All-star Analyst Turnover, Investment Bank Market Share, and the Performance

of Initial Public Offerings

Jonathan Clarke

Georgia Institute of Technology

Craig Dunbar*

University of Western Ontario

Kathleen Kahle University of Pittsburgh

This version: September 2002

Preliminary: please do not quote with author permission

Abstract This paper examines the impact of all-star analyst turnover on initial public offering market share and the performance of initial public offerings. We find that investment banks losing all-stars do not experience a significant change in market share. In contrast, the bank acquiring the all-star analyst experiences a significant increase in market share of 1.25%. In response to losing a star analyst, investment banks begin to compete more on price by cutting the abnormal spread and take on more speculative issuers. The banks gaining the all-stars also take on more speculative issuers and become more aggressive by issuing research reports earlier and more often. Key words: IPO, all-star analyst, market share, underpricing, IPO performance JEL Classification: G24, G32 + We gratefully acknowledge the contribution of I/B/E/S International Inc. for providing earnings per share forecast data, available through the Brokers Estimate System. This data has been provided as part of broad academic program to encourage earnings expectations research. * Corresponding author. Mailing address is University of Western Ontario, Richard Ivey School of Business, London, Ontario, N6A 3K7 Canada. Phone number is (519) 661-3716. Fax number is (519) 661-3959. E-mail address is cdunbar@ivey.uwo.ca.

1

1. Introduction

There is substantial anecdotal evidence suggesting that top-rated analysts are essential for

attracting investment-banking business. For example, Kessler (2001) notes, “The battle for top analysts

and IPOs becomes circular since the higher an analyst ranks in the polls, the easier it is for their firm to

win IPOs, and the more quality companies the firm brings to market and does banking business with, the

higher the analysts rank in the polls.” The importance of top-rated analysts seems to be reflected in their

pay. For example, noted telecommunications analyst Jack Grubman received $25 million dollars in cash

and stock to stay with Salomon Smith Barney in 1998. Despite the abundant claims in the popular press,

however, there is little direct evidence that all-star analysts influence initial public offering market share.

In this paper, we directly examine the impact of analyst reputation on investment bank initial public

offering market share by examining a sample of analysts named to Institutional Investor’s All-American

Research Team who subsequently switch investment banks. We examine whether the gain (loss) of an

all-star analyst helps (hurts) investment banks in the IPO market, and whether the salaries of these

analysts are justified by the benefits to the firms. We also examine how the gain or loss of an all-star

affects investment bank and analyst behavior.

Two papers in the literature are related to our study. Dunbar (2000) argues that the presence of

strong analysts is likely to be attractive to firms wishing to conduct initial public offerings, because

analyst reputation can certify to potential investors that a deal is not overpriced. In empirical tests,

Dunbar finds that the percentage change in an investment bank’s Institutional Investor (II) ranking has a

positive effect on changes to its market share. Krigman, Shaw, and Womack (2001) examine a sample of

firms that switch underwriters following their initial public offering. They find that obtaining

Institutional Investor all-star coverage is one of the most important motives for issuers to switch

underwriters. They argue that issuers place significant value on high-quality research coverage and are

willing to pay in the form of higher underwriting fees.

2

We expand on the work of Dunbar (2000) and Krigman et al. (2001) in five important ways.

First, rather than treating all-star standing as static as in Krigman et al., our analysis allows us to examine

the dynamic relation between market share and all-star analysts. Second, we allow the market share

relationships to be different for banks gaining and losing stars. Dunbar (2000) does not allow for this

potential asymmetry in his analysis. Third, we examine market share at both an aggregate and industry

level. Since analysts specialize by industry, we would expect the power of our analysis to be enhanced by

considering industry effects. Fourth, we consider the impact of all-star analyst turnover on bank and

analyst behavior in the market for IPOs. Specifically we examine whether gaining or losing a star affects

the pricing and performance of IPOs. We also consider whether gaining or losing a star affects earning

forecasts made by analysts. In doing so, we attempt to determine whether banks and analysts alter their

behavior in an attempt to preserve or expand IPO market share around the time of the turnover. Finally,

we examine the factors affecting changes in IPO market share for banks gaining and losing all-star

analysts in a regression framework, as in Dunbar (2000). Our analysis considers a number of factors not

previously considered, including measures of analyst behavior. Also, whereas Dunbar examines market

share changes at a fixed point (calendar year end), we consider changes after a significant shock to the

bank, potentially providing a more powerful test of market share changes.

Contrary to popula r belief, we find that investment banks losing all-stars do not experience a

significant decline in either industry level or aggregate market share following the departure of the star.

However, losing an all-star does have an impact on the bank’s pricing and performance in the IPO

market. There is evidence that banks losing a star compete for business by cutting their fees in IPOs.

Analysts in the bank losing the star also become more aggressive by issuing forecasts earlier after the

star’s departure. There is no evidence that the investment bank changes its objectivity in an effort to

retain market share, however. Earnings forecasts relative to the consensus do not change significantly

following the departure of the all-star.

We find that acquiring an all-star analyst has a positive impact on an investment bank’s market

share. The bank’s overall market share increases a statistically significant 1.260 percentage points.

3

Based on average annual IPO proceeds of $24.6 billion over the 1985 to 2000 period and an assumed 7

percent spread, the improvement to market share translates to an approximate $22 million increase in fees

for the bank. This provides some justification for the huge salaries received by some prominent analysts.

The gains to market share are largest when the bank has few existing all-star analysts and when the bank’s

pre-acquisition market share is less than the bank losing the all-star. When the investment bank is able to

acquire an analyst named to Institutional Investor’s First, Second, or Third Team, industry market share

increases a statistically significant 4.326 percentage points.

Acquiring an all-star also has a significant impact on the pricing and performance of initial public

offerings. Banks are more likely to make pos itive price adjustments in the IPO pricing process after they

acquire a star analyst. This is consistent with the star using his or her reputation to acquire more positive

information during the IPO marketing process (Benveniste and Spindt, 1989). There is a significant

decrease in the long-run performance of issuers taken public by banks acquiring a star. This is consistent

with banks becoming aggressive by selecting issuers of more questionable quality in an attempt to expand

IPO market share. Analysts at the bank gaining the star begin recommending firms earlier and more

often. The volatility of forecasts relative to consensus increases, also suggesting that banks take on more

speculative deals. Average earnings forecasts, relative to consensus, do not change, however, suggesting

that aggressiveness does not affect the analyst’s objectivity.

Building on the analysis in Dunbar (2000), we examine the factors affecting market share

changes around analyst turnover. Like Dunbar, we relate market share changes to measures of IPO

pricing and performance including abnormal first-day returns, one-year abnormal returns, and abnormal

underwriting fees for the bank in the year prior to the move. A new measure of performance considered

in our analysis is the price adjustments made by banks in the IPO pricing process. We also consider a

number of measures of analyst performance including the number of forecasts, the time to first forecast,

the percentage of offerings where the lead bank is first to make forecasts, and the forecast level relative to

consensus.

4

Our evidence reverses several findings by Dunbar (2000). For banks losing a star analyst, we

find that market share changes are positively related to past mean abnormal underpricing. Banks leaving

more money on the table are rewarded with increased share in the IPO market. For both banks losing and

gaining a star, market share changes are positively related to volatility of abnormal long-run returns on

past IPOs. This suggests that banks taking on more speculative issues are also rewarded with increased

IPO market share.

Market share changes for both banks losing and gaining all-star analysts are positively related to

the volatility of price adjustments made on past IPOs. Banks able to extract more information (both

positive and negative) from investors in the IPO pricing process attract more future business. Market

share changes are also positively related to various measures of analyst aggressiveness. For example,

analysts making earlier forecasts and more numerous forecasts gain market share. For banks losing an

all-star analyst, market share changes are positively related to mean levels of earnings forecasts relative to

consensus. Banks that make higher earnings forecasts on firms they take public are rewarded with greater

future business. Finally, for banks gaining an all-star, volatility of earnings forecasts relative to consensus

is positively related to market share changes. This is consistent with banks increasing their presence in

the IPO market by taking on more speculative issuers.

The organization of the remainder of this paper is as follows. In Section 2, we develop our basic

hypotheses regarding the impact of all-star analyst turnover on investment bank market share. In section

3 we summarize the all-star turnover data. In section 4 we examine market share effects for bank gaining

and losing an all-star analyst. In section 5 we examine the pricing and performance of IPOs and analyst

forecast activity around analyst turnover. In section 6 we identify the factors affecting IPO market share

for banks gaining and losing an all-star. Finally, we summarize the paper in Section 7.

5

2. Investment Bank Market Share and All-Star Turnover

The primary focus of our analysis is on the effects of all-star turnover on the IPO market share of

banks losing and gaining the star. In this section we develop three alternative theories, which we use to

relate star turnover and market share. First, the certification hypothesis predicts that market share for a

bank gaining (losing) a star analyst should significantly increase (decrease). An extensive literature

examines the impact of investment bank reputation on the IPO market. Insiders have better information

regarding the true value of their firm and an incentive to offer securities when they are overvalued. In this

market environment, investors will only participate in IPOs if they can purchase shares below what they

believe the shares should be worth. Booth and Smith (1986) argue that IPOs can be priced closer to their

“intrinsic value” if insiders credibly certify that they are not selling overpriced securities. One

certification mechanism is to hire a reputable investment bank to manage the offering. Banks are credible

third party certifiers because they lose future business if their certification is found to be inaccurate (e.g.

investors will avoid purchasing shares from banks that systematically sell overpriced securities).

Conversely, banks that enhance their reputation through a record of accurate certification should see

increases gain market share.

The presence of an All-American analyst is likely to have significant impact on the reputation of

the investment bank. Michaely and Womack (1999) note that analysts play an active role in underwriting

new issues. Adding an all-star analyst should increase investors’ confidence that an offering is not

overpriced since more is at stake (mispricing damages both the bank and analyst’s reputation). This

enhanced reputation should allow the bank to more effectively compete for business. Banks losing an all-

star should lose market share since less reputation is at stake in any issue, diminishing the bank’s

certification. 1

1 Analysts in banks having an underwriting relationship with the firm are also more likely to make earnings forecasts and recommendations to buy an IPO’s shares in the first few months after the IPO. The market generally responds positively to this coverage, and Stickel (1992) finds that the reaction is most positive for analysts on II’s All-

6

Other theories suggest a less positive role for investment banks in the market for IPOs. Our

market power hypothesis builds on Baron and Holmstrom (1980) and Baron (1982) who argue that the

market for underwriter services is less than perfectly competitive. Issuers choose amongst a small set of

banks that they perceive are capable of marketing their offering successfully. Banks acquiring a star

analyst improve their status among issuers, leading to greater future deal flow. Conversely, banks losing

a star analyst see their status among issuers fall, leading to lower future deal flow.2

Finally, the investment bank aggressiveness hypothesis argues that banks compete for business in

the IPO market through aggressively seeking out issuers regardless of quality. While the theories

underlying the certification hypothesis argue that reputable banks should avoid speculative offers (see

Carter and Manaster, 1990, Chemmanur and Fulgheiri, 1994), banks concerned with their market share

may expand the range of firms they attempt to take public. Banks may also become more aggressive in

the services they provide to issuers in an attempt to gain business (e.g. make earlier and more frequent

analyst forecasts and recommendations).

Banks competing for a star analyst may become more aggressive after the competition is resolved

and the analyst moves. The bank losing the star will attempt to preserve their position in the market to

offset any perceived loss of stature due to the star’s departure. The bank gaining the star will be more

aggressive to justify the high cost involved in winning the competition (i.e. high analyst salaries). The

effect on market share is uncertain but both banks could see increases.

3. Analyst Turnover Data

Our empirical tests focus on a sample of Institutional Investor All-American analysts (all-stars)

who switch investment banks between 1988 and 1999. Following Hong, Kubik, and Solomon (2001), we

American Research Team. These arguments suggest that losing an all-star will have a significantly negative impact on an investment bank’s market share, while gaining an all-star should improve market share. 2 At a first blush, the certification and market power make identical predictions regarding the effect on market share of analyst turnover. In later sections we identify different predictions of these theories with respect to bank and analyst behavior around the star turnover.

7

use the I/B/E/S detail file to determine whether an analyst moves to a different brokerage house.3 The

detail file assigns each individual analyst a numerical code, making it possible to track forecasts of EPS

across time even if the analyst switches firms. For each analyst in our sample who switches investment

banks, we are able to identify the date of the analyst’s last forecast at her old firm and first forecast at her

new firm. Typically the I/B/E/S database only identifies each analyst and her employer by a unique

numerical code. However, we were granted access to the Broker Code Key, which allows us to identify

the last name and first initial of each analyst in the database and the identity of their employer. This

additional information allows us to identify those analysts that were named to Institutional Investor’s All-

American Research Team in a given year. We consider only those cases of turnover where the analyst

was an all-star in the year prior to or the year of the switch. We further eliminate cases where the switch

was due to the merging of two investment banks. For example, we eliminate four cases in which an all-

star switched from Kidder Peabody to Paine-Webber in 1994. The final sample consists of 222

observations.4 Using the industry classifications reported for each all-star in II as a guide, we then assign

each of these analysts to one of the 48 Fama-French Industries. This allows us to match the analyst

turnover sample with our initial public offering sample based on industry. Our turnover sample only

includes cases where an all-star switched from one investment bank to another. It does not include cases

where the all-star left the business.

Although many rankings of individual analysts are published each year, the choice of II’s All-

American Team seems particularly appropriate.5 Hong, Kubik, and Solomon (2001) and Nocera (1997)

note that sell-side analysts generally aspire to be II All-Americans. Stickel (1992) finds that members of

the II All-American Team supply more accurate earnings forecasts than other analysts when forecasts are

matched by the corporation followed and by the date of brokerage house issuance. This contemporaneous

3 Over the 1988 to 1999 period an average of 3150 analysts from 238 brokerage houses submitted forecasts to I/B/E/S each year. 4 The median length of time between the analyst’s first forecast with her new employer and the last forecast with her previous employer is 24 trading days. We eliminate cases where the elapsed time is greater than 100 trading days. 5 See Li (2002) for a discussion of differences between the Institutional Investor and Wall Street Journal ranking of analysts.

8

advantage is complemented by a timing advantage; All-Americans supply forecasts more often than other

analysts. Stock returns immediately following large upward forecast revisions suggest that All-Americans

impact prices more than other analysts. However, there is virtua lly no difference in returns following

large downward revisions. Nevertheless, the collective results suggest a positive relation between

reputation and performance, and, assuming that All-Americans are better paid, between pay and

performance.

Selection to the All-American team is almost entirely based on survey data. II sends out a

questionnaire to the directors of research and chief investment officers of various money management

institutions and also to other sell-side analysts. They rank each analyst based on the following six

dimensions: accessibility and responsiveness, earnings estimates, useful & timely calls, stock selection,

industry knowledge, and written reports. Scores for each analyst are calculated by taking the number of

votes awarded by each survey respondent and weighting them by the size of the respondent’s firm. The

results are published each year in the October issue of the magazine. The methodology has not changed

significantly over our sample period, although the number of survey respondents has increased steadily

over the sample period.

Table 1 provides some descriptive statistics on analyst turnover. Panel A provides statistics on

analyst turnover by year, both for all analysts and for all-stars. Whereas percent turnover per year has

remained fairly constant for all analysts, averaging roughly six to nine percent per year between 1988-

1999, turnover has increased among the all-stars. In the late 1980s and early 1990s, all-star turnover was

only about two percent per year. By the late 1990s, turnover was averaging close to ten percent per year

for the all-stars. This would seem to indicate that competition for all-stars has increased in the 1990s.

Panel B of Table 1 examines all-star analyst turnover by II ranking. Within our sample, 14% of

the turnover is within First Team analysts, 16% is by Second Team analysts, 25% is by Third Team

analysts, and 45% is by Runners-up analysts. This would seem to indicate that it is more difficult to lure

First and Second Team analysts away from their firms.

9

4. IPO Market Share

Our sample of initial public offerings consists of all completed deals between 1986 and 2000 that

are recorded in Securities Data Corporation’s New Issues database. Following Dunbar (2000) and Bates

and Dunbar (2002), we exclude offerings by closed-end funds, real estate investment trusts, ADRs, and

unit offerings. For each remaining offering, we obtain data from SDC on the offering date, the book

manager(s) of the offering, the gross domestic proceeds raised in the offering, excluding overallotments,

the offering price, and the underwriter spread. The final sample consists of 5,253 initial public offerings.

Gaining or losing an all-star analyst could impact both the investment banks overall market share

and its share of IPOs within the industry covered by the all-star. We, therefore, compute market share

both for the bank as a whole and at the industry level. Specifically, we compute the sum of gross

proceeds (on global shares excluding over allotments) for which the underwriter was also the book

manager. To account for mergers in the investment banking industry, we gather data from SDC on all

combinations during the period. If the book manager recently emerged from a merger, the gross proceeds

of all offerings by any precedent bank are added together. In cases with multiple book managers, equal

credit is given to each bank. Market share is then defined as the sum of gross proceeds for the bank

divided by the sum of gross proceeds for all IPOs over the sample period. We repeat this measurement at

the industry level. Industry market share is computed as the sum of gross proceeds on all IPOs in the

same Fama-French 48 industry group taken public by the bank during the time period, divided by the sum

of gross proceeds on all IPOs issued in that industry over the same period.

Table 2 examines changes in investment bank market share surrounding all-star analyst turnover.

For the bank gaining the star analyst, we compute both industry and overall (across all industries) market

share in the year before and the year after the analyst issues her first recommendation with the new bank.

For the bank losing the star, we perform similar calculations but use one year prior to and one year

following the date of the analyst’s last recommendation with her old bank. Because deal activity is

10

substantially higher in the latter half of our sample, we also present results separately for the post-1994

period.

Contrary to the certification and market power hypotheses, losing an all-star appears to have little

adverse effect on an investment bank. The bank’s market share in the year prior to the analyst’s departure

is 3.035% percent, on average. This increases an insignificant 0.37% following the departure of the all-

star. Similar results hold at the industry level. The bank’s average market share in the all-star’s industry

prior to the switch is 3.555%. This increases an insignificant 0.045% following the star’s departure.

Similar results hold in the post-1994 period.6 The investment bank losing the analyst seems justified,

therefore, in letting the all-star depart. This could reflect self-selection in the data, since if the bank

thought that losing the star would hurt, it might have fought harder to keep her. The evidence is also

consistent with the aggressiveness hypothesis, which predicts that banks losing a star would fight to

preserve market share.

In contrast to the above results, the bank gaining the all-star experiences a sharp increase in

market share both at the industry level and at the overall bank level. The investment bank experiences a

statistically significant 1.246% (t=2.36) increase in market share in the year following the arrival of the

all-star. At the industry level, market share increases 1.812%. While the industry market share change is

larger in magnitude than the overall market share change, it is not statistically significant (t=1.58). In the

post-1994 period, changes in both measures of market share are statistically significant.

Interestingly, the number of initial public offerings in the analyst’s industry increases

significantly in the post-turnover period. In the year following the departure of the all-star, there is an

approximately 40% increase in the number of initial public offerings in the star’s industry. This is

consistent with investment banks being able to predict “hot” industries and securing all-stars in those

industries in anticipation of the increase in deal activity.

6 In all subsequent tests, we also examine the post-1994 period separately. However, since our results do not differ substantially from the full sample results, we do not report them.

11

While these results seem to suggest that star analysts, on average, increase the market share of the

banks to which they move, it is important to note that causality could be an issue. For example, the above

results are also consistent with the all-star jumping to an investment bank that is on the rise, from a bank

that has plateaued. In order to address this issue, we examine the sequence of six-month market share

numbers beginning 18 months prior to the all-star switching firms. The bank that the analyst switches to

is doing considerably better in months –18 to –12 than the bank that the analyst switches from. However,

this difference is insignificant in months –6 to 0. This result is not consistent with analysts moving to

banks that are on the rise.

Table 3 presents market share results for the bank losing the all-star, conditioned on several

variables. We stratify the sample by (1) whether the all-star was replaced by another all-star, (2) the

number of total all-stars at the bank prior to the star’s departure, (3) the Institutional Investor ranking of

the analyst, and (4) the relative reputation of the analyst’s previous and new employer. Replacement of

the star provides some indication of the significance of the departure. It is also consistent with an

investment banking market that is heating up. We find that the bank losing the star sees a statistically

significant reduction in its market share of 1.811% if the star is replaced. Interestingly, industry market

share increases (although not significantly). This suggests that the bank attempts (successfully) to remain

competitive in the stars industry but this focus (and perhaps an overall drop in stature) is damaging

overall. Self-selection again could affect the interpretation of this result. Only in those cases where the

analyst departure is damaging to the bank and its ability to gain business will the bank attempt to replace

her. The replacement decision may be beneficial for the bank if its market share would have been even

worse without making the replacement.

While we consider the number of stars at the bank, we have no strong prior beliefs regarding the

relation between number of stars and market share. Gaining or losing a star may be more significant for

banks with few analysts since bank capabilities are more dramatically affected whereas the loss or gain of

a star at a bank with numerous stars may not be noticeable. On the other hand, banks with more stars are

presumably more reputable so changes to capabilities, even if small, can have a dramatic effect on

12

perceived bank quality (see Chemanur and Fulghieri, 1994). We find, however, no relation between

number of stars and market share changes for banks losing a star.

Gaining or losing a ranked analyst (i.e. first, second, or third team) is likely to have a more

significant effect on investment banks. We find no significant effect on market share changes, however,

when turnovers are broken down by all-star ranking.

We also consider market share changes broken down by relative pre-move reputation of the banks

gaining and losing the star. The certification hypothesis should predict that analyst moves would be more

significant when the two banks are of relatively similar stature.7 Losing a star to a bank that already was

far more reputable or less reputable should not have a significant impact on a banks ability to compete for

business. The market power hypothesis would make a similar prediction. The bank aggressiveness

hypothesis could predict the opposite for banks losing a star. When losing a star to a bank of similar

stature, the losing bank may become even more aggressive, reducing possible effects to market share. We

measure pre-move reputation using the Carter-Manaster ranking. We group all-star turnovers based on

whether pre-move Carter-Manaster rankings are similar (within 1 ranking) or different.8 We find no

evidence of significant market share changes (at the aggregate or industry level) for these groups

Table 4 presents the various market share stratifications for the bank gaining the all-star. Market

share gains are significant if the star analyst is not replaced at the original bank and insignificant if the

star is replaced by another star analyst at the original bank. This is consistent with increased competitive

pressure negating the gains from acquiring the star. Analyst quality is also an important determinant of

the market share gains accruing to the bank acquiring the all-star. Acquiring an analyst who was ranked

on Institutional Investor’s First, Second, or Third Team, increases overall bank market share by 1.603

percentage points and industry market share by 4.326 percentage points. Both numbers are statistically

7 Much of the existing liturature assumes that reputation is a function solely of actions taken by a given bank (e.g. Chemmanur and Fulghieri, 1994, Dunbar, 2000, Krigman Womack and Shaw, 2001). In this view, relative stature should not have an impact on market share. Investment banks compete for a fixed number of offerings, however, suggesting that reputation should be viewed in a relative sense. Mispricing by one bank might not damage its reputation if others perform more poorly, for example. 8 This cutoff is admittedly arbitrary (it was initially selected so that subsamples are of approximately equal size). Our findings are not significantly affected, however, when other cutoffs are used.

13

significant. In contrast, acquiring an analyst who finished as a Runner-up does not have any significant

impact on market share at either the industry or aggregate level. Finally, market share gains are positive

and significant (at the 10% level) if the bank acquiring the star started had a similar reputation to the bank

losing the star, consistent with the certification and market power hypotheses.

5. Bank and Analyst Performance in the IPO Market

In this section we examine the pricing and performance of IPOs as well as analyst forecast

activity around star turnovers. While the evidence on market share appears to be primarily consistent

with the investment bank aggressiveness hypothesis, an examination of bank and analyst behavior should

yield a deeper understanding of how banks losing stars are able to preserve market share and how banks

acquiring a star expand market share.

5.1 IPO pricing and performance - hypotheses

For banks losing or gaining stars, we examine the underwriter spread, the initial (first day)

returns, the price adjustments and the one-year post offering abnormal return for issuers taken public by

the banks over the year before and after the analyst turnover.

The certification hypothesis discussed previously argues that banks with greater reputation can

price IPOs closer to intrinsic value, resulting in lower first day returns. If adding a star analyst adds to the

reputation of the firm, underpricing on IPOs should decrease after the star joins the bank. Similarly, if

losing a star reduces reputation, initial returns should increase after a star leaves the firm.

In contrast, the market power hypothesis predicts that initial returns should increase for banks

gaining a star analyst and decrease for firms losing the star. Loughran and Ritter (2002b) note that if

underwriters are given discretion in allocating shares, they will not always act in the best interest of the

issuing firm. The underwriter might purposely leave more money on the table than necessary, and then

14

allocate these shares to favored clients. Adding an all-star might give more market power to the bank

relative to the issuer. This gain in market power could lead to an increase in underpricing. Since analyst

turnover occurs prior to an increase in IPO activity (see table 2), issuers will be more willing to accept

mispricing (Loughran and Ritter, 2002a, use prospect theory to formally make this point). On the other

hand, banks losing a star should lose market power, leading to lower initial returns.

The investment bank aggressiveness hypothesis yields no obvious predictions regarding IPO

initial returns. As noted previously, banks both losing and gaining a star analyst may be more aggressive

in selecting issuers, resulting in a pool of firms that are more speculative and of lower average quality.

Since initial returns are higher for more speculative issuers, we could observe higher returns for both

banks after the turnover. Controlling for issuer characteristics, it is not obvious whether underpricing

would be abnormally higher, however.

Dunbar (2000) finds that banks cutting fees (underwriter spread) realize increases to their market

share.9 Krigman, Shaw and Womack (2001), however, find that fees play no significant role in the

underwriter choice decision. The certification hypothesis would predict lower investment banking fees

for firms losing a star and higher fees for firms gaining a star. Booth and Smith (1986) argue that firms

may be willing to accept more positive first-day returns when using a less reputable investment bank if

the bank reduces its fees. Conversely, banks with greater reputation can charge higher fees (in part as

compensation for the greater reputation it places at risk in each offering). The market power hypothesis

would make similar predictions. Banks gaining the star have more bargaining power and can, therefore,

extract higher fees from issuers. Banks losing a star lose bargaining power, resulting in lower fees. The

bank aggressiveness hypothesis would predict that both banks would cut fees as part of an aggressive

strategy to increase (or preserve) market share.

Benveniste and Spindt (1989) develop a model to explain the well-known “partial adjustment”

phenomenon. They note that revisions to the offer price prior to the offer date are a product of

9 Pepall and Richards (2001) formally argue that a firm without a star may be so disadvantaged relative to the rival employing the star, that the firm must cut its price aggressively in order to compete.

15

information gathered by underwriters from investors during the pre-issue period. In their model, investors

must be persuaded to truthfully reveal their private demand for an issue. When an investment bank learns

that demand for an issue is higher than expected, the offer price is raised, but not to the full market value.

The remaining adjustment comes in the form of underpricing, which compensates investors for supplying

information. Hanley (1993) finds evidence consistent with this story. She finds that underwriters prefer

to compensate investors for revealing information by allocating a small number of greatly underpriced

shares, rather than a large amount of somewhat underpriced shares. Hanley also finds that partial

adjustments are more positive for reputable banks, consistent with the notion that reputable banks have

greater ability to illicit truthful positive information since they have a greater expected future deal flow

(the ability to exclude investors from future deals provides additional incentives for truthtelling). The

certification hypothesis would, therefore, predict that price adjustments increase (decrease) after gaining

(losing) a star analyst due to the change in bank reputation that results. Since reputable banks should be

better able to uncover both positive and negative information from investors, the volatility of price

adjustments should also increase (decrease) after gaining (losing) a star.10

The market power hypothesis would predict that partial adjustments are more positive for banks

after they acquire an all-star analyst. Loughran and Ritter (2002b) argue that issuers are complacent

about IPO underpricing in cases where their wealth is increased relative to expectations. Banks with

greater market power can manage those expectations in part by initially suggesting low filing prices.

When the bank increases the price during the bookbuilding process, the issuer becomes more complacent

and does not complain if the resulting underpricing is in large. Banks losing a star should lose their

market power resulting in lower price adjustments. The market power hypothesis makes no obvious

prediction regarding the volatility of price adjustments (banks gaining a star should have more positive

adjustments but fewer negative price adjustments).

10 While more reputable banks can better elicit positive and negative information, they also select less speculative firms to take public. It is not obvious which effect will dominate.

16

The investment bank aggressiveness hypothesis would predict greater volatility of price

adjustments for both banks acquiring and losing a star analyst (by taking on more speculative issuers,

banks are more likely to learn both negative and positive information during bookbuilding). No obvious

predictions emerge regarding average price adjustments, however.

The long-run performance of issuers may also be affected by all-star analyst turnover. In the

certification model of Chemmanur and Fulghieri (1994), two types of firms attempt a public offering:

firms that have good prospects after the offering and firms that have poor prospects after the offering.

Investment banks evaluate firms, and market only those firms that they believe have good prospects. A

bank’s reputation evolves based on its ability to accurately screen for good performers. Taking a firm

public that has good prospects enhances reputation, whereas taking a firm public that has poor prospects

hurts reputation. Empirically, firms with good (poor) prospects should have positive (negative) abnormal

long-run performance. If all-star analysts have a superior ability to screen the quality of deals, then the

certification hypothesis would predict that post-IPO performance of firms taken public by a bank losing

(gaining) an all-star should be more negative (more positive) than prior to the move. Chemmanur and

Fulghieri (1994) and Carter and Manaster (1990) also argue that more reputable banks should be more

selective in their choice of issuers, resulting in a pool of firms that are less speculative. Since returns for

less speculative firms should be more predictable, the vola tility of post-IPO returns should decrease

(increase) for banks gaining (losing) the star reflecting the change of firm mix resulting from the change

in reputation.

The market power hypothesis would predict that firms acquiring a star should see a decline in

average post-IPO performance. The evidence presented in Baker and Wurgler (2002) offers an

alternative view. They find evidence that a firm’s capital structure evolves as the collective outcome of

past attempts to time the market. All-stars might possess superior market-timing skills that help clients go

public during a window of opportunity when the stock is overvalued. This market timing would result in

more negative long-run performance for IPOs taken by public by the investment bank following the

addition of an all-star. Conversely, banks losing a star might lose some of this market-timing ability

17

resulting in a positive change in long-run performance. The market power hypothesis makes no obvious

prediction regarding the volatility of long-run performance, however.

The bank aggressiveness hypothesis would predict that banks both losing and gaining a star

become more aggressive in the selection of banks. Average quality of issuers should decline, manifesting

itself as lower long-run performance. Also, since the average risk of issuers should increase, volatility of

long run-returns should also increase for both banks.

5.2 IPO pricing and performance – Empirical methods and results

For those firms with CRSP data, underpricing is defined as: 100*(P1st day close – Poffer)/Poffer, where

P1st day close is the closing price at the end of the first-day of trading and Poffer is the offering price from

SDC. Following Dunbar (2000), our measure of the underwriting spread is defined as 100*(SP/Poffer),

where SP is the gross spread per share in the offering and Poffer is the offering price. Price Adjustment for

each IPO is defined as the final offering price minus the average of the high and low initial filing prices

all divided by the average of the high and low initial filing prices.

Beatty and Ritter (1986), Beatty and Vetsuypens (1995) and Dunbar (2000) note that there exists

normal, or predictable, variation in the above-mentioned IPO performance variables. In addition to

unadjusted performance measures, therefore, we also focus on abnormal measures. To identify the

normal first day return, we carry out separate regressions each year of the first-day return on various firm

and market condition variables suggested in the literature (see Appendix A for details). The abnormal

first-day return is then defined as the actual percentage return minus the predicted first day return, using

the estimated regression results for year the issuer goes public. Our measure of abnormal fees and

abnormal price adjustments are similarly defined.

One-year buy-and-hold returns are measured for each issuer on CRSP from the end of the first

month of trading through the following twelve months. In order to measure abnormal returns each IPO is

matched with a portfolio of public firms based on the issuer’s market capitalization and book-to-market

18

ratio. Market capitalization is computed as the IPO price multiplied by the number of shares outstanding

after the IPO. The number of shares outstanding after the IPO is primarily obtained from CRSP. In cases

where SDC notes that the IPO shares are dual class, the shares outstanding are obtained from SDC or

EDGAR.11 Book-to-market is computed as the book value of equity per share prior to the IPO plus the

value of primary shares in the offering divided by the firm’s market capitalization. At the end of the first

month of trading, the IPO firm is matched to a size and book-to-market portfolio.12 The abnormal return

for the IPO is computed as the buy-and-hold return on the issuer for the following twelve months net of

the buy-and-hold return on the size and book-to-market portfolio.13

Our analysis of IPO pricing and performance for banks losing and acquiring a star analyst is

summarized in Table 5. For the bank gaining the all-star, we find no significant change in abnormal

underpricing. This is not consistent with certification (which would predict a negative change) or market

power (which would predict a positive change). The abnormal spread is positive and statistically

significant in the year prior to the move and does not change significantly following the arrival of the all-

star. There is no evidence that banks gaining a star cut fees to more aggressively compete for business.

Abnormal one-year post IPO performance is statistically significantly positive prior to the arrival of the

all-star and falls a statistically significant 11.1% following the arrival of the all-star. This is consistent

with the investment bank taking on a greater number of marginal deals to improve market share. It is also

consistent with banks using additional market power to market IPOs during windows of opportunity with

investors are likely to overvalue shares. Price adjustments are not unusual prior to analyst turnover, but

become more positive afterward. This is consistent with certification.

For the investment bank losing the all-star, we find little to suggest that there is a decline in

reputation. Abnormal underpricing is significantly positive prior to the departure of the all-star, which is

consistent with greater market power. This measure does not change significantly following the departure

11 Loughran and Ritter (2002b) find the CRSP misstates number of shares outstanding when the firm has more than one class of shareholders. 12 We use the Fama-French 100 portfolio cutoffs (ten market capitalization by ten book-to-market). See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html for details. 13 If the issuer delists prior to 12 months, the abnormal return calculation ends at the month of delisting. We also considered other long-run return windows (up to three years) and found similar results to those presented here.

19

of the star, suggesting little change to a bank’s market power or reputation. Abnormal price adjustment is

positive and significant prior to the departure of the star, which indicates an ability to extract positive

information. There is no significant change after the star’s departure. Abnormal compensation is

significantly positive pre-move and declines significantly following the departure of the star. Thus, banks

respond to the loss of the star by cutting spreads and competing more on price, consistent with the bank

aggressiveness hypotheses. It is also consistent with the view that banks having reduced reputation are

not able to charge increased fees.

5.3 Analyst forecasts – hypotheses

For banks losing or gaining stars, we examine several measure of analyst activity on issues taken

public by the bank, including the proportion of IPOs where the analyst is first to make a forecast on the

issuing firm, the number of forecasts made on the issuing firm over the year after it goes public, the days

until the first forecast is made by the lead bank analyst and the level of earnings forecasts, in relation to

consensus, for the lead bank analyst on IPOs taken public by the analyst’s bank.

The certification hypothesis makes no obvious predictions regarding the timing and frequency of

analyst forecasts. Graham (1999) shows theoretically that career concerns lead analysts with more

reputational capital at stake to be more conservative in their earnings forecasts. The intuition behind the

model is that analysts with high reputation issue more conservative forecasts to protect their status and

pay level. Banks acquiring (losing) a star should, therefore, reduce (increase) their earnings forecasts

relative to consensus, on average. Volatility of forecasts relative to consensus also provides information

on the riskiness of firms taken public. Banks acquiring (losing) a star should be more (less) selective in

their choice of firms leading to a reduced (increased) volatility of earnings forecasts relative to consensus.

Issuers appear to prefer active aftermarket activity by their investment banks. Krigman, Shaw,

and Womack (2001) find that aftermarket activities including analyst forecasts most help to explain

underwriting switching decisions. Banks seeking to solidify or improve their market power after they

20

acquire a star should provide more earnings forecasts and provide them earlier in the aftermarket for firms

they take public. The forecasts themselves may not be different before and after the move, however.

The bank aggressiveness hypothesis makes similar predictions regarding the timing and

frequency of forecasts for both banks associated with the analyst turnover. To more aggressively compete

for business, banks both losing and gaining a star analyst should issue forecasts earlier and more

frequently. Unlike the market power hypothesis, the bank aggressiveness hypothesis predicts that

earnings forecasts should also become more positive, relative to consensus, for both banks. Finally, since

both banks should be willing to underwrite IPOs for more speculative offerings, the volatility of forecasts

relative to consensus should increase following the analyst turnover.

5.4 Analyst forecasts – Empirical methods and results

We consider a number of proxies designed to capture the aggressiveness of research analysts.

First forecast is a dummy variable taking the value of one if the lead investment bank is the first bank to

issue a forecast for an initial public offering and zero otherwise. Number of recommendations is the

number of recommendations issued by the lead underwriter in the year following the issue date. Days to

first recommendation is the number of calendar days from the initial public offering’s issue date to the

first forecast by the lead investment bank. We define analyst forecast error as the difference in the

earnings per share estimate of the lead investment bank relative to the corresponding consensus estimate

scaled by the stock price. Each of the above measures is calculated for each IPO in our sample using data

from the I/B/E/S files.

In Table 6, we examine the impact of all-star turnover on various measures of analyst

performance. For the bank gaining the all-star, we find an increase in analyst aggressiveness. Analysts

issue a statistically significant 1.1 more forecasts, on average, than before the arrival of the star and also

issue forecasts 8.60 days sooner. The higher frequency of forecasts and the increase in timeliness does

not seem to affect forecast quality. The forecast error is slightly negative pre-move indicating that the

21

lead investment bank is slightly more conservative than the consensus estimate and it does not change

significantly after the arrival of the star. There is a significant increase in the standard deviation of the

forecast error. This is consistent with results in the previous section suggesting that the investment bank

takes on more speculative deals following the arrival of the star analyst.

For the bank losing the all-star, we find that the remaining analysts also become somewhat more

aggressive. In the year following the departure of the star, they issue their first forecasts on initial public

offerings for which they were the lead underwriter 11.8 days earlier than before the all-star departed.

There is no significant change in the number of recommendations issued for each IPO, nor does the

forecast error change significantly following the departure of the star. The standard deviation of the

forecast error does not change significantly following the departure of the star, suggesting that the bank

does not take on more speculative deals.

6. Factors affecting IPO market share for banks with analys t turnovers

In this section we examine in a multivariate regression framework the factors affecting

investment bank market share changes around all-star analyst turnovers. Following Dunbar (2000) we

regress the change in market share from the year prior to the analyst turnover for the bank to the year after

the turnover year on various factors that capture the bank’s performance on IPOs in the year prior to the

move. Our regressions build on Dunbar’s in three ways. First, we estimate separate regressions for banks

losing and gaining the star. Since the market share effects of turnover are different for these two groups

of banks, as reported earlier, we would expect the relations between market share changes and IPO

performance factors to be different. Second, we include more IPO performance factors than considered

previously. Third, we also include variables that capture the change in IPO performance factors in our

regressions. Evidence in section 5 suggests that bank and analyst performance changes after the star

turnover. Including change in performance factors allows us to determine whether that changed behavior

had an impact on market share.

22

In our regressions, we include all performance measures considered previously as independent

variables. Specifically, we include as independent variables in our regressions the mean abnormal first

day returns, the mean abnormal spreads, the mean abnormal price adjustments, the standard deviation of

abnormal price adjustments, the mean abnormal one-year returns, the standard deviation of abnormal one-

year returns, the percentage of first forecasts, the mean number of recommendations per IPO, the mean

days to first recommendation per IPO, the mean forecast error and the standard deviation of forecast error.

All these variables are measured over the year prior to the analyst turnover. We also include in our

regressions the changes in these variables from the year prior to the turnover to the year after the turnover.

Based on the discussion in sections 5.1 and 5.3, the basic alternative hypotheses would make

different predictions regarding the relation between bank and analyst performance measures and IPO

market share changes. The certification hypothesis would predict a positive relation market share

changes and mean abnormal spreads, mean abnormal price adjustments, standard deviation of abnormal

price adjustments and mean abnormal one-year returns. Banks with less reputation must charge lower

fees to compete. Banks acquiring more positive and negative information prior to pricing of IPOs

enhance their reputation and market share. Similarly, banks enhance their reputation and market share by

bringing public firms that perform better post-IPO. The certification would also predict a negative

relation between market share changes and mean abnormal first day returns, standard deviation of

abnormal one-year returns, mean forecast errors and the standard deviation of forecast errors. Banks

leaving too much money on the table or bringing forward more speculative offerings damage their

reputation and market share. Aggressive analyst behavior (by issuing more biased recommendations) also

hurts reputation and market share.

The market power hypothesis would predict a positive relation between market share changes and

mean abnormal first day returns, mean abnormal spreads, mean abnormal price adjustments, percentage

of first forecasts, and mean number of recommendations per IPO. Banks with increasing market power

should see increases to their market share. Leaving more money on the table, charging higher fees,

realizing positive price increases in the IPO process and making earlier and more frequent analyst

23

earnings forecast are all evidence of increased market power. The market power hypothesis would also

predict a negative relation between market share changes and mean abnormal one-year returns and days

to first forecast. Again, early forecasts are evidence of increased market power. Similarly, the ability to

take firms public during periods where they are more likely to be overvalued (resulting in negative post-

IPO performance) is evidence of market power.

The investment bank aggressiveness hypothesis would predict a positive relation between market

share changes and mean abnormal first day returns, standard deviation of abnormal price adjustments,

standard deviation of abnormal one-year returns, the percentage of first forecasts, mean number of

recommendations per IPO, mean forecast error and the standard deviation of forecast error. Aggressive

banks are more likely to take on more speculative offerings (proxied by the volatility measures and

abnormal first day returns) and make earlier, more frequent and more positive earnings forecasts on the

firms they take public. This aggressiveness in turn allows the bank to expand its market share. The

aggressiveness hypothesis would predict a negative relation between market share changes and mean

abnormal spreads, mean abnormal one-year returns and days to first recommendation. Aggressive banks

will try to exploit windows of opportunity where issuers can issue overvalued shares. They will also

make earlier recommendations. Finally, more aggressive banks would be more likely to cut fees to attract

market share.

In addition to the bank and analyst performance variable s noted above, we include various

dummy variables to capture the nature of the analyst turnover. These variables emerge from the analysis

in Section 4. For example, we include dummy variables capturing whether the star is replaced at the bank

losing the star, whether the bank had more than ten all-stars prior to the move, whether the star analyst

was ranked first, second or third team by Institutional Investor, and whether the banks had similar pre-

move reputation (proxied by Carter-Manaster rank).

In Table 7 we present the various regression model estimates for the bank gaining the all-star.

Given the high correlations among some independent variables, we report univariate and multivariate

model results. We only report our most parsimonious models (i.e. we do not report univariate regressions

24

for variables that are not significant and do not include those variables in our multivariate models. In

model (1), we find that the market share changes are positively related to mean abnormal one-year

returns, consistent with the certification hypothesis. The coefficient on the change in mean abnormal one-

year return is not significant, indicating that the market does not respond to changes in IPO performance

after the star analyst turnover. In model (2) we find that market share changes are significantly positively

related to the standard deviation of pre-move abnormal one-year return. This is more consistent with

market power and bank aggressiveness. It suggests that banks taking on more speculative offerings

expand their market share. The coefficient on the change in standard deviation of abnormal one-year

return is not significant, indicating that the market does not respond to changes in the volatility of IPO

performance after the star analyst turnover. Specification (3) indicates that the standard deviation of

abnormal price adjustment has a significantly positive effect on market share changes, consistent with

certification and aggressiveness. Again, the coefficient on the change in standard deviation of abnormal

price adjustment is not significant, indicating that the market does not respond to changes in the volatility

of price adjustments after the star analyst turnover. In specification (4), we find a positive (but

insignificant) relation between market share changes and the proportion of IPO where the lead bank

makes the first post-IPO earnings forecast. We also find a significantly positive relation between the

change in the proportion of IPOs where the lead bank makes the first post-IPO forecast and market share

changes. This is consistent with market power and aggressiveness. As banks move to increase their

frequency of first forecasts, they are rewarded with higher market shares. In model (5) we find a

significantly positive relation between mean forecast errors and market share changes. We also find a

significantly positive relation between changes in mean forecast errors and market share changes. This

evidence is consistent with the bank aggressiveness hypothesis. Banks that are more positive in their

forecasts (and become more positive after gaining the star) gain market share. Thus, while we find that

banks on average do not become more aggressive in the forecasts (see Table 6), those banks that do

become more aggressive are rewarded with increases to their market share.

25

In regressions (6) to (8) of Table 7, we estimate multivariate regression models using the

significant factors identified in regressions (1) to (5). We do not include the IPO performance measures

together given their high correlations (when included together, all coefficients become insignificant). The

evidence in regressions (6) to (8) generally is consistent with that from prior models. The mean pre-move

abnormal one-year return becomes insignificant as does the change in mean forecast errors, however.

In Table 8, we present similar regression model estimates for the bank losing the all-star. As in

Table 7, given high correlations among some independent variables, we report univariate and multivariate

model results. We also only report our most parsimonious models (i.e. we do not report univariate

regressions for variables that are not significant and do not include those variables in our multivariate

models). In model (1), we find that market share changes are negatively related to a dummy variable

taking the value one if the star is replaced. This is consistent with prior evidence in Table 3. In model

(2), we find that market share changes are significantly positively related to mean abnormal initial returns,

consistent with the market power and aggressiveness hypotheses. It is also inconsistent with evidence

reported by Dunbar (2000). The coefficient on the change in mean abnormal initial return is also

significantly positive, indicating that the market responds to changes in IPO initial performance after the

star analyst leaves. In model (3) we find that market share changes are significantly positively related to

the standard deviation of pre-move abnormal one-year return (and changes to this volatility measure).

This is consistent with the market power and bank aggressiveness hypotheses. It suggests that banks

taking on more speculative offerings expand their market share.

Specification (4) indicates that the standard deviation of abnormal price adjustment (and its

changes) has a significantly positive effect on market share changes, consistent with certification and

aggressiveness. In specification (5), we find a significantly positive relation between market share

changes and the mean number of forecasts made by bank prior to losing the star. This is consistent with

the market power and aggressiveness hypotheses. The coefficient on the change in mean number of

forecasts is not significant, indicating that the market does not respond to changes in forecast volume after

the star analyst turnover. Model (6) shows similar findings for the mean days until first recommendation.

26

Model (7) shows that the change in standard deviation of forecast errors is significantly positively related

to market share changes. This suggests that banks changing their behavior after losing a star by taking on

more speculative offerings are rewarded with increased market shares, consistent with the aggressiveness

hypothesis.

In regressions (8) and (9) of table 8, we estimate multivariate regression models using the

significant factors identified in regressions (1) to (7). We do not include the standard deviation of

abnormal one-year return and the standard deviation of abnormal price adjustments together given their

high correlations (when included together, all coefficients become insignificant). The evidence in

regressions (8) and (9) generally is consistent with that from prior models. The dummy variable taking

the value one if the star is replaced, mean abnormal underpricing, mean number of forecasts, mean days

to first forecast and standard deviation of forecast errors are not significant in these specifications,

however.

Overall, the evidence on factors affecting investment bank market share changes after a star

analyst turnover is primarily consistent with the market power and aggressiveness hypotheses. Banks

taking on more speculative offerings and those making quicker, more frequent and more positive earnings

forecasts are more likely to see their market share improve. While some evidence is supportive of the

certification hypothesis, there is some significant inconsistent evidence. Significantly, mean abnormal

initial returns are positively related to market share changes.

7. Conclusions

This paper examines the impact of all-star analyst turnover on initial public offering market share.

Using a sample of 222 Institutional Investor All-American analysts who switch investment banks between

1988 and 1999, we find that investment banks losing all-stars do not experience a significant decline in

either industry level or aggregate market share following the departure of the star, while acquiring an all-

star significantly improves a bank’s share of the initial public offering market. These results provide

27

some justification for the high salaries received by some research analysts. The average gain in IPO

market share an investment bank captures by a acquiring an all-star analyst corresponds to roughly an

annual increase of $22 million in fees.

Both losing and acquiring an all-star has a significant impact on the performance of their initial

public offerings. The bank acquiring the all-star becomes more aggressive in attracting business by

taking on more speculative deals and issuing forecasts for initial public offerings earlier and more often.

The bank losing the all-star attempts to compete on price by cutting fees. Analysts at the bank also

become more aggressive by issuing forecasts sooner after an IPO.

Building on the analysis in Dunbar (2000), we examine the factors affecting market share

changes around analyst turnover. Our evidence reverses several findings by Dunbar (2000). For banks

losing a star analyst, we find that market share changes are positively related to past mean abnormal

underpricing. Banks leaving more money on the table are rewarded with increased share in the IPO

market. For both banks losing and gaining a star, market share changes are positively related to volatility

of abnormal long-run returns on past IPOs. This suggests that banks taking on more speculative issues

are also rewarded with increased IPO market share. Overall our evidence suggests that bank

aggressiveness is rewarded with increased market share.

References

Baker, Malcolm and Jeffrey Wurgler, 2002, Market timing and capital structure, Journal of Finance 57, 1-32. Baron, David, 1982, A model of the demand for investment banking advising and distribution services for new issues, Journal of Finance 37, 955-976. Baron, David and Bengt, Holmstrom, 1980, The investment banking contract for new issues under asymmetric information: Delegation and the incentive problem, Journal of Finance 35, 1115-1138. Barry, C., Muscarella, C., Peavy, J., Vetsuypens, M., 1990. The role of venture capital in the creation of public companies: Evidence from the going public process. Journal of Financial Economics 27, 447-471. Bates, Thomas, and Craig Dunbar, 2002, Investment bank reputation, market power, and the pricing and performance of IPOs, Working paper, University of Western Ontario. 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, and Michael Vetsuypens, 1995, Underpricing, overpricing, and reputation: Are underwriters penalized for IPO mispricing? Unpublished working paper, Southern Methodist University, Dallas, TX. Benveniste, Lawrence M, Alexander P. Ljungqvist, William J. Wilhelm and Xiaoyun Yu, 2002, Evidence of Information Spillovers in the Production of Investment Banking Services, Journal of Finance, forthcoming. Benveniste, L. and P. Spindt, 1989, How investment bankers determine the offer price and allocation of new issues, Journal of Financial Economics 24, 343-361. Booth, James R. and Lena Chua, 1996, Ownership dispersion, costly information, and IPO underpricing, Journal of Financial Economics 41, 291-310. Booth and Smith, 1986, Capital raising, underwriting and the certification process, Journal of Financial Economics 15, 261-281. Bradley, Daniel, and Brad Jordan, 2002, Partial adjustment to public information and IPO underpricing, Journal of Financial and Quantitative Analysis forthcoming. Carter, R., Dark, F., Singh, A., 1998. Underwriter reputation, initial returns, and the long-run performance of IPO stocks. Journal of Finance 53, 285-311. Carter, Richard, and Stephen Manaster, 1990, Initial public offerings and underwriter reputation, Journal of Finance 45, 1045-1067. Chemmanur, T. and P. Fulghieri, 1994, Investment bank reputation, information production, and financial intermediation, Journal of Finance 49, 57-79.

Cooper, Rick, Theodore Day, and Craig Lewis, 2001, Following the leader: a study of individual analysts’ earnings forecasts, Journal of Financial Economics 61, 383-416. DuCharme, Larry L., Shivaram Rajgopal, and Stephan E. Sefcik (2001) “Why was Internet IPO Underpricing so Severe?” unpublished University of Washington working paper. Dunbar, Craig, 2000, Factors affecting investment bank initial public offering market share, Journal of Financial Economics 55, 3-41. Fama, Eugene and James MacBeth, 1973, Risk, return, and equilibrium: empirical tests, Journal of Political Economy, 81, 607-636. Graham, John, 1999, Herding among investment newsletters: Theory and evidence, Journal of Finance 54, 237-268. Hanley, K., 1993, The underpricing of initial public offerings and the partial adjustment phenomenon, Journal of Financial Economics 34, 231-250. Hansen, Robert and Paul Torregrosa, 1992, Underwriter compensation and corporate monitoring, Journal of Finance, 47, 1537-1555. Hong, Harrison, and Jeffrey Kubik, 2002, Analyzing the analysts: career concerns and biased earnings forecasts, forthcoming, Journal of Finance. Hong, Harrison, Jeffrey Kubik, and Amit Solomon, 2000, Security analysts: career concerns and herding of earnings forecasts, Rand Journal of Economics 31, 121-144. James, Christopher, 1992, Relationship-specific assets and the pricing of underwriter services, Journal of Finance, 47, 1865-1885. Johnson, J. and R. Miller, 1988, Investment banker prestige and the underpricing of initial public offerings, Financial Management 17, 19-29. Kessler, Andy, 2001, We’re all analysts now, Wall Street Journal, July 30, A18. Krigman, Laurie, Wayne Shaw, and Kent Womack, 2001, Why do firms switch underwriters? Journal of Financial Economics 60, 245-284. Laderman, Jeffrey, 1998, Wall Street’s spin game: stock analysts often have hidden agenda, Business Week , October 5, 148. Li, Xi, 2002, Career concerns of equity analysts: compensation, termination, and performance, working paper, Vanderbilt University. Ljungqvist, Alexander P. and William J. Wilhelm, 2002a, IPO Pricing in the Dot-Com Bubble, Journal of Finance, forthcoming. Ljungqvist, Alexander P. and William J. Wilhelm, 2002b, IPO Allocations: Discriminatory or Discretionary?, Journal of Financial Economics, forthcoming.

Loughran, Tim and Jay Ritter, 2002a, Why don’t issuers get upset about leaving money on the table in IPOs, Review of Financial Studies 15, 413-443. Loughran, Tim, and Jay R. Ritter, 2002b, Why has IPO underpricing changed over time? Unpublished University of Florida working paper. Lowry, Michelle, and G. William Schwert, 2002, Biases in the IPO Pricing Process, Penn State University and University of Rochester working paper. Megginson, William, and Kathleen Weiss, 1991, Venture capitalist certification in initial public offerings, Journal of Finance 46, 879-903. Michaely, Roni, and Kent Womack, 1999, Conflict of interest and the credibility of underwriter analysts recommendations, Review of Financial Studies 12, 653-686. Nocera, Joseph, 1997, Who really moves the market? Fortune, October 27, 90-110. Pepall, Lynne and Daniel Richards, 2001, Reach for the stars: A strategic bidding game, Economica 68, 489-504. Rock, Kevin, 1986, Why new issues are underpriced, Journal of Financial Economics 15, 187-212. Stickel, Scott, 1992, Reputation and performance among security analysts, Journal of Finance 47, 1811-1836.

Table 1 All-star analyst turnover by year and by team ranking Our sample consists of 222 cases where an all-star analysts switches investment banks between 1988 and 1999. Data on # of analysts, # of institutions, and turnover or obtained from the I/B/E/S detail file. We define all-star analysts as those finishing either first, second, third, or runner-up on Institutional Investor’s annual ranking of analysts. We consider only turnover cases where an analyst was an all-star in either the year prior to or the year of her switch. In Panels B, we calculate the frequency of turnover by position on the Institutional Investor All-American Research Team. Panel A: Analyst Turnover by Year

# of

Analyst

# of All-

Stars

# of

institutions

Turnover

%

Turnover

All-star turnover

Percent All-star turnover

1988 2618 325 172 154 5.88% 7 2.15% 1989 2841 368 183 249 8.76% 23 6.25% 1990 2648 336 187 149 5.63% 9 2.68% 1991 2440 331 191 151 6.19% 9 2.72% 1992 2269 353 192 114 5.02% 7 1.98% 1993 2479 389 221 166 6.70% 16 4.11% 1994 2876 371 226 219 7.61% 23 6.20% 1995 3141 262 231 248 7.90% 20 7.63% 1996 3528 267 261 282 7.99% 21 7.87% 1997 3997 272 308 349 8.73% 27 9.93% 1998 4410 322 351 404 9.16% 26 8.07% 1999 4543 344 329 367 8.08% 34 9.88% Panel B: All-Star analyst turnover by Institutional Investor team ranking Place # of Analysts % of sample First Team 31 13.96% Second Team 35 15.77% Third Team 55 24.77% Runner Up 101 45.50% Total 222

Table 2. Initial public offering market share around all-star analyst turnover This table reports descriptive statistics for the change in market share around all-star analyst turnover. The sample consists of 222 cases where an Institutional Investor All-Star left one investment bank for another. Market share is defined as the sum of gross proceeds (not including the overallotment option) for the investment bank in a given period divided by the sum of gross proceeds on all IPOs over the same period. Industry market share is similarly defined, where only issues in the all-star analyst’s Fama-French industry are considered. We calculate aggregate and industry-level market share for one-year prior and one-year after the departure of the all-star analyst. Number of IPOs in industry is defined as the number of issues in the same Fama-French industry over the year prior to the analyst’s departure .

bank moving from bank moving to

mean t-stat %

positive num obs mean t-stat

% positive

num obs

full sample period 1988 to 1999 Market share 1 year prior to move 3.035 222 3.021 222 Change in market share (post move - pre move) 0.369 0.74 52.7 222 1.246 2.36 57.2 222 Industry market share 1 year prior to move 3.555 209 3.417 208 Change in industry market share (post move - pre move) 0.045 0.04 76.0 192 1.812 1.58 80.6 196 Number of IPOs in industry 1 year prior to move 22.932 222 22.856 222 % change in number of industry IPOs (post move - pre move) 39.870 4.49 53.6 192 44.569 4.40 56.6 196 Post 1994 Market share 1 year prior to move 3.056 128 3.540 136 Change in market share (post move - pre move) 1.368 1.87 53.9 128 1.420 1.89 55.9 136 Industry market share 1 year prior to move 4.865 125 1.908 131 Change in industry market share (post move - pre move) -0.274 -0.16 73.9 111 2.738 2.13 79.3 121 Number of IPOs in industry 1 year prior to move 31.375 128 30.103 136 % change in number of industry IPOs (post move – pre move) 33.933 3.08 51.4 111 36.762 2.85 53.7 121

Table 3. Market share changes for bank losing all-star conditioned on various variables This table shows changes in market share for the bank losing the all-star analyst stratified by various variables. The Carter-Manaster rank is the Carter-Manaster (1990) ranking on a 0-9 scale for the book manger of the IPO. If an underwriter always appears in the highest bracket of the underwriting section of the prospectus, it is assigned the top ranking of 9 on a 0-9 scale. Market share is defined as the sum of gross proceeds (not including the overallotment option) for the investment bank in a given period divided by the sum of gross proceeds on all IPOs over the same period. Industry market share is similarly defined, where only issues in the all-star analyst’s Fama -French industry are considered.

Sample

mean market share 1

year prior to move

mean change in market share (post move - pre

move)

t-statistic (change =

0) num obs

mean industry market

share 1 year prior to

move

mean change in industry

market share (post move - pre move)

t-statistic (change =

0) num obs

Star is not replaced at bank 2.783 1.003 1.71 172 3.793 -1.039 -0.96 143 Star is replaced at bank 3.902 -1.811 -2.14 50 2.798 3.207 1.29 49 Bank had more than 10 stars in year prior to move 4.091 0.518 0.70 136 5.175 0.117 0.07 116 Bank had 10 or fewer stars in year prior to move 1.365 0.134 0.24 86 1.047 -0.066 -0.09 76 Star was ranked First, Second, or Third by Institutional Investor 2.457 1.023 1.65 122 2.894 1.403 1.01 102 Star was ranked "Runner up" by Institutional Investor 3.741 -0.429 -0.53 100 4.318 -1.494 -0.99 90 Absolute difference in Carter-Manaster Ranks for banks >=1 2.251 0.482 0.74 110 3.621 -0.725 -0.69 94 Absolute difference in Carter-Manaster Ranks for banks <1 3.805 0.259 -0.77 112 3.489 0.783 0.45 98

Table 4 Market share changes for bank gaining all-star conditioned on various variables This table shows changes in market share for the bank gaining the all-star analyst stratified by various variables. The Carter-Manaster rank is the Carter-Manaster (1990) ranking on a 0-9 scale for the book manger of the IPO. If an underwriter always appears in the highest bracket of the underwriting section of the prospectus, it is assigned the top ranking of 9 on a 0-9 scale. Market share is defined as the sum of gross proceeds (not including the overallotment option) for the investment bank in a given period divided by the sum of gross proceeds on all IPOs over the same period. Industry market share is similarly defined, where only issues in the all-star analyst’s Fama -French industry are considered.

Sample

mean market

share 1 year prior to

move

mean change in market share (post move - pre

move)

t-statistic (change =

0) num obs

mean industry market

share 1 year prior to

move

mean change in industry

market share (post move - pre move)

t-statistic (change = 0)

num obs

Star is not replaced at original bank 3.202 1.357 2.17 172 3.479 2.634 1.76 146 Star is replaced at original bank 2.400 0.863 0.94 50 3.224 -0.589 -0.57 50 New bank had more than 10 stars in year prior to move 3.874 1.637 2.3 161 4.447 1.548 1.03 141 New bank had 10 or fewer stars in year prior to move 0.772 0.213 0.59 61 0.754 2.488 1.77 55 Star was ranked First, Second, or Third by Institutional Investor 3.309 1.603 2.22 122 2.808 4.326 2.31 106 Star was ranked "Runner up" by Institutional Investor 2.671 0.811 1.05 100 4.128 -1.150 -1.02 90 Absolute difference in Carter-Manaster Ranks for banks >=1 2.373 1.148 1.46 110 2.451 2.050 1.33 98 Absolute difference in Carter-Manaster Ranks for banks <1 3.658 1.342 1.78 112 4.422 1.573 0.92 98

Table 5. Changes in IPO performance variables around all-star analyst turnover This table shows changes in underpricing, spread, price adjustment, and performance around all-star analyst turnover for both the bank moved to and the bank moved from. Underpricing is defined as: 100*(P1st day close – Poffer)/Poffer, where P1st day close is the closing price at the end of the first-day of trading and Poffer is the offering price from SDC. Price adjustment is the IPO offer price divided by the average of the high and low initial filing price. In order to measure one-year abnormal performance, we use the CRSP NYSE/AMEX value-weighted index, with dividends, for IPOs that initially list on the New York or American Stock Exchanges. We use the Nasdaq composite index for all other IPOs. Returns are calculated to the end of the one-year IPO anniversary or until the issuing firm stops trading. Spread is defined as 100*[SP/P Offer], where SP is the gross spread per share in the offering. Abnormal measures are the residuals from regressions of these variables on various deal characteristics.

Bank losing star Bank gaining star Mean t-stat Mean t-stat mean IPO initial return 1 year prior to move 15.438 11.57 20.914 10.15 mean change in IPO initial return (post move - pre move) 7.233 3.38 8.249 4.20 mean abnormal IPO initial return 1 year prior to move 0.856 1.86 0.473 0.54 mean change in abnormal IPO initial return (post move - pre move) 0.330 0.34 -0.064 -0.05 mean IPO spread 1 year prior to move 6.798 396.90 6.759 273.58 mean change in IPO spread (post move - pre move) -0.024 -1.19 -0.019 -0.83 mean abnormal IPO spread 1 year prior to move 0.100 6.90 0.074 4.87 mean change in abnormal IPO spread (post move - pre move) -0.034 -2.00 -0.018 -1.13 mean 1-year abnormal return for IPOs 1 year prior to move 0.138 5.28 0.141 5.12 mean change in 1-year abnormal return (post - pre) -0.063 -1.23 -0.111 -2.87 standard deviation of 1-year abnormal return for IPOs 1 year prior to move 0.647 17.09 0.715 18.40 mean change in standard deviation of 1-year abnormal return (post - pre) -0.030 -0.66 -0.009 -0.18 mean IPO % price adjustment 1 year prior to move 0.152 35.72 0.158 32.66 mean change in IPO % price adjustment (post move - pre move) 0.014 2.08 0.032 5.06 mean abnormal IPO % price adjustment 1 year prior to move 0.005 1.66 -0.002 -0.60 mean change in abnormal IPO % price adjustment (post move - pre move) 0.000 0.02 0.015 2.82 standard deviation of abnormal IPO % price adjustment 1 year prior to move 0.690 21.50 0.756 21.18 mean change in standard deviation of abnormal IPO % price adjustment (post move - pre move) 0.002 0.05 -0.021 -0.45 Number of observations 190 190

Table 6 Changes in analyst performance variables around all-star turnover This table shows changes in analyst performance variables around all-star analyst turnover for both the investment bank gaining the all-star and the investment bank losing the all-star. All analyst performance variables are calculated using data from the I/B/E/S/ files. Number of recommendations per year is the number of research reports issued for a company in the year following the issue date. Days until first recommendation is the number of calendar days from the issue date until the lead investment bank issues its first research report. Mean forecast error is defined as the difference between the EPS estimate of the IPO lead bank and the consensus EPS estimate scaled by stock price.

bank losing star bank gaining star

mean t-stat Mean t-stat

proportion of IPOs in year prior to move where IPO lead bank analyst makes first forecast 0.470 22.40 0.485 24.42

change in proportion of IPOs where IPO lead bank analyst makes first forecast (year post move - year pre move) 0.028 1.30 -0.002 -0.12

mean number of recommendations per IPO in year prior to move by IPO lead bank analyst 16.849 33.10 17.273 33.78

change in mean number of recommendations per IPO by IPO lead bank analyst (year post move - year pre move) 0.692 1.43 1.100 2.88

mean days until first recommendation by IPO lead bank analyst in year prior to move 72.444 20.12 65.350 20.30

change in mean days until first recommendation by IPO lead bank analyst (year post move - year pre move) -11.806 -4.76 -8.591 -3.32

mean forecast error for IPO lead bank analyst in year before move -0.277 -4.57 -0.205 -5.93

change in mean forecast error for IPO lead bank analyst (year post move - year pre move) 0.041 0.54 0.033 0.60

standard deviation of forecast error for IPO lead bank analyst in year before move 1.118 10.66 0.868 10.52

change in standard deviation of forecast error for IPO lead bank analyst (year post move - year pre move) -0.060 -0.45 0.383 3.32

Number of observations 153 168

Table 7 Regression analysis of change in market share for bank gaining the all-star analyst This table shows the results from simple regressions of the change in market share on various control variables for the bank gaining the all-star. In order to measure one-year abnormal performance, we use the CRSP NYSE/AMEX value-weighted index, with dividends, for IPOs that initially list on the New York or American Stock Exchanges. We use the Nasdaq composite index for all other IPOs. Returns are calculated to the end of the one-year IPO anniversary or until the issuing firm stops trading. Price adjustment is the IPO offer price divided by the average of the high and low initial filing price. Mean forecast error is defined as the difference between the EPS estimate of the IPO lead bank and the consensus EPS estimate scaled by stock price. T-statistics are reported in parentheses.

Intercept

Mean pre-move IPO

1-yr abnormal

return

Change in mean IPO 1-yr abnormal return (post -

pre)

Std. deviation of pre-move

IPO 1-yr abnormal return

Change in std. deviation of

IPO 1-yr abnormal return

(post -pre)

Std. deviation of pre-move abnormal

price adjustment

Change in std. deviation of abnormal

price adjustment (post -pre)

Proportion of IPOs taken public by

bank over year pre move where

bank’s analyst made first forecast

change in proportion of IPOs taken public by bank where bank’s analyst made first

forecast (year post move - year pre move)

mean forecast error for IPO

lead bank analyst in year before move

change in mean forecast error for

IPO lead bank analyst (year post move - year pre

move) Num obs R2

(1) 0.734 4.454 0.124 184 0.032

(1.05) (1.90) (0.08)

(2) -1.805 4.416 0.851 184 0.056

(-1.41) (2.86) (0.67)

(3) -1.815 4.272 0.995 184 0.039

(-1.21) (2.36) (0.73)

(4) -1.251 5.286 6.178 168 0.023

(-0.59) (1.40) (1.95)

(5) 2.386 4.501 2.404 168 0.026

(2.97) (2.01) (1.76)

(6) -0.240 4.211 -0.002 3.603 5.921 3.961 1.875 166 0.075

(-0.10) (1.59) (-0.00) (0.91) (1.83) (1.77) (1.32)

(7) -3.266 4.460 0.687 4.201 6.401 3.963 1.209 166 0.099

(-1.31) (2.62) (0.49) (1.09) (2.01) (1.79) (0.83)

(8) -3.019 4.146 0.730 3.959 6.429 3.852 1.221 166 0.081

(-1.15) (2.02) (0.48) (1.01) (1.99) (1.72) (0.83)

Table 8 Regression analysis of change in market share for bank losing the all-star analyst This table shows the results from simple regressions of the change in market share on various control variables for the bank losing the all-star. In order to measure one-year abnormal performance, we use the CRSP NYSE/AMEX value-weighted index, with dividends, for IPOs that initially list on the New York or American Stock Exchanges. We use the Nasdaq composite index for all other IPOs. Returns are calculated to the end of the one-year IPO anniversary or until the issuing firm stops trading. Price adjustment is the IPO offer price divided by the average of the high and low init ial filing price. Mean forecast error is defined as the difference between the EPS estimate of the IPO lead bank and the consensus EPS estimate scaled by stock price

Intercept

Dummy = 1 if star is replaced

with another star after the move

Mean abnormal

under-pricing for

IPOs in year prior to move

Change in mean

abnormal under-pricing (post –

pre)

Std. deviation of pre-

move IPO 1-yr

abnormal return

Change in std.

deviation of IPO 1-

yr abnormal

return (post -pre)

Std. deviation of pre-move

abnormal price

adjust-ment

Change in std.

deviation of

abnormal price

adjust-ment

(post -pre)

mean number of

forecasts per IPO in year

prior to move by IPO

lead bank analyst

change in mean

number of forecasts per IPO by lead bank analyst

(year post move - year pre move)

mean days to first forecast per IPO in

year prior to move by IPO

lead bank analyst

change in mean days to first

forecast per IPO by IPO lead bank analyst

(year post move - year pre

move)

standard deviation of

forecast error for IPO lead bank analyst

in year before move

change in standard

deviation of forecast error for IPO lead bank analyst (year post

move - year pre move)

Num obs R2

(1) 1.003 -2.813 222 0.025

(1.79) (-2.38)

(2) 0.314 0.291 0.128 162 0.063

(0.47) (2.20) (1.90)

(3) -3.511 6.062 4.148 162 0.077

(-2.75) (3.63) (2.62)

(4) -4.212 6.638 3.772 162 0.073

(-1.47) (1.87) (2.47)

(5) -4.029 0.283 0.061 153 0.039

(-1.90) (2.41) (0.51)

(6) 2.997 -0.033 -0.017 153 0.022

(2.16) (1.76) (-0.60)

(7) -0.450 1.157 1.075 153 0.019

(-0.41) (1.43) (1.68)

(8) -5.397 -2.117 0.212 0.071 4.205 4.174 0.169 -0.041 0.011 -0.001 0.259 0.956 153 0.131

(-1.22) (-1.32) (1.49) (0.96) (1.81) (2.23) (0.94) (-0.28) (0.42) (-0.02) (0.29) (1.40)

(9) -5.938 -2.317 0.213 0.060 4.092 3.742 0.120 0.007 0.012 0.004 0.280 1.004 153 0.127

(-1.31) (-1.45) (1.50) (0.81) (1.61) (2.05) (1.05) (0.05) (0.49) (0.13) (0.31) (1.43)

1

Appendix A

In order to measure abnormal first day returns, abnormal price adjustments and abnormal

underwriter spreads we estimate different regression models which are then used to predict “normal”

outcomes for these variables. For first day returns, we regress first day returns against a number of

different variables motivated by the exiting IPO literature. We carry out separate regressions each year

to allow for time series variation in the impact of each variable on initial returns. The abnormal initial

return for a given IPO is then defined as the actual return subtract predicted initial return using the

estimated model from the same year as the IPO. The process is similar for price adjustments and

spreads. In Table A1 we summarize the regression analysis. For each measure (initial returns, price

adjustments and spreads), we report the average coefficient estimate in the 16 annual regressions (from

1985 to 2000) for each independent variable. The t-statistic reported in the table is based on the

standard deviation of the time series of 16 estimates measures for each coefficient estimate (this

approach is similar to the bootstrap approach used by Fama and MacBeth, 1973; see Lowry and

Schwert, 2002, for details on this approach). The R2’s reported are the average of the 16 annual

regressions.

Independent variables in the initial return regressions are motivated based on the existing

literature. We include two independent variables capture the relationship between information

revelation and first day return. Price Adjustment is defined as the final offering price minus the average

of the high and low initial filing prices all divided by the average of the high and low initial filing

prices. Price Adjustment + takes the same value as Price Adjustment when Price Adjustment is

positive, and 0 otherwise. This specification allows for an asymmetric relationship between price

adjustments and initial returns and has been used in a number of studies (Lowry and Schwert, 2002,

Bradley and Jordan, 2002, Ljungqvist and Wilhelm, 2002a and Ljungqvist and Wilhelm, 2002b).

Following Bradley and Jordan (2002), we include Overhang as an independent variable in our

initial return regressions measured as the number of shares outstanding after the IPO net of the number

2

of shares offered in the IPO all divided by the number of shares offered in the IPO. The number of

shares in the IPO is obtained from SDC and includes global tranches but exclude the overallotment

option. Overhang attempts to reflect the fraction of shares that cannot freely trade immediately after the

IPO. We include the variable Venture Capital Backing, which takes the value 1 if the IPO is venture

capital backed (as indicated by SDC) and 0 otherwise, as an independent variable in our analysis as a

measure of external certification (see Barry, Muscarella and Vetsuypens, 1990, and Megginson and

Weiss, 1991). We include a number of proxies for the ex ante uncertainty regarding the value of the

firm going public including High Tech, a dummy variable taking the value 1 if the issuer is classified as

a high technology firm by SDC (see Bradley and Jordan, 2002, Lowry and Schwert, 2002, and

Ljungqvist and Wilhelm, 2002a); NYSE, a dummy variable taking the value 1 if the issuer lists on the

New York Stock exchange research (Lowry and Schwert, 2002); AMEX, a dummy variable taking the

value 1 if the issuer lists on the American Stock Exchange (Lowry and Schwert, 2002); Firm Standard

Deviation, defined as the standard deviation of stock returns from days +21 to +50 relative to the IPO

(see Johnson and Miller, 1988, Carter Dark and Singh, 1998, and Lowry and Schwert, 2002).

We include three measures of investment bank reputation and certification in our analysis.

Carter and Manaster Ranking is obtained from Carter and Manaster (1990) as updated by Carter, Dark

and Singh (1998) and more recently by Loughran and Ritter (2002b). These rankings are on a 0 to 9

scale, with 9 being the most reputable. Market Share is measured for each bank. For each IPO we

examine all IPOs in the year leading up to the offer (including the IPO). We compute the sum of gross

proceeds (on global shares excluding over allotments) for which the underwriter was also the book

manager. To account for mergers in the investment banking industry, we gather data from SDC on all

combinations during the period. If the book manager recently emerged from a merger, the gross

proceeds of all offering by any precedent bank are added together. For example , offerings by Salomon

Smith Barney, all IPOs by Salomon Bros. and Smith Barney in the prior year are included in the

calculation of Salomon Smith Barney’s market share. In cases with multiple book managers, equal

credit is given to each bank. Market Share is then defined as the sum of gross proceeds for the bank,

3

divided by the sum of gross proceeds for all IPOs over the sample period. Industry Market Share is

computed as the sum of gross proceeds on all IPOs in the same Fama-French 48 industry group taken

public by the firm over the year prior to the IPO, divided by the sum of gross proceeds on all IPOs

issued in that industry over the same period (see Carter and Manaster, 1990, and Bates and Dunbar,

2002).

We also control for market return prior to the IPO. We include Market Return, defined as the

compound return from day –50 to –2 relative to the IPO on the CRSP value weighted index, as an

independent variable in our analysis. To allow for non-linearities in the relation between market

returns and initial IPO returns, we also include Market Return + as an independent variable where

Market Return + takes the same value as Market Return whenever it is positive, and 0 otherwise (see

Loughran and Ritter, 2002b, and Lowry and Schwert, 2002). As an additional measure of pre-IPO

market activity we include Lagged Average Underpricing, the mean first-day returns on all IPOs on

days –60 to –1 relative to the offering, as an independent variable (see Loughran and Ritter, 2002b,

Lowry and Schwert, 2002, and Bradley and Jordan, 2002). We also include Lagged Industry Average

Underpricing as an independent variable in our analysis, defined as the average underpricing over the

year prior to the offering of IPOs in the same industry (see Benveniste, Ljungqvist, Wilhelm and Yu,

2002, and DuCharme, Rajgopal and Sefcik, 2001). Industry is defined using Fama and French’s, 1997,

48 industry group classification scheme, which is based on primary SIC codes obtained from SDC.

Finally we include two measures of IPO intensity. Number of Prior IPOs is the number of IPOs from

days –60 to –1 relative to the offering and Number of Prior Industry IPOs is the number of IPOs over

the year prior to the offering in the same Fama-French 48 industry group (see Booth and Chua, 1996,

and Benveniste, Ljungqvist, Wilhelm and Yu, 2002).

Independent variables for the price adjustment regressions are similar to those for the initial

returns regressions with a few exceptions (see Lowry and Schwert, 2002, Ljungqvist and Wilhelm,

2002). We exclude price adjustment variables for obvious reasons. We also exclude measures of past

4

underpricing and IPO intensity measured at the industry level as these variables are never significant in

year-by-year regressions.

Independent variables in the underwriter spread regressions include many of the variables from

the IPO initial return regressions based on existing studies of IPO spreads (e.g. Hansen and Torregrosa,

1992, James, 1992 and Dunbar, 1998). Specifically we include overhang as a measure of insider

retention, Venture Capital Backing, Carter-Manaster Ranking, Market Share and Industry Market share

as measure of third party certification, and High Tech, NYSE, AMEX and Firm Standard Deviation as

measures of issuer riskiness. We also inc lude IPO proceeds (offering price multiplied by shares

offered, in millions) and the logarithm of proceeds in our regressions. Hansen and Torregrosa (1992)

find that spreads are significantly related to offering size but are U-shaped (i.e. they experience

diseconomies of scale at some point).

5

Table A1 Regressions of Percentage First Day Returns, Price Adjustments and Underwriter Spreads on Various Independent Variables The dependent variables are defined as follows. IPO initial return defined as 100*(P1-P0)/P0 where P1 is the first-day closing stock price or bid-ask average (from CRSP) and P0 is the IPO offer price. Price Revision is the IPO offer price divided by the average of the high and low initial filing price (stated as a percentage). IPO percentage spread is the gross underwriter spread per share dividend by the offering price per share (states as a percentage). Independent variables are defined as follows. IPO proceeds is the offering price multiplied by the number of shares sold (including global shares, excluding the overallotment option) in millions of dollars. Logarithm of IPO proceeds is the natural logarithm of IPO proceeds. Price revision is as defined above. Price Revision + is the IPO offer price divided by the average of the high and low initial filing price if positive and zero otherwise. Overhang is (S1-S)/S where S1 is the shares outstanding after the IPO and S is the shares offered in the IPO. Venture Capital Backing equals 1 if the issue is venture capital-backed and 0 otherwise. High Tech equals 1 if the issue is classified as a high-tech by SDC and 0 otherwise. NYSE equals 1 if the IPO lists on the NYSE and 0 otherwise. AMEX equals 1 if the IPO lists on the American Stock Exchange and 0 otherwise. Firm Std. Deviation equals the standard deviation of daily stock returns for the issuing firm from days 21 to 50 relative to the IPO. Market Share is the sum of gross proceeds (not including the overallotment option) over the year prior to the IPO of all offerings where the IPO book manager is the book manager (equal credit given if there is more than one manager) divided by the sum of gross proceeds on all IPOs over the same period. Industry Market Share is similarly defined where only issues in the IPO firm’s Fama -French industry are considered. Carter-Manaster Rank is the Carter-Manaster (1990) ranking on a 0-9 scale for the book manager of the IPO (the maximum rank if there is more than one book manager). Market Return equals the buy and hold CRSP value-weighted index return from days -50 to -2 relative to the IPO. Market Return + equals the buy and hold CRSP value-weighted index return from days -50 to -2 relative to the IPO if positive, 0 otherwise. Market Std. Deviation is the standard deviation of daily returns for the CRSP value-weighted index from days -50 to -2 relative to the IPO. Lagged Avg. Underpricing is the average initial return for issues on days -60 to -1 relative to the IPO. Number of Prior IPOs is the number of issues from days -60 to -1 relative to the IPO. Lagged Avg. Industry Underpricing is the average initial return for issues from the same Fama -French Industry over the year prior to the IPO. Number of Prior Industry IPOs is the number of issues in the same Fama -French industry over the year prior to the IPO. Estimates reported are an average of year-by-year regression coefficients and t-statistics are based on the standard deviation of the times-series of coefficient estimates. R2 is the average value of the R2 from year-by-year regressions.

6

Table A1 Regressions of Percentage First Day Returns, Price Adjustments and Underwriter Spreads on Various Independent Variables (continued)

Dependent variable is the

IPO initial return Dependent Variable is Price

Revision Dependent variable is the

IPO percentage spread Intercept 3.854 0.26 6.037 1.01 10.361 53.38 IPO proceeds 0.000 -1.19 Logarithm of IPO proceeds -0.529 -12.46 Price Revision 31.866 10.00 Price Revision + 33.319 4.07 Overhang 1.273 3.21 0.928 3.77 -0.031 -4.61 Venture Capital Backing 1.420 0.86 -0.304 -0.51 -0.048 -1.62 High Tech 0.432 0.59 2.302 2.17 0.025 0.70 NYSE -3.924 -2.48 -2.007 -1.76 0.116 2.21 AMEX -4.896 -2.46 -6.603 -2.61 0.028 0.40 Firm Std. Deviation 1.585 4.39 0.020 0.06 0.028 3.18 Industry Market Share -0.015 -0.54 0.036 2.23 0.002 2.78 Market Share 0.257 2.03 0.420 3.69 0.021 3.91 Carter-Manaster Rank -0.855 -1.62 -0.486 -2.51 -0.206 -8.65 Market Return 1.741 1.02 -4.650 -0.92 Market Return + -0.945 -0.55 5.521 1.08 Market Std. Deviation 9.246 0.99 -8.881 -1.77 Lagged Avg. Underpricing -0.340 -2.30 -0.141 -2.27 Number of Prior IPOs -0.162 -2.96 0.072 2.00 Lagged Avg Industry Underpricing 0.127 3.54 Number of Prior Industry IPOs 0.026 0.72 R2 0.339 0.185 0.712