Two essays on underwriter compensation for initial public offerings

MISSING IMAGE

Material Information

Title:
Two essays on underwriter compensation for initial public offerings
Physical Description:
vi, 76 leaves : ill. ; 29 cm.
Language:
English
Creator:
Chen, Hsuan-Chi
Publication Date:

Subjects

Subjects / Keywords:
Finance, Insurance and Real Estate thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Finance, Insurance and Real Estate -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1999.
Bibliography:
Includes bibliographical references (leaves 73-75).
Statement of Responsibility:
by Hsuan-Chi Chen.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 030471365
oclc - 42451513
Classification:
lcc - LD1780 1999 .C516
System ID:
AA00017709:00001


This item is only available as the following downloads:


Full Text










TWO ESSAYS ON UNDERWRITER COMPENSATION
FOR INITIAL PUBLIC OFFERINGS














By

HSUAN-CHI CHEN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1999














To My Wife, Hsiu-Chuan (Jenny) Yeh














ACKNOWLEDGMENTS


This dissertation was inspired by initial conversations with Jay Ritter. I am deeply

indebted to Jay Ritter, who provided guidance, support, and inspiration throughout the

entire process. I also wish to express my sincere thanks to Mark Flannery, Bruce

Foerster, Jason Karceski, Tracy Lewis, and Michael Ryngaert for many helpful

comments. I/B/E/S kindly allowed the use of their data on analyst forecasts. Finally, I

thank my wife, Hsiu-Chuan, for her encouragement throughout the course of this

enterprise.














TABLE OF CONTENTS

page

ACKNOWLEDGMENTS.............................................................. .... iii

ABSTRACT....... ...................................... ................................. v

CHAPTERS

1 INTRODUCTION.......................................................................... 1

2 THE SEVEN PERCENT SOLUTION............................................... 3

Introduction. ........ .......................................... ....... ................... 3
The Facts..... ....... .................................................. .......... 7
Explanations for High Spreads...................................................... 14
Explanations for the Clustering of Spreads at Seven Percent........................ 21
Possible Reasons for the Increased Clustering of Spreads over Time.............. 29
Summary and Conclusions.......................................................... 30

3 COMPETITION AND COLLUSION IN THE IPO MARKET..................... 53

Introduction................................ ........................ .. ........ ................ 53
An Implicit Collusion Model for the IPO Market.................................... 55
Conclusions........................................................... ................. 65

4 SUMMARY AND CONCLUSIONS................................................. 67

APPENDIX: PROOFS OF THE MAIN RESULTS IN CHAPTER 3................. 69

REFERENCES............................................................................ 73

BIOGRAPHICAL SKETCH............................................................. 76














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

TWO ESSAYS ON UNDERWRITER COMPENSATION
FOR INITIAL PUBLIC OFFERINGS


By

Hsuan-Chi Chen

August 1999


Chairman: Jay R. Ritter
Major Department: Finance, Insurance and Real Estate

Gross spreads received by underwriters on initial public offerings (IPOs) in the

U.S. are much higher than in other countries. Furthermore, I find that in recent years at

least 90 percent of deals raising between $20-80 million have spreads of exactly 7.0

percent, three times the proportion of a decade earlier. Investment bankers readily admit

that the IPO business is very profitable, and that they avoid competing on fees because

they "don't want to turn it into a commodity business." This dissertation examines

several features of the IPO underwriting business that result in a market structure where

spreads are high and proposes some explanations for the high frequency of clustering at

seven percent.

I also propose a game-theoretic model, as one possible explanation, for the pattern

of clustering spreads for moderate-size offerings and of lower spreads for large offerings








in the IPO market. I show that investment bankers, while acting noncooperatively, may

charge above-competitive spreads in equilibrium. The evidence of underwriting spreads

from the 1990s is consistent with the implications of this model.

The dissertation suggests that although issuing firms face high and, for moderate-

size deals, nonnegotiable spreads, issuers can still bargain on additional co-managers and

receive more extensive analyst coverage.














CHAPTER 1
INTRODUCTION

"Going public" is an important decision in the life cycle of a company. It provides

a firm with avenues to raise external capital for investment and future growth and at the

same time creates a liquid outlet for the original shareholders to diversify their

investments. One issue regarding initial public offerings (IPOs) of equity is the direct

cost (a gross spread) paid by IPO firms to underwriters who play an important role to

assist firms to go public. See, among others, Barry et al. (1991), Beatty and Welch

(1996), James (1992), Lee et al. (1996), and Ritter (1987).

The dissertation is composed of two essays that study gross spreads and a

possible pricing strategy used by underwriters. Chapter 2, the first essay, examines gross

spreads received by underwriters in the U.S. for the period 1985-1998. Gross spreads in

the U.S. are much higher than in other countries. I examine several features of the IPO

underwriting business that result in a market structure where spreads are high.

Furthermore, in recent years at least 90 percent of deals raising between $20-80

million have spreads of exactly 7.0 percent, three times the proportion of a decade earlier.

Investment bankers readily admit that the IPO business is very profitable, and that they

avoid competing on fees because they "don't want to turn it into a commodity business."

I provide several alternative explanations for the clustering of gross spreads at 7.0 percent

and the increased clustering of spreads over time.

Chapter 3, the second essay, proposes a game-theoretic model, as one possible





2

explanation, for the pattern of clustering spreads for moderate-size offerings and of lower

spreads for large offerings in the IPO market. I show that investment bankers, while

acting noncooperatively, may charge above-competitive spreads in equilibrium.

The dissertation concludes with a summary in Chapter 4.














CHAPTER 2
THE SEVEN PERCENT SOLUTION


Introduction


On September 18, 1996, International Network Services went public, selling

2,500,000 shares at $16.00 each. The issuing firm paid a commission to its investment

bankers, also known as a gross spread or underwriting discount, of $1.12 per share, or 7.0

percent. International Network Services was not alone in paying a 7.0 percent spread. In

1995-1998, for the 1,111 initial public offerings (IPOs) raising between $20 and $80

million in the U.S., over 90 percent of issuers paid gross spreads of 7.0 percent. Not 6.9

percent or 7.1 percent. Exactly 7.0 percent. It is widely accepted that there are fixed

costs associated with issuing securities, leading to economies of scale in the costs of

issuing debt, equity, and hybrid securities. For above moderate-size offerings, however,

no economies of scale are evident.

The clustering of spreads at 7.0 percent has not always been present. There is

much more clustering at 7.0 percent now than a decade ago, although the average spread

on IPOs has not changed during this period. In the 1985-1987 period, only about one-

quarter of moderate-size IPOs had spreads of exactly 7.0 percent, in contrast to the more

than 90 percent incidence that has prevailed in recent years.

When one examines IPOs outside of the U.S., such as in Japan, Hong Kong, or

Europe, one finds that the spreads are approximately half the level of those in the U.S.








And within the U.S., when one examines bond, convertible bond, and seasoned equity

offerings, one finds that the spreads do not show pronounced clustering on one number.l

This chapter examines several possible explanations for the high average spreads

on IPOs in the U.S., the striking fact that so many issuers pay exactly 7.0 percent, and the

increase in clustering during the past decade. I argue that the spreads for most deals

above $30 million are above competitive levels. One reason for this opinion is that, as

already mentioned, the spreads are much higher than elsewhere in the world. A second

reason is that if spreads were determined primarily by costs, the average spreads on $80

million deals would be lower than on $20 million deals. But they are not. And if costs

were the main determinant of spreads, $40 million deals by risky companies would have

higher spreads than $40 million deals by relatively easy-to-value firms. But they do not.

A third reason for the opinion is that investment bankers readily concede that the spreads

are high, "The fact is, we'd be cutting our own throats to compete on price."2

Ideally, we would like to have cost information to directly test whether the

underwriter spreads that are charged are equal to costs, including a competitive rate of

return on capital employed. Unfortunately, this information is proprietary. More

importantly, there would be problems in interpreting the numbers, since many of the costs

are hard to allocate, and the costs of operating a gold-plated operation are higher than for

a bare-bones operation. In other words, the costs are endogenous. High spreads induce


1 See the appendix table in Lee et al. (1996).

2 Attributed to the anonymous head of underwriting for an investment bank in Roger
Lowenstein's April 10, 1997 Wall Street Journal column.








underwriters to compete for business by adding additional services.

There are several features of the IPO underwriting market that are conducive to

high spreads. The importance of underwriter prestige results in a pecking order where

few issuers will turn down a "bulge bracket" underwriter for a less prestigious one, even

if it means paying higher fees. The importance of analyst coverage limits the number of

viable competitors for a given deal, and leads issuing firms to choose a lead underwriter

at least partly on the basis of characteristics unrelated to the fees charged.

Explanations that I present for the high average spreads, and the high frequency of

7.0 percent spreads, include the possibility of implicit or explicit collusion among

investment bankers. Among game theorists, the term "implicit collusion" is used to

denote an outcome in which sellers keep prices above competitive levels without

explicitly colluding. In the implicit collusion explanation, individual underwriters realize

that by undercutting spreads to win a deal, competitors will respond by charging lower

spreads in the future, resulting in a lower present value of profits. The self-interest of

each individual investment banker results in higher spreads than if the fees were at

competitive levels. The logic is based on the Dutta and Madhavan (1997) model of a

noncooperative game used to explain the high bid-ask spreads on Nasdaq stocks.

I also discuss several other, more innocuous, reasons for why IPO spreads are

high. These reasons include the necessity of having high spreads in order to induce

underwriters to do a credible job at certifying the quality of an offering. Also, while

spreads are the primary direct compensation of underwriters, the money "left on the

table" via the short-run underpricing of IPOs is an important indirect compensation, for

underwriters are able to allocate this money to their favored clients. Alternatively, low








spreads would encourage greater underpricing as a way of reducing the expected costs of

stabilizing IPOs, increasing the indirect costs to issuing firms. Another reason for high

spreads is that the prospect of future underwriting profits induces an underwriter to be

more concerned about its reputation, and therefore the "certification" function of

underwriting is enhanced. Yet another reason is that high spreads induce underwriters to

compete for business on the basis of analyst coverage, which enhances the liquidity of a

company's stock.

There are some similarities between patterns of gross spreads on IPOs and those

of Nasdaq bid-ask spreads. Prior to the publicity generated by Christie and Schultz

(1994), Nasdaq market makers avoided odd-eighth quotes. In other words, quotes at $10,

$10.25, and $10.50 were far more common than quotes at $10.125, $10.375, and

$10.625. Various parties have argued that the avoidance of odd-eighth quotes facilitated

either implicit or explicit collusion to keep the bid-ask spreads wide. (The evidence of

explicit collusion was strong enough to result in a lawsuit settlement of approximately $1

billion paid by market makers.3) In the IPO market, the avoidance of spreads that are not

exactly 7.0 percent could facilitate either explicit collusion or implicit collusion, in that it

is readily observable (from the prospectus) whether one underwriter is charging a fee that

is "too low." On the other hand, there is a tendency to have at least some clustering at

integers in other markets, including the London gold market (Ball et al., 1985) and

interest rates paid on bank deposits (Kahn et al., 1999) in spite of no requirement to avoid

noninteger prices.


3 See Wall Street Journal, November 10, 1998.








Explanations for the increased clustering over time are harder to come by. It is

possible that 7.0 percent has arisen as a focal point partly because issuers have placed

relatively little attention on fees, and underwriters find it easy to justify a spread by

pointing to previous deals done at the same spread.

The structure of the remainder of this chapter is as follows: Section 2 presents the

facts on the distribution of gross spreads on IPOs in the U.S. over the 1985-1998 period.

Section 3 discusses features of the IPO market that may facilitate high spreads. Section 4

discusses alternative explanations for the clustering of spreads at 7.0 percent. Section 5

discusses possible reasons for the increased clustering of spreads over time. Section 6

concludes this chapter.

The Facts


Data


I examine the spreads on 3,203 firm commitment IPOs from January 1985 to

December 1998 covered in the New Issues database of Securities Data Company (SDC).

Closed-end funds, American Depository Receipts (ADRs), real estate investment trusts

(REITs), and unit offerings are excluded from the sample. The sample is restricted to

equity IPOs with proceeds of at least $20 million, because the compensation for

underwriting smaller offerings is much higher due to the diseconomies of scale, and these

deals may be accompanied with underwriter warrants.4 Throughout the chapter, the


4 See Barry et al. (1991) for a description of underwriter warrants and related regulation.








proceeds exclude overallotment options, and are expressed in terms of dollars of 1997

purchasing power adjusted using the U.S. GDP implicit price deflator.5 It is worth noting

that in the U.S., buyers of IPOs pay no brokerage commission.

To examine the analyst coverage of IPO firms, I collected data from I/B/E/S

(Institutional Brokers Estimate System). Using broker translation codes provided by

I/B/E/S, I combined the SDC and I/B/E/S detailed history datasets and identified analyst

coverage provided by managing underwriters and unaffiliated brokerage firms. Of the

2,968 issuing firms in the period 1985-1997, 2,911 IPOs have some level of research

coverage tabulated by I/B/E/S within one year after the offering. The sample ends in

1997 because of this requirement.

Empirical Evidence of the Clustering of Underwriting Spreads at Seven Percent

Table 1 reports the number of offerings by calendar year, offering size, and gross

spread. The aggregate percentage numbers for each calendar year are plotted in Figure 1.

Inspection of Table 1 and Figure 1 reveals that the proportion of IPOs with a 7.0 percent

spread shows an upward trend until 1995, when the proportion stabilized at around 77

percent. For deals with proceeds of $20 million up to $80 million, which are referred to

as moderate-size deals, the increasing concentration at 7.0 percent is especially

noteworthy. This is illustrated in Figure 2.

At the bottom of Table 1, and in Figures 3 and 4, I report the distributions after

aggregating the years into three time periods: 1985-1987, 1988-1994, and 1995-1998. I



5 An overallotment option gives the underwriter the right, but not the obligation, to
purchase additional shares from the issuer. The spread per share on these incremental
shares is the same as for the rest of the issue.








also report the numbers by proceeds category: moderate-size deals, and large deals. In

1985-1987, 26 percent of the moderate-size IPOs occurred at 7.0 percent, with 46 percent

at lower spreads and 28 percent at higher spreads. By 1995-1998, only five percent of

these deals occurred at lower spreads, and only four percent occurred at higher spreads.

Fully 91 percent of these moderate-size IPOs paid a spread of exactly 7.0 percent.

While the clustering of spreads has increased over time, in Table 2, I show that the

average spread has remained virtually unchanged over the last 14 years.6 This includes

not only high-volume periods, but also the low-volume years following the October 1987

stock market crash. During the 1988-1990 period when there was relatively little equity-

issuing activity, presumably there was plenty of excess capacity in the underwriting

business.

In Figures 5 and 6, I show scattergrams of the relation between spreads and the

logarithm of, respectively, either expected proceeds (Figure 5) or actual proceeds (Figure

6), for IPOs from 1998. Expected proceeds are computed as the midpoint of the file price

range multiplied by the number of shares that are anticipated to be sold at the time of

filing the registration statement. Inspection of these figures shows the strong clustering of

spreads at 7.0 percent for deals with proceeds of $20 million (log(20) = 3.00) to $80

million (log(80) = 4.38). In Figure 7, I show the scattergram of the relation between

spreads and proceeds for the subset of the deals with proceeds of less than $100 million,



6 The low value-weighted spread for large IPOs in 1998 is attributable to three very large
IPOs (all over $2 billion) at gross spreads of approximately four percent. The spreads on
the Conoco, Fox Entertainment, and Infinity Broadcasting IPOs were in line with
predictions, given the size.








without logging the proceeds.

As reported in Table 1, of the 1,111 moderate-size IPOs in 1995-1998, 56 had

spreads of less than 7.0 percent and 45 had spreads of more than 7.0 percent, according to

SDC. When we inspect these 101 IPOs that apparently deviated from the 7.0 percent

spread norm, it turns out that 38 of them had spreads different from 7.0 percent because

of "rounding errors." Although there is no requirement that spreads be expressed as

pennies per share, if a firm has an offer price of $13.50 per share, the spread tends to be

either 94 cents (6.963 percent) or 95 cents (7.037 percent), rather than the 94.5 cents that

would make the percentage spread equal to 7.0 percent. If we classify these cases as 7.0

percent spreads, there are only 63 of the 1,111 moderate-size IPOs with non-seven

percent spreads. On closer inspection, several of these 63 were ADRs of foreign firms

that were misclassified by SDC, several were Canadian companies, 11 were IPOs where

the spreads were lower than 7.0 percent, but the expected proceeds from the preliminary

prospectus indicated an intention to raise at least $80 million, and five were IPOs where

the spreads were higher than 7.0 percent, but the expected proceeds from the preliminary

prospectus was less than $20 million. (The spread is generally negotiated before the

preliminary prospectus is issued.) This leaves only about 40 out of 1,111 moderate-size

IPOs with non-seven percent spreads. Thus, in 1995-1998, the clustering of spreads at

7.0 percent is actually even more extreme than the 91 percent of moderate-size deals

shown in Table 1. When the above reclassifications are made, it appears that at least 95

percent of IPOs with expected proceeds of $20-80 million paid 7.0 percent spreads.

Pricing Behavior of Individual Investment Bankers

Figure 3 shows that 91 percent of IPOs with proceeds of $20-80 million paid








exactly 7.0 percent as their underwriting spread in 1995-1998. To investigate the pricing

behavior of the most active underwriters, I calculate the market share for each investment

banker. The market share for each bank is defined as the sum of gross proceeds (in 1997

dollars) raised by the bank acting as the lead underwriter, divided by the sum of gross

proceeds (in 1997 dollars) raised in all IPOs, during 1995-1998. Figure 8 displays the

clustering behavior of the twelve IPO underwriters with the largest market shares.7 It

shows that the pricing behavior of these prominent investment bankers is quite uniform.

Spreads in Other Countries

Spreads on IPOs in other countries are much lower than in the U.S. For example,

the March 19, 1997 IPO of Cambridge Antibody Technology in the United Kingdom,

underwritten by Kleinwort, raised $66 million (U.S.) with a 3.36 percent spread

(composed of a 3.0 percent commission plus a 150,000 pound fee), according to SDC.8

In Japan, IPO spreads are typically in the range of 3-3.5 percent of proceeds for moderate-

size deals. In Taiwan, spreads are even lower, although potential buyers of IPOs also pay

a commission, unlike in the U.S.



7 The market shares in 1995-1998 are highly correlated with the market shares from
earlier in the 1990s reported in Beatty and Welch (1996, Table 3).

8 In 1998, the British government's Monopolies and Mergers Commission launched an
inquiry into possible price-fixing on equity offerings in the U.K. (see The Economist,
June 27, 1998). The focus of the British inquiry appears to be on the standard two
percent fee that is charged for "sub-underwriting" seasoned equity offerings. In the U.K.,
seasoned equity offerings are typically rights offerings priced at a discount of about 10
percent to the market price, with the price set several weeks before the exercise date. The
sub-underwriting fee is the price of the put option that the issuer has implicitly purchased.
Plausible estimates of the put value are substantially less than two percent.








There are certainly differences in the regulations and associated costs for

underwriting IPOs in different countries. One reason that is frequently advanced for

higher costs of underwriting in the U.S. is the lawsuit potential. But the reality is that

auditing firms and investment bankers have been successful in their strategy during the

1980s of fighting securities fraud lawsuits with such ferocity that plaintiffs attorneys

usually do not even bother suing them anymore (see Beatty and Welch, 1996). Instead,

directors and officers (D&O) insurance pays settlements in the typical securities class

action lawsuit, with underwriters untouched.

Spreads on Small and Large IPOs

For deals below $20 million, spreads are typically even higher than 7.0 percent, as

shown in Figure 7. This is consistent with substantial fixed costs in underwriting.

Revenue must be sufficient to cover the cost of writing a prospectus and conducting a

roadshow, as well as cover the costs of "busted" deals and the costs of prospecting for

business. The conventional wisdom is that the large, prestigious investment banking

houses have costs that are so high that they do not find it profitable to do these small

deals, which are primarily sold to a retail, rather than an institutional, clientele. These

small deals frequently include compensation for the underwriters that includes warrants to

buy stock in the issuing company, in addition to the stated gross spread (see Barry et al.,

1991). I do not analyze these small deals.

For large deals, which I define as IPOs with proceeds of $80 million or more,

average spreads are below 7.0 percent, and there is little clustering. Inspection of Figures

5 and 6 shows that economies of scale are clearly displayed for these large IPOs: the

bigger the deal, the lower the spread tends to be.








Spreads on Seasoned Equity Offerings

Given the clustering of IPO spreads, a natural question is whether the same

pattern is observed for the spreads on seasoned equity offerings (SEOs). Figures 9 and 10

show the spreads on SEOs, also known as follow-on offerings, from 1998. Figure 9 for

SEOs is analogous to Figure 6 for IPOs, showing spreads versus the natural logarithm of

proceeds. Figure 10 for SEOs is analogous to Figure 7 for IPOs, showing spreads versus

proceeds on deals below $100 million. Inspection of Figures 9 and 10 shows that there is

a slight tendency to prefer spreads that are at integers or half-integers (4.5 percent, 5.0

percent, 5.5 percent, etc.), but that this tendency is fairly mild. Figure 9 suggests that

there are economies of scale in conducting SEOs, with a nearly linear relation between

the log of proceeds and spreads. Most importantly, there is no tendency to cluster on a

single number for SEOs. For SEOs of a given size, there is considerable dispersion in the

spreads paid on different deals. A comparison of the figures for IPOs and SEOs suggests

that there is something special about 7.0 percent spreads when it comes to IPOs.

Summary

In sum, the facts show that there is pronounced clustering of gross spreads at

exactly 7.0 percent for almost all IPOs raising $20-80 million. This concentration

increased gradually during the 1980s and 1990s, to the point where in recent years well

over 90 percent of moderate-size IPOs in the U.S. have paid 7.0 percent spreads. In other

countries, spreads are substantially lower than in the U.S. The patterns suggest that gross

spreads are competitive for deals below $20 to $30 million, but increasingly profitable on

larger deals. There is no clustering on a single number that is observable for SEOs.








Explanations for High Spreads


At its most general level, this chapter is asking the following question: What

features of the IPO market structure are conducive to an equilibrium in which fees are

high?

When going public, an issuing firm typically conducts a "beauty contest" to

choose a lead manager (also known as the book manager) and one or two co-managers.

The conventional wisdom is that underwriter prestige and analyst reputation are of

paramount importance in this decision. Underwriters do not commit to a specific offer

price at the time an underwriting agreement is signed. Thus, competition on the basis of

valuations is muted.

Analyst Coverage

An implicit understanding is that the managing underwriters of an IPO will each

assign a securities analyst to cover the company and produce research reports and issue

buy recommendations for the stock (see Power, 1993; Rajan and Servaes, 1997; Dunbar,

1998; Michaely and Womack, 1998). For a small firm (in 1998 a firm with a market

capitalization of equity of $250 million is too small to be included in the Russell 2000),

there is a presumption that the stock price is affected by analyst coverage and whether

there are buy recommendations on the stock. "Buy" recommendations may be especially

important after the lock-up provision has expired, and insiders want to sell some of their

stock in the open market.9 In other words, the objective function of managers at the time



9 A lockup provision restricts pre-issue shareholders from selling shares while it is in
effect, without the explicit written permission of the managing underwriter. A typical
lockup provision is for 180 calendar days after the IPO.








of the offering includes raising money at the time of the offering, and raising money in

future open-market insider sales. Other shareholders benefit, too, from the enhanced

liquidity of their shares that is a consequence of the analyst coverage.

In the 1980s, many IPOs did not have co-managers. Today, almost all IPOs have

one, two, three, or even more co-managers. A reason for this growth in the number of co-

managers is that the issuing firm is essentially buying additional analyst coverage at no

incremental expense (since the underwriting fees will be 7.0 percent of proceeds whether

or not there are co-managers). In Table 3, I report the number of managers for moderate-

size (Panel A) and large IPOs (Panel B). Panel A shows that in 1985-1987, 37 percent of

moderate-size IPOs were solely managed, whereas in 1995-1998, only four percent were.

In Table 4, I report the number of analyst forecasts within one year of an IPO

reported by I/B/E/S. In Panel C, I report two regressions where the number of analysts

making an earnings forecast within a year of the IPO is the dependent variable. The top

row reports the results of a pooled time series-cross section regression. The second row

reports the average coefficients from a time series of thirteen regressions, each of which

uses the IPOs from one calendar year during 1985-1997. The R2 reported for this second

regression is the average of the thirteen R2 values, and the t-statistics are based upon the

time-series standard deviation of the coefficients. Both regressions find that more

analysts follow an IPO if it is larger, and if it has a higher first-day return. These results

are consistent with Rajan and Servaes (1997, Table II). Of particular interest, however, is

the finding that an incremental co-manager adds 0.36-0.55 net analysts, holding the

proceeds and first-day return constant.

Securities analysts are beneficiaries of this system. Analysts with good








reputations (as measured, for instance, by the annual Institutional Investor magazine all-

star rankings) can command a high salary and bonuses. Analysts who help bring in equity

financing business also stand to receive large bonuses (Siconolfi, 1992; Smith, 1996;

Raghavan, 1997). Krigman et al. (1998) report survey evidence that issuers cite analyst

coverage as a main determinant for choosing underwriters.

The importance of analyst coverage represents a potential barrier to entry for new

underwriters. Without having a well-regarded analyst, issuers will be skeptical about the

ability of an underwriter to successfully tout the stock in the aftermarket, or place it

initially. Furthermore, by emphasizing industry expertise, the IPO underwriting business

becomes one of differentiated products, reducing the number of viable competitors for

any given deal.

Underwriter Prestige

Underwriter prestige is a second important criterion for choosing managers.

There is a perception that the "certification" of a prestigious underwriter is very valuable

to an issuing firm. While underwriter prestige can and does change over time, the most

prestigious underwriters today include Goldman Sachs, Morgan Stanley Dean Witter, and

Merrill Lynch. As long as issuing firms choose a lead underwriter primarily on the basis

of analyst and investment banker reputations, there is little incentive for underwriters to

charge differential gross spreads, for the elasticity of demand is not high with respect to

the fees charged. Competition to be a lead manager is thus focused on the intermediary's

"quality" rather than the fees charged.

For intermediaries to have an incentive to certify the value of the item being sold,

a stream of future quasi-rents must be anticipated (Beatty and Ritter, 1986; Booth and








Smith, 1986). Quasi-rents are the cash flows above marginal cost that can be viewed as a

return on the prior investment in establishing a reputation. This framework would

suggest that above-competitive spreads are needed to give underwriters an incentive to

turn down deals that may be attractive in the short-run, but would be bad for investors in

the long-run.

While above-competitive underwriter compensation is needed to induce

investment bankers to perform a certification function, it is not clear why this should

result in spreads of 7.0 percent for essentially all deals, whether they are $20 million or

$80 million. Given the economies of scale that exist in the cost structure, the 7.0 percent

pricing structure results in substantial profits on deals at the high end of the moderate-size

range.

Underwriting Syndicates

Once an issuer chooses a book manager and co-managers, the lead manager

invites other underwriters into the underwriting syndicate. Typically, the syndicate is

split into several brackets (see Carter et al., 1998 for a description), depending upon how

many syndicate members there are. Because the fees are shared among the syndicate

members, at first glance the resulting revenue sharing might be viewed as conducive to a

reduced competitive environment.

Historically, syndicates existed partly for regulatory capital requirement and risk-

sharing purposes, and partly to facilitate the distribution of an issue. This was

particularly relevant when the lead underwriter did not have a significant retail or

institutional distribution network, and had limited capital. Today, there is little reason to

form a syndicate to perform the traditional economic roles of risk sharing, distribution,








and meeting capital requirements. Not surprisingly, syndicate size as measured by the

number of participating firms has fallen over time, even as the number of co-managers

has grown. Underwriters such as Merrill Lynch, with their large institutional and retail

distribution networks, do not need other investment bankers to assist in distributing a

given issue. And with their large capital bases, risk sharing would seem to be important

only for the very largest issues.

In recent years almost all IPOs have not only had a book manager, but also one or

two co-managers. Table 5 presents a hypothetical example with Goldman Sachs as lead

underwriter and BT Alex. Brown as the co-manager. We assume that there are thirteen

other members of the underwriting syndicate, where the two managers underwrite

900,000 shares each, the seven members of the major bracket underwrite 100,000 shares

each, and the six members of the submajor bracket underwrite 50,000 shares each. The

total number of shares is 2,800,000, before a 15 percent overallotment option, and gross

proceeds at a $12.00 offer price would be $33,600,000 net of the overallotment option.

We assume a gross spread of 7.0 percent, which has three components, as shown in Panel

A of Table 5. Of the 84 cent gross spread, we assume a selling concession of 48 cents,

an underwriting fee of 19 cents, and a management fee of 17 cents.10 A syndicate

member would receive 48 cents for each share whose sale is attributed to that member. In

general, there is no necessary relation between the number of shares underwritten and the



10 A management fee goes to managing underwriters for managing the deal. An
underwriting fee goes to managing underwriters and syndicate members for assuming the
risk of purchasing securities from the issuer. A selling concession is given to managing
underwriters, syndicate members, and selling group members for distributing securities to
investors.








selling credits earned. Normally, the vast majority of the shares sold will be credited to

the book manager, as illustrated in Panel B of Table 5. All 15 syndicate members would

receive the underwriting fee of 19 cents per share, minus underwriting and stabilization

expenses, for each share underwritten. The managing underwriters would receive 17

cents on every share sold by any member, with the split between the lead and co-

managing underwriters usually tilted in the lead manager's favor.

In a typical IPO, the vast majority of revenue and profits goes to the book

manager, as illustrated in Panel C of Table 5. The book manager receives at least a

proportionate share of the management fee revenue, the majority of the selling concession

revenue, and part of the net underwriting fee revenue. This last item is typically a small

number, and may even be negative if stabilization expenses are high.

There are certain ongoing expenses that lead and co-managers have, such as the

pay of analysts and corporate finance employees, so the revenue figures are not the same

as profit figures. Furthermore, part of the revenue is merely a competitive return on the

capital required by regulators to underwrite securities. But the example in Table 5 shows

how lucrative it can be to be a lead manager or co-manager on a large IPO. Thus,

although the fees are shared among syndicate members, there is still fierce competition to

be a lead manager.

It is noteworthy that new entrants to the IPO underwriting market have not tried to

gain market share by cutting spreads. The two most prominent new entrants in IPO

underwriting in the 1990s have been Deutsche Bank Securities (formerly Deutsche

Morgan Grenfell (DMG)) and Friedman Billings Ramsey. Both firms have charged 7.0

percent spreads on moderate-size IPOs in the mid-1990s.








The Cost of Price Support

Underwriters frequently stabilize, or support, the price of an IPO immediately

after it has gone public. Price support involves the practice of being a net buyer of shares,

which are retired if the underwriter overallocated the issue. An implicit commitment to

stabilize an IPO can be viewed as the underwriter writing a put option with an exercise

price equal to the offer price. The anticipated value of this put option is, in principle, one

of the determinants of gross spreads. The value of this put option can be reduced by

decreasing the offer price relative to the expected aftermarket price. Thus, if spreads

were lower, underwriters might be tempted to set lower offer prices, in order to reduce the

value of the puts that they are writing. The indirect cost of this underpricing might be far

larger than the savings in direct costs on the spreads.

A desire by issuers to reduce the amount of money left on the table by paying high

spreads is plausible. This does not explain, however, why there should be clustering at

7.0 percent. Furthermore, an issuer which wanted to increase incentives for stabilization

could insist on a change in the components of the gross spread, reducing the selling

concession and/or management fee, and increasing the underwriting fee. In any case,

Aggarwal (1998) reports that empirically the costs of stabilization to underwriters are

relatively small.

Spreads Signal High Quality Underwriting

In many other markets, such as medical specialties, consulting, and legal advice,

professionals find that charging a low price for their services signals lower quality, and

results in lost business. This phenomenon is present in markets where clients are unable

to easily evaluate quality on either an ex ante or ex post basis, and where repeat dealings








are uncommon. In these markets, sellers of services tend to charge high prices and at

least some market participants have excess capacity (which they attempt to hide because

its existence might signal low quality). An underwriter charging a low fee might raise

concerns about its willingness to engage in price stabilization, provide analyst coverage,

exercise care in helping to write a prospectus, aggressively market a deal, etc. The quality

of all these underwriting services is difficult to contractually specify, and difficult to

evaluate. As a result, in equilibrium, no underwriter charges a lower fee because the

lower fee will not generate additional clients.

IPO underwriting seems to share many of the characteristics of markets where

price signals quality. But while any given firm will generally go public just once, limiting

repeat business and learning, this is too myopic a view. Reputation effects overcome

some of the information asymmetries: if Goldman Sachs decided to cut its spreads, few

issuers would suspect that it had become a low-quality underwriter.

Explanations for the Clustering of Spreads at Seven Percent


Explicit Collusion

One possible explanation for the clustering of spreads at 7.0 percent is collusion.

If underwriters compete for business on the basis of spreads that they charge, competition

will drive the spread on any given deal to the cost of providing the services, including

compensation for expected risks that are borne by the underwriter. If underwriters agree

to form a cartel, they can increase their profits. On every deal, a mechanism to decide

how much to charge would be needed. One possible arrangement is to agree to always

charge the same fees (7.0 percent), with the profits shared. The existence of syndicates








would seem to be an excellent way to share the profits.

With literally scores of people involved in setting spreads at different investment

banking firms, the ability to explicitly collude and keep it a secret strains credibility.

Legal liability is also a deterrent.11 And while underwriting syndicates could in principle

be used for sharing profits, in practice the lead underwriter grabs the lion's share of

profits.

Implicit Collusion

While explicit collusion is difficult to coordinate, it is possible that investment

bankers are of the opinion that if they compete aggressively on the basis of fees, spreads

will be driven down to the point where there will be little money for year-end bonuses for

the individuals involved. Thus, the individuals involved may act strategically to avoid

turning IPO underwriting into a "commodity business." Implicit collusion requires that

each underwriter realizes that high spreads result in lots of money for year-end bonuses

that would be jeopardized if spreads get driven down to competitive levels due to

cutthroat competition. As long as the present value of the future cash flows resulting

from high spreads is greater than the short-term gains associated with undercutting the

competition to win a deal, each underwriter will avoid cutting its spread.

The logic behind the implicit collusion explanation for high spreads can be

formalized in a noncooperative dynamic game (see Chapter 3), based on the Dutta and

Madhavan (1997) model. With IPOs, it is quite plausible that underwriters fear that



11 NASD Notice to Members 98-88, issued in October 1998, reminds underwriters that
there is no standard level of underwriter compensation, and that coordination among
members on the gross spreads charged is explicitly prohibited. This notice is partly
motivated by the pattern of clustering at 7.0 percent spreads on IPOs.








quoting a lower spread will set off a price war that will drive gross spreads down on

future deals. After all, many individuals in the business state that they do not want to

charge a lower spread because they "don't want to turn it into a commodity business." It

is hard to think of stronger evidence to support the proposition that the participants are

thinking strategically. In other words, they are forecasting the spreads that will prevail in

the future based upon what is done today, and acting accordingly. Thus, each underwriter

may decide to keep its spread above competitive levels, even without explicit collusion.

The implicit collusion argument also offers an explanation for why the spreads on

deals above $80 million are generally lower than 7.0 percent. Since there are economies

of scale in the costs of underwriting IPOs, deals above $40 or $50 million are for the most

part extremely profitable at a 7.0 percent spread. If the profits on a deal are too large,

each underwriter has an incentive to undercut the competition, even if it means

jeopardizing all of the future profits from high spreads. In order to forestall a price war

from breaking out, underwriters must limit the economic profits earned on any given deal

to a "reasonable" level (see Rotemberg and Saloner, 1986; Dutta and Madhavan, 1997;

Chapter 3). Beyond a certain level of proceeds size, spreads of 7.0 percent are

unsustainable.

The implicit collusion explanation for high average fees and clustering at 7.0

percent raises the following question: Why, since sellers in every industry prefer high

prices to low prices, is an implicit collusion equilibrium sustainable with IPO spreads and

not in most other businesses? The answer is that customers view the fees charged as just








one of a set of characteristics on which to choose an underwriter.12 If customers were

more focused on fees, the implicit collusion equilibrium would be harder to sustain

relative to the competitive equilibrium.

Resolving Agency Costs with Multiple Principals

There are many other markets for intermediary services which display strong

clustering of fees at integers. Probably the most obvious example is brokerage

commissions on residential real estate, where in many cities, almost all transactions are

done at 6.0 percent. The real estate market involves both listing agents (representing

sellers) and buying agents. Each property has a vector of characteristics and different

buyers have different tastes, resulting in a time-intensive matching problem. In real

estate, agents representing buyers observe the fees being offered on a property through the

multiple listing service books, and they will steer clients away from properties that do not

compensate the agents well. Because real estate agents are representing multiple clients,

charging a uniform commission eliminates the incentive of agents to spend a

disproportionately low amount of effort on properties offering lower commissions

(Williams, 1998). Also, there is less homogeneity of real estate percentage fees than it

appears. In fact, real estate fees can be negotiated in two ways. First, on properties that

ex ante appear easy to sell, a seller can get a listing agent to rebate part of the listing fee.

Second, when a buyer makes a bid on a property, the seller can respond with a reservation

price (net of commission) where the difference is too small to give the brokers their full



12 Krigman et al. (1998) report survey evidence that several factors are often considered
important when issuers choose underwriters. Those factors include underwriter's
reputation and status, underwriter's industry expertise and connections, and quality and
reputation of analysts.








commission. Rather than start all over with a new buyer or a new house, the agents can

be expost "squeezed." In other words, there is a hold-up problem.

There are several important distinctions between the markets for intermediaries in

real estate and for IPOs. With financial securities, the objective function of both buyers

and sellers is dependent primarily upon a single attribute, the cash payoffs. So unlike real

estate, there is not a time-intensive problem of matching desired characteristics. Thus, a

uniform commission so that one client is not favored over another client is not needed

with IPOs.

Other Components of Compensation

If underwriter compensation is composed of more than just the spread, the

clustering of spreads at 7.0 percent may give a misleading view of the degree of

clustering of the total compensation. On small offerings, underwriter warrants,

"nonaccountable expense allowances," and other additional underwriter compensation is

common. But for deals above $20 million, industry practice is to have the gross spread

represent all of the underwriter's compensation. Inspection of a random sample of

prospectuses in 1997 for IPOs with proceeds of $20-25 million found no cases of

nonaccountable expense allowances boosting underwriter compensation. It should also

be noted that almost all IPOs have a 15 percent overallotment option, so there is almost

no time-series or cross-sectional variation on this dimension.

One dimension on which there is substantial cross-sectional variation, however, is

the degree of short-run underpricing. On average, the first-day return on IPOs is 10-15

percent (see Lee et al., 1996, for example), which is an indirect cost of going public. The

first-day returns represent profits to investors (and an opportunity cost to issuers) that are








approximately twice as large as the direct fees received by underwriters. For example, on

a $40 million IPO with a 7.0 percent spread and a first-day return of 14 percent, the direct

fees are $2.8 million, and the money "left on the table" is $5.6 million. In this example,

investment bankers have $5.6 million in profits to hand out to favored clients, such as

clients who are willing to overpay for other services.

Since in the U.S. the offer price is typically not set until the day before trading

commences, ex ante negotiation of the gross spread does not eliminate the potential for

hold-up problems. This is because the number of shares and the offer price are not

negotiated until after almost all costs (other than stabilization costs) are sunk. Because it

is thus not clear whether the ex ante or ex post proceeds should be used to examine

whether there is a substitution between spreads and underpricing, I report results both

ways.

In Table 6, I divide the IPOs from 1985-1998 into three subperiods and proceeds

categories based upon (i) the actual gross proceeds (Panel A), and (ii) the expected gross

proceeds, calculated using the midpoint of the file price range (Panel B). Within each

category, I then report the average first-day return for IPOs with spreads (i) below 7.0

percent, (ii) equal to 7.0 percent, and (iii) above 7.0 percent. Visual inspection of Panels

A and B fails to find any obvious monotonic tradeoff between the spread and average

first-day returns. One pattern that does emerge, however, is the higher underpricing

during 1995-1998 than during 1985-1987.

In Panel C, I report pooled cross section-time series regression results for

moderate-size IPOs with the percentage first-day return as the dependent variable. The

explanatory variables include the logarithm of actual gross proceeds, the percentage offer








price revision, a dummy for a gross spread below 7.0 percent, a dummy for a gross spread

above 7.0 percent, and yearly fixed effects dummy variables. Consistent with Hanley's

(1993) empirical work on partial adjustment, revisions in the offer price reliably predict

the first-day return. For both the entire 14 year sample period, and for the 1995-1998

subperiod, IPOs with spreads above or below 7.0 percent have a lower first-day return

than those with spreads of exactly 7.0 percent, although the patterns are not always

statistically reliable. The lower first-day return on IPOs with low spreads is consistent

with the hypothesis that lower spreads are found on low-risk (and hence low-return) IPOs.

Though the negative coefficients on the dummy variables for above 7.0 percent spreads

are consistent with the notion that reduced underpricing is a substitute for a high spread,

the statistical significance is not present in the subperiod.

Another form of compensation for the lead underwriter is the profits from future

market making activity. Ellis et al. (1998) report that the lead underwriter usually is the

most active market maker in the first sixty days of post-IPO trading for a sample of 312

Nasdaq-listed IPOs from 1996-1997. They calculate that the lead underwriter makes

money on this market making activity, although the numbers are modest relative to the

fees from the gross spread.

There is no regulatory constraint that forces spreads to equal 7.0 percent. National

Association of Securities Dealers (NASD) Rule 2710 prohibits a member from

participating in a public offering with unfair or unreasonable underwriter compensation,

where NASD Regulation's Corporate Financing Department has direct responsibility for

the review of underwriter compensation. There is no evidence that this rule is a binding

constraint for these moderate-size and large IPOs.








Resale Price Maintenance

As discussed in Section 3, the conventional wisdom is that future analyst coverage

is an important consideration for firms going public. Thus, investment bankers are

intermediaries who are selling a bundle of services: the IPO underwriting itself, and

future analyst coverage. By paying above-competitive underwriting fees, issuing

companies induce underwriters to offer more analyst coverage in their attempt to compete

for the profitable business. The logic is the same as that of the "resale price maintenance"

literature, where a producer wants distributors to offer a minimum service level (Telser,

1960). By setting a minimum price at which the products can be resold, the producer

induces the intermediaries to offer more services than they otherwise would.

While high spreads on average can be viewed as compensation for inducing

underwriters to provide future analyst coverage, this does not explain why 7.0 percent

spreads on $80 million deals are the norm, just as they are on $20 million deals. The

clustering at 7.0 percent for almost all deals within a very large range of proceeds

suggests that the larger and safer deals are providing substantial economic profits to the

underwriters involved.

Presumably, some issuers would prefer to pay high fees and "purchase" a high

level of services. Other issuers would prefer fewer services in return for lower fees. But

the lack of dispersion of spreads suggests that this choice is not available to issuers.

Cross-subsidization

One possible reason for the clustering of spreads at 7.0 percent is that

underwriters have difficulty knowing their exact costs on a given deal, and they price

their services in a manner whereby some issuers subsidize other issuers. As long as








underwriters break even on average, there is little reason to change this policy. The

problem with this argument is that, given the economies of scale that exist, it does not

take a rocket scientist to realize that $80 million deals are more profitable than $20

million deals if the percentage spread is the same on both. An underwriter could increase

its profitability by concentrating on the larger deals.

Possible Reasons for the Increased Clustering of Spreads over Time


As shown in Table 1 and Figures 1-4, the clustering of spreads has increased

substantially in the last decade. This raises the question, why?

One possibility is that, as in almost all markets, learning has occurred over time.

In the mid and late 1970s, IPO volume in the U.S. (and almost all other countries) was

virtually nonexistent, with the number of deals per year less than the number in many

weeks during the 1990s. As IPO volume picked up in the 1980s, four boutiques (L.F.

Rothschild, Unterbeg, Towbin; Robertson Stephens; Hambrecht & Quist; and Alex.

Brown) specializing in IPO underwriting captured a large share of a growing market. In

the early 1980s, "bulge bracket" investment bankers such as Goldman Sachs, Morgan

Stanley, and Merrill Lynch did relatively few IPOs. By the mid-1980s, the bulge bracket

firms started to get more involved, and L.F. Rothschild went out of business after a falloff

in IPO volume after the 1987 market crash.

In the 1980s, there was more heterogeneity of spreads, less concentration of

underwriters, and sole managers were more common. While the average spread on IPOs

has not changed, even as clustering has increased, this is not evidence that spreads are at

competitive levels. In many related markets, such as the fees on mergers and








acquisitions, investment banker fees have fallen from the mid-1980s to the mid-1990s.

The fact that IPO spreads have not fallen is consistent with the existence of a market

structure that is conducive to implicit collusion.

Several other reasons may explain the increased clustering over time. First,

precedent is important. It is easier to justify a given spread to a client if an underwriter

can point to other recent deals at the same (or higher) spread. Charging a higher than 7.0

percent spread might have become increasingly unattractive as competitors used it to

dissuade a potential client from going with an expensive underwriter. Also, in the 1980s

it may have been more common to negotiate the spread at the pricing meeting

immediately before an offering, at the same time that the offering price and number of

shares to be issued are negotiated (see Uttal, 1986 for a description of the negotiation of

the spread in the Microsoft IPO).

Summary and Conclusions


This chapter presents evidence of the clustering of gross spreads on IPOs at 7.0

percent, with the concentration of 7.0 percent spreads increasing during the 1990s. For

offerings with proceeds of $20-80 million (in dollars of 1997 purchasing power), at least

90 percent of IPOs during 1995-1998 had spreads of exactly 7.0 percent. For

comparison, only 26 percent of moderate-size IPOs in 1985-1987 had 7.0 percent spreads.

There is widespread agreement that fixed costs exist in underwriting IPOs, yet investment

bankers charge the same 7.0 percent spread on $20 million deals as they do on $80

million deals. The average spread on IPOs has remained virtually constant during the

1985-1998 period, in contrast to declining fees for mergers and acquisitions, etc. Spreads








on U.S. IPOs are roughly twice as high as in other countries.

I argue that for most IPOs with gross proceeds larger than $30 million, spreads are

above competitive levels in the U.S. The high average spread and the concentration of

spreads at 7.0 percent is consistent with implicit collusion on the part of investment

bankers. In other words, even though investment bankers are acting independently,

average spreads are above competitive levels. Several features of the IPO underwriting

market are conducive to spreads above competitive levels. The importance of analyst

coverage and buy recommendations, and the perceived importance of underwriter

prestige, facilitate high spreads.

If gross spreads are above competitive levels, investment bankers have an

incentive to use nonprice competition to attract deals. Although issuing firms face high

and, for moderate-size deals, nonnegotiable spreads, issuers can still bargain on another

dimension. In particular, by insisting on additional co-managers, issuing firms receive

more extensive analyst coverage. I show that the number of co-managers has increased

over the last decade, and that adding an additional co-manager adds between 0.36 and

0.55 net analysts following the stock. Highly ranked analysts have benefited, as their

compensation has been bid up as underwriters use the implicit promise of favorable

coverage and buy recommendations to compete for business. Investment bankers are also

able to use analyst coverage as a means for product differentiation, relaxing price

competition. There is a further aspect to industry specialization by analysts that is

relevant. To the degree that the only underwriters that would be viable competitors are

those with a well-regarded analyst, an investment banker that undercuts spreads on IPOs

in one industry cannot expect to gain market share in IPOs from other industries. To the








degree that the broader market is split into sub-markets, the gains from undercutting the

spread for an underwriter are limited to the submarket, increasing the sustainable spread

(see Dutta and Madhavan, 1997, pp. 260-261). In this respect, industry specialization by

investment bankers is analogous to payment for order flow on Nasdaq stocks.

The evidence is consistent with the implicit collusion explanation. Investment

bankers realize that if one investment banker tries to win business by cutting spreads, the

underwriting industry is likely to move to an equilibrium with low spreads, and lower

compensation for corporate finance employees.

With approximately $40 billion per year in IPO volume in recent years, if spreads

on average have been one percent too high, issuing firms have paid excess investment

banking fees of $400 million per year. If spreads are above competitive levels, it is clear

that analysts and corporate finance specialists at investment banking firms have benefited.

But who has been hurt? To the degree that investors have incorporated excessive fees

into their forecasts of future corporate expenses, and priced the securities appropriately,

they have not been hurt. But the founders of the companies, and the founders of other

companies that might have gone public but never did, have suffered because the costs of

raising external capital are excessive.

In contrast to the pattern for IPOs, there is little clustering of spreads on follow-on

offerings, and economies of scale are evident for all proceeds sizes. There is some

evidence that gross spreads on follow-on offerings have come down a little in recent years

(see Beatty et al., 1998; Gande et al., 1998). Trade journals have attributed this to

competition from commercial banks, which are trying to enter the underwriting business.

It remains to be seen how big an impact this has on the gross spreads on follow-on








offerings, and whether this competition has any effects on the gross spreads on IPOs. In

the year after Nationsbank, Bankers Trust, BankAmerica, and BancBoston bought

investment banking firms that specialized in IPOs, there does not seem to have been any

impact in the IPO market. Another source of competition may emerge from the

innovation of internet technology. For example, new underwriters E*Offering and W. R.

Hambrecht threaten to undercut the 7.0 percent fee that is now standard.13 Only time will

tell whether this changes the gross spreads that prevail in the IPO underwriting industry.

































13 See the Wall Street Journal article by Buckman (1999) and the article of Fortune
magazine by Tully (1999).










Table 1: Number of IPOs by Year, Proceeds, and Gross Spread, 1985-1998

The sample consists of 3,203 firm commitment initial public offerings (IPOs) with proceeds of at
least $20 million before the exercise of the overallotment option. Securities Data Co. is the
source of the data. Closed-end funds, REITs, ADRs, and unit offerings are excluded. The
amount of proceeds is expressed in terms of dollars of 1997 purchasing power, using the U.S.
GDP implicit price deflator. There are three categories of gross spreads expressed as a
percentage of proceeds: below 7%, 7%, and above 7%.


Proceeds Gross Spread
Year (Millions) Below 7% 7% Above 7% All
20-79.99 38 18 22 78
1985 80-up 16 1 3 20
All 54(55%) 19(19%) 25(26%) 98(100%)
20-79.99 92 38 51 181
1986 80-up 42 8 8 58
All 134(56%) 46(19%) 59(25%) 239(100%)
20-79.99 51 49 39 139
1987 80-up 29 4 3 36
All 80 (46%) 53 (30%) 42 (24%) 175 (100%)
20-79.99 16 22 10 48
1988 80-up 13 3 0 16
All 29(45%) 25(39%) 10(16%) 64(100%)
20-79.99 13 32 8 53
1989 80-up 13 2 0 15
All 26(38%) 34(50%) 8(12%) 68(100%)
20-79.99 13 37 14 64
1990 80-up 12 0 0 12
All 25 (33%) 37 (49%) 14(18%) 76(100%)
20-79.99 26 109 27 162
1991 80-up 46 2 0 48
All 72(34%) 111(53%) 27(13%) 210(100%)
20-79.99 25 155 18 198
1992 80-up 62 4 1 67
All 87(33%) 159(60%) 19(7%) 265(100%)
20-79.99 26 223 18 267
1993 80-up 68 10 0 78
All 94 (27%) 233 (68%) 18(5%) 345 (100%)
20-79.99 18 153 16 187
1994 80-up 35 5 0 40
All 53 (23%) 158 (70%) 16(7%) 227 (100%)
20-79.99 14 248 12 274
1995 80-up 53 12 0 65
All 67(20%) 260(77%) 12(3%) 339(100%)
20-79.99 21 348 14 383
1996 80-up 74 26 2 102
All 95 (20%) 374(77%) 16(3%) 485(100%)
20-79.99 15 256 16 287
1997 80-up 59 30 1 90
All 74 (20%6) 286 (76%) 17(4%) 377(100%)
20-79.99 6 158 3 167
1998 80-up 45 23 0 68
All 51(22%) 181 (77%) 3 (1%) 235 (100,%)









Table 1. Continued


20-79.99 181(46%) 105(26%) 112(28%) 398(100%)
1985-87 80-up 87(76%) 13(12%) 14(12%) 114(100%)
All 268(52%) 118(23%) 126(25%) 512(100%)
20-79.99 137(14%) 731(75%) 111 (11%) 979(100%)
1988-94 80-up 249(90%) 26(10%) 1( 0%) 276 (100%)
All 386(31%) 757(60%) 112( 99%) 1255(100%)
20-79.99 56 (5%) 1010(91%) 45( 4%) 1111(100/)
1995-98 80-up 231 (71%) 91(28%) 3 (1%) 325 (100%)
All 287 (20%) 1101(77%) 48( 3%) 1436(1000%)








Table 2: Average Gross Spreads (%) by Year and Size of IPOs, 1985-1998
The sample consists of 3,203 firm commitment initial public offerings (IPOs) in 1985-1998 with
proceeds of at least $20 million before the exercise of the overallotment option. Closed-end
funds, REITs, ADRs, and unit offerings are excluded from the sample. The amount of proceeds
is expressed in terms of 1997 dollars, using the U.S. GDP implicit price deflator. IPOs with
proceeds of at least $20 million but less than $80 million are designated as moderate-size IPOs,
others are designated as large IPOs. VW spread denotes value-weighted spread with proceeds
being the weight. EW spread denotes equally weighted spread. Numbers in parentheses
represent the standard deviation.


Moderate-size IPOs


SVW EW
Year
spread spread
1985 6.88 6.96
(0.47)
1986 6.89 6.94
(0.40)
1987 6.96 7.00
(0.40)
1988 6.85 6.92
(0.27)
1989 6.95 6.97
(0.12)
1990 6.99 7.03
(0.43)
1991 6.99 7.01
(0.18)
1992 6.99 7.01
(0.26)
1993 6.98 6.99
(0.16)
1994 6.94 6.97
(0.21)
1995 6.98 6.98
(0.13)
1996 6.98 6.99
(0.15)
1997 6.95 6.98
(0.30)
1998 6.97 6.98
(0.18)


Large IPOs All IPOs

VW EW VW EW
spread spread spread spread
6.08 6.38 6.40 6.84


5.83

5.55

6.01

5.89

5.83

5.82

5.78

5.75

5.94

5.82

5.74

6.01

5.05


(0.64)
6.33
(0.71)
6.08
(0.80)
6.14
(0.64)
6.16
(0.59)
5.98
(0.62)
6.03
(0.80)
6.04
(0.57)
6.14
(0.59)
6.09
(0.53)
6.15
(0.63)
6.38
(0.63)
6.36
(0.70)
6.14
(0.92)


6.20

6.04

6.33

6.30

6.45

6.28

6.24

6.23

6.44

6.34

6.27

6.37

5.52


(0.56)
6.79
(0.56)
6.82
(0.63)
6.72
(0.52)
6.80
(0.45)
6.86
(0.60)
6.79
(0.58)
6.77
(0.56)
6.80
(0.47)
6.82
(0.45)
6.82
(0.44)
6.86
(0.41)
6.83
(0.50)
6.74
(0.64)


r f










Table 3: Trend in the Number of Managers in Underwriting Syndicates, 1985-1998

The sample consists of 3,203 firm commitment IPOs in 1985-1998 with proceeds of at least $20
million (1997 purchasing power) before the exercise of the overallotment option. Closed-end
funds, REITs, ADRs, and unit offerings are excluded from the sample. IPOs with proceeds of at
least $20 million but less than $80 million are designated as moderate-size IPOs, others are
designated as large IPOs. If there is a sole manager on a deal, the number of managers is one. If
there is one co-manager on a deal, the number of managers is two.

Number of Managers
Years 1 2 3 4 All Mean # of Median # of
or more Managers Managers
Panel A: Moderate-size IPOs
N 146 184 55 13 398 1.9 2
1985-1987 % of IPOs 37% 46% 14% 3% 100%
% of Proceeds 36% 44% 16% 4% 100%
N 137 648 182 12 979 2.1 2
1988-1994 % of IPOs 14% 66% 19% 1% 100%
% of Proceeds 12% 65% 21% 2% 100%
N 46 581 415 69 1,111 2.5 2
1995-1998 % of IPOs 4% 52% 38% 6% 100%
% of Proceeds 4% 47% 41% 8% 100%
N 329 1413 652 94 2,488 2.2 2
All % of IPOs 13% 57% 26% 4% 100%
% of Proceeds 12% 53% 30% 5% 100%
Panel B: Large IPOs
N 12 48 22 32 114 2.9 2
1985-1987 % of IPOs 11% 42% 19% 28% 100%
% of Proceeds 8% 32% 22% 38% 100%
N 27 86 93 70 276 2.9 3
1988-1994 %ofIPOs 10% 31% 34% 25% 100%
% of Proceeds 8% 26% 29% 37% 100%
N 5 47 107 166 325 3.8 4
1995-1998 %ofIPOs 1% 15% 33% 51% 100%
% of Proceeds 1% 8% 25% 66% 100%
N 44 181 222 268 715 3.3 3
All % of IPOs 6% 25% 31% 38% 100%
% of Proceeds 5% 18% 26% 51% 100%









Table 4: Analyst Following by Offering Size, Subperiod and Number of Managers
in the Underwriting Syndicate, and Regression Analysis, 1985-1997

The sample consists of 2,911 firm commitment IPOs in 1985-1997 with proceeds of at least $20
million before the exercise of the overallotment option and covered by I/B/E/S. The analyst
following information is from IIB/E/S. Closed-end funds, REITs, ADRs, and unit offerings are
excluded from the sample. The amount of proceeds is expressed in terms of 1997 dollars, using
the U.S. GDP price deflator. IPOs with proceeds of at least $20 million but less than $80 million
are designated as moderate-size IPOs, others are designated as large IPOs. Panel A and B report
the summary statistics of analyst following. Number of full coverage represents number of IPOs
in which all managers issue earnings forecasts within one year after the offering. The first row in
Panel C reports the results of pooled cross-sectional and time-series regression analysis. The
amount of proceeds is measured in millions of dollars. The sample size is reduced because of the
requirement of first-day return which is defined as the percentage return form the offering price
to the closing price of the first trading day. Numbers in parentheses are t-statistics calculated
using White (1980) robust standard errors. The second row reports average parameter values
from yearly cross-sectional regressions using the approach of Fama and MacBeth (1973).

Number of Managers
Years 1 2 3 4 or more All
Panel A: Moderate-size IPOs
1985-1987 1. Number of IPOs 133 177 52 12 374
2. Number of Full Coverage 94 106 13 0 213
3. Mean Manager Forecasts 0.71 1.45 1.90 1.50 1.25
4. Mean Unaffiliated 1.75 2.00 1.96 1.67 1.90
Forecasts
1988-1994 1. Number of IPOs 131 647 181 12 971
2. Number of Full Coverage 114 569 126 6 815
3. Mean Manager Forecasts 0.87 1.86 2.60 3.17 1.88
4. Mean Unaffiliated 1.68 1.96 2.18 6.92 2.03
Forecasts
1995-1997 1. Number of IPOs 40 517 332 46 935
2. Number of Full Coverage 31 480 287 27 825
3. Mean Manager Forecasts 0.77 1.92 2.86 3.39 2.28
4. Mean Unaffiliated 1.27 1.39 1.61 2.41 1.51
Forecasts
Panel B: Large IPOs
1985-1987 1. Number of IPOs 11 46 19 31 107
2. Number of Full Coverage 8 28 4 3 43
3. Mean Manager Forecasts 0.73 1.48 1.37 2.16 1.58
4. Mean Unaffiliated 3.82 5.30 2.79 2.90 4.01
Forecasts
1988-1994 1. Number of IPOs 27 85 92 68 272
2. Number of Full Coverage 26 70 60 31 187
3. Mean Manager Forecasts 0.96 1.80 2.48 3.44 2.36
4. Mean Unaffiliated 6.44 4.06 4.78 5.24 4.83
Forecasts
1995-1997 1. Number of IPOs 4 40 85 123 252
2. Number of Full Coverage 3 38 60 63 164
3. Mean Manager Forecasts 0.75 1.95 2.64 3.81 3.07
4. Mean Unaffiliated 2.00 2.97 3.89 3.58 3.56
Forecasts









Table 4. Continued


Panel C: Dependent Variable: Number of Forecasts within One Year After IPO
Number of First-day Adjusted
Intercept Ln(Proceeds) Managers Return (%) R2 Sample Size
Pooled -4.36 1.84 0.55 0.02 0.210 2,844
CS-TS (-8.05) (10.69) (3.63) (5.88)
Fama- -4.02 1.84 0.36 0.05 0.227 13
MacBeth (-6.92) (12.61) (1.74) (3.99)








Table 5: How the Fees Are Shared in a Typical Syndicate

This table presents a representative example of an IPO, showing how the shares being
sold are allocated to the members of the syndicate for the purpose of compensating them.
Most of the information contained here would be publicly disclosed, but the split of the
management fee between the book manager (lead manager) and the co-manager is not
publicly disclosed. The net underwriting fee is also not publicly disclosed. In Panel C, it
is assumed that the overallotment option is exercised in full, with the number of shares
underwritten increased by 15 percent above the Panel A numbers for all syndicate
members.

Panel A: Underwriting information contained in prospectus and registration
statement


Gross proceeds:
Offer price:
Shares offered:
Gross spread:
Management fee:

Underwriting fee:
Selling concession:


Underwriting


$33.6 million
$12.00 per share
2,800,000 (plus 15% overallotment option of 420,000 shares)
84 cents (7 %) $2,704,800 total including overallotment option
17 cents (by convention, 20% of gross spread, rounded up to the
nearest penny)
19 cents
48 cents

Number of shares
underwritten


Goldman Sachs (book manager)
BT Alex Brown (co-manager)

Bear Steams
Deutsche Bank Securities
Donaldson Lufkin Jenrette
Lehman Brothers
Merrill Lynch
Morgan Stanley
Salomon Smith Barney

BancBoston Robertson Stephens
CIBC Oppenheimer
A. G. Edwards
Friedman Billings Ramsey
Hambrecht & Quist
NationsBanc Montgomery
Total


900,000
900,000

100,000
100,000
100,000
100,000
100,000
100,000
100,000

50,000
50,000
50,000
50,000
50,000
50,000
2,800,000








Panel B: Allocation of shares


2,800,000 deal size
+420,000 overallotment option
3,220,000
-100,000 10% of non-managing underwriters' underwriting commitment (initial
retention)
-700,000 managers' initial retentions (which are attributed to the book manager and
co-manager on an un-even basis, possibly a 70-30 split, as assumed)
-50,000 to friends of the company (handled by lead manager)
-50,000 to company employees (handled by lead manager)
2,320,000 "institutional pot", allocated to institutional investors by book manager

Of the institutional pot, 30% gets allocated in this example evenly among the managers,
and 70% is the "jump ball", or competitive portion, almost all of which will typically be
attributed to the lead manager (let's assume 1,500,000 of the 1,624,000 shares in the jump


ball). So the sales credits are as follows:
Lead: 490,000
50,000
50,000
348,000
1,500,000
2,438,000
Co-manager: 210,000
348,000
50,000
608,000
Other underwriters: 100,000
74,000
174,000


from 70% of 700,000 initial retention
from friends of company
from company employees
from 15% of institutional pot
from jump ball

from 30% of 700,000 initial retention
from 15% of institutional pot
from jump ball

from initial retention
from jump ball


Panel C: Allocation of fees
Fees: underwriting fees of 190 x 3,220,000 = $611,800 minus assumed syndicate costs
of $450,800 (including stabilization costs) = net of $161,000, or 50 per share
Amount of revenue (net of $450,800 syndicate costs)
Managers Underwriting Selling concession Shares
Underwriter fees @ 170 fees @ 50 net @ 480/share Total credited
Lead $273,700 $51,750 $1,170,240 $1,495,690 2,438,000
Co-manager $273,700 $51,750 $291,840 $617,290 608,000
100,000 share 0 $5,750 $8,352 $14,102 121,800
bracket on average on average in total
50,000 share 0 $2,875 $4,176 $7,051 52,200
bracket on average on average in total








Table 6: Average First-day Returns by Offering Size, Subperiod, and Gross
Spreads, 1985-1998

In Panel A, the sample consists of 3,122 firm commitment IPOs in 1985-1998 with actual
proceeds of at least $20 million (in 1997 purchasing power) before the exercise of the
overallotment option. Closed-end funds, REITs, ADRs, and unit offerings are excluded
from the sample. The first-day return is calculated as the percentage return from the
offering price to the closing price of the first trading day. The number of IPOs is shown
in parentheses. In Panel B, the sample consists of 3,271 firm commitment IPOs in 1985-
1998 with expected proceeds (in 1997 purchasing power) of at least $20 million dollars.
The amount of expected proceeds is computed using the midpoint of file price range
multiplied by the number of shares quoted in the preliminary prospectus. Panel C reports
regression results for IPOs with actual proceeds of at least $20 million but below $80
million. The amount of actual proceeds is measured in millions of dollars. Degree of
offer price revision is defined as the percentage change from the midpoint of file price
range to the final offer price. Numbers in parentheses are t-statistics using White (1980)
robust standard errors. Coefficients for year dummy variables are reported at the bottom
of the Panel.

Amount of Proceeds (in millions of dollars)
Years Gross Spread 20-39.99 40-59.99 60-79.99 80-up
Panel A: Actual Proceeds
1985-1987 <7.0% 2.5% ( 81) 4.5% ( 52) 4.2% ( 40) 6.2% ( 79)
=7.0% 6.6% ( 73) 9.9% ( 23) 5.1% ( 3) -0.3% (13)
>7.0% 7.3% ( 92) 1.9% ( 13) -0.5% ( 3) 0.7% (12)
All 5.5% (246) 5.5% ( 88) 3.9% ( 46) 4.7% (104)
1988-1994 <7.0% 8.5% ( 43) 7.7% ( 50) 7.5% ( 40) 6.2% (239)
=7.0% 11.7% (465) 16.6% (191) 13.6% ( 54) 11.3% (24)
>7.0% 11.6% ( 88) 12.3% (14) 10.7% ( 8) 32.4% ( 1)
All 11.4% (596) 14.6% (255) 11.0% (102) 6.8% (264)
1995-1998 <7.0% 10.3% ( 20) 6.3% (11) 9.6% (24) 12.3% (226)
=7.0% 18.5% (581) 24.8% (298) 21.0% (123) 26.8% (90)
>7.0% 14.7% ( 33) 21.9% ( 9) 19.1% ( 3) 9.1% ( 3)
All 18.0% (634) 24.1% (318) 19.1% (150) 16.4% (319)
Panel B: Expected Proceeds
1985-1987 <7.0% 4.7% ( 71) 3.1% ( 63) 2.4% ( 32) 5.2% ( 80)
=7.0% 6.1% ( 78) 8.4% ( 21) 4.0% ( 4) -0.4% (14)
>7.0% 5.0% (97) 1.1% (15) 1.6% ( 3) 0.7% (12)
All 5.3% (246) 3.9% ( 99) 2.5% ( 39) 3.9% (106)
1988-1994 <7.0% 9.8% ( 45) 9.1% ( 50) 7.3% ( 49) 5.5% (228)
=7.0% 12.5% (528) 9.2% (195) 5.6% ( 49) 4.9% (24)
>7.0% 9.1% (100) 7.7% ( 20) 14.6% ( 7) 0.0% ( 0)
All 11.8% (673) 9.1% (265) 7.0% (105) 5.4% (252)
1995-1998 <7.0% 9.3% (19) 12.4% (12) 10.5% ( 25) 11.9% (218)
=7.0% 21.7% (690) 17.9% (272) 17.4% (118) 9.1% (75)
>7.0% 8.5% ( 36) 19.8% ( 13) 17.6% ( 4) 13.7% ( 4)
All 20.8% (745) 17.8% (297) 16.2% (147) 11.2% (297)









Table 6. Continued


Panel C: Dependent Variable: Percentage First-day Returns for Moderate-size IPOs
Logarithm Degree of Dummy for Dummy for
Sample of Actual Offer Price Gross Spread Gross Spread Adjusted Number of
Period Proceeds Revision (%) Below 7% Above 7% R2 Observations
1985-1998 -2.37 0.60 -2.58 -2.54 0.24 2428
(-1.95) (16.91) (-2.74) (-2.48)
1995-1998 -2.99 0.68 -7.62 -3.15 0.20 1100
(-1.29) (11.86) (-2.99) (-1.22)

(continued)
Year 85 86 87 88 89 90 91 92 93 94 95 96 97 98

1985-1998 16.0 18.3 17.8 18.0 18.0 18.6 20.1 20.7 21.4 21.8 26.8 23.4 24.3 36.0


1995-1998


28.6 26.0 26.9 38.4













100.0%

90.0%

80.0%

70.0%

60.0%

50.0% -

40.0%

30.0% -

20.0%

10.0%

0.0%
85 86 87 88 89 90 91 92 93 94 95 96 97 98


* Below 7%
b7%
I Above 7%


Year


Figure 1. Gross spread distribution. The sample consists of 3,203 firm
commitment IPOs from 1985 through 1998 with proceeds of at least $20 million
(expressed in terms of dollars of 1997 purchasing power) before the exercise of
the overallotment option. Closed-end funds, REITs, ADRs, and unit offerings are
excluded from the sample. There are three categories of gross spreads expressed
as a percentage of proceeds: below 7%, 7%, and above 7%. The percentage of
IPOs in each category are from Table 1.












100.0%

90.0%

80.0%

70.0% -
0
9: 60.0%
5 6% Below 7%
SC 50.0% a 7%
| CAbove 7%
40.0% -- -

30.0%

20.0%

10.0% -

0.0% II44I
85 86 87 88 89 90 91 92 93 94 95 96 97 98

Year


Figure 2. Gross spread distribution for moderate-size IPOs. The sample consists of
2,488 firm commitment IPOs from 1985 through 1998 with proceeds of at least $20
million but less than $80 million (expressed in terms of dollars of 1997 purchasing
power) before the exercise of the overallotment option. Closed-end funds, REITs, ADRs,
and unit offerings are excluded from the sample. There are three categories of gross
spreads expressed as a percentage of proceeds: below 7%, 7%, and above 7%. The
percentage of IPOs in each category are from Table 1.










100.0%
90.0%
80.0%
70.0%
0
- 60.0% -
50.0%
p-

30.0%
C-
1 40.0%
0--
20.0%
20.0%
10.0% -
0.0%


85-87


I


ii


88-94


* Below 7%
=7%
[ Above 7%


7i


95-98


Years


Figure 3. Gross spread distribution for moderate-size IPOs from 1985 through 1998.
The sample consists of 2,488 firm commitment IPOs from 1985 through 1998 with
proceeds of at least $20 million but less than $80 million (expressed in terms of dollars of
1997 purchasing power) before the exercise of the overallotment option. Closed-end
funds, REITs, ADRs, and unit offerings are excluded from the sample. There are three
categories of gross spreads expressed as a percentage of proceeds: below 7%, 7%, and
above 7%. The percentage of IPOs for each of the three periods are from Table 1.










100.0%
90.0%
80.0%
70.0%
60.0%
50.0%
40.0%
30.0%
20.0%
10.0%
0.0%


85-87


88-94


U Below 7%
m7%
Lo Above 7%


95-98


Years


Figure 4. Gross spread distribution for large IPOs from 1985 through 1998. The
sample consists of 715 firm commitment IPOs from 1985 through 1998 with proceeds of
at least $80 million (expressed in terms of dollars of 1997 purchasing power) before the
exercise of the overallotment option. Closed-end funds, REITs, ADRs, and unit offerings
are excluded from the sample. There are three categories of gross spreads expressed as a
percentage of proceeds: below 7%, 7%, and above 7%. The percentage of IPOs for each
of the three periods are from Table 1.










11.0 -
10.0 -.a
9.0
.6 8.0
a. 7.0-
S 6.0 -
4 5.0
4.0


II


me a _._ I

*--- -- I --
.M *


-- 0


-1
0 **


3.0
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5
Log (expected proceeds, Smillion)

Figure 5. Scatter diagram relating IPO expected proceeds and gross spreads. The
sample consists of 330 firm commitment IPOs in 1998 with actual proceeds of at least $5
million before the exercise of the overallotment option. Closed-end funds, REITs, ADRs,
and unit offerings are excluded from the sample. The amount of expected proceeds from
the preliminary prospectus is measured in millions of dollars and then the natural
logarithm is taken. Two IPOs with spreads of 11.11% and 2.97% are not shown in the
diagram. A $20 million IPO has a log of 3.00, and an $80 million IPO has a log of 4.38.


-


1.01


1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5
Log (proceeds, $million)


7.0 7.5 8.0 8.5


Figure 6. Scatter diagram relating actual proceeds and gross spreads. The sample
consists of 330 firm commitment IPOs in 1998 with nominal proceeds of at least $5
million before the exercise of the overallotment option. Closed-end funds, REITs, ADRs,
and unit offerings are excluded from the sample. The amount of proceeds is measured in
millions of dollars and then the natural logarithm is taken. Two IPOs (with proceeds of
$5.0 million and $1.3 billion) with spreads of 11.11% and 2.97% are not shown in the
diagram. A $20 million IPO has a log of 3.00, and an $80 million IPO has a log of 4.38.


11.0
10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0


/ I


/ I A A I A A


_ __


t-7- t


-i-T .


-0 0 1


i













IU.U '_ _
9.0 m
-8.0 -
7.0 ------- -- --- --- --- -- ------- enmsmmesmn a sness es-my**- ---**
w 8*0 *
IV 0 __


S6.0 .
6.0- -- --- --- ----- ------------------
0 5.0 --
4.0
3.0
0 10 20 30 40 50 60 70 80 90

Proceeds, $million


Figure 7. Scatter diagram for IPOs with proceeds of at least $5 million but less than
$100 million in 1998. The sample consists of 278 firm commitment IPOs in 1998 with
nominal proceeds of at least $5 million but less than $100 million before the exercise of
the overallotment option. Closed-end funds, REITs, ADRs, and unit offerings are
excluded from the sample. One IPO (with proceeds of $5.0 million) with the spread of
11.11% is not shown in the diagram.









Figure 8. Gross spread distribution for moderate size IPOs underwritten by main
investment bankers from 1995 through 1998.
The sample is firm commitment IPOs underwritten by twelve main investment bankers
from 1995 through 1998 with proceeds of at least $20 million but less than $80 million
(expressed in terms of dollars of 1997 purchasing power) before the exercise of the
overallotment option. Closed-end funds, REITs, ADRs, and unit offerings are excluded
from the sample. There are three categories of gross spreads expressed as a percentage of
proceeds: below 7%, 7%, and above 7%. Salomon and Smith Barney merged in late
1997. Deals led by Salomon Smith Barney are attributed to Salomon. Morgan Stanley
includes Morgan Stanley Dean Witter, Robertson Stephens includes BA Robertson
Stephens and BancBoston Robertson Stephens, Alex Brown includes BT Alex Brown,
and Montgomery includes Nationsbanc Montgomery.











Morgan Stanley


95 96 97 98
Year


40-


'(
0
E
20

0 -


--"









95 96 97 98
Year


Lehman Brothers












95 96 97 98
Year


95 96 97 98
Year


CS First Boston











95 96 97 98
Year


Smith Barney

40 --


E <7%/
40
S20 7%
E -_3-- >7%
2 LI.1 -


95 96 97 98
Year


Robertson Stephens


0
,..
00

0 20

1


9596 9798
Year


40

0

o
20

E


40



-20
E
z


* <7%
*/7%
0>7%


M <7%,
0 7%0
>7%P/o


40


0



2


M B<'7%
0n7%
0 >7%


95 96 97 98
Year


Salomon Brothers

40


S- a <7%
20 707%
E 0 >7%


0
95 96 97 98
Year


AlexBrowm


40


20


* <7%
207%
0>7%


95 96 97 98
Year


c3

" 20



0


40


0
20



0


E<7%/

D0>7%


Montgomery


0



2
2


* 7%
D0>7%


95 96 97 98
Year


* <7%

D >7%/


Bear Steams


40

0 -

-20 7
E >7%



95 96 97 98
Year


Merrill Lynch


Goldman Sachs


























1.5 2.0 2.5 3.0 3.5


4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5


Log (proceeds, $million)



Figure 9. Scatter diagram for SEOs with proceeds of at least $5 million in 1998. The
sample consists of 366 seasoned equity offerings (SEOs) with proceeds of at least $5 million
before the exercise of the overallotment option. Closed-end funds, REITs, ADRs, and unit
offerings are excluded from the sample. The amount of proceeds is measured in millions of
dollars and then the natural logarithm is taken. One SEO (proceeds of $268.3 million) with a
spread of 0.69% is not shown in the diagram. This was from Flextronics International, a
company from Singapore.

nn 0


0 *
~


Il 10"..


-6 +7--0;


p^
0%f.


U-.
*


1*-I


0I
-0


LUi
0 10 20 30 40 50 60 70 80 90
Proceeds, $million


Figure 10. Scatter diagram for SEOs with proceeds of at least $5 million but less
than $100 million in 1998. The sample consists of 228 seasoned equity offerings (SEOs)
with proceeds of at least $5 million before the exercise of the overallotment option. Closed-end
funds, REITs, ADRs, and unit offerings are excluded from the sample.


0* *


: i-IM __i 0 a


0
00
1 00*
io 0 -0 %0- 4


o o*














CHAPTER 3
COMPETITION AND COLLUSION IN THE IPO MARKET


Introduction


The investigation of Chapter 2 gives the following facts. Over 90% of U.S.

corporations that went public using firm commitment offerings to raise proceeds ranging

from $20 million to $80 million in 1995-1998 paid exactly 7.0% of the proceeds to

investment bankers. For large offerings that raised more than $80 million, average gross

spreads are below 7.0% and there is little clustering on a single number.

Chapter 2 has discussed several possible explanations for the 7.0 percent

clustering. The purpose of this chapter is to offer a game-theoretic model, as one possible

explanation, for the above pricing pattern of investment bankers. The model

demonstrates that investment bankers, though competing on spreads, can still sustain

profitable equilibria in a stochastic and infinite horizon setting. In the setup of the model,

the perfect competition equilibrium is one of the candidates of equilibria. However, the

perfect competition outcome is not in the best interests of investment bankers.

Investment bankers prefer outcomes with higher profits if they can find some pricing

strategies that constitute profitable equilibria. In the dynamic game literature (e.g.,

Rotemberg and Saloner, 1986; Dutta and Madhavan, 1997) such pricing behavior is

termed "implicit collusion" because investment bankers, while acting in their own








interest, set spreads above the competitive level and earn abnormal profits without any

formal agreement about how to charge in the IPO market.

In the most profitable noncooperative equilibrium (call it "the second-best

equilibrium"), investment bankers can sustain a high spread for a wide range of offerings

and they have to charge lower spreads for sufficiently large offerings to keep other

underwriters from undercutting. Also, the empirical pattern is consistent with the

prediction of this equilibrium. The notion underlying the second-best equilibrium is

similar to those of Rotemberg and Saloner (1986) and Dutta and Madhavan (1997). In an

oligopolistic pricing model over the business cycle, Rotemberg and Saloner show that

firms set relatively low prices during periods of high demand in equilibrium. Dutta and

Madhavan offer a model to explain market maker's pricing behavior on Nasdaq. When

expected volume is less than a critical level, dealers set prices equal to those set under

overt collusion. If the volume shock is above the critical level, dealers narrow the bid-ask

spreads to prevent individual dealers from defecting from the proposed equilibrium

spreads. The model in section 2 is based on the framework of these papers.

The model suggests that investment bankers are able to sustain a high spread for

offerings smaller than some critical level and they have to charge lower spreads for


1 In the same spirit, Andrew D. Klein, founder of Wit Capital that distributes IPO shares
over the Internet, has the following comment: "Out of mutual self-interest, most bankers
have lived by a not-needed-to-discuss code that they wouldn't cut spreads." See Business
Week, November 9, 1998. Also, an anonymous head of underwriting for an investment
bank says of the gross spread, "The fact is, we'd been cutting our own throats to compete
on price." See Wall Street Journal, page Cl, April 10, 1997.









sufficiently large offerings to prevent other investment bankers from undercutting. The

prediction is consistent with the pattern observed in Figure 6 of Chapter 2. Another

implication of the model is that the critical offering size increases and it is easier to

implicitly collude for investment bankers when more IPOs are brought to the market

during a fixed calendar period (for example, a year). The cost of defecting from the

"implicit collusion" equilibrium is high in the more active IPO market since investment

bankers give up more expected profits during a fixed calendar period if they undercut

spreads. The implication is also consistent with the fact that it is a high IPO volume

period from 1995 through 1998.

Section 2 presents the model and analyzes the second-best strategy. Some caveats

about the model are also given. Section 3 concludes the chapter. Proofs of propositions

are delegated to the appendix.

An Implicit Collusion Model for the IPO Market


Basic Setup

Consider M symmetric investment bankers providing underwriting service in an

infinite-horizon setting. At the start of each period, one firm needs the underwriting

service of one investment banker to help it to go public. The new issuing firm is

characterized by gross proceeds raised before the fee charged by its underwriter. Let P,

be the gross proceeds or offering size at time t. We assume the amount of proceeds for

issuing firms is independent and identically distributed from period to period and regard it

as the random variable P that has domain [P,P] and has a distribution F(P). After

observing the size of an issuer in each period, M investment bankers simultaneously









quote their spread levels at which they are willing to provide underwriting service. Since

the offering size of new issuing firms may be different from period to period, this game is

a dynamic game rather than a repeated game in which the proceeds or economic

environment is constant in each period. There is no private information, so we have a

dynamic game with complete information.

We assume the expected profit function of investment bankers as a group is

r(S; P,) and the function is strictly concave in the underwriter spread S and increasing

in the proceeds P, 2 Formally, the monopoly spread is

5" =argmax(S ; P,).
S

When investment bankers form a cartel and charge a monopoly spread S", they are able to

earn the maximal profit from the perspective of the whole investment banking industry.

For analytic simplicity, we assume that there are no fixed costs of underwriting. Instead,

the underwriting technology has constant returns to scale as represented by total costs on

a deal that are proportional to proceeds, total costs = cP.

When investment bankers set the spread equal to the marginal cost c, they earn no

economic profits. It is easy to see that when each investment banker charges an

underwriting spread at the marginal cost c in each period no matter what spreads other



2 One example of expected profit function is x(S; Pt) = (S) (S P c Pt), where 4(S) is the
probability that the issuing firm will accept the deal under the spread of S. Also,
assuming that the function *(S) is strictly decreasing and (S). "(S) < 2('(S))2 can assure
the existence of the monopoly spread.









investment bankers charge, this pricing behavior constitutes a subgame-perfect Nash

equilibrium. First, investment bankers will not charge below c since they are going to

make losses and will be better off by choosing not to participate in the game or simply

charging c. Second, when all investment bankers charge above c, one investment banker

can make more profits by undercutting the prevailing spread. Finally, after any previous

history of play, the strategy requires each investment banking firm to set its spread equal

to c in every future period regardless of its rivals' behavior. Since each investment banker

earns at most zero when its opponents set their spreads equal to c, and it earns exactly this

amount by setting its spread equal to c in each future period, the strategy of "pricing at

marginal cost" in all periods is a subgame perfect Nash equilibrium.

Although "pricing at marginal cost" is a sustainable equilibrium of the pricing

game played by investment bankers, the competitive outcome does not coincide with the

best interests of investment bankers as a whole group. Investment bankers may prefer to

form a cartel to capture the monopoly rent. But they cannot reach such cooperative

agreements easily for two reasons. First, explicit price fixing agreements violate the

antitrust law. Second, investment bankers have incentives to deviate from cooperative

agreements by undercutting spreads when they observe some large firms are going to go

public. Therefore, explicit collusion cannot be sustained for all periods, even if

investment bankers understand it is beneficial for them to cooperate in the long run.

In the dynamic context, we still can find some equilibria in which investment

bankers act noncooperatively and make positive profits. We are especially interested in

looking for the "second-best" equilibrium, which is defined as a noncooperative









equilibrium that has the highest possible profit for the whole investment banking industry

among noncooperative equilibria.3 We call this the implicit collusion equilibrium.

The Second-best Equilibrium

Assume the current realization of the IPO proceeds is P'. Let J(P) denote the total

expected present value of future profits to the whole investment banking industry under

the second-best strategy and n(S'; P) be the current profits to investment bankers as a

group from charging the spread S'. Similar to the argument of dynamic programming, the

value function J(P) for the second-best strategy can be expressed in terms of current-

period profits plus the present value of future expected profits, as

J(P') = max [r(S';P') + pE(J(P))] (3-1)
S.

1
s.t. [7(S'; P) + pE(J(P))] > n(S; P') (3-2)
M

for all Si
transactions) is the one-period discount factor for all M investment bankers.

Equation (3-1) is known as the Bellman equation. Constraint (3-2) is the

incentive compatibility condition. It means that the expected profits from following the

second-best strategy are greater than one-shot profits from deviating from the strategy and

charging a lower spread. Specifically, an investment banker sets spreads equal to the

suggested spreads according to the second-best strategy and expects the tacit pricing to


3 As described in Dutta and Madhavan (1997), the "first-best" equilibrium refers to overt
collusion since it yields the highest profits for the investment banking industry. Note that
overt collusion is not an equilibrium of the noncooperative pricing game discussed in this
chapter.








continue through the future. From the assumption of symmetric investment bankers, the

investment banker's expected profits are decomposed as follows: he receives an

immediate profit of t(S';P') and he also receives the expected discounted
M

continuation payoff of pE(J(P)) .4 For the R.H.S. of (3-2), if the investment banker
M

undercuts the spread to capture the market, the investment banker earns 1r(S; P') for the

current period and no profits for future periods, expecting other investment bankers

thereafter to price at the marginal cost, which is the worst possible punishment. The

optimization problem (3-1) means that investment bankers maximize their aggregate

intertemporal profits subject to the incentive compatibility constraint (3-2). This is

simply due to the definition of the second-best equilibrium.

From our assumptions, the expected value of future profit to the whole investment

banking industry can be expressed as


E(J(P)) = J (S(P); P)dF(P). (3-3)


If investment bankers expect future profits to be zero whether they cooperate at time t or

not, the pricing game at time t will essentially be a one-shot game in which the well-

known equilibrium is the competitive equilibrium. Lemma 3.1 characterizes the




4 The cooperative bargaining feature of the underwriting syndicate is not pursued here.
While investment bankers share revenues through the underwriting syndicate, in practice,
the lead manager grabs the lion share of revenues. Table 5 of Chapter 2 shows how the
gross spread is split in a typical syndicate. Therefore, assuming each investment banker
acts independently to try to lead the deal would be appropriate.








competitive equilibrium via the expected continuation payoffs in the second-best

equilibrium.

Lemma 3.1: The unique noncooperative equilibrium is the competitive equilibrium if and

only if the expected continuation payoffs in the second-best equilibrium are zero, i.e.,

E(J(P)) =0.

Lemma 3.1 implies that if there exists a second-best equilibrium other than the

competitive equilibrium, the expected continuation payoffs must be strictly positive.

Since the time horizon of the pricing game is infinite, the extent of patience for

investment bankers will affect the magnitude of the expected continuation payoffs. We

now focus on the expected continuation payoff.

Note that the monopoly spread S" maximizes the profit function, and at worst

investment bankers get zero profits when they charge at c. The per-period payoff for each

underwriter is nonnegative and bounded by ir(Sm";P) and hence E(J(P)) is also

nonnegative and bounded above. When investment bankers are more patient, i.e., the

discount factor p is larger and the future is appreciated more, investment bankers are less

willing to undercut gross spreads to give up future profits. Intuitively, the expected

discounted continuation payoff K=pE(J(P)) under any strategy will increase in the

discount factor.

The following lemma, which is a variant of lemma 4 in Dutta (1995), captures the

above intuition and facilitates the proofs of the later main propositions.

Lemma 3.2: Consider any strategy a and its associated outcome. Suppose, at a fixed

discount factor p1, the expected continuation payoff E(J(P; p )) is nonnegative after








each finite history and suppose further that the per-period payoff for each underwriter is

bounded Then, the expected discounted continuation payoff under a, after each finite

history, is nondecreasing in the discount factor, i.e., K2=p2E(J(P;p2)), KK=pE(J(P;p,)),

K2 > K, whenever p2 > p,.

Since we are interested in the noncooperative equilibrium with the highest

possible profits for the whole investment banking industry, it is beneficial to examine the

extent to which the collusive spread is sustainable. When the collusive spread S" is

sustainable given current offering size P, it follows from (3-2) that

K 1
->(I )r-(S "; P). (3-4)
M M

The L.H.S. of (3-4) is the expected continuation profits to an investment banker given

that no other investment bankers deviate from the second-best strategy. K thwarts an
M

individual investment banker from charging a lower spread to capture the whole IPO

market since it is the expected profits forgone in the future if someone defects. However,

rc(S";P) is increasing in P and as a result, inequality (3-4) may not hold for a large

offering size. This means the collusive spread can no longer be sustainable for large

offerings since any investment banker has an incentive to undercut the spread to grasp the

business of large offerings. Therefore, investment bankers will set spreads lower for

larger offerings to keep an individual investment banker from cheating. It follows that

investment bankers will sustain the collusive spread for small offerings and charge less

for sufficiently large IPO firms.








On the other hand, Lemma 3.2 states that the expected continuation payoff K is

higher when investment bankers are more patient. This makes (3-4) hold more easily and

collusive pricing is sustainable in a wide range of offering size. In contrast, if investment

bankers are impatient, they undercut spreads more easily and act more competitively.

Then, constraint (3-2) may not hold and the competition equilibrium arises.

With the help of Lemma 3.1 and Lemma 3.2, Proposition 3.1 characterizes the

second-best equilibrium.

Proposition 3.1: There exists a constant poe[I-I/M, 1) such that when p>po, the second-

best strategy for any investment banker is as follows: 1) if the offering size is below a

unique critical proceeds Pc(p), where P Pc(p)_P charge the collusive spread S; 2) if

the offering size is above the critical proceeds (P>Pc(p)), charge an underwriting spread

S*(P), where S*(P)
underwriting spread at c in each period onwards.

When
underwriting spread at c in each period.

Proposition 3.1 shows that if investment bankers are sufficiently impatient (with

low discount factor), the competitive equilibrium is the only equilibrium. If investment

bankers are patient enough, their pricing behavior is like "implicit collusion" for offerings

below some critical level of offering size, and they will charge lower spreads to prevent

defection of individual investment banker for larger issues. Note that "implicit collusion"

in the second-best equilibrium is distinct from "overt collusion". For overt collusion,

investment bankers charge a unique profit-maximizing spread for all new issues (this









assumes that no economies of scale exist). However, in the second-best equilibrium,

investment bankers recognize their dynamic interdependence of their respective pricing

behavior and charge different spreads for different size of IPOs.

We define "the start of each period" as when one new issuing firm needs

underwriting services and will go public. That is, each period represents one transaction.

To address empirical questions, it is interesting to analyze the effect of changes in the

frequency of going public during a fixed calendar period. When more IPOs are brought

to the market in a fixed calendar period, the discount factor used between transactions

becomes larger (the discount rate becomes smaller because it is for a shorter time period)

assuming that the discount rate used during a fixed calendar period is unchanged. Hence,

investment bankers act as if they are more patient in the "hot" market than in the "cold"

market. By Lemma 3.2, the expected discounted continuation payoffs will increase.

Intuitively, as investment bankers recognize relatively more profitable opportunities in

the "hot" market, they are less willing to undercut spreads to give up future profits. Thus,

it is easier for investment bankers to implicitly collude in the hot IPO market. This may

seem counter to the Rotemberg and Saloner (1986) result that it is more difficult to

collude during booms. The difference in results is because Rotemberg and Saloner are

defining a boom as high current volume relative to the present value of future volume.

Here, the present value of future volume increases relative to the current transaction in a

hot market. Proposition 3.2 characterizes the nature of changes in the frequency of going

public.









Proposition 3.2: Given the second-best strategy, when more IPOs are brought to the

market during a fixed calendar period, it is easier to implicitly collude for investment

bankers, i.e., Pc increases.

Chapter 2 shows that there were roughly 360 IPOs with proceeds of at least $20

million per year during 1995-1998. This average number of IPOs is far greater than that

for the period 1991-1994, which was about 260 IPOs per year. Therefore, the period

1995-1998 has seen a relatively hot IPO market. Consistent with the prediction of

Proposition 3.2, 91% of moderate-size IPOs have paid the same gross spread of 7.0

percent and this partly reflects reluctance of investment bankers to undercut spreads

during the relatively hot period.

Caveats

The above dynamic game-theoretic model is based upon certain assumptions that

deviate from the institutional framework of the IPO underwriting business. First, it

assumes that there are constant returns to scale (the marginal cost of underwriting is c)

rather than being characterized by a fixed cost and declining average costs. Second, the

model assumes no capacity constraints: a given underwriter can capture all of the

underwriting business in a period if it charges the lowest spread, no matter how large the

deal. In reality, for all but the largest underwriters, either lowering the offer price or

forming a syndicate to share the risks and market a large deal would be required.

Furthermore, there are constraints on grabbing numerous deals in an active market

because writing a prospectus in a manner that will be approved by the S.E.C. takes time,

as does scheduling and conducting a road show. In other words, there are increasing

marginal costs of underwriting in the short run for any given underwriter.









Third, the model assumes, as is typical in models of dynamic games with

complete information, that in the second-best equilibrium a potential price-cutter assumes

that other underwriters will respond by setting spreads equal to marginal cost for all

future deals. This threat, with its maximum penalty to a price-cutter, achieves maximal

deterrence. It is not at all clear how time-consistent this threat is, however. If the

deterrence fails, and one underwriter cuts its spreads, it is not obvious that ex post the best

response is to cut all future spreads to marginal cost, rather than earn some economic

profits on a smaller market share with spreads still above marginal cost. This time-

consistency problem is not unique to my model, but is instead a generic problem in game

theory.

Fourth, the model assumes that penalties occur entirely in the price dimension;

future spreads are set equal to marginal cost. In reality, other underwriters have an

opportunity to impose a quantity penalty by excluding a price-cutter from future

syndicates.

I conjecture that the effect of all of these deviations from the formal model above

is to have an equilibrium where the sustainable spread is above marginal cost, but below

the monopoly spread. In other words, the qualitative results from the noncooperative

dynamic game that I have modeled above are likely to hold. Investment bankers realize

that they don't want to turn IPO underwriting into a "commodity business."


Conclusions


This chapter presents a game theoretic model to explain the pattern of clustering

spreads for a wide range of offerings and of lower spreads for large offerings in the IPO








market. In a stochastic and infinite horizon setting, investment bankers while acting

noncooperatively may charge underwriting spreads above competitive levels in

equilibrium. The model implies that investment bankers are able to sustain a high spread

for offerings smaller than some critical level and they have to charge lower spreads for

sufficiently large offerings to prevent other investment bankers from undercutting.

Another implication of the model is that when more IPOs are brought to the market

during a fixed calendar period, it is easier to implicitly collude for investment bankers.

The cost of defecting from the "implicit collusion" equilibrium is high in the more active

IPO market since investment bankers give up more expected profits during a fixed

calendar period if they undercut spreads. Interestingly, the evidence of underwriting

spreads from the 1990s is consistent with the implications of the implicit collusion model.













CHAPTER 4
SUMMARY AND CONCLUSIONS

This dissertation has examined gross spreads on IPOs in the U.S. and several facts

emerge. First, gross spreads received by underwriters on IPOs in the U.S. are much

higher than in other countries. Second, in recent years at least 90 percent of deals raising

between $20-80 million, which are referred to moderate-size IPOs, have spreads of

exactly 7.0 percent. Third, the clustering of spreads at 7.0 percent for moderate-size IPOs

has increased over time. For comparison, only 26 percent of moderate-size IPOs in 1985-

1987 had 7.0 percent spreads. But the average spread on moderate-size IPOs has

remained virtually constant during the 1985-1998 period.

There is widespread agreement that fixed costs exist in underwriting IPOs, yet

investment bankers charge the same 7.0 percent spread on $20 million deals as they do on

$80 million deals. I argue that for most IPOs with gross proceeds larger than $30 million,

spreads are above competitive levels in the U.S. The high average spread and the

concentration of spreads at 7.0 percent is consistent with implicit collusion on the part of

investment bankers. In other words, even though investment bankers are acting

independently, average spreads are above competitive levels. Investment bankers realize

that if one investment banker tries to win business by cutting spreads, the underwriting

industry is likely to move to an equilibrium with low spreads, and lower compensation for

corporate finance employees. Several features of the IPO underwriting market are

conducive to spreads above competitive levels. The importance of analyst coverage and








buy recommendations, and the perceived importance of underwriter prestige, facilitate

high spreads.

If gross spreads are above competitive levels, investment bankers have an

incentive to use nonprice competition to attract deals. Although issuing firms face high

and, for moderate-size deals, nonnegotiable spreads, issuers can still bargain on another

dimension. In particular, by insisting on additional co-managers, issuing firms receive

more extensive analyst coverage. I show that the number of co-managers has increased

over the last decade, and that adding an additional co-manager adds between 0.36 and

0.55 net analysts following the stock.

Several commercial banks entered the underwriting business during 1997-1998.

But in the year after Nationsbank, Bankers Trust, BankAmerica, and BancBoston bought

investment banking firms that specialized in IPOs, there does not seem to have been any

impact in the IPO market. Another source of competition may emerge from the

innovation of internet technology. For example, new underwriters E*Offering and W. R.

Hambrecht threaten to undercut the 7.0 percent fee that is now standard. Only time will

tell whether this changes the gross spreads that prevail in the IPO underwriting industry.













APPENDIX
PROOFS OF THE MAIN RESULTS IN CHAPTER 3

Proof of Lemma 3.1:

The part of necessity is obvious. To prove sufficiency, suppose there exists one

equilibrium other than the competitive equilibrium. There must be some pricing rule that

specifies the underwriting spread above the competitive level as the part of the

equilibrium strategy other than the competitive equilibrium strategy. From (3-1), this will

yield positive expected continuation payoffs to investment bankers as a group, a

contradiction. Q.E.D.



Proof of Proposition 3.1:

Step 1: Based on the work of Abreu (1988), we know that the equilibrium strategy

can be specified as two parts: (1) the initial path and (2) the punishment paths, one for

each investment banker.

The initial path is followed in the event that no player deviates from this path in

the past. And some punishment is imposed in the event that some player deviates from

the initial path or from any punishment path. Further, the optimal punishment is the

pricing rule that yields the lowest expected profit. Obviously, we can take the optimal

punishment to be "pricing at marginal cost in each period." Now, we need to construct

the initial path. Let K=pE(J(P)) be the present value of expected continuation payoffs to

investment bankers as a group under the initial path. Suppose that the current size of the









IPO is P and K is greater than zero. From the incentive compatibility condition (3-2), the

collusive price is sustainable if and only if

K 1
-- > (1- )(Sm"; P)
M M

Since 7t(S; P) is strictly increasing in P, the above inequality may not always hold.

Define critical offering size Pc as the solution to

K= (1 )(S ; P), (A-l)
M M

if a solution exists. Note that Pc depends on p since K=pE(J(P)). If

K I
-<(1- )7(S";P), then define Pc=P. On the other hand, if
M M

K 1
> (- -)7(S"; P), then define Pc= P.
M M

Construct the initial path as follows: (1) For P<-Pc, price at the collusive spread

S". (2) For P>Pc, price at S* which solves

K 1
= (1- -)i(S*; P). (A-2)
M M

Note that S*
The construction of the initial path and the punishments gives the second-best

strategy. To make this clear, observe that the incentive compatibility condition is always

satisfied and the pricing strategy is deviation-proof, if investment bankers follow this

second-best strategy from the next period onward. Furthermore, the strategy achieves the

highest profit among the sustainable set of prices. From the second-best strategy and (3-

1), it follows that









(S,; P) +K for P I Pc
J(P) =f (A-3)
''1 A+K forP>Pc

Combination of (3-4) and (A-3) implies

MK
E(J(P)) < for all P. (A-4)
M-1

Substituting K=pE(J(P)) into (A-4), we have the inequality p>(l-1/M) since we assume

K>0 at this point. Therefore, if p<(l-1/M), then K=0 and the unique equilibrium is the

competitive equilibrium from Lemma 3.1.

Step 2: In step 1, we suppose K>0. Here, we show that the expected discounted

continuation payoffs are positive for sufficiently large discount factors. In fact, there

exists po<1 such that K>0 for p>po. To show this, we construct another equilibrium for

an appropriate discount factor: choose any Po> P and for P
spread according to the collusive pricing strategy while for P>Po, price at the marginal

cost. Lemma 3.1 tells us the underwriting spread for P>Po is the equilibrium price. Let

Ko be the expected lifetime payoff to this strategy. We have

Ko = "- 0 r(Sm";P)dF(P). (A-5)


The strategy is incentive compatible for P
K1
KO- (1 )(S"; P) (A-6)
M M

for P
holds. Since the profits to this strategy are strictly positive from the right hand side of(A-









6), it follows that the second-best profits are strictly positive for p=po. Hence, Lemma

3.2 implies that K>O for p2po.

Using step 1 and step 2, we have established that there exists a discount factor

poe[l-1/M, 1) such that: (1) for p
equilibrium and K=0; (2) for p>po, the second-best pricing strategy exists and is described

as in step 1 and K>0. Q.E.D.



Proof of Proposition 3.2:

Let R denote the discount rate over a fixed calendar period (for instance, a year)

with n IPOs brought to the market. Therefore, in the model, there are n periods during

the fixed calendar period and the discount factor between any two IPOs will become

1!
1+R

and we have

dp, p
dp" = PIn(l + R)> 0.
dn n2



When the frequency of going public increases during a fixed calendar period, the discount

factor applied to the next IPO, pn increases monotonically. From Lemma 3.2, K

increases as p, increases and in turn Pc increases, this follows directly from the

definition of Pc. Hence, as n increases, it is easier to implicitly collude among

investment bankers and the second-best profits also increase. Q.E.D.














REFERENCES


Abreu, Dilip. (1988). "On the theory of infinitely repeated games with discounting."
Econometrica 56, 383-396.

Aggarwal, Reena. (1998). "Stabilization activities by underwriters after new offerings."
Working paper, Georgetown University, Washington, DC.

Ball, Clifford, Walter Torous, and Adrian Tschoegl. (1985). "The degree of price
resolution: the case of the gold market." Journal of Futures Markets 5, 29-43.

Barry, Christopher, Chris Muscarella, and Michael Vetsuypens. (1991). "Underwriter
warrants, underwriter compensation, and the costs of going public." Journal of
Financial Economics 29, 113-135.

Beatty, Randolph and Jay Ritter. (1986). "Investment banking, reputation, and the
underpricing of initial public offerings." Journal of Financial Economics 15, 213-232.

Beatty, Randolph, Rex Thompson, and Michael Vetsuypens. (1998). "Issuance costs and
regulatory change in the investment banking industry." Working paper, Southern
Methodist University, Dallas, TX.

Beatty, Randolph and Ivo Welch. (1996). "Issuer expenses and legal liability in initial
public offerings." Journal of Law and Economics 39, 545-602.

Booth, James and Richard Smith. (1986). "Capital raising, underwriting, and the
certification hypothesis." Journal of Financial Economics 15, 261-281.

Buckman, Rebecca. (1999). "Internet brokerage firms click into online stock
underwriting." Wall Street Journal, January 28, Cl.

Carter, Richard, Frederick Dark, and Ajai Singh. (1998). "Underwriter reputation, initial
returns, and the long-run underperformance of IPO stocks." Journal of Finance 53,
285-311.

Christie, William and Paul Schultz. (1994). "Why do NASDAQ market makers avoid
odd-eighth quotes?" Journal of Finance 49, 1813-1840.

Dunbar, Craig. (1998). "Factors affecting investment bank initial public offering market
share." Journal of Financial Economics, forthcoming.










Dutta, Prajit. (1995). "Collusion, discounting and dynamic games." Journal of Economic
Theory 66, 289-306.

Dutta, Prajit and Ananth Madhavan. (1997). "Competition and collusion in dealer
markets." Journal of Finance 52, 245-276.

The Economist. (1998). "Overcharging underwriters." June 27th issue.

Ellis, Katrina, Roni Michaely, and Maureen O'Hara. (1998). "When the underwriter is the
market maker: An examination of trading in the IPO aftermarket." Working paper,
Cornell University, Ithaca, NY.

Fama, Eugene and James MacBeth. (1973). "Risk, return, and equilibrium: Empirical
tests." Journal of Political Economy 81, 607-636.

Gande, Amar, Manju Puri, and Anthony Saunders. (1998). "Bank entry, competition, and
the market for corporate securities underwriting." Journal of Financial Economics,
forthcoming.

Hanley, Kathleen Weiss. (1993). "The underpricing of initial public offerings and the
partial adjustment phenomenon." Journal of Financial Economics 34, 231-250.

James, Christopher. (1992). "Relationship-specific assets and the pricing of underwriter
services." Journal of Finance 47, 1865-1885.

Kahn, Charles, George Pennacchi, and Ben Sopranzetti. (1999). "Bank deposit rate
clustering: Theory and empirical evidence." Journal of Finance, forthcoming.

Krigman, Laurie, Wayne Shaw, and Kent Womack. (1998). "Why do firms switch
underwriters?" Working paper, Dartmouth College, Hanover, NH.

Lee, Inmoo, Scott Lochhead, Jay Ritter, and Quanshui Zhao. (1996). "The costs of raising
capital." Journal of Financial Research 19, 59-74.

Lowenstein, Roger. (1997). "Street's incredible unshrinking spread." Wall Street Journal,
April 10, Cl.

Michaely, Roni and Kent Womack. (1998). "Conflict of interest and the credibility of
underwriter analyst recommendations." Review of Financial Studies, forthcoming.

National Association of Securities Dealers. (1998). Notice to Members 98-88.

Power, William. (1993). "Why hot, new stocks get booster shots." Wall Street Journal,
February 10, Cl.










Raghavan, Anita. (1997). "How one top analyst vaults "Chinese Wall" to do deals for
firm." Wall Street Journal, March 25, Al.

Rajan, Raghuram and Henri Servaes. (1997). "Analyst following of initial public
offerings." Journal of Finance 52, 507-529.

Ritter, Jay. (1987). "The costs of going public." Journal of Financial Economics 19, 269-
281.

Rotemberg, Julio and Garth Saloner. (1986). "A supergame-theoretical model of price
wars during booms." American Economic Review 76, 390-407.

Siconolfi, Michael. (1992). "At Morgan Stanley, analysts were urged to soften harsh
views." Wall Street Journal, July 14, Al.

Smith, Geoffrey and Paula Dwyer. (1998). "Coincidence or collusion? Two academics
question the standard 7% IPO fee." Business Week, November 9, 1998, p163.

Smith, Randall. (1996). "Firms vie for spot in AT&T's huge IPO." Wall Street Journal,
February 1, C1.

Telser, Lester. (1960). "Why should manufacturers want fair trade?" Journal of Law and
Economics 3, 86-105.

Tully, Shawn. (1999). "Can the net revolutionize IPOs?" Fortune (March 15, 1999), 35-
36.

Uttal, Bro. (1986). "Inside the deal that made Bill Gates $350,000,000." Fortune (July 21,
1986), 343-361.

Wall Street Journal. (1998). "U.S. judge approves record settlement in Nasdaq lawsuit."
November 10, C19.

White, Halbert. (1980). "A heteroskedasticity consistent covariance matrix estimator and
a direct test for heteroskedasticity." Econometrica 48, 817-838.

Williams, Joseph. (1998). "Agency and brokerage of real assets in competitive
equilibrium." Review of Financial Studies 11,239-280.














BIOGRAPHICAL SKETCH


Hsuan-Chi Chen received a bachelor's degree in mathematics in 1989 from

National Taiwan University, and a Master of Business Administration degree in 1993

from Fu Jen University. He taught at Fu Jen University during 1993-1995 and entered the

Ph.D. program in finance at the University of Florida in 1995.








I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.


R Ritter, Chair
Joseph B. Cordell Eminent Scholar,
Finance, Insurance and Real Estate

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.


MaW'. Flannery
Barnett Banks Eminent Schiola Finance,
Insurance and Real Estate

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.


Michael D. Ryngaert
Associate Professor of Finance,
Insurance and Real Estate

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.


Tracy R. Lewis
James W. Walter Eminent Scholar
in Entrepreneurship

This dissertation was submitted to the Graduate Faculty of the Department of
Finance, Insurance and Real Estate in the College of Business Administration and the
Graduate School and was accepted as partial fulfillment of the requirements for the
degree of Doctor of Philosophy.

August 1999
Dean, Graduate School



















LT>
IT o






















UNIVERSITY OF FLORIDA
1 I 111111111111 lll 111111 1 1111111
3 1262 08555 0878




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EGDZGW8WH_L72JO9 INGEST_TIME 2013-10-24T20:34:18Z PACKAGE AA00017709_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES