e., both
the sellers and the buyers have private signals, is beyond the scope of the present
paper. Note that Chemmanur (1993) analyzes two-way information asymmetry, in
the context of fixed-price offers.
• In our framework, it is possible to exclude investors who report bad signals,
except when all of them do so. This directly reflects our assumption that each large
IPO AUCTIONS
17
investor has the potential to absorb the whole issue. If instead we assumed, like
Benveniste and Spindt (1989) that the bidding capacity of the informed investors
was below S, the threat to be excluded from the allocation would be less strong.
Correspondingly, quantity discrimination would be a less powerful tool, and it
would be necessary to rely on underpricing to a larger extent. In that context, the
optimal price schedule would entail underpricing in several states, as in Benveniste
and Spindt (1989).
• In our model, underpricing is decreasing in the number of bidders, and the
outcome of the mechanism goes to the competitive outcome (no underpricing) as
the number of bidders goes to infinity. Note, however, that it would not be very
reasonable to assume infinitely many bidders in our framework. In practice, there
is only a limited number of professional investors who, like the large traders we
consider in the model, have private information about the value of the shares and
have the capacity to absorb the whole issue.
With the pricing schedule presented in Proposition 1, underpricing arises when
the IPO price is high. This suggests that underpricing and IPO prices are likely to
be positively correlated. Now, with this pricing schedule, the covariance between
underpricing and IPO prices cov(v − p, p) simplifies to
π N (1 − π N )(v N − p(N ))( p(N ) − E(v | b)).
This is positive if
E(v) − E(v | b)) > kπ (1 − π ) N (v1 − v0 ),
which is likely to hold, especially when k is relatively low. This positive covariance
can be interpreted in terms of underreaction of prices to demand. Note that, if
no large investor could absorb the whole issue, and correspondingly there was
underpricing in more states of the world, as in Benveniste and Spindt (1989),
underreaction would also arise.
Benveniste and Spindt (1989) and Benveniste and Wilhelm (1990) provide an
interpretation of the Book Building method along the lines of this type of optimal
mechanism. The indications of interest transmitted by investors to the investment
bank are similar to messages about signals, and their reflection in the price adustment and discretionary allocation set by the intermediary are similar to the direct
mechanism. Consistent with this theory, Hanley (1993) found positive correlation
between price adjustment and underpricing in Book Building data.
3. THEORETICAL ANALYSIS OF FIXED PRICE OFFERS, MARKET
CLEARING AUCTIONS, AND MISES EN VENTE
3.1. Fixed Price Auction
Our analysis in this section is similar to Rock (1986). The differences are the
following. First, unlike Rock (1986), we do not impose the condition that informed
18
BIAIS AND FAUGERON-CROUZET
investors do not have enough funds to buy the whole IPO. This is more realistic,
given the relatively small size of IPO issues, compared to the large bidding power
of institutional investors. Second, we consider N informed investors, with different
and imperfect signals, rather than a single, perfectly informed investor.
In the case of fixed price offers, the only choice of the seller is to set the price, p.
Since the uninformed agents cannot absorb the whole issue, and since the seller
must sell the S shares, he must ensure that the informed agents are willing to bid
for the shares, even when they have observed bad signals. Hence, the price must
be set to satisfy the individual rationality constraint of the informed agent with a
bad signal.
We assume that informed investors can costlessly bid for up to S shares. If they
want to bid for a larger amount, up to Q̄ > S, they incur cost c. As only S shares
are for sale, such very large bids are expected not to be fully executed. They can
be optimal, however, to the extent that they enable the investor to obtain a larger
share of underpriced and overbid issues. The cost c can be thought of as the cost
of immobilizing funds during the period of the IPO (Keloharju (1993) notes that
in Finland this period can be quite long).
Consider the candidate equilibrium where investors with good signals bid for
a large amount ( Q̄), while investors with a bad signal bid for a lower amount (S)
and retail investors also participate to the IPO. In this context, the execution rate
when there are n good signals is
τn =
S
.
S(1 − k + N − n) + n Q̄
Setting the expected gain of the informed with a bad signal bidding for S shares
to 0
E(Sτ (v − p) | b) = 0,
pins down the IPO price. Correspondingly, we obtain the following proposition:
PROPOSITION 2. If K 1 ≤ c ≤ K 2 (where K 1 and K 2 are constants defined in
the proof ), then in the fixed price offer the highest possible IPO price is
p=
N −1
λl v(l),
l=0
where,
πl τl
λl = N −1
,
l=0 πl τl
and the informed agent with a good signal bids for Q̄, while the informed agent
with a bad signal bids for S, and the uninformed agent bids for S(1 − k).
IPO AUCTIONS
19
First note that consistent with empirical evidence (Koh and Walter (1989), Levis
(1990), Keloharju (1993)) in the equilibrium described in Proposition 2, demand is
positively correlated with underpricing. Indeed the latter is equal to vn − p, while
the former is equal to: S(1 − k + N − n) + n Q, so that both are increasing in n.
Second, manipulating the 0 profit condition of the informed investors with a bad
signal, the IPO price can be expressed as
p = E(v | b) +
cov(τ, v − p | b)
.
E(τ | b)
As the execution rate is decreasing in the number of good signals (n), the covariance
is negative. Hence, the IPO price is lower than the expectation of the value of the
asset conditional on a bad signal (E(v | b)). Underpricing pricing is necessary to
convince the investors with bad signals to participate to the offer, in spite of the
winner’s curse problem they face (in the same spirit as in Rock (1986)). Indeed
they obtain worse execution when the share is worth a lot, and many investors with
good signals place large bids, resulting in low execution rates.
Further note that the expected profit of the uninformed agent is greater than that
of the investors with bad signals:
E(τ (v − p)) > E(τ (v − p) | b) = 0.
Hence retail investors earn positive expected profits in the fixed price offer.
These results emphasize that the lack of adjustment of prices to demand leads to
large rents, left both to the informed and the uninformed agents, and consequently
large underpricing.
3.2. Uniform Price Walrasian Auction
In this mechanism, the seller sets a reservation price p, the investors submit
¯ and demand. In this
demand functions, and the IPO price is set to equate supply
market clearing uniform price auction, as in the analyses of Wilson (1979) and
Back and Zender (1993), there is scope for tacit collusion between investors, as
stated in the next proposition:
PROPOSITION 3. For any reservation price p ≥ E(v | b), there exists a
¯ are constant whatever the
Bayesian Nash equilibrium where investors’ demands
realization of the signals, and the resulting IPO price is equal to the reservation
price. The equilibrium demand of each investor is, at price p,

 N S+ 1 − σ ( p − p)
¯
S
1
 with σ ≤
.
N (N + 1) E(v | g) − p
¯
The intuition for this result is the following. The demand functions have a relativelly small slope (σ ). Hence the residual supply function faced by each investor
20
BIAIS AND FAUGERON-CROUZET
is rather inelastic: It takes a big price increase to increase the residual supply. This
large price impact makes it unattractive for each investor to attempt to increase
her purchases. Our theoretical model provides an alternative interpretation for the
thought-provoking empirical findings of Kandel et al. (1997) on the uniform price,
market clearing IPO mechanism used in Israel. In particular, Kandel et al. (1997)
find that (i) there is significant underpricing, and (ii) the (absolute value of the)
slope of the demand schedules is low, i.e., there is a flat, around the IPO price.
This is consistent with our theoretical result that the slope of the demand curve σ
must be low in the tacit-collusion Bayes-Nash equilibrium.
Because of the strategic complementarities between the actions of the bidders
in this auction, there exist multiple equilibria, some of which do not involve tacit
collusion. Yet, it is likely that the bidders will focus on the tacit collusion Nash
equilibrium presented in the proposition, since it is the most advantageous for
them.
To cope with tacit collusion within this mechanism while satisfying the individual rationality condition of the informed investor with a bad signal, it is best for
the seller to set p to E(v | b).
¯ subsection, note that although underpricing is less severe in this
To conclude this
auction than in the fixed price offer, it is still quite significant, due the possibility
of tacit collusion it offers to the bidders.
3.3. The Mise en Vente
The Mise en Vente is an auction-like IPO procedure commonly used in France.
It operates as follows. Five days prior to the IPO the quantity offered and the
reservation price are set jointly by the bank, the broker, and the firm. On the
day of the IPO, investors submit limit orders to their brokers. The latter transmit these orders to the stock exchange. The total demand function is computed
and graphically plotted by the auctioneer, who is a Bourse official. As a function of this demand, the auctioneer sets the IPO price. As in the Book Building
method, there is no formal explicit algorithm mapping demand into prices. But
price adjustment in the Mise en Vente exhibits strong empirical regularities, as
shown in the next section. Eligible orders, above the IPO price, obtain prorata
execution.
To illustrate this description, consider the Mise en Vente of Partouche. On March
29, 1995, 500,000 shares were offered. The reservation price was 185. The total
demand, expressed at all prices, amounted to 8.4 million shares, i.e., 16.29 times
the supply. Figure 1 plots the demand expressed at each price. The IPO price was
set to 200, which corresponds to a percentage price adjustment of 8.1%. (Note
that the price at which supply would have been equal to demand was 220, i.e.,
the IPO price was deliberately set below market clearing). Eligible orders, placed
above 200, obtained prorata execution. The first secondary market price, set after a
tâtonnement process which lasted two days, was equal to 215, which corresponds
to 7.5% underpricing.
FIG. 1. Buy orders placed at the different prices for the Mise en Vente of Partouche, March 29, 1995.
IPO AUCTIONS
21
22
BIAIS AND FAUGERON-CROUZET
Denote D( p) the cumulated demand stemming from limit orders placed at prices
equal to or higher than p. The optimal Mise en Vente arising in our framework is
described in the next proposition:
PROPOSITION 4.
If
c < k(1 − π ) N −1
N −1
Q̄ − S
S
N
(N − 1) Q̄ + S the optimal mechanism can be implemented with a Mise en Vente where (i) the reservation price is v0 , (ii) investors place limit orders at prices p ∈ {v0 , v1 , . . . , v N −1 ,
p N }, (iii) the price schedule is
if D( p N ) < S, then p = v0 , while if D( p N ) ≥ S,
then p = max p[ D( p) ]+ , min{ p, s.t., D( p) = D( p N )} ,
N
(where [.]+ means that if the number within brackets is not an integer it is rounded
up to the next integer), (iv) the equilibrium strategies of the investors are to demand
Q shares at price pn for investors with good signals, to demand S shares at price
v0 for investors with bad signals, and to demand S(1 − k) shares at price v0 for
retail investors.
In the optimal Mise en Vente, consistent with the workings of the actual mechanism, prices increase with total demand, as the latter is equal to
n Q̄ + (N − n + 1 − k)S.
In equilibrium, in the optimal Mise en Vente, investors with good signals place
more aggressive demands, as they are more eager to purchase the shares. Consequently, the higher the value of the asset, the better the private signals, the higher
the demand, and the higher the IPO price. This theoretical result is consistent with
the empirical findings by Derrien and Womack (1999) that this auction-like IPO
method efficiently incorporates market information into IPO prices.
Yet, the price set in this mechanism does not clear market in order to satisfy
the incentive compatibility constraint. Correspondingly, there is oversubscription
at the IPO price.
Our analysis suggests that the Mise en Vente and the Book Building methods
have similar incentive properties and can reach similar outcomes. This contrasts
with the discussion in Benveniste and Wilhelm (1990) that, with uniform prices
and even-handed allocations, information revelation should be impossible. The
point is that, in the Mise en Vente, while the pro-rata allocation rule is indeed
even-handed, it leads to investors with different signals being treated differently
because they have placed different bids.
IPO AUCTIONS
23
Consistent with our theoretical result, a recent empirical study of the Book Building method in the United Kingdom, by Cornelli and Goldreich (1998), shows that
in this IPO institution: (i) the IPO price is deliberately set below market clearing,
so that there is oversubscription at the IPO price, (ii) good deals are underpriced,
and (iii) rationally anticipating this, investors inflate their demand and apply for
more shares than they actually desire to buy. All these features are exhibited in our
theoretical model of the Mise en Vente. This highlights that, in spite of institutional
differences, Book Building and Mise en Vente can have similar properties.
In the uniform price market clearing auction, there exists a tacit collusion Nash
equilibrium whereby the bidders place demands such that the IPO price is always
equal to the reservation price. The issue arises, therefore, if such tacit collusion
equilibria exist in the Mise en Vente. The following proposition states that it is not
the case.
PROPOSITION 5.
If
E(v | g) − v1 >
1
(E(v | g) − v0 ),
N +1−k
then tacit collusion (such that the IPO price is equal to the reservation price
irrespective of the signal) is not a Bayesian Nash equilibrium in the optimal Mise
en Vente.
In contrast with the case of the market clearing uniform price auction, tacit
collusion can be unravelled in the case of the Mise en Vente. This is because in the
Mise en Vente prices respond less strongly to demand than in the market clearing
case. This is conducive to outbidding, since the investors can increase their market
share, by raising their demand, without affecting the price too much.
The condition stated in Proposition 5 is not very demanding. For example, in the
linear parametrization assumed in Benveniste and Spindt (1989), where vn = αn,
the condition holds.
4. EMPIRICAL ANALYSIS OF THE MISE EN VENTE
To assess the empirical relevance of our theoretical analysis we confront it to a
sample of 92 Mises en Vente which took place between 1983 and 1996. Our data
corresponds to firms listed on the “Second Marché” an intermediary tier of the
stock market created in France in 1983 for growth companies and for which the
listing requirements are less stringent than for the first tier (the “Cote Officielle”
or Official List). Practically all IPOs between 1983 and 1996 have taken place
on the “Second Marché” (only a few exceptions, including privatizations, are on
the Official List). This contrasts with the United Kingdom where IPOs are more
frequently on the Official List than on the Unlisted Securities Market (see Levis
(1990)). In Finland, there are IPOs on the Helsinki Stock Exchange or on the
24
BIAIS AND FAUGERON-CROUZET
TABLE 2
Summary Statistics on 92 Mises en Ventes, 1983–1996
Variable
Average
Standard deviation
Underpricing =
ln(stock market
clearing price/IPO price)
demand )
ln( Total
Supply
Supply (number of shares)
Price adjustment =
ln(IPO price/
reservation price)
Sales year before IPO∗
Age at time of IPO∗
13%
16.52%
55
159262
17.36%
74
141401
10.66%
FF 635, 805, 979
28.32
1, 377, 289, 794
21.7
∗
Statistics based on 68 observations between 1983 and 1994.
OTC market (see Keloharju (1993)). In 1996, an additional tier of the French
equity market, the “Nouveau Marché,” was created. It aims at attracting younger
companies to the Bourse. Because the companies listing on this market are quite
different from those listing on the “Second Marché” they are not included in the
present analysis.
For the IPOs in our sample we observe:
• the IPO price, the reservation price, and the secondary market clearing price,
• the number of shares sold,
• the total number of shares demanded,
• the age of the firm at the time of IPO, and the sales during the year prior to
IPO (these variables were available only for the period up to 1994).
Table 2 reports summary statistics on these variables. In particular, note that the
firms going public in France are relatively older than their U.S. counterparts (a
feature shared with firms going public in other European countries). Note also that
underpricing is on average equal to 13%, a figure very similar to those observed in
the United States in the context of the Book Building procedure, while on average
the price adjustment is 17.36%.
Our theoretical analysis implies that the adjustment of the IPO price over the
reservation price should increase with total demand. To test this hypothesis, we
regressed price adjustment on the strength of demand and estimated the following
regression
ln(P j /R j ) = a + b ln(D j /S j ) + e j ,
where P j is the price of IPO j, R j is the initial reservation price for this IPO,
D j is the total demand (i.e., sum of the quantities of all the orders placed at
all prices in this IPO) and S j is the number of shares sold. The results are in
Table 3. The first column in Table 3 presents the simple OLS results, for which
the slope is significantly positive. The second and third columns present estimates
25
IPO AUCTIONS
TABLE 3
IPO price
Regression of Price Adjustment (ln( reservation
price )) on Demand and Control
Variables (t Statistics Are in Parentheses)
Sample period
Number of observations
Estimation method
Constant
demand )
ln ( Total
Supply
ln (sales year before IPO)
ln (age at time of IPO)
∗
1983–1996
1983–1996
1983–1994∗
92
OLS
92
White correction
for heteroschedasticity
−0.03
(−1.88)
0.053
(72.8)
—
68
OLS
−0.027
(−1.47)
0.061
(11.79)
—
—
—
−0.1
(−0.79)
0.06
(12.1)
0.0008
(0.13)
0.01
(1.77)
Statistics based on 68 observations between 1983 and 1994.
obtained after correcting for heteroskedasticity, using the White adjustment, and
after controlling for exogenous variables such as the age of the firm and the level of
its sales. The estimates obtained in these regressions are similar to those obtained
in the simplest version of the test, and the coefficient of total sales is significantly
positive in the three specifications. Figure 2 illustrates this analysis by plotting
price adjustment (measured as the logarithm of the ratio of the IPO price to the
initial reservation price) against total demand (measured as the logarithm of the
ratio of total demand to supply).
Our maintained hypothesis is that the Mise en Vente implements the optimal
mechanism, like the Book Building method. Under this hypothesis, it should exhibit empirical regularities which are characteristic of this mechanism, and which
are also observed in the Book Building method. One such empirical regularity is
the underadjustment of prices to demand, which manifests itself in positive covariance between price adjustment and underpricing, and which was documented
empirically in the case of the Book Building method by Hanley (1993). To test
this hypothesis, we regressed underpricing on price adjustment, as
ln(v j /P j ) = a + b ln(P j /R j ) + e j ,
where v j is the first market clearing price after the IPO, P j is the price of
IPO j, and R j is the initial reservation price for this IPO. The results are in
Table 4. As in the previous regression we estimated a simple OLS specification, as
well as two other specifications, one correcting for heteroskedasticity and the other
controlling for age and sales. Consistent with the underadjustment hypothesis, the
slope is significantly positive, in all three specifications. Figure 3 illustrates this
analysis by plotting underpricing (measured as the logarithm of the ratio of the
first market clearing price to the IPO price) against price adjustment.
FIG. 2. OLS regression of price adjustment on strength of demand, 92 Mises en Vente, 1983–1996.
26
BIAIS AND FAUGERON-CROUZET
FIG. 3. OLS regression of underpricing on price adjustment, 92 Mises en Vente, 1983–1996.
IPO AUCTIONS
27
28
BIAIS AND FAUGERON-CROUZET
TABLE 4
clearing price
Regression of Underpricing (ln ( stock market
)) on Price Adjustment and
IPO price
Control Variables (t Statistics Are in Parentheses)
Sample period
1983–1996
1983–1996
1983–1994∗
Number of observations
Estimation method
92
OLS
68
OLS
−0.03
(−1.15)
0.91
(6.92)
—
92
White correction for
heteroschedasticity
−0.5
(−3.28)
0.77
(80.3)
—
ln (age at time of IPO)
—
—
demand )
ln ( Total
Supply
—
—
Constant
IPO price
ln ( reservation
price )
ln (sales year before IPO)
∗
0.32
(1.04)
0.97
(3.2)
−0.01
(−0.87)
−0.03
(−1.7)
−0.00024
(−0.009)
Statistics based on 68 observations between 1983 and 1994.
5. CONCLUSION
Our analysis suggests that it is important to organize IPO auctions so that
the IPO price reflects the information held by investors. If this is not the case,
as in fixed price offers, underpricing is bound to be large and very little information can be generated about the value of the stock during the IPO process.
Yet, our analysis suggests that the optimal auction may not be the (a priori natural) Walrasian uniform price auction. In such an auction, the strong reaction
of prices to demand can lead to tacit collusion between bidders, as in Wilson
(1979), which leads to large underpricing. For optimal information elicitation
to be immune to tacit collusion, some form of underadjustment of prices to
the informational content of demand is required. Such information elicitation
can be achieved in similar ways by apparently very different institutions: the
Book Building method used in the United States and the Mise en Vente used in
France.
While our analysis suggests that the Dutch auction used by Open IPO may not
be optimal, it raises—but does not resolve—the issue how to translate into explicit
computerizable rules the mapping of demand into prices that occurs in the Mise
en Vente or the Book Building.
While the present paper emphasizes the similarities between the Book Building
method and the Mise en Vente auction, there are some important potential differences between the two systems. In the Mise en Vente, bids are anonymous, while
in the Book Building the investment banker observes (and has entire discretion to
condition on) the identity of the investors sending indications of interest. Lack of
IPO AUCTIONS
29
investor’s anonymity and discretionary allocation in the Book Building can have
positive consequences, as they can be used to enhance the ability to extract information from the bidders. Indeed, Benveniste and Spindt (1989) analyze how long
term nonanonymous relationship can be used efficiently in the IPO process, while
Sherman (1999) analyzes the superiority of the book building over auction mechanisms (see also Sherman (2000)). On the other hand, if the investment banker is not
acting in the best interest of the seller, discretionary allocations can have negative
consequences. In particular, they can worsen winner’s curse problems for retail
investors and consequently reduce their participation in auctions. In this context,
the anonymity and nondiscretionary rules used in auctions can be attractive. This
may well be why Open IPO insists that its Dutch auction mecanism “levels the
playing field” and “ensures that all bidders are on equal footing.” Mechanisms other
than the Dutch auction could potentially be used to “level the playing field.” Biais
et al. (2002) analyzed the case where the investment bank and the institutional
investors can collude (so that the investment bankers are tempted to treat large
institutional investors more favorably than retail investors). They showed theoretically and econometrically that the Mise en Vente can be an efficient mechanism
in this context.
Wilhelm (1999) discussed that, while traditional investment banking was
based on relationships with a small number of large investors, Internet technology could potentially allow one to sell unseasoned shares to a large number of
relatively small investors, with no or litttle relationship with the investment bank.
An interesting avenue of further research could be to investigate the consequences
of this evolution for the optimal pricing and allocation of IPOs. In particular,
it would be interesting to analyze price discovery, information elicitation, and
strategic issues in the context of Internet–based sales to a large diffuse investor
base.
APPENDIX: PROOFS
Proof of Proposition 1. As is typical in such two-type mechanism design problems, the incentive compatibility condition of the informed investor with bad news
is redundant, and only the incentive compatibility condition of the informed agent
with a good signal is binding.
The constraint that all shares be sold is
∀l = 0, . . . N − 1, (l + 1)q(g; l) + (N − l − 1)q(b; l + 1) + qu (l + 1) = S,
which can be rewritten as
∀l, q(g; l) =
S − qu (l + 1) − (N − (l + 1))q(b; l + 1)
.
l +1
30
BIAIS AND FAUGERON-CROUZET
Substituting this value of q(g; l) in the incentive compatibility condition becomes
N −1
πl
l=0
≥
S − qu (l + 1) − (N − (l + 1))q(b; l + 1)
(vl+1 − p(l + 1))
l +1
N −1
πl q(b; l)(vl+1 − p(l)).
l=0
To relax this condition, minimize qu (l + 1), for l = 0, . . . N − 1, by setting it
equal to 0, and minimize q(b; l), l = 0, . . . N , by setting: q(b; l) = 0, for l > 0,
and q(b; 0) = Sk
. Thus the incentive compatibility condition simplifies to
N
N −1
πl
l=0
S
Sk
(vl+1 − p(l + 1) ≥ π0 (v1 − p(0)).
l +1
N
Treating this constraint as an equality, p(N ) can be written as a function of the
other prices
π N −1 p(N ) = π N −1 v N +
N −2
l=0
N
πl
(vl+1 − p(l + 1))
l +1
− π0 k(v1 − p(0)).
Now, the objective of the mechanism designer is to maximize (under the incentive
compatibility and individual rationality conditions) the expected proceeds, which
can be written as
N −1
µn p(n) = µ N p(N ) +
n=0
N −2
µn p(n),
n=0
where µn denotes the probability that n signals out of N are good. Noting that µ N =
ππ N , and substituting the incentive compatibility condition into the objective, the
latter becomes
π π N −1 v N +
N −2
l=0
+
N −1
N −2
πl
πl
N
N
vl+1 −
p(l + 1) − π0 k(v1 − p(0))
l +1
l +1
l=0
µn p(n) + µ0 p(0).
n=1
N
Noting that µn+1 = ππn n+1
, and that because of the participation constraint:
31
IPO AUCTIONS
p(0) = v0 , this can be rewritten as
µN vN +
N −1
µn vn −
n=1
+
N −2
N −2
µn+1 p(n + 1) − π π0 k(v1 − v0 )
n=0
µn+1 p(n + 1) + µ0 v0 .
n=0
Noting that π0 = (1 − π ) N −1 , the expected proceeds are equal to
E(vn ) − kπ (1 − π ) N −1 (v1 − v0 ).
Note that prices do not appear in this expression. Hence, prices are indeterminate,
as long as the incentive compatibility condition is satisfied. One possible solution
is to set
p(l) = vl , ∀l = 1, . . . , N − 1,
and
p(N ) = v N −
1−π
π
N −1
k(v1 − v0 ).
Note that, for this solution, the quantity allocated to the retail investors qu (n), n =
1, . . . N − 1, is no longer relevant since it is multiplied by 0 in the incentive
compatibility condition. Thus, for the price solution above, the optimal mechanism does not require that the retail investors receive 0 shares for vn , n = 1, . . .
N − 1.
Proof of Proposition 2. This participation constraint of the informed investor
with a bad signal, E(Sτ (v − p) | b) = 0, can be rewritten more explicitly as
N −1
πl [v(l − p]τl = 0.
l=0
Hence, the IPO price can be written as
p=
N −1
l=0
N −1
πl τl
vl =
λl vl .
N −1
l=0
l=0 πl τl
32
BIAIS AND FAUGERON-CROUZET
The investor with a bad signal is better off bidding for S shares than for Q̄ if
c > Q̄
N −1
l=0
S
πl [v(l) − p]
S(1 − k + N − (l + 1)) + (l + 1) Q̄
.
Denote K 1 the right-hand side of this inequality.
Now consider the informed investor with a good signal. She is better off bidding
for Q̄ than for S shares if
N −1
S
Q̄
πl [vl+1 − p]
S(1 − k + N − (l + 1)) + (l + 1) Q̄
l=0
N −1
S
>S
πl [vl+1 − p]
.
S(1 − k + N − l) + (l) Q̄
l=0
−c
That is, if
N −1
πl [vl+1 − p]S
l=0
×
Q̄
S
−
S(1 − k + N − l) + l Q̄ + Q̄
S(1 − k + N − l) + l Q̄ + S
> c.
Denote K 2 the left-hand side of the inequality. Note that it is indeed positive,
since
Q̄
S
>
.
S(1 − k + N − l) + l Q̄ + Q̄
S(1 − k + N − l) + l Q̄ + S
Proof of Proposition 3. To establish that the strategies stated in the proposition
form an equilibrium, we need to prove that the investor with a good signal, anticipating that the other investors follow the equilibrium strategy, prefers to purchase
S/(N + 1) shares at price p, rather than bidding more aggressively, to increase her
¯ does not want to do so, then a fortiori, the uninformed
market share. Indeed, if she
investor or the informed investor with bad news do not want to undercut, while
they are still willing to buy a constant amount at p.
Anticipating that the other investors follow their¯ equilibrium strategies, she faces
the residual supply curve:
S−N
S
− σ ( p − p) .
N +1
¯
IPO AUCTIONS
33
Her task is to choose the price p which maximizes her expected profit, ( p),
where
S
( p) =
+ N σ ( p − p) (E(v | g) − p),
N +1
which is equal to the product of the residual supply curve by the unit profit. To
prove that she finds it optimal to purchase S/(N + 1) shares at price p, we need to
¯
show that this is decreasing for all prices above p. Now
¯
∂( p)
S
= N σ (E(v | g) − p) −
+ N σ ( p − p)
p
N +1
is decreasing in p. Hence, to show that it is negative for all prices larger than or
equal to p, we only need to show that it is negative for p, i.e.,
¯
¯
S
N σ (E(v | g) − p) −
< 0,
N
+1
¯
that is
1
S
.
N (N + 1) E(v | g) − p
¯
Proof of Proposition 4. First, consider the case of the informed investor with
a good signal. Her equilibrium expected gain is
σ <
π N −1 (vn − p)
S
− c.
N
Could she obtain more by deviating from her equilibrium strategy? First, suppose
she demands a lower amount q, still at price pn . If q > S, her expected gain is
strictly lower than in equilibrium, since she just obtains less shares. If q ≤ S, then
her expected gain is
π N −1 (vn − p)S
S
(N − 1) Q̄ + S which is lower than her equilibrium gain if
π
N −1
N −1
( Q̄ − S)
(vn − p)S
> c.
N (N − 1) Q̄ + S
Second, suppose she places a bid at a lower price than pn , say P. If P < pn−1 , the
IPO price is pn−1 , while if P ≥ pn−1 , then the IPO price is above P. In both cases
she obtains 0 shares.
34
BIAIS AND FAUGERON-CROUZET
The retail investors and the informed investor with a bad signal obtain 0-expected
gains in equilibrium and it is straightforward to see that they cannot obtain positive
profits by deviating.
Proof of Proposition 5. Consider a candidate equilibrium, whereby tacit collusion would prevail, and in which the IPO price would always equal v0 . In such
an equilibrium the investors would bid as much as possible while not pushing the
price above v0 and still deterring the other investor from competing away market
shares. To avoid pushing the price above v0 , they place bids for (slightly less than)
S
shares at p N and for S(1 − 1/N ) at v0 . The expected profit from such tacit
N
collusion after observing a good signal is
S
[E(v | g) − v0 ].
N +1−k
If, she expects the others to follow this strategy, the informed investor placing a
bid for S shares at p1 = v1 expects to gain
S[E(v | g) − v1 ].
Hence, she prefers to follow that strategy rather than to implicitly collude if
E(v | g) − v1 >
1
[E(v | g) − v0 ].
N +1−k
REFERENCES
Allen, F., and Faulhaber, G. (1989). Signaling by underpricing in the IPO market, J. Finan. Econ. 23,
303–323.
Beatty, R. P., and Ritter, J. (1986). Investment banking, reputation and the underpricing of initial public
offerings, J. Finan. Econ. 15, 213–232.
Belletante, B., and Paliard, R. (1993). Does knowing who sells matter in IPO pricing: The French
Second Market Experience, Cah. Lyon. Rech. Gestion 14, 42–74.
Benveniste, L., and Spindt, P. (1989). How investment bankers determine the offer price and allocation
of new issues, J. Finan. Econ. 24, 343–361.
Benveniste, L., and Wilhelm, W. (1990). A comparative analysis of IPO proceeds under alternative
regulatory environments, J. Finan. Econ. 28, 173–207.
Benveniste, L., and Wilhelm, W. (1997). Initial public offerings: Going by the boook, J. Appl. Corp.
Finan., Spring.
Biais, B., Bossaerts, P., and Rochet, J. C. (2002). An optimal IPO mechanism, Rev. Econ. Stud.,
forthcoming.
Brennan, M., and Franks, J. (1995). Underpricing, ownership and control in IPO of equity securities
in the UK, mimeo, London Business School.
Chemmanur, T. (1993). The pricing of IPOs: A dynamic model with information production, J. Finance
48(1), 285–304.
IPO AUCTIONS
35
Cornelli, F., and Goldreich, D. (1998). Book building and strategic allocation, working paper, London
Business School.
Derrien, F., and Womack, K. (1999). Auctions versus Book Building and the control of underpricing
in hot IPO markets, Working paper, Dartmouth College.
Dubois, M. (1989). Les entreprises qui s’introduisent en Bourse sont-elles sous évaluées? J. Soc. Stat.
Paris 130(1), 4–616.
Grinblatt, M., and Hwang, C. (1989). Signalling and the pricing of new issues, J. Finance 44(2),
393–420.
Hanley, K. (1993). The underpricing of initial public offerings and the partial adjustment phenomenon,
J. Finan. Econ. 231–250.
Hanley, K., and Wilhelm, W. (1995). Evidence on the strategic allocation of IPO, J. Finan. Econ.
239–257.
Hanley, K., Lee, C., and Seguin, P. (1996). The marketing of closed end fund IPOs: Evidence from
transactions data, J. Finan Intermediation 5, 127–159.
Husson, B., and Jacquillat, B. (1989). French new issues, underpricing, and alternative methods of
distribution, in “A Reappraisal of the Efficiency of Financial Markets” (R. Guimaraes et al., Eds.).
Springer Verlag, Berlin.
Husson, B., and Jacquillat, B. (1990). Sous-évaluation des titres et méthodes d’introduction au second
marché (1983–1986), Finance 11(1).
Ibbotson, R. (1975). Price performance of common stock new issues, J. Finan. Econ. 235–272.
Ibbotson, R., Sindelar J., and Ritter, J. (1988). Initial public offerings, J. Appl. Corp. Finan. 1(4), 37–45.
Jacquillat, B., and Mac Donald, J. G. (1974). L’efficacité de la procédure française d’introduction en
Bourse, Rev. Banque 31–39.
Jacquillat, B., Mac Donald, J. G., and Rolfo, J. (1978). French auctions of common stock: New issues,
J. Bus. Finan. 2, 305–322.
Kandel, S., Sarig, O., and Wohl, A. (1997). The demand for stocks: An analysis of IPO auctions,
Working paper, Tel Aviv University.
Keloharju, M. (1993). The winner’s curse, legal liability, and the long run price performance of initial
public offerings in Finland, J. Finan. Econ. 251–277.
Koh, F., and Walter, T. (1989). A direct test of Rock’s model of the pricing of unseasoned issues,
J. Finan. Econ. 251–272.
Leleux, B. (1993). Post–IPO performance: A French appraisal, Finance 14(2), 79–105.
Levis, M. (1990). The winner’s curse problem, interest costs and the underpricing of initial public
offerings, Econ. J. 76–89.
Loughran, T., Ritter, J., and Rydqvist, K. (1994). Initial public offering: International insights, PacificBasin Finan. J. 2, 165–199.
Michaely, R., and Shaw, W. (1993). The pricing of initial public offerings: Tests of adverse selection
and signalling theories, Rev. Finan. Stud. 7(2).
Mirat, B. (1983). Les Introductions sur le Second Marché, Banque 1127–1134.
Mirat, B. (1984). Les Introductions sur le Second Marché (1983–1984), Banque 1027–1035;
1148–1156.
Ritter, J. (1984). The hot issue market of 1980, J. Bus. 57, 215–240.
Rock, K. (1986). Why new issues are underpriced, J. Finan. Econ. 15, 187–212.
Sherman, A. (1999). Global trends in IPO methods: Book Building versus auctions, working paper,
University of Minnesota.
Sherman, A. (2000). IPOs and long term relationships, Rev. Finan. Stud. 13(3), 697–710.
Smith, C. (1986). Investment banking and the capital acquisition process, J. Finan. Econ. 15, 3–29.
36
BIAIS AND FAUGERON-CROUZET
Spatt, C., and Srivastava, S. (1991). Preplay communication, participation restrictions, and efficiency
in initial public offerings, Rev. Finan. Stud. 4, 709–726.
Tinic, S. (1988). Anatomy of initial public offerings of common stock, J. Finance 789–822.
Welch, I. (1989). Seasoned offerings, imitation costs, and the underpricing of initial public offerings,
J. Finance 44, 421–449.
Wilhelm, W. (1999). Internet investment banking: The impact of information technology on
relationship banking, J. Appl. Corp. Finan. Spring.
Wilson, R. (1979). Auctions of shares, Quart. J. Econ. 675–698.