.k 5
(.3
iiia. A...
(.1 P

2.: t :1.
.::u.. ‘1

3:
I .5. 2.3.5. »‘
1.5:.1‘
s... .l.
.47 a 2' I:

E... 3...!
m , . 3
.

:A . .
.951...

.nﬂzi. kn MA. 743?.

$11 1
a? 3!

7.353.”
iii»? in

.4...

1.1:!

.1450... 1
. Inn:
7

.

.1... ..
3 :J.
=31»;v

m."

.
\i‘ \10..7V.dﬂuﬂ.ﬂW
.. $¥§$ﬂ 3.: .3:
im§%
haw... ﬁt}... -.
bu...

3
on
.me
oouoxuiu
1. :
s... a
a:
v

.t. a“... 135.9 Err. .
uuw.tnrno,Wr¢13nﬂ+nrl.. hum“

, Bu,

1‘1: 1
”ha 4.3.1:
3.1.. Hula:

 

 

 

LIBRARY
Michigan State
University

This is to certify that the
dissertation entitled

Three Essays on Initial Public Offerings and Market
Information

presented by

William C. Johnson

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Finance

 

 

Maﬁor Pfofessor's Signature

?cwim. PD”?

 

two? S”; 206%

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

- , -.—.----.— -0— a-

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DateDue.indd-p.1

THREE ESSAYS ON INITIAL PUBLIC OFFERINGS AND MARKET
INFORMATION

By

William C. Johnson

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Finance

2006

ABSTRACT

THREE ESSAYS ON INITIAL PUBLIC OFFERINGS AND MARKET
INFORMATION

By

William C. Johnson

This dissertation examines market information and how different types of information
impact initial public offerings. The ﬁrst essay examines a group of IPO underwriters that
also manage institutional funds from 1993 through 1998. We ﬁnd that there is a
statistical difference between the average fund holdings (1.24% of shares outstanding) of
IPOs for institutions that underwrite a particular IPO and the fund holdings (0.92%) of
institutions that do not underwrite the same IPO. We provide evidence that underwriters
use their institutional funds as vehicles to help them earn more equity underwriting
business. We also show that IPO underwriters use their superior information to earn
annualized excess returns 7.6% above non-underwriters, beneﬁting their institutional
investors. In the second essay we study the discretionary disclosure of [PO ﬁrm earnings
in unofﬁcial press releases during the ﬁrst 25 days of trading, the quiet period. This
information release is unique in that the market has no priors concerning management
earnings release so management has full discretion over the release of this information.
Not surprisingly, the probability of releasing earnings during the quiet period is positively
related to the abnormal operating performance of the ﬁrm. The probability of earnings

release is also related to the volatility in the overall market implying that managers are

timing their earnings releases. We ﬁnd that firms are able to attract greater analyst
coverage, particularly from unafﬁliated analysts when they release earnings earlier ie.,
during the quiet period. In the third essay We examine the relationship between ﬁrm
news and volatility over time. Contrary to the negative relationship between
idiosyncratic news and ﬁrm volatility predicted by current theories, we document a
positive relationship between ﬁrm level news and ﬁrm level volatility. We conﬁrm, as
conjectured in prior literature, an increase in idiosyncratic IPO volatility over the period
from 1973 through 2003. We ﬁnd the increase in volatility is over twice as large for
IPOs as for ﬁrms matched to the IPOs based on size and book-to-market ratio. We also
develop a comprehensive database consisting of 129,737 observations of news
disclosures by recent IPO and matching ﬁrms, and we use the database to document the
distribution of news day returns. Keeping the media sources of news constant, we ﬁnd
that the number of news citations per ﬁrm has risen sharply over the last three decades,
and the stock price reaction to news has become signiﬁcantly larger. Our ﬁnal
contribution presents evidence that the increased idiosyncratic volatility for both [PCs
and matching ﬁrms is signiﬁcantly related to an increase in news in recent decades. In
fact, an increase of just one extra news day per month adds 0.6% to the monthly standard

deviation for IPOs, or 2.1% per year.

Cepyright by
William C. Johnson

2006

To my loving and supportive wife.

ACKNOWLEDGEMENTS

I would like to express my appreciation to the members of my dissertation
committee, Dr. Geoffrey Booth, Dr. Marilyn Johnson, and Dr. Jennifer Marietta-
Westberg, for their helpﬁrl comments and guidance. I would particularly like to express
my thanks to Dr. Jennifer Marietta-Westberg for all her help and assistance throughout
my time at Michigan State University.

I would also like to thank Dr. Charles Hadlock for his guidance in the process of
going from a graduate student to a professor. I thank the other professors of the ﬁnance
department who have been instrumental in my development as an academic.

I thank all the graduate students who I have learned ﬁ'om at Michigan State
University: William Gerken, Umit Gurun, Iordanis Karagiannidis, Hyun-Seung Na,
Joshua Pierce, Gwinyai Utete, and Sangho Yi. The path to getting my PhD. has been
much richer from our discussions and disagreements.

Finally, my deepest gratitude goes to my sons Samuel and William and my wife
Ginger for their understanding during all the late nights spent working. Without their

sacriﬁces, this would never have been possible.

TABLE OF CONTENTS

 

 

LIST OF TABLES ................................................................................... ix
LIST OF FIGURES ................................................................................. xi
ESSAY l. UNIVERSAL BANKING ASSET MANAGEMENT AND STOCK
UNDERWRITING
1.1 Introduction ................................................................................... l
1.2 Motivation for Institutional Purchases of IPOs .......................................... 6
1.2.1 The Quid Pro Quo Hypothesis .................................................... 7
1.2.2 The Superior Information Hypothesis ........................................... 8
1.3 Data and Institutional IPO Holdings ..................................................... 10
1.4 IPO Holdings and Performance .......................................................... 14
1.4.1 Testing the Quid Pro Quo Hypothesis .......................................... 14
1.4.2 Testing the Superior Information Hypothesis ................................. 21
1.4.3 Robustness Tests .................................................................. 25
1.5 Conclusions ................................................................................. 26
APPENDD( 1.TABLES OF ESSAY 1 .......................................................... 28
ESSAY 2. VOLUNTARY EARNINGS DISCLOSURE DURING THE QUIET
PERIOD
2.1 Introduction ................................................................................. 41
2.2 Background ................................................................................. 45
2.3 Hypothesis Development ................................................................ 48
2.4 Data and Preliminary Analysis ......................................................... 50
2.4.1 Data ................................................................................ 51
2.4.2 Sample Statistics ................................................................. 51
2.4.3 Quiet Period Returns ............................................................ 52
2.5 Tests of Hypotheses ...................................................................... 53
2.5.1 Probability of Quiet Period Earnings Release ............................... 53
2.5.2 Determinants of the Stock Price Response to Earnings
Announcements .................................................................. 58
2.5.3 Analyst Coverage and Quiet Period Earnings Release ..................... 61

vii

2.6 Conclusions ................................................................................ 67

APPENDIX 2.TABLES OF ESSAY 2 ......................................................... 7O

ESSAY 3. THE EFFECT OF NEWS ON VOLATILITY: A STUDY OF IPOS

3. 1 Introduction ................................................................................ 87
3.2 IPO and Matching Firm Sample ........................................................ 93
3 .3 IPO Volatility .............................................................................. 94
3 .4 News Flow and Volatility ............................................................... 96
3.4.1 News Measurement ............................................................. 96
3.4.2 Estimating the Relationship between News and Volatility ............... 102
3.4.3 Granger Causality .............................................................. 108
3 .5 Identifying Exogenous Shocks to News ............................................. 110
3 .6 Robustness Tests ........................................................................ 115
3.6.1 Sample Evaluation ............................................................. 115
3.6.2 Restricting Sources to the New York Times and
Dow Jones Publications ....................................................... 117
3.7 Conclusions .............................................................................. 117
APPENDD( 3.TABLES OF ESSAY 3 ........................................................ 121
APPENDIX ADATA CONSTRUCTION FOR ESSAY 3 ................................. 134
APPENDDI B. FIGURES ...................................................................... 136
BIBLIOGRAPHY ................................................................................. 142

viii

Table 1.1

Table 1.2
Table 1.3

.Table 1.4

Table 1.5
Table 1.6
Table 1.7

Table 1.8

Table 1.9

Table 1.10

Table 2.1
Table 2.2
Table 2.3

Table 2.4

Table 2.5

Table 2.6

Table 2.7

LIST OF TABLES

Quarterly Holdings, Analyst Recommendations, and Secondary

Stock Offerings for Ofﬁce Marg IPO date: November 2, 1994. . . 2.9
Summary Statistics ............................................................. 31
Percentage of IPO Shares Held ............................................... 32
Contemporaneous and Prior Quarterly Excess Stock Returns

for IPOs Bought by Institutions .............................................. 33
Stock Buys Pre- and Post-Secondary Stock Offerings .................... 34
Probit Regressions for Determinants of Institutional Buys ............... 35

Probit Regressions for Determinants of SEO Underwriter Identity... ..36

Multivariate Regression for Determinants of Future Quarterly
Excess Stock Returns ......................................................... 38

Future Abnormal Stock Returns for IPOs Bought by Institutions ...... 40

Future Quarterly Excess Stock Returns for [PCs Bought by

Institutions (Negative Prior Returns Only) ................................. 41
IPOs by their Release of Earnings during the Quiet Period ............. 71
IPOs Sample Characteristics ................................................ 73
Market Adjusted Mean Returns Around Earnings Releases ............ 75
Logistic Regression Results with Quiet Period Earnings Release

as the Dependent Variable ................................................... 76
Multivariate Regression of Earnings Announcement Day Returns

for IPO ﬁrms Releasing Earnings during the Quiet Period .............. 79
Summary Statistics for Analyst Coverage ................................. 82

Logistic Regression Results with Analyst Coverage as the
Dependent Variable ........................................................... 84

ix

Table 2.8

Table 3.1
Table 3.2

Table 3.3

Table 3.4

Table 3.5

Table 3.6

Table 3.7

Table 3.8

Table 3.9

Multinonrial Logistic Regression Results with Analyst Coverage
as the Dependent Variable .................................................... 86

Mean Monthly Standard Deviation in the First Five Trading Years... 122
Factiva News Sources ......................................................... 123
Newspaper and Newswire Citations in the First Five Trading Years. 124

Excess Returns and Standard Deviation of Returns on News

and Non-News Days ......................................................... 125

Summary Statistics for IPOs and Matching Firms ...................... 127
Regressions of Monthly Standard Deviation on the Monthly

Number of News Hits and Control Variables ............................. 129
Granger Causality ............................................................. 131
Changes in Monthly News Hits and Volatility,

Pre- and Post-Exogenous News Shocks ................................... 132
Articles Related to the Factiva Search Firms ............................. 133

Figure 1.1

Figure 2.1

Figure 3.1

Figure 3.2

Figure 3.3

LIST OF FIGURES

Underwriter and Non-Underwriter IPO Holdings .. ............................... 137
Cumulative market-adjusted returns: Returns around the

Earnings Announcement made during the quiet period .................... 138
Monthly volatility over time for [PCs issued from 1973-1998 ........... 139

News Day return probability distribution across decades for
sampled IPO ﬁrms ............................................................... 140

Relationship between number of news days per month and
Monthly volatility ............................................................... 141

ESSAY 1. UNIVERSAL BANKING, ASSET MANAGEMENT, AND STOCK
UNDERWRITING

1.1 Introduction

In recent years, researchers have produced a great deal of work analyzing
conﬂicts of interest within universal banks and other ﬁnancial institutions. With the
increase in universal banking in the United States since the 1999 Gramm-Leach-Bliley
Act, the potential for conﬂicts of interest within a ﬁnancial institution has certainly
increased. However, little research has been conducted concerning the conﬂict of two
particular activities within the same institution: asset management and securities
underwriting. The current paper investigates the stock holdings of ﬁnancial institutions
with underwriting divisions. In particular, we are interested in determining if fund
investors suffer or beneﬁt from institutions with both asset management and underwriting
divisions.

There is anecdotal evidence in the press that asset management divisions of
investment banks may feel pressure to hold recent IPOs brought forward by the same
bank’s underwriting division. A March 12, 2003 article in the Wall Street Journal
describes such an event at Deutsche Bank. An underwriting executive at Deutsche
phoned the chief investment ofﬁcer at Deutsche’s asset management unit and asked him
to buy some of the struggling media company Vivendi Universal which Deutsche had
helped bring public. The chief investment ofﬁcer was told to “be a team player.”
Evidently a shouting match broke out between the two of them when the request was

refused.

The SEC has expressed concern about this issue as well. A copy of a speech by
SEC Director Stephen M. Cutler on September 9, 2003 says that “...an asset manager
might feel pressured to invest in companies that its investment banking afﬁliate had
underwritten. But certainly a ﬁrm’s advisory clients would be interested — not to
mention, troubled — to learn that their portfolios were viewed by some as a tool for
attracting investment banking business to the ﬁrm.”

The literature in the area of universal banking focuses on two main conﬂicts of
interest: bank lending/underwriting conﬂicts and sell side analysis/underwriting conﬂicts.
For instance, Puri (1996, 1999) shows that bank lenders/underwriters can be effective in
certifying underwritten securities, implying that the universal bank adds value to the
underwriting process. Michaely and Womack (1999) provide evidence that sell side
analysts working for the underwriter of an IPO give biased recommendations compared
to non-underwriter analysts. In the context of SEO issues, Dugar and Nathan (1995) and
Lin and McNichols (1998) ﬁnd that analysts employed by SEO underwriters release more
favorable earnings forecasts and stock recommendations than non-afﬁliated analysts.
Ljungqvist, Marston, and Wilhelm (2005) ﬁnd that analyst initiations by investment bank
afﬁliated analysts have no relationship with whether or not the investment bank will win
future underwriting jobs. Agrawal and Chen (2004) study stock recommendations and
forecasts made by analysts at brokerages, investment banks, and combination
brokerage/investment banks. Their results are consistent with brokerages providing more
frequent and less accurate forecasts, possibly to increase the number of transactions by

their stock clients.

Only Ber, Yafeh, and Yosha (2001) explicitly consider the dynamic relationship
between stock underwriting and fund management over time. However, their paper
focuses on the Israeli markets which could be quite different ﬁom the US. securities
markets. They ﬁnd that institutional portfolios show poor one-year post IPO stock
performance when their asset management division buys IPOs brought public by their
own underwriting unit. For a static look at ﬁmd holdings, Ritter and Zhang (2005) use
the ﬁrst quarter institutional holdings in IPOs as a proxy for the amount of shares
institutions self-allocate at the time the IPO goes public. Their focus is on what drives
institutions to allocate shares from the initial offering to their asset management division.

Our paper is unique in that it is the ﬁrst paper examining US. markets to take a
dynamic look at the potential beneﬁts and conﬂicts of interest inherent in an institution
with both underwriting and asset management divisions. In contrast to Ber, Yafeh, and
Yosha (2001), our paper shows that there are some beneﬁts to utilizing asset managers
who also have underwriting divisions. We ﬁnd evidence that underwriter/asset managers
utilize their underwriting relationship to beneﬁt institutional investors. However, we also
provide evidence that underwriters use their assets under management to gain future
underwriting business.

Our focal point is a group of IPO underwriters that also manage institutional
funds from 1993 through 1998. We consider only institutions that both underwrite stocks
and have asset management divisions. For each IPO, we refer to “underwriters” as those
investment bank/asset managers that underwrite the IPO and hold shares in a given
quarter. We refer to “non-underwriters” as those investment bank/asset managers that

did not underwrite the given IPO, but do hold shares in the given quarter. Underwriter

purchases of their own IPOs occur frequently. Over 50% of the IPOs in our sample are
purchased in underwriter institutional ﬁrnds. Also, 68 of the 141 investment banks we
analyze purchase stock in their own IPOs.

We test two hypotheses. The ﬁrst we refer to as the Quid Pro Quo Hypothesis.
This states that underwriters will utilize their institutional funds as a vehicle to earn more
underwriting business. The second hypothesis is the Superior‘Information Hypothesis.
Here we test whether underwriters use the information they discover during the IPO
underwriting process to earn superior ﬁrture returns in their managed ﬁmds.

We ﬁrst document that there is a statistical difference between the stock holdings
(1.2% of shares outstanding) of [PCs for institutions that underwrite a particular IPO and
the stock holdings (0.9%) of institutions that do not underwrite the same IPO, averaging
holdings across the ﬁrst eight quarters aﬁer the IPO. Consistent with Wermers (1999),
we ﬁnd that non—underwriter institutional holders are on average momentum traders in
that they purchase stocks after they have gone up by 3.4% in the previous quarter. In
contrast, underwriters are not momentum traders; underwriters purchase their own IPOs
after a statistically insigniﬁcant previous quarterly return of 0.9%. This difference in
trading behavior is consistent with the Quid Pro Quo Hypothesis.

For each of our IPO ﬁrms, we also document the trading patterns for underwriters
and non-underwriters in the quarters surrounding subsequent SEOs and in-house analyst
coverage initiations. Underwriters sell their holdings in quarters after they have initiated
analyst coverage or have underwritten the ﬁrm’s secondary offering. Non-underwriters

do not sell after these events. These results show that underwriters could be trading for

reasons other than to maximize institutional investor returns, also consistent with our ﬁrst
hypothesis.

We next document that the stock purchases of non-underwriters earn a future
stock return that is not statistically different from zero, but large purchases by
underwriters provide a statistically signiﬁcant excess return of 1.9% over the subsequent
quarter or 7.7% per year. This ﬁnding implies that underwriters utilize superior
information gleaned from the underwriting relationship with the ﬁrm to make decisions
about the stocks they purchase.

If underwriters learn superior information about the ﬁrm during the IPO, we do
not expect that the underwriter will be able to leverage that information under all
conditions. We ﬁnd that the superior future returns by underwriters are dependent on the
information environment for the IPO and the underwriter reputation rank. We classify
ﬁrms as to whether they have analyst coverage or not. When ﬁrms have no analyst
coverage, underwriters earn statistically signiﬁcant future quarterly returns of 1.8%. If
ﬁrms do have analyst coverage, underwriters are not able to earn superior future returns.
We also ﬁnd that high rank underwriters earn signiﬁcant quarterly returns of 2.3%, where
low rank underwriter returns are not signiﬁcantly different from zero.

We show that IPO underwriters are able to exploit their superior information;
SEO underwriters do not earn statistically signiﬁcant future returns. We conﬁrm the
positive ﬁrture abnormal retruns for underwriters in both cross-sectional regressions and
through calculating buy-and-hold abnormal returns based on a size and industry matching

ﬁrm technique.

The remainder of the paper is organized as follows. Section I discusses the
theoretical motivation for the paper and proposes hypotheses to test. Section 11 describes
our underwriter and institutional holdings sample. In section IH we examine IPO
holdings of institutional fund managers and test our motivating hypotheses. In the

paper’s last section, we offer implications of our ﬁndings and conclusions.

1.2 Motivation for Institutional Purchases of IPOs

We characterize the actions taken by the asset management division of an
investment banking ﬁrm as value decreasing or value increasing activities. Value
decreasing activities generally take the form of a conﬂict of interest between the asset
manager’s ﬁduciary responsibilities to their institutional investors and the incentive to
purchase stocks that the institution has underwritten. If an institutional asset manager has
sufﬁcient liquid assets, perhaps he can stabilize stock prices by making a series of large
stock purchases using the assets of the institutional fund. It is also possible that the
institutional trader may hold the stock as a quid pro quo for the ﬁrm that went public.
The underwriter/institutional trader might simply hold the stock to send a signal to the
IPO ﬁrm that the underwriter is attempting to market the security. In the extreme form,
value decreasing activities would include “stufﬁng,” where underwriters of poor quality
securities are disproportionately purchasing and holding these low quality stocks in their

own asset management division. This motivates our ﬁrst hypothesis.

1.2.] Hypothesis I Quid Pro Duo: Underwriter/asset managers will utilize their
institutional funds as a vehicle to attract more underwriting business from the

newly public ﬁrms.

The empirical implications of Hypothesis I are as follows. On average,
underwriters will purchase more of their own poorly performing IPOs than
non-underwriters will. We know from Wermers (1999) that on average, institutional
traders are momentum traders. Hypothesis I predicts that underwriters will trade less on
momentum than non-underwriters.

Hypothesis I also predicts certain changes in portfolio holdings around the time of
secondary stock offerings and analyst initiations. If investment bank asset management
divisions hold stocks that their ﬁrm has underwritten in an effort to attract or keep their
business as the SEO underwriter, we would expect to see the IPO underwriters hold large
blocks of stock before the SEO. At the same time, the underwriter is likely to liquidate
the stock position after the SEO once the ﬁrm has chosen the underwriter for its
secondary offering. In the Quid Pro Quo Hypothesis, we view in-house analyst
recommendations as possible substitutes for purchases by the asset management division.
Quid Pro Quo predicts that the underwriter might hold the stock until its in-house analyst
makes a recommendation, and then liquidate its position.1

Value increasing activities generally involve utilization of superior information

obtained as the underwriter for a ﬁrm. We know that institutional investors have the

 

' We examine quid pro quo activities within an investment bank, but Hoberg and Seyhun (2005) examine
another type of quid pro quo activity that involves investment banks. They ﬁnd evidence that lead IPO
underwriters and venture capitalists collaborate. Venture capitalists tolerate higher underpricing to beneﬁt
the IPO underwriters, and IPO underwriters provide positive price support prior to the expiration of lock-up
periods. Many venture capitalists sell their holdings as soon as the lock-up provisions expire.

7

ability to utilize superior information to earn abnormal returns [see Ben Dor (2003), Field
and Lowry (2004), and Krigman, Shaw, and Womack (1999)]. The underwriter, through
the process of taking a ﬁrm public and through its own due diligence will have access to
more information than the general public and possibly more than other institutional
investors. Although the asset management division and the underwriting division are
separate entities, it seems possible, if not likely, that they would share information about
ﬁnns.2 The underwriter is in a good position to provide this superior information to the
asset management division of the institution. Also, many underwriter analysts provide
ﬁrm coverage for the IPO. It is not unreasonable to expect the information gleaned by
the analyst or the underwriting division to be utilized by the institutional investing arm of
the underwriter. Such trading on information should result in higher ﬁrture returns for the

underwriters.

1.2.2 Hypothesis 11 Superior Information: Underwriter/asset managers will purchase
stocks that have superior future return prospects to benefit their institutional

investors

Whereas the Quid Pro Quo Hypothesis can be tested by looking at past returns,

the Superior Information Hypothesis can be tested by looking at future returns. If

 

2 It is fairly common for the asset management division of a ﬁrm and the underwriting division for the same
frrm to be in different buildings. This type of physical separation would help to reduce the amount of
interaction between the underwriting division and the asset management division. However, one major
investment bank executive we interviewed stated that employees oﬁen move across divisions within an
investment bank and such movement would invariably result in personal contacts existing across divisions,
even without the ﬁrm explicitly encouraging these relationships. It is possible that these relationships serve
as one mechanism to allow the sharing of information from the underwriting to the asset management
divisions.

subsequent to all stock purchases, the stock value increases, then the underwriter is said
to be trading on superior information.

What might motivate the underwriting division and the asset management
division to interact? The asset management branch of the institution is likely to be
compensated largely as a ﬁmction of assets under management. The literature has
generally shown that there is a positive correlation between past returns and future ﬁrnd
ﬂows for institutional funds. As such, it is generally not in the asset management’s best
interest to hold stocks that are expected to perform poorly. The asset management
division is likely to be receptive to the Superior Information sharing discussed in
Hypothesis II, but is not likely to favor the Quid Pro Quo theory of Hypothesis 1.

In contrast, the underwriting division is compensated for the number of security
issuances the institution performs in the form of IPOs, 81308, and other investment
banking activities. Underwriters that take ﬁrms public successfully (i.e., with higher
future stock returns) are more likely to attract new clients. Thus, the underwriting
division would likely be in favor of the activities implied in Hypothesis 1. The
underwriter could be indifferent between providing information to the asset management
division of the institution as in Hypothesis H.

We see that the two institutional branches are motivated to engage in different
activities: Quid pro quo for the underwriter and information sharing for the asset
manager. In equilibrium it is possible that the underwriter will share information with the
asset manager in exchange for the asset manager doing some degree of price stabilization

or quid pro quo holdings of recent IPOs.

We should note that fund managers are prohibited by SEC rules from buying
securities to beneﬁt anyone but fund investors. In fact, until 1997, fund managers were
limited to purchases of 4% or less of shares outstanding in stock deals where the banking
afﬁliate participated. Under Wall Street pressure, the SEC increased this ruling to 25%

of shares outstanding after 1997.3

1.3 Data and Institutional IPO Holdings

Our data source for [PCs from 1993 through 1998 is the Securities Data Company
(SDC) new issues database. We examine IPOs prior to the internet/bubble period to
avoid the inﬂuence of this period in our data sample.4 Since 1978 the SEC has required
institutional fund managers with discretionary holdings of more than 10,000 shares or
$200,000 to report their holdings on a quarterly basis.5 By institution name, we merge
the IPO information from SDC to the Thompson Financial database of 13f ﬁlings for
institutional holdings. We consider only underwriters that are listed in the SDC database
as lead underwriters.

We include quarterly holdings of [PCs for the ﬁrst eight quarters the ﬁrm is
publicly traded.6 We examine only the ﬁrst eight quarters because we think it is unlikely

that underwriters would be able to leverage their superior information over the market for

 

3 SEC Release IC-22775.

4 Loughran and Ritter (2004) document a regime shift in several important control variables for our sample
moving from 1993-1998 to post-1998 data.

5 Similar to Field and Lowry (2005) we ﬁnd that a large percentage of our data involves holdings below
10,000 shares and $200,000. We ﬁnd that 11,901 data points or 25% of our dataset is below the threshold
of 10,000 shares or $200,000.

6 Quarter 1 holdings is the ﬁrst quarter end from the date of the IPO. A ﬁrm that goes through its IPO on
September 29, 1996 will have quarterly holdings recorded on September 30, 1996. A fum that goes
through its IPO on July 1, 1996 will also have its ﬁrst quarterly holdings listed on September 30, 1996.

10

longer than two years. However, we do feel that quid pro quo activities may occur for
more than eight quarters. In an effort to balance tests between these two hypotheses, we
settle on an eight quarter horizon. We wish to look at analyst coverage and subsequent
SEO activity as well, which precludes us from just looking at one or two quarters after
the IPO.

The quarterly holdings from Thompson are already aggregated across all fund
holdings within an institution. Our screening criteria yield a sample of 6,441
observations where IPO underwriters hold their own IPO in their managed funds, and
28,233 observations where institutional holders underwrite securities, but do not
underwrite the speciﬁc IPO they hold. There are a total of 2,412 distinct IPOs.

To control for analyst coverage by the institutional investors, we utilize Thomson
Financial’s I/B/E/S database to obtain sell side analyst coverage of IPOs from 1993-2000.
We merge this database into our holdings database based on the institution names. We
use information about which ﬁrm the initiating analyst works for as well as how many
quarters after the IPO the analyst initiates coverage of the IPO.

Since we are interested in controlling for SEO underwriting relationships, we use
the Securities Data Company (SDC) secondary issues database as well. Once again, we
merge the information about when a stock goes through its secondary offering and who
the underwriter is into the original IPO and institutional holdings data. Finally, stock
return information comes from CRSP.

The format of the holdings data can be seen in Panel A of Table 1.1. This table
contains the institutional holdings of Ofﬁce Max for the ﬁrst six calendar quarters ﬁ'om

the November 2, 1994 IPO date. We examine institutional holders that underwrite Ofﬁce

11

Max’s IPO, as well as institutional holders who do underwrite IPOs, but were not part of
the underwriting syndicate for Ofﬁce Max. Quarter one holdings are for the ﬁrst
calendar quarter end after the IPO date, in this case, December 31, 1994. There are
institutions that hold stock in the ﬁrm for as little as one quarter (HSBC Asset
Management holds 41,850 shares in quarter ﬁve) or as many as six quarters (William
Blair & Company).

In Table 1.1 Panel B, we report the analyst coverage information for Ofﬁce Max.
There are two analysts in our dataset that initiate coverage of Ofﬁce Max during the ﬁrst
eight quarters it publicly trades and hold stock sometime during that window. Morgan
Stanley Group Inc, an underwriter for the stock, initiates coverage in quarter two. Bear
Stearns &> Co, a non-underwriter, also initiates coverage in quarter two. Finally, the
secondary stock offering information for the company is contained in Table 1.1 Panel C.
This table shows that the ﬁrm undergoes a secondary stock offering in quarter three and
all three IPO underwriters are also SEO underwriters.

In Table 1.2 Panel A, we report the characteristics of IPO holdings for both
underwriters and non—underwriters. This table shows that underwriters hold slightly
smaller ﬁrms with an average market capitalization of $1,020 million versus $1,230
million for non-underwriters. Underwriters hold IPOs with lower monthly turnover than
non-underwriters. We deﬁne monthly turnover as the total monthly volume for a given
IPO divided by its shares outstanding at the end of the same month. The [PCs that
underwriters hold are less likely to be venture capital-backed than are the
non-underwriter holdings. We use the investment bank reputation rankings from

Loughran and Ritter (2004) available on Jay Ritter’s website,

12

bear.cba.uﬂ.edu/ritter/index.html, and average them across the ﬁrm’s holdings in the
same year. Firms who purchase their own IPOs have a higher ranking (8.4) than
non-underwriter purchases (7.1). ~ All differences between underwriters and
non-underwriters are signiﬁcant at the 1% level.

Underwriters frequently buy their own IPOs. In Table 1.2 Panel B, we report that
underwriters purchase stock in 52% of the 2,412 [PCs that came public between 1993
and 1998. About one-fourth of the IPO sample goes through an SEO within the ﬁrst
eight quarters after the IPO, and over 40% of IPOs have analyst coverage initiated by
their underwriter within eight quarters of the IPO. Also, the number of IPOs purchased
by underwriters is not driven by just a few underwriting/asset management ﬁrms. Of the
141 ﬁrms who both underwrite and manage assets, 68 of them purchase their own IPOs.

We now turn to institutional holdings of IPOs averaged across the ﬁrst eight
quarters of trading, reported in Table 1.3 Panel A. From 1993-1998, average holdings of
IPOs for all non-underwriter funds are 0.92% of total shares outstanding. The average
underwriter holdings are 1.24% of total shares outstanding, for a statistically signiﬁcant
difference (p-value of 0.00) of 0.32% between the two groups.7 The table also
summarizes the average IPO holdings over the early and late 19908. Underwriter
holdings have increased from 1.19% in the 1993-1995 time period to 1.30% from
1996-1998. Non-underwriter holdings have dropped from 0.94% to 0.90% over the same
time period.

Table 1.3 Panel B reports the percentage of shares held grouped by quarters.

Underwriters hold more shares than non-underwriters across all quarter groupings. There

 

7 The share holdings of non-underwriters average 137,200 shares with a market value of $3.3 million. The
share holdings of underwriters average of 245,600 shares with a market value of $6.2 million.

13

is a decrease in the percentage of holdings across quarters for both underwriters and
non-underwriters as time passes from the IPO. Underwriter holdings drop from 1.33% in
quarters one through four to 1.11% in quarters ﬁve through eight. Non-underwriter
holdings have a similar decline. Figure 1.1 shows the general trend of institutional

holdings for underwriter and non-underwriters every quarter for the ﬁrst eight quarters

1.4 IPO Holdings and Performance

1.4.1 Testing The Quid Pro Quo Hypothesis

We now move to testing Hypothesis 1: Quid Pro Quo. The prior literature
documents that institutional holders oﬁen follow a short-term momentum strategy
(Wermers 1999). But if underwriters are trying to stabilize prices or gain more
underwriting business from IPO ﬁrms, they may purchase shares in [PCs that are recent
poor performers.

For each IPO, we form quarterly retmns less the CRSP equally weighted index
before and after the 13f quarterly ﬁling dates. We do not calculate a return for what we
call quarter zero, however, as each IPO will have a different period of time from the IPO
date to the required ﬁrst ﬁling.8 We also classify the institutional holder as a buyer if the
net change in shares from the previous quarter is positive for a given IPO. Out of 34,674
total observations, 75% show an increase in holdings over the previous quarter. The 13f
ﬁlings report shares held on the last day of the quarter, but we have no way of knowing

whether those shares were purchased at the beginning, middle, or end of the quarter.

 

8 A ﬁrm that goes through its IPO on September 29, 1996 will only have one day of returns in quarter zero.
In contrast, a ﬁrm that goes through its IPO on July 1, 1996 will have nearly a full quarter of returns.

14

Examining prior returns (quarter t -1) of stocks purchased in the future (quarter t)
avoids timing ambiguity in that purchases made in quarter t are clearly occurring after
prior retums. However, it is possible that institutional traders only consider stock price
movements over the short term (less than one quarter) in their purchase decisions. We
would not be able to detect any relationship between ﬁrm stock purchases and past stock
' returns in this case. By also examining contemporaneous returns and purchases for
quarter t, we make our analysis timelier, but it is not possible to disentangle cause and
effect. We choose to report both contemporaneous (quarter t) and prior (quarter t - 1)
returns that correspond to changes in quarter t share holdings.

Based on Table 1.4 Panels A and B, we see that non-underwriters in general are
momentum traders, consistent with prior research. Non-underwriters purchase stocks
with a contemporaneous quarterly excess return of 3.38%, and a prior quarterly excess
return of 3.40%. Underwriters, in contrast, make purchases subsequent to much lower
positive returns. Underwriters purchase stocks with a contemporaneous return of just
0.4% and a prior return of 0.9%. The difference in underwriter and non-underwriter prior
quarterly returns is -2.5%, statistically signiﬁcant at the one percent level.

We now continue to examine the Quid Pro Quo Hypothesis by testing to see if
underwriting institutions are liquidating their stock positions after an SEO. If the ﬁrm is
only holding the stock in order to get the SEO business, then we might expect to see
investment banks that become SEO underwriters buying the stock ahead of the SEO, but
selling it after the SEO.

In Table 1.5 Panel A, we compare the net purchases (measured in percent of

shares outstanding) of IPOs in quarters pre- and post-SEO. We see that all institutions,

15

regardless of if they are the original IPO underwriter, are buying the stock before the
SEO. IPO underwriters who are also the SEO underwriter purchase 0.20% of shares
outstanding in the quarter before the ﬁrm goes through an SEO. Likewise, IPO
non-underwriters who underwrite the subsequent SEO are buying 0.12% of shares
outstanding.

Recall that institutional traders are, on average, momentum investors and thus buy
shares as stock prices rise. SEOs are typically led by stock price run-ups so it is not
surprising to see momentum investors buying before the SEO. However, in the quarter
after the SEO, only the IPO underwriters are liquidating stock. They are selling 0.08% of
shares outstanding for a statistically signiﬁcant difference from before to aﬁer the SEO of
0.27%. Note that this ﬁgure is not only statistically signiﬁcant, but is also very
economically signiﬁcant. With an average level of holdings of $5.6 million, this results
in a decrease of nearly $1.75 million in stock holdings in the quarter after the SEO. In
contrast, there is no difference between the pre- and post-SEO trades for institutions that
did not underwrite the original IPO but did underwrite the SEO. There is also no
difference in pre- and post-SEO trades for those institutions that did not become the SEO
underwriter.

Consistent with the Quid Pro Quo hypothesis, we see that IPO underwriters who
are also the SEO underwriters change their behavior aﬁer the SEO. In contrast, all the
other institutional groupings are on average still purchasing shares after the SEO. It may
be that IPO underwriters are liquidating their higher than average holdings after an SEO

in an effort to diversify their fund’s holdings. In any case, it is clear that the underwriters

l6

are behaving differently from the non-underwriters in a way that could be a breach of
their ﬁduciary responsibility

We now use a probit regression framework to ﬁnd the determinants of
institutional buys. This framework allows us to incorporate the effects of prior returns on
buying behavior, as well as the effects of other institutional services such as analyst
coverage or SEO underwriting. The dependent variable is a dummy variable taking on a
value of one (zero otherwise) if an institution purchases a stock during a particular
quarter. We control for prior returns, ﬁrm size and institutional rank, as well as including
dummy variables for analyst coverage and SEO underwriting in the previous quarter.
These dummy variables measure whether the given institution initiates coverage or
underwrites the SEO; they not measure if other institutions provide analyst coverage or
underwrite SEOS. Following Petersen (2005), we report p-values using clustered
standard errors in all regressions throughout the paper in an effort to control for serial
correlation among manager observations.

Table 1.6 gives the determinants of stock purchases for an underwriter and a
non-underwriter, respectively. The unconditional probability for an underwriter making a
purchase over the ﬁrst eight quarters that a stock trades is 65%, and 78% for a
non-underwriter.

The regressions provide information consistent with both underwriters and
non-underwriters acting as momentum traders in that their coefﬁcients on the previous
excess retruns are positive. Thus, the higher the previous quarterly return, the higher the
probability that the underwriter and the non-underwriter will purchase the stock. This

result is statistically signiﬁcant at the one percent level although the economic

l7

signiﬁcance is questionable. For instance, for the underwriters, if the previous quarterly
excess retrun of the stock were to increase one standard deviation, the probability of
purchasing the stock over the subsequent quarter would increase by 2.2%. Likewise for
non-underwriters, the results are statistically signiﬁcant, but economically questionable.
An increase in the previous quarterly excess return by one standard deviation results in an
increase in the probability of a non-underwriter stock purchase by 2.6%.

The coefﬁcient for an analyst initiation in the previous quarter provides a negative
and statistically signiﬁcant result of -0.19 (p-value of 0.02). This implies that if the
underwriter initiated analyst coverage in the previous quarter, the underwriter is 7.4%
less likely to buy stock in the current quarter. We consider two interpretations for this
result to be valid. The underwriter could be selling the stock because the informational
advantage possessed prior to analyst initiation is no longer present. It is also possible that
the underwriter is selling the stock because they have fulﬁlled their implicit obligation to
their clients by holding the stock and/or providing analyst coverage. In contrast, analyst
initiations have no effect on when non-underwriters purchase the stock.

Both probit regressions also include dummy variables for previous quarter SEOs.
The resulting coefﬁcient is statistically signiﬁcant and negative for the underwriter. This
means that the underwriter is 9.0% less likely to purchase the stock in the quarter after an
SEO compared to another quarter. In contrast, previous SEOs have no statistically
signiﬁcant effect on non-underwriter purchases. These results are consistent with the
purchase and sales data contained in Table 1.5. After becoming the SEO underwriter, the
IPO underwriter starts selling the stock. If an IPO non-underwriter becomes the SEO

underwriter, it has no effect on their decision to trade the stock.

18

To this point, we have several stylized facts that support the Quid Pro Quo
Hypothesis. First, underwriters hold more shares of stock for the ﬁrst eight quarters that
the stock publicly trades. Our univariate results show that underwriters trade less on
momentum than non-underwriter institutional traders do. In addition, IPO underwriters
are selling their stock after they provide additional services to the ﬁrms they have
underwritten. Once the underwriter initiates analyst coverage, they are more likely to sell
the stock that they hold. Also, in the quarter after an SEO, the underwriter is more likely
to sell stock. Non-underwriter purchases are not signiﬁcantly inﬂuenced by either
analyst coverage or SEO underwriting.

Now we wish to test if the actions taken by underwriters are successful in gaining
additional business for the underwriter. If the underwriter is not more likely to gain
future business by initiating analyst coverage on stocks, or buying and/or holding large
amounts of stock, then the facts we have observed might simply be coincidental. One of
the main ways underwriters can gain additional business ﬁom the ﬁrms they take public
is ﬁom subsequent equity offerings. We now study if the actions taken by equity
underwriters are successful in gaining them future underwriting business with the ﬁrms
they take public in the IPO.

We use a probit regression framework where the dependent variable is a dummy
variable taking on a value of one (zero otherwise) if the institution is the underwriter of a
secondary stock offering within the ﬁrst eight quarters of trading. We use similar control
variables as in our previous regressions. The variables of most interest are the

underwriter ﬂag, the analyst initiations, and the buys by IPO underwriters and

19

 

non-underwriters. These regressions use only 1,415 data points because we analyze only
quarters when an SEO occurs.

The ﬁrst regression in Table 1.7 shows that the underwriter ﬂag is large and
statistically signiﬁcant with a coefﬁcient of 2.17. This result is not surprising as
Ljungqvist and Wilhelm (2005) have reported that for 64% of ﬁrms, the ﬁrm
underwriters also underwrite subsequent SEOs. If 'a ﬁrm has analyst coverage from an
investment bank in the previous quarter, then that investment bank is 30.5% more likely
to be the SEO underwriter for the ﬁrm. This result is statistically signiﬁcant at the one
percent level. We are not suggesting causality, simply that SEO underwriters oﬁen
initiate coverage for the ﬁrm prior to the underwriting of the new security. This is
another type of quid pro quo that we do not investigate here. Note that these results are
in contrast to Ljungqvist, Marston, and Wilhelm (2005) who show that aggressive analyst
coverage does not attract future underwriting business.

If a non-underwriter purchased the stock just in the previous quarter, there is no
signiﬁcant impact on whether they become the SEO underwriter. If the IPO underwriter
purchased stock in the previous quarter, there is also not a signiﬁcant impact on
becoming the SEO underwriter. It may be that underwriters of the IPO build up their
stock levels slowly over time so that our regression is not able to detect a relationship
between purchases in the previous quarter and the identity of SEO underwriters. To this
end, we run the second regression of Table 1.7 with institutional holdings as an
explanatory variable. We feel that the level of holdings might be more important as they

are the outcome of purchases over a longer period of time than simply one quarter.

20

These results show that the stock holdings of non-underwriters are not statistically
signiﬁcant determinants of SEO identity. However, IPO underwriters who hold more
stock are rewarded for their higher holdings through a positive and statistically signiﬁcant
probability of obtaining future business as the stock underwriter (coefﬁcient of 0.08 with
a p-value of 0.02). In other words, a one standard deviation increase for the level of
underwriter holdings provides an increase of 2.5% in the “probability of being the
underwriter for subsequent stock offerings. Note that this result obtains even aﬁer we
control for the IPO underwriter identity, given that underwriting relationships are
persistent. It appears that underwriter activities such as initiating analyst coverage and
holding higher levels of stock subsequent to the IPO are in fact rewarded by future
business with the ﬁrm in the form of SEO underwriting. We should also mention that
higher institutional rank makes it signiﬁcantly more likely the ﬁrm will become the SEO
underwriter.

1.4.2 Testing The Superior Information Hypothesis

We now turn to an analysis of Hypothesis II: Superior Information. For every 13f
ﬁling on quarter t, we calculate future returns for quarter t + 1. The ﬁrst data column in
Table 1.8 gives the results of a regression with ﬁrture quarterly excess returns as the
dependent variable. The explanatory variables here include the prior quarterly return (to
account for return autocorrelation), a dummy variable for institutional large buys, a
cross-dummy for an underwriter large buy, the logarithm of market equity, the rank of the
institution holding the stock, a dummy variable for analyst initiations by the institution in

the previous quarter, and a dummy variable for the underwriter. We deﬁne a large buy as

21

one in which the underwriter is both buying in the quarter and has stock holdings greater
than the median across all observations.

We ﬁnd that stock returns are positively autocorrelated in the short term as
predicted by the past literature. The institutional large buy variable does not have
signiﬁcant predictive ability for future returns, although the underwriter large buy
variable does. An underwriter buy predicts higher and statistically signiﬁcant future
quarterly returns of 1.92% or 7.68% per year. 9 Also, the results show that the larger the
size of an IPO, the better the future performance. If an institution has a higher
underwriter ranking, this does not appear to appreciably affect the subsequent returns of
the institution’s purchases. Analyst coverage initiation has no predictive ability for future
returns.

The evidence presented in the “All Firms” column of Table 1.8 shows that
underwriter/asset mangers make superior proﬁts through their trading activities where
non-underwriter institutional traders do not. However, we wish to determine when
underwriters might have a particular informational advantage. As underwriters perform
their due diligence, we may expect to see them make informed trades when they have the
highest informational advantage. Firms with no analyst coverage might provide the best
ability for underwriters to trade on their special knowledge. To this end, we run another
regression using only ﬁrms that have no analyst coverage by any institution. In the “No
Analyst Coverage” column of Table 1.8, we ﬁnd that ﬁrms with no analyst coverage have

a statistically signiﬁcant return of 1.78% after underwriter large purchases. Note that

 

9 If we deﬁne large buys as the top 30‘h percentile of purchases and small buys as the bottom 30“I percentile
of purchases, this increases the strength and statistical signiﬁcance of our results. Our results are not
appreciably changed if we deﬁne the size of a large buy for each institution or for the population as a
whole.

22

 

there is no analyst coverage for the ﬁrm so the analyst coverage explanatory variable is
no longer applicable for this regression. The signiﬁcance of the other explanatory
variables is consistent with the previous regression. When there is no analyst coverage,
underwriters are able to use their special information gleaned from the due diligence
process to make superior trades in the asset management division.

We then test the sample of ﬁrms that do have analyst coverage. The “Analyst
Coverage” column in Table 1.8 reports that the underwriter buy variable is no longer
statistically different from zero. This result implies that underwriters are not able to
outperform in their large buys once there is analyst coverage for the frrrns. This result is
consistent with the idea that underwriter managers have an informational advantage only
as long as there is limited public scrutiny of a ﬁrm.

We also examine performance differences between high and low rank
underwriters. High rank underwriters are deﬁned as those underwriters with a ranking at
the time of stock holdings of at least 8.1. The fourth and ﬁfth data columns provide the
ranking results. High rank underwriters earn statistically signiﬁcant excess returns of
2.34% per quarter on their large purchases. Low rank underwriters, however, do not earn
signiﬁcant excess returns.

Finally, we test to see if SEO underwriters are capable of making superior trades
in their asset management portfolios. We ﬁnd that unlike IPO underwriters, SEO
underwriters earn a statistically insigniﬁcant return on their large purchases. 7 The
coefﬁcient on the large SEO underwriter buys is 1.38% with a p-value of 0.35. Thus, it

appears that IPO underwriters have an informational advantage that they can exploit until

23

there is analyst coverage. In contrast, SEO underwriters do not appear to have any
informational advantage that allows them to make superior trades. '0

We have shown that underwriter asset managers have the ability to make stock
purchases prior to positive future abnormal returns. However, it is possible that these are
purchases of riskier stock that should in fact have higher expected future returns.
Following the recommendations in Lyon, Barber, and Tsai (1999), we control for ﬁrm
risk by creating abnormal returns using a matching ﬁrm for each IPO. Each quarter, we
select a matching ﬁrm in the same industry (2-digit SIC code) as the IPO,ll with the
closest market capitalization to the IPO. We require that the matching ﬁrms be
“seasoned” in that they are at least ﬁve years removed from their initial public offering.
We calculate matching ﬁrm and IPO returns over identical quarters, and then subtract the
matching ﬁrm return from the IPO return. If a matching ﬁrm is delisted during our
measurement window, we simply select another size and industry-matched ﬁrm at that
point.

In Table 1.9, we report one-, two-, and three-month buy-and-hold abnormal
returns for each underwriter grouping. Panel A gives the abnormal returns according to
large institutional buys, while Panel B reports small institutional buys. We separate
purchases in this manner under the hypothesis that underwriters will most exploit their
superior information through their large purchases. If underwriters are trading consistent

with the Quid Pro Quo hypothesis, it would likely be through smaller transactions.

 

'0 In untabulated results, we also test if underwriter excess returns are clustered by institution, by industry,
or by year. We ﬁnd no statistical evidence that excess returns from underwriter large purchases are driven
by just a few institutions, a few industries, or occur in only a few years. Also, underwriters’ superior
returns are not conﬁned to just the ﬁrst few quarters.

n We match on industry as opposed to book-to-market because book values can be difﬁcult to obtain in the
early quarters after an IPO.

24

For the large purchases in Panel A, the underwriters outperform the
non-underwriters over both one- and two-month horizons. For example, over two
months, the underwriters outperform non-underwriters by 1.65% (p-value of 0.01). Over
three months, there is not a signiﬁcant difference between the two groups. It is possible
that the three-month returns are not signiﬁcant because the institutional trades become
public halfway through this return window. The trades of institutions are made available
(through their 13f ﬁling) to the public 45 days after the end of the quarter. Once the
institutional trading information becomes available, it would be much more difﬁcult for
the institutions to earn abnormal returns. Our three-month returns are based on 45 days
when the trading information would not be available, and 45 days when it would be.

In Panel B, we report abnormal returns after small institutional purchases.
Returns are marginally different across the two groups at the one-month horizon, but not
at the two- or three-month horizon. The results of Table 1.9 are consistent with
undermiters using large purchases to exploit their superior information learned during
the IPO underwriting process.

1.4.3 Robustness Tests

We provide evidence consistent with underwriters purchasing stocks aﬁer stock
price declines. Our tests show results that are consistent with Hypothesis 1: it appears
that institutional asset managers are using their fund assets as a quid pro quo for ﬁrms
going through IPOs. However, it is also possible that this result is consistent with
Hypothesis H. If underwriters know that stocks are undervalued due to their superior
information gathered as the underwriter, then it is only natural to expect ﬁrms to purchase

stocks aﬁer a price decline. For this result to be consistent with Hypothesis H, we would

25

expect these stocks to subsequently rebound in prices. In other words, if the underwriter
is buying the stocks because they are underpriced, then the subsequent stock performance
should be positive.

To this end, we examine the future returns of stocks purchased aﬁer the stock
price has declined. We show in Table 1.10 that stocks purchased by the underwriters
after down quarters have returns of -0.22% over the subsequent quarter. Non-underwriter
purchases made after quarters of negative returns have returns of -0.21% in the
subsequent quarter. The difference between these values is not statistically signiﬁcant.

We also split our sample by small and large buys. We ﬁnd that small underwriter
buys after negative performance result in future returns of -0.81% on average where
non-underwriters have future returns of -0.02%. For large buys, the ﬁgures for
underwriters and non-underwriters are 0.35% and -0.41%, respectively. None of the
differences between underwriters and non-underwriters are statistically signiﬁcant at a
meaningful level. These results imply that underwriters are not purchasing stocks after

price declines in anticipation of higher future returns on these stocks.

1.5 Conclusions

There is a clear potential for conﬂicts of interest within an institution that
underwrites securities and also manages institutional funds. At the same time, the
combined underwriter/asset manager may possess superior information allowing the
institution to make proﬁts above those possible for other institutional traders. We test

two main hypotheses relating the beneﬁts and costs of underwriter/asset managers: the

26

Quid Pro Quo Hypothesis and the Superior Information Hypothesis. We ﬁnd evidence
that underwriters utilize their institutional holdings to stabilize prices or otherwise
provide a quid pro quo to ﬁrms they have brought public. We also ﬁnd strong support for
the statement that asset managing institutions are trading on their informational
advantage as underwriters, earning an annualized excess return of 7.7%.

The Quid Pro Quo Hypothesis constitutes a severe breach of trust for asset
managers. One question of signiﬁcance is can these activities take place in an efﬁcient
market equilibrium? In an economy with perfect information, it is clear that the
underwriters of securities could never stabilize prices using institutional assets as this
action would be penalized by the market. Institutions engaging in such activity would be
severely punished by investors removing their assets from the management of that
institution. Thus, institutional investors would not engage in price stabilization. If,
however, the institutional asset manager can exploit superior information from the
underwriter sometimes and provide quid pro quo for the underwritten securities at other
times, it seems likely that both price stabilization and information sharing will occur in
equilibrium. Our results are consistent with this conjecture. We ﬁnd that while assets
under management are utilized to attract more investment banking business, the assets

also have positive abnormal proﬁts in the large trades of the institution.

27

APPENDIX 1.

TABLES OF ESSAY 1

28

 

 

 

 

Scam mmo.mm oomdm 8mg; oomdm oomﬁm 2: 25%—2 Emmet mmD
05 32.85 582.3
332 mmcﬁm 8de $53 Smwm 2: Bangui 33:33.5
Elﬁn o3? 05 .2303 can;
con. _ .2 900 “coaowmcmz 88655qu
and: 883% 255535 3:232
25.? 25.2. oooﬁv ooodv covdm cotom EoEommeaE 8mm< boEowEoZ
oooﬁ _ 2v.— oov€ nomé 2: 00 a. £653 Etc—2
25.83 8183 83mm 893.3 3655 .oaeom
on”. _ v EoEomeWZ 8mm< 0mm:
o0 a. msomm $8200
o0 8 98M 8:5
:88m 7.5 325 £020
oocdoo ooodon o0 anowﬁaz «58200
cog; 80.2 mmmxev coma: co a. 8.88m Hum
.05 ~8>c<
OE 5:2 850 he muoztﬁoucaéoz
wbmdm oowém 086m omwdm octmm 8me .00 d .55 SEE?
oawK—o; “SQ—man... 82$ www.mF 2:63 2:32 2: 9.80 tho—Sam Hamass.)—
ooowm 892v o0 a. 8385 be; :80
on: SSE 8E0 no.“ £053.53:

8 3 e0 8 mo 5

 

520585 1822535 3 $520: buotmsd n< 35$

:3 .N 3:80.52 "85. Oh—
ma—z 35¢ .Su mwnmuobc xoeum baa—83m can .mnecacaoﬁaeuom warm—«5... .3563: bunt-30

“A use“.

29

.83 CB 2: Bus $53 5:23 .96 05 a Cg one .5530 .3383 Dam 05 Sea 888
coca—8&5 0mm .mmmn 88m Ecuauozou can 383:5 5358 v5 ..ﬁoemim nomEoF—t Sea 8% $522. 28925 chides £5
gong—o Sim 2: B 88% Be: on :5 .On: :03» 05 Busing: “on on 85 Enema—«E “833:3 32882: 805 we ..EBEZoE:
ASP. 8 Home“ 03 .8526 nozw a E 8.35 Ben 98 OB 05 0:53.25 35 Eowaag «83259 33582: 3.08 we ..EoEEovEr.
2 Show 63 .0& some Sm .95 65 85 aeoEBSm «2 3.8.83 85 35 5636 8080358 «83 an 96: cm? .38 3895.623
805. .waﬁ can 33 3233 average: 32 we 33va 0% Dam 05 E 35336 2m 35 Eougovg 2:888 63

0mm .00 a. ~35 82:52
0mm 05 955 thou—Em 5&qu
0mm 00 Q 8585 85>? BBQ

 

0mm 532 850 How muozbiocaD

 

 

 

 

 

 

e0 3 5 ...o No 8
ammv mwctoto x85 >328ka “U ESE
Em wcobm 00 um mEmBm Sem
0% 682 SEC no.“ muozgouabéoz
Em weobm 05 @380 zoESm aawuoE
On: 522 BED new muozgoueb
co 3 3 8 No 5

 

320:0.85 :30:ng an mcoumwnoEEooou “ma—SE um 35$

 

983 E 2.3.

30

Table 1.2
Summary Statistics

We examine underwriters that are classiﬁed in the SDC IPO database as lead underwriters between 1993
and 1998. For each IPO, we refer to “underwriters” as those investment bank/asset managers that
underwrite the IPO and hold shares in a given quarter. We refer to “non-underwriters” as those investment
bank/asset managers that do not underwrite the given IPO, but do hold shares in the given quarter. We
acquire reputation rankings from Jay Ritter’s website bear.cba.uﬂ.edu/ritter/index.html. Measurement of
the ﬁrst eight quarters of trading begins at the ﬁrst calendar quarter after the IPO date. Two-tailed
p-values are given in brackets.

Panel A: Averages across ﬁrst eight quarters of tradirg

 

 

 

 

 

 

Institutional Institutional UW - Non-UW
Average Holder = UW Holder = Non-UW [p-value]
IPO Market Value -210.0
(millions) 1’019'9 1’229'9 [0.00]
IPO Monthly .44
Turnover % 13'0 17'0 [0.00]
. 1.1
IPO Offer Price 16.5 15.4 [0.00]
IPO VC-Backed % 27.7 34 0 '6'3
' [0.00]
Institutional Holder 8 4 7 6 0.7
Reputation Rank ' ' [0.00]
N 6,441 28,233
Panel B: Number of observations
Institutional Institutional Total
Holder = UW Holder = Non-UW Sample
Number of '
Distinct IP Os 1,255 2,340 2,412
Number of
Distinct lPOs Issuing 401 675 675
SEO
Number of
Distinct [PCs with 721 506 992
Analyst Coverage
Number of
Distinct Institutions 68 140 141
Number of
Distinct Institutions 32 39 42

Providing Coverage

31

Table 1.3
Percentage of IPO Shares Held

We examine underwriters that are classiﬁed in the SDC IPO database as lead underwriters between 1993
and 1998. These underwriters must also have an asset management division that ﬁles quarterly 13f
statements with the SEC. For each IPO, we refer to “underwriters” as those investment bank/asset
managers that underwrite the IPO and hold shares in a given quarter. We refer to “non-underwriters” as
those investment bank/asset managers that do not underwrite the given IPO, but do hold shares in the given
quarter. Measurement of the ﬁrst eight quarters of trading begins at the ﬁrst calendar quarter aﬁer the IPO
date. Both panels report the percentage of shares held, deﬁned as the number of shares stated in the
quarterly 13f ﬁlings (obtained from Thomson Financial) divided by the number of shares outstanding
(obtained from CRSP). Two-tailed p-values are given in brackets.

Panel A: Average percentage holdings across ﬁrst eight quarters of trading

 

 

Institutional Institutional UW -
IPOs Holder = Holder = Non-UW
Issued in N UW N Non-UW [p-value]
0.32
1993-1998 6,441 1.24 28,233 0.92 [0.00]
1993-1995 3 595 1.19 14 758 0 94 0'25
' ’ ’ ' [0.00]
1996-1998 2 846 1 30 13 475 0 90 0'40
’ ' ’ ' [0.001

 

Panel B: Average percentage holdings grormed by quarters

 

 

Institutional Institutional UW -
Quarters Holder = Holder = Non—UW
Post-IPO N UW N Non-UW [p-value]
1993—1998
1—4 3,783 1.33 14,175 1.01 0.31
[0.00]
0.29
5 — 8 2,658 1.11 14,058 0.83 [0.00]

 

32

 

Table 1.4
Contemporaneous and Prior Quarterly Excess Stock Returns
for IPOs Bought by Institutions

For Panel A, we calculate quarterly excess stock returns in quarter t for all institutional IPO purchases
occurring in quarter t. For Panel B, we calculate quarterly excess stock returns in period t - 1 for all
institutional IPO purchases occurring in quarter t. We classify institutional buyers as those institutions
whose quarterly change in holdings from the previous quarter is greater than zero for a given IPO. All
excess returns are less the CRSP equally weighted market index. Two-tailed p-values are given in
brackets.

Panel A: Contemporaneous quarterly excess returns

 

 

 

 

 

Institutional

Holder N Return %
UW 4,175 0.40
Non-UW 21,881 3.38
UW —

Non-UW [€35]
[p-value] '
Panel B: Prior quarterlﬁxcess returns
Institutional

Holder N Return %
UW 3,519 0.90
Non-UW 19,145 3.40
UW -

Non-UW [:38]
[IL-value] '

 

33

Table 1.5
Stock Buys Pre- and Post-Secondary Stock Offerings

Percentage of shares bought is deﬁned as the number of shares purchased from the previous quarter divided
by the number of shares outstanding. Pre-SEO refers to quarter 1 - l conditional on an SEO occurring in
quarter t. Post-SEO refers to quarter t + 1 conditional on an SEO occurring in quarter t. Two-tailed
p-values are given in brackets.

Panel A: Percentage of shares bought before and after an SEO by SEO underwriter

 

 

Pre SEO -

Institutional . Post SEO

Holder N Pre-SEO N Post-SEO [p—value]
0.27
IPO UW 228 0.20 274 -0.08 [0.04]
IPO N UW 73 0 12 107 0 13 ‘0'02
on- . . [0.85]

 

Panel B: Percentage of shares bought before and after an SEC by SEO non-underwriter

 

 

Pre SEO -

Institutional Post SEO

Holder N Pre-SEO N Post-SEO [p-value]
0.21
IPO UW 49 0.33 39 0.13 [0.35]
IPO N UW 973 O 20 1 438 0 15 0'05
0‘" ' ’ ' [0.19]

 

34

Table 1.6
Probit Regressions for Determinants of Institutional Buys

The dependent variable is a dummy variable taking on a value of one (zero otherwise) if the institution
purchases the stock in quarter t. Explanatory variables from quarter t - 1 include the quarterly excess stock

return (Return t-l), the logarithm of the market capitalization of the stock (Ln(ME) t-l)’ the reputation
rank of the investment bank (Institutional Holder Rank t-l): a dummy variable taking on a value one (zero
otherwise) if the institution initiates analyst coverage in the quarter (Analyst Coverage t-l)» and a dummy
variable taking on a value of one (zero otherwise) if the ﬁrm issues an SEO in the quarter (SEO Offering t-

1). All excess returns are less the CRSP equally weighted market index. P-values are given in brackets,

and reﬂect controls for heteroskedasticity and serial correlation among manager observations (clustered
standard errors). '

 

 

Institutional Institutional
Buyer = Buyer =
UW Non-UW
0.002 0.001
Return t-l [0.00] [0.00]
-0.001 -0.090
Ln(ME) t-l [0.93] [0.00]
Institutional Holder 0.017 .0009
Rank H [0.77] [0.70]
Analyst -0. 193 0.153
Coverage H [002] [021}
SEO -0.243 -0.049
Offering H [0.00] [0.30]
Interce t 0.258 1.327
P [0.61] [0.00]
N 5,484 24,822

 

35

Table 1.7

Probit Regressions for Determinants of SEO Underwriter Identity
The regression includes only the quarters in which a ﬁrm issues a secondary equity offering. The
dependent variable is a dummy variable taking on a value of one (zero otherwise) if the institution
underwrites an SEO in quarter t. Explanatory variables from quarter t - 1 include the quarterly excess stock

return (Return t-l): the logarithm of the market capitalization of the stock (Ln(ME) t-l), the reputation

rank of the investment bank (Institutional Holder Rank t-l), a dummy variable taking on a value of one

(zero otherwise) if the institutional holder is the IPO underwriter (IPO UW ﬂag), a dummy variable taking
on a value one (zero otherwise) if the institution initiates analyst coverage in the quarter (Analyst

Coverage t-1), a dummy variable taking on a Value of one (zero otherwise) if the institution purchased
stock in the quarter (Institutional Buy t-1), the percentage shares the institution holds in the quarter

(Institutional Holdings t-1), and two cross-variables between the underwriter ﬂag and the institutional buy

and institutional holding variables. All excess returns are less the CRSP equally weighted market index.
P-values are given in brackets, and reﬂect controls for heteroskedasticity and serial correlation among
manager observations (clustered standard errors).

36

Table 1.7

 

Return [-1

Ln(ME) t-l

Institutional
Holder Rank [-1

IPO UW ﬂag

Analyst

Coverage [-1

Institutional
BUY t-l

IPO UW *
Institutional

Buy t-l
Institutional
Holdings [-1

IPO UW *
Institutional

Holdings [-1

Intercept

N

0.001
[0.68]

-0.027
[0.64]

0.138-

[0.00]

2.171
[0.00]

1.246
[0.00]

-0024
[0.84]

0.012
[0.96]

-2.408
[0.00]

1,415

0.001
[0.69]

-0.036
[0.52]

0.138
[0.00]

2.110
[0.00]

1.236
[0.00]

-0044
[0.35]

0.076
[0.02]

-2332
[0.00]

1,415

 

37

Table 1.8
Multivariate Regression for Determinants of Future Quarterly
Excess Stock Returns

The dependent variable is quarterly excess stock returns in quarter t for all institutional IPO buys or sells
occurring in quarter t - I . Explanatory variables from quarter 1 — 1 include the quarterly excess stock return

(Return [-1), the logarithm of the market capitalization of the stock (Ln(ME) [-1), the reputation rank of

the investment bank (Institutional Holder Rank t-1), a dummy variable taking on a value of one (zero
otherwise) if the institutional holder is the IPO underwriter (IPO UW ﬂag), a dummy variable taking on a
value of one (zero otherwise) if the institutional holder is the SEO underwriter (SEO UW ﬂag), a drunmy
variable taking on a value one (zero otherwise) if the institution initiates analyst coverage in the quarter

(Analyst Coverage [-1), a dummy variable taking on a value of one (zero otherwise) if the institution

makes a large stock purchase in the quarter (Institutional Large Buy t-1), , and cross-variables between the
IPO and SEO underwriter ﬂags and the institutional large buy variable. A large buy is deﬁned as those
institutions that hold more shares than the median and also buy in a given quarter. All excess returns are
less the CRSP equally weighted market index. Analyst coverage (no analyst coverage) observations are
IPOs that have at least one (zero) analyst that have covered the ﬁrm since the IPO date. High (low)
underwriter rank observations have a ranking at the time of stock holdings of at least (not more than) 8.1.
P-values are given in brackets, and reﬂect controls for heteroskedasticity and serial correlation among
manager observations (clustered standard errors).

38

Table 1.8

 

 

No High Low
Analyst Analyst Rank Rank SEO
All Firms Coverage Coverage UW UW UW
0.03 0.03 0.04 0.03 0.03 0.03
Return t—1 [0.00] [0.00] [0.00] [0.00] [0.01] [0.00]
Institutional -0.08 -014 0.34 -O.36 0.59 0.23
Large Buy 1-1 [0.86] [0.80] [0.63] [0.51] [0.36] [0.59]
IPO UW *
Institutional 1.92 1.78 1.72 2.34 0.63
[0.00] [0.04] [0.13] [0.00] [0.73]
Large Buy [-1
SEO UW *
Institutional 1 '38
[0.35]
Large Buy [-1
0.96 0.17 1.38 0.82 1.17 0.97
LniME) t-1 [0.00] [0.30] [0.00] [0.00] [0.00] [0.00]
Institutional 0.02 0.06 0.02 0.38 0.30 0.00
Holder Rank 1-1 [0.84] [0.71] [0.91] [0.44] [0.17] [0.97]
Analyst -044 0.26 1.34 -1.06 1.09
Coverage t-l [0.72] [0.86] [0.38] [0.63] [0.41]
-094 —1.13 -0.85 -104 -159
1P O W Flag [0.03] [0.08] [0.24] [0.04] [0.06]
SEO UW Flag [’3 $6]
m t -5.67 -2.19 -7.57 -8.02 -8.32 -577
creep [0.00] [0.18] [0.00] [0.04] [0.00] [0.00]
N 33,979 22,164 11,815 25,028 8,951 33,979

 

39

Table 1.9
Future Abnormal Stock Returns for IPOs Bought by Institutions

Panel A reports the quarterly excess stock returns in quarter t for all institutional IPO large purchases
occurring in quarter t - 1 . A large buy is deﬁned as those institutions that hold more shares than the median
and also buy in a given quarter. Panel B reports the quarterly excess stock returns in quarter t for all
institutional IPO small purchases occurring in quarter t - I . A small buy is deﬁned as those institutions that
hold fewer shares than the median and also buy in a given quarter. Abnormal returns are calculated using a
matching ﬁrm technique. For each IPO, a matching fum is chosen in the same industry (2-digit SIC code)
with the closest market capitalization for the given quarter. Matching ﬁrm and IPO returns are calculated
over identical periods. Abnormal returns reﬂect the IPO return minus the matching ﬁrm return. One-tailed
p-values that test whether the return difference is greater than zero are given in brackets. .

Panel A: Abnormal returns (%) for large institutional buys

 

 

 

 

 

N l-Month 2-Month 3-month
IPO UW 2,088 -0.06 0.40 0.95
IPO Non-UW 10,505 -0.93 -1.25 0.11
111138 EX'UW 0.87 1.65 0.84
[p-value] [0.05] [0.01] [0.17]
Panel B: Abnormal returns (%) for small institutional buys
N l-Month 2-Month 3-month
IPO UW 2,001 0.51 -0.13 0.01
IPO Non-UW 10,651 -0.21 -0.69 0.14
11.1283“, 11w 0.72 0.56 -0.14
0‘“ [0.09] [0.21] [0.44]

[p-value]

4O

Table 1.10
Future Quarterly Excess Stock Returns for IPOs Bought by Institutions
(Negative Prior Returns Only)

We calculate quarterly excess stock returns in quarter 1 for all institutional IPO purchases occurring in
quarter t - 1, but consider only observations where the excess return in quarter 1 — l is negative. A large buy
is deﬁned as those institutions that hold more shares than the median and also buy in a given quarter. A
small buy is deﬁned as those institutions that hold fewer shares than the median and also buy in a given
quarter. All excess returns are monthly returns less the CRSP equally weighted market index. One-tailed
p-values that test whether the return difference is greater than zero are given in brackets.

 

Return % Return %

 

. . N Return % for for
Instrtutronal Holder for all Buys for all Buys Small Bug Large Bgys
IPO UW 2,244 -0.22 -0.81 0.35
IPO Non-UW 10,817 -0.21 -0.02 -0.41
338 13:: _UW -000 -079 0.76
[p-value] [0.50] [0.7 8] [0.21]

 

41

ESSAY 2. VOLUNTARY EARNINGS DISCLOSURE DURING THE QUIET

PERIOD

2.1 Introduction

Managers can release information through regulated ﬁlings with the SEC (for
instance, 10Q reports), through information intermediaries such as stock analysts, or
through discretionary disclosures using press releases. Of these media, discretionary
press releases are the least studied and one of the most interesting methods of information
dissemination by a manager. But direct press releases are difﬁcult to study due to the
large range of information that can be released through this medium and the lack of
structure to a discretionary press release. In addition, it is often unclear if managers are
releasing information to the public because they have established a track record of doing
so and they are trying to maintain their record, or they have some good news to release to
the market. To simplify the analysis of discretionary disclosure, we choose a unique
setting to study information disclosure. We choose a subset of ﬁrms that have high
information asymmetry, a limited amount of public information about operating
performance, and no track record of earnings releases: ﬁrms conducting initial public
offerings.

In the context of an initial public offering, we have a unique opportunity to study
under what conditions information is released through a discretionary press release. This
is due to the fact that ﬁrm disclosure is largely regulated by the securities exchange

commission during the quiet period that follows an initial public offering. We use the

42

quiet period setting as a natural experiment to examine the conditions under which the
manager releases information through a discretionary press release. SEC regulation
limits ﬁrms from any activities that would be seen as “marketing the security,”

12 Using the quiet period to

prohibiting the release of any forward-looking statements.
study discretionary disclosure is important because it limits the type and amount if
information that can and will be released by the ﬁrms studied. What managers have
discretion over during the quiet period is the release of historical earnings of the ﬁrm
prior to the customary quarterly ﬁlings. The release of this earnings information during
the quiet period is the focus of this paper.

We ﬁnd that ﬁrms releasing earnings information during the quiet period tend to
be larger more mature ﬁrms with more prestigious underwriters. This is not surprising in
that larger ﬁrms are more likely to have better internal controls to allow the early release
of earnings. Quiet period earnings releasing ﬁrms also tend to have performed better
over the quarter for which they are releasing earnings. However, fnms are also more
likely to release earnings aﬁer higher market volatility the ﬁrst ﬁve days the stock trades.
This result implies that managers consider more than just the performance of their ﬁrm in
the release of earnings information. In other words, it appears that managers are timing
the market.

On average, the ﬁrms releasing earnings during the quiet period have a positive

stock price response of 2.04% around the earnings release date. We ﬁnd a positive

 

'2 The U. S. Securities and Exchange Commission release #5180 says that “ . . .any publication of
information by a company in registration other than by means of a statutory prospectus should be limited to
factual information and should not include such things as predictions, projections, forecasts or opinions
with respect to value.” This disclosure limitation applies to the whole period “. . .during which dealers must
deliver a prospectus [to investors in the ﬁrm].” In our sample period, this time lasts from the time the ﬁrm
has an agreement with the underwriter until 25 days aﬁer the IPO date.

43

relationship between earnings release abnormal returns and the volatility of the stock but
no relationship with the abnormal operating performance of the ﬁrm. The fact that the
market reacts not to the superior earnings performance of the ﬁrm, but to the ﬁrm
volatility once again implies that managers are timing the market in their press releases.
Managers release earnings information when their stock price is more volatile and the
market responds with a higher increase in stock prices. An increase in IPO ﬁrm volatility
of 1.0% over the ﬁrst ﬁve days the IPO trades results in an increase in market response to
the discretionary disclosure announcement of 1.5%. Finally we predict and ﬁnd that
ﬁrms releasing earnings during the quiet period are more able to attract analyst coverage
compared to ﬁrms not releasing earnings during the quiet period. The increase in analyst
coverage comes from analysts not afﬁliated with the underwriter of the IPO.

This paper contributes to the literature in several ways. First, the paper tests the
determinants of disclosure and the market response to this disclosure. There are many
theoretical explanations for disclosure, most revolving around information asymmetry
reduction. This paper attempts to determine what causes ﬁrms to release earnings. Also,
it is not clear if market reactions to discretionary disclosure are reactions to the
information contained in the disclosure or reactions to the fact that a disclosure was
made. The old adage of no news is good news is clearly not true in the case of ﬁrm
disclosure. But is it the case that any news is good news or is it simply the case that only
good news gets released? This paper attempts to disentangle these two effects.

The paper also goes to the very start of managerial decision making concerning
information release. As pointed out by Core (2001), studies of discretionary disclosure

are actually equivalent to studies of disclosure policy. This is because an act of

44

discretionary disclosure is a direct function of and a strong proxy for the disclosure policy
of the ﬁrm. But in the current paper, they are one and the same thing. An IPO ﬁrm has
no disclosure policy until it decides to release or not release earnings. As such, the
current research studies the genesis of a ﬁrm’s disclosure policy.

Finally, the paper attempts to separate the endogenous relationship between
analyst coverage and information asymmetry. We study a subset of ﬁrms that have no
analyst coverage not because analysts choose not to cover the ﬁrms, but because SEC
regulation forbids coverage.13 This allows us to better understand the information
conditions under which analysts might be likely to cover ﬁrms.

The remainder of the paper is organized as follows. The next section discusses
the previous literature relating to information disclosure and how it might affect [PO
ﬁrms. Section II develops the hypotheses we will test and provides a framework for our
analysis. We describe the data and do some preliminary analysis in Section 111. Section
IV conducts the major tests of our hypotheses and reports our ﬁndings. In Section V we

discuss our ﬁndings and conclude the paper.
2.2 Background

Managers can disclose information for several reasons but most of these reasons
revolve around the idea that the manager is better informed about the value and prospects
of her ﬁrm than the market. Initial research by Leland and Pyle (1977) and Myers and

Majluf (1984), among others, implies that there is an adverse selection problem for ﬁrms

 

‘3 Technically, the SEC regulation only forbids the coverage of underwriter afﬁliated analysts of the ﬁrm.
However, unafﬁliated analysts typically wait for the end of the quiet period to initiate coverage as well.

45

selling securities. As such, ﬁrms will release information about the value of the ﬁrm to
decrease information asymmetry when they wish to sell new securities, either stock or
bonds. However, these explanations for disclosure do not account for the fact that ﬁrms
release information at times when they are not selling stock or bonds.

Diamond (1985) develops a model that shows that ﬁrms will release information
even in the absence of explicit regulatory requirements to do so. Although ﬁrm disclosure
reduces the incentives to acquire private information, there are two beneﬁts to ﬁrm
disclosure shared by the market participants. First, individual investors no longer need to
expend resources on acquiring information. Second, there is an improvement in risk-
sharing among the market participants when ﬁrms disclose information. These results
imply that investors will want the managers to release information in a discretionary
manner.14

This line of thought is extended by Lang and Lundholrn (1996) to consider the
relationship between information disclosure and analyst coverage. They show that ﬁrms
with more precise disclosure have more analyst coverage. Their results imply that it is
costly for analysts to provide coverage for a ﬁrm and less costly if there is more
information available about the ﬁrm. [But Healey and Palepu (2001) point out that this
relationship is endogenous. Further, the effect of voluntary disclosure on analyst
following is not clear. Analysts may prefer less disclosure in that it allows them to
distribute information directly from managers creating a higher demand for their services.

In the context of equity offerings, two papers discuss the role of disclosure. Lang

and Lundholrn (2000) discuss the effect of ﬁrm disclosures prior to ﬁrms condusting

 

‘4 Healey and Palepu (2001) point out that information is a public good that is paid for by investors but
shared by current and potential investors. This freeriding by potential investors implies that information
will actually be underproduced by the company.

46

 

seasoned equity offerings. They ﬁnd that ﬁrms that increase their disclosure prior to
equity issuance have high stock price run-ups prior to the equity issuance announcement.
However, the market discounts this increase in disclosure and these stocks tend to decline
more than ﬁrms that do not increase their disclosure prior to equity issuance. Schrand
and Verrecchia (2004) look at the disclosure of private ﬁrms prior to their initial public
offerings and ﬁnd that ﬁrms with more disclosure have lower underpricing in the IPO.
This study documents the disclosure practices of a subset of IPOs prior to their going
through the IPO process and argues that the lower underpricing of the IPO is caused by
the lower level of information asymmetry for the ﬁrms releasing more information.

Our work is unique, however, in that it examines the relationship between
disclosure and economically important factors beyond the underpricing of the IPO. In
particular, we are interested in the relationship between disclosure and the market
reaction to this disclosure. We consider the possibility that managers strategically release
news when it beneﬁts their ﬁrm the most and the market reacts to this strategic release of
news appropriately. Consider a ﬁrm whose operating performance is only average
compared to industry peers. This ﬁrm may have few incentives to disclose information
in that the market will correctly anticipate the mediocrity of the ﬁrm’s performance.
However, if the ﬁrm has mediocre performance, but that ﬁrm’s stock or the general stock
market has been going down recently, then even a mediocre ﬁrm may have incentives to
release earnings simply to assure investors that the ﬁrm is performing acceptably well. In
contrast, if a ﬁrm’s operating performance is mediocre, but the ﬁrm’s stock or the general
market have been doing very well recently, then the managers may withhold information

from the market. We also look directly at the relationship between disclosure and the

47

IPO ﬁrm’s ability to attract ﬁnancial analysts. As pointed out by Lang and Lundholm
(1996) information disclosure by the ﬁrm may make it less costly for analysts to cover
the ﬁrm and ultimately attract more analyst coverage. This is important because and
increase in analyst coverage may help to reduce the ﬁrm’s information asymmetry and

ultimately, the cost of capital for the ﬁrm.

2.3 Hypothesis Development

As has already been discussed, most disclosure incentives for managers deal with
the alleviation of information asymmetry problems. The literature has tested in the past if
good news generally comes earlier than bad news (see Begley and Fischer (1998) and
Chambers and Penman (1984) for example). The implications for IPO ﬁrms releasing
earnings during the quiet period is simply that ﬁrms with better operating performance
will release information during the quiet period. This idea leads to our ﬁrst hypothesis

for why ﬁrms release earnings during the quiet period.

Hypothesis 1: Superior Performance Hypothesis: F irms with operating
performance above the market expectations for the ﬁrm will release

earnings information earlier than other ﬁrms.

This hypothesis states that ﬁrms release earnings during the quiet period because

they outperform the general market expectations for that ﬁrm. The only consideration

that managers make in whether to release earnings or delay the release of earnings is the

48

performance of their ﬁrm relative to expectations. Firms that delay the release of
earnings until after the expiration of the quiet period according to the Superior

Performance Hypothesis are ﬁrms that perform poorly.

The Superior Performance Hypothesis has several empirical implications that are
V easily tested. First, if IPO ﬁrms perform well relative to expectations, then they should
be more likely to release earnings during the quiet period. Also, the stock price response
to the news release should be positively related to the operating performance of the ﬁrm:
if the ﬁrm releases more positive information on the release date, the stock price response

should be higher.

However, it is also possible that managers use more than just the subset of
information about their ﬁrm operating performance in making the decision of when to
release earnings. For instance, managers might release earnings information when their
stock performs poorly to signal that their operating performance if not as bad as the
market price implies. Managers may also release earnings when the overall market
performs poorly as a way to show that their ﬁrm is performing better than the general

market. This conjecture motivates our second hypothesis.

Hypothesis II: The Market Timing Hypothesis: Managers consider
information about their own operating performance as well as their own
stock returns, stock volatility, market returns, and market volatility in

determining when to release earnings.

49

 

The Market Timing Hypothesis has some testable implications beyond those of
the Superior Performance Hypothesis. Both hypotheses predict that ﬁrms with higher
operating performance will have a higher probability of releasing earnings during the
quiet period. However, the Market Timing Hypothesis also predicts that quiet period
release will occur for reasons other than high operating performance. For instance, if the
stock performance is very negative or very volatile despite average operating
performance, the ﬁrm will release earnings simply to signal that the ﬁrm is of average
quality. 15 In addition, if the market return is highly negative or very volatility despite
average operating performance, the ﬁrm will release earnings to signal its quality. Thus,
we see that the Market Timing Hypothesis has several implications for the probability of

early earnings release above the Superior Performance Hypothesis.

The Superior Performance Hypothesis implies that the return on the day the
earnings information is released will be related to the Operating performance of the ﬁrm.
However, the Market Timing Hypothesis predicts that there should be a relationship
between the stock response to earnings release and other factors. For instance, if the
stock market in general or the IPO stock in particular have been performing poorly (either
through high volatility or through low returns) then the earnings announcement return

should be larger than if the ﬁrm has not had poor past stock performance.

2.4 Data and Preliminary Analysis

 

'5 There is no reason to expect a ﬁrm with poor stock performance and poor operating performance to
release earnings during the quiet period. In such a situation, it would seem that the market already
anticipates the poor operating performance of the ﬁrm without any information release from management.

50

2.4.1. Data

Our data source for IPOs ﬁom 1993 through 1998 is the Securities Data Company
(SDC) new issues database. We begin our sample period in 1993 so that we have analyst
coverage for the IPOs16 and end our sample in 1998 to avoid any shifts in regime for
important control variables as documented in Loughran and Ritter (2004). Stock return
information is collected from the Center for Research in Security Prices (CRSP). Firms
must meet teh following criteria. The IPOs must have an offer price midpoint of at least
$5 per share. Unit offerings, closed-end funds, Real Estate Investment Trusts (REITs),
partnerships, non-US. operating companies (as deﬁned by CRSP), and American
Depository Receipts (ADRs) are excluded from the sample universe. Our ﬁnal sample
includes 2,075 IPOs.

Firms are considered to release earnings during the quiet period if the IPO
releases earnings information using a non-SEC ﬁling avenue (such as a press release)
during the quiet period. To determine this, we review all articles released in the Dow
Jones newswire (using the Factiva Database) from the IPO date until the end of the quiet
period 25 days later. Information not related to the ﬁrm earnings or earnings information
released on the day of the quarterly ﬁling are not considered as earnings information
released during the quiet period through discretionary means. Of the 2,075 ﬁrms
sampled, 132 released earnings during the quiet period in a press release.

2.4.2 Sample Statistics

Table 2.1 provides a breakdown for ﬁrms that release earnings during the quiet

period. The table is split out into two categories: ﬁrms not releasing earnings during the

quiet period and ﬁrms releasing earnings information during the quiet period. Panel B

 

‘6 I/B/E/S data coverage begins in 1993.

51

 

shows that ﬁrms releasing earnings during the quiet period are not clustered in any
particular year and Panel C shows that the quiet period earnings releasing ﬁrms are not
disproportionately clustered in any particular industry.

In Table 2.2 we report the characteristics of the IPOs that do not release earnings
during the quiet period and compare them to ﬁrms that do release earnings during the
quiet period. We can see, for instance, that quiet period earnings releasing ﬁrms have a
market capitalization of $424 million on average where non-releasers have a market
capitalization of $204 million. We can also see that earnings releasers offer a greater
number of shares in the IPO, have higher ranked underwriters, and have higher offer
prices.

2.4.3 Quiet Period Stock Returns

To determine if the release of earnings information during the quiet period has
any effect on the ﬁrms releasing earnings, we look at the short-term returns in the
marketplace. If earnings release does not elicit a return response in the market, then it
seems unlikely that the release of the earnings is adding any new information to the
market. Table 2.3 Panel A provides evidence that the earnings release is providing new
information to the marketplace. In particular, we see that the two day excess return (IPO
return less the equally weighted index) for the ﬁrms releasing earnings during the quiet
period is a positive and statistically signiﬁcant 2.04%. Figure 2.1 illustrates the
cumulative abnormal returns around the date of the earnings release for ﬁrms releasing
earnings during the quiet period.

In Table 2.3 Panel B we summarize the cumulative abnormal returns (IPO return

less the equally-weighted index) over the whole quiet period (25 calendar days) for non-

52

quiet period releasing ﬁrms and quiet period releasing ﬁrms. We see that in general,
quiet period releasing ﬁrms have an abnormal return of 1.46% whereas non-releasing
ﬁrms have an abnormal return of 0.53%. This difference is not statistically signiﬁcant.
We also look at the cumulative abnormal returns over the ﬁrst ﬁve trading days starting
from the IPO date. We see that the quiet period releasing ﬁrms have a CAR of -0.15%
where the non-releasers have a CAR. of -0.06%. Once again, this difference is not
statistically signiﬁcant.

We also summarize the excess stock volatility for the whole quiet period and over
the ﬁrst ﬁve trading days for non-quiet period releasing ﬁrms and quiet period releasing
ﬁrms. We can see that over the whole quiet period non-quiet period releasing ﬁrms have
an excess return volatility of 3.6% compared to quiet period releasing ﬁrms that have a
volatility of 3.9%. This difference is statistically signiﬁcant at the 10% level. The
volatility of non-quiet period releasing fums over the ﬁrst ﬁve trading days is 3.5%
compared to quiet period releasing ﬁrms that have a volatility of 3.85%. This difference

is statistically signiﬁcant at the 5% level.

2.5 Test of Hypotheses

2.5.1 Probability of Quiet Period Earnings Release

We are interested in determining what causes a manager to release earnings
during the quiet period. The ﬁrst set of tests we conduct consists of looking at the
determinants of quiet period earnings release to try to distinguish between the Superior

Performance and the Market Timing hypotheses. Several of our explanatory variables

53

(Market volatility, Market return, IPO CAR, and IPO volatility) are calculated over the
ﬁrst ﬁve trading days that the stock trades. As such, we eliminate all earning releases
during the quiet period that occur prior to day 6. This eliminates two observations from
our sample.

Since we are interested in what inﬂuences ﬁrms to release earnings during the
quiet period, we use a logistic regression with the dependent variable being a dummy
variable taking a value of one if the ﬁrm releases earnings information during the quiet
period and zero otherwise. Our initial regression contains several explanatory variables
that might be important in determining the probability of releasing information during the
quiet period. We control for IPO underpricing because ﬁrms with higher underpricing
might have more uncertainty and therefore, be more likely to release information during
the quiet period. We control for the proceeds of the IPO, a proxy for ﬁrm size known at
the time of the IPO. Larger ﬁrms are more likely to have the internal controls that allow
them to release earnings information earlier than smaller ﬁrms. We also control for
underwriter rank using a durrrrny variable taking a value of one if the underwriter rank is
9 and zero otherwise. This underwriter ranking system was originally developed by
Carter and Manaster (1990) and has been updated in Appendix 3 of Loughran and Ritter
(2004). The rank of the underwriter might be important because ﬁrms may use a quiet
period earnings release to compensate for not having a highly ranked underwriter to
certify their ﬁrm value.

We control for the revision of the offering (offer price minus the offer range
midpoint divided by the offer range midpoint) because ﬁrms with higher revisions will

have higher demand in the marketplace. Presumably, the market may have lower

54

expectations for information release if the ﬁrm is has more investors interested in buying
the stock. Finally, we control for venture capital backing with a dummy variable taking a
value of one if the IPO is backed by a venture capitalist and zero otherwise. This control
is important as venture capitalists often have substantial controlling interests in IPOs and
may desire the ﬁrm to release earnings in a certain way to help create information
momentum.17

The results for our baseline regression are contained in Table 2.4. This table
shows that the probability of releasing earnings during the quiet period is positively
related to the ﬁrm underpricing. This value is statistically signiﬁcant at the 1% level and
implies that an increase in underpricing of one standard deviation or 26% results in an
increased probability of earnings period release of 1.3%. Note that the unconditional
probability of quiet period earnings release is 6.3% so the effect of underpricing is quite
economically signiﬁcant. This result for underpricing is likely related to information
asymmetry. Firms with higher information asymmetry are more likely to have higher
underpricing and to beneﬁt more from disclosure.

Likewise, we can see that the probability of a quiet period earnings release is
positively related to the size of the offering and statistically signiﬁcant at the 1% level.
This implies that an increase in ﬁrm proceeds by one standard deviation results in an
increase in the probability of releasing earnings of 1.6%. This result is not surprising
either in that larger ﬁrms with better internal controls are likely to have a better control
over the internal process of generating the information for a press release. Thus, they are

more likely to make earnings announcements during the quiet period.

 

'7 See Aggarwal, Krigman, and Womack (2002) for a further discussion of information momentum.

55

We now extend the model to look at other determinants of earnings release during
the quiet period. Model (2) adds the market volatility for the ﬁrst ﬁve days the stock
trades as an explanatory variable in the regression. This variable is the standard deviation
of the value weighted index for the ﬁrst ﬁve days the stock is publicly traded. Note that
the coefﬁcient on market volatility is positive and statistically signiﬁcant. The magnitude
of the variable implies that a 1% increase in the stock market volatility implies a 2.48%
increase in the probability of releasing earnings during the quiet period. This result is not
only statistically signiﬁcant, but is very economically signiﬁcant as well. The positive
relationship between market volatility and the probability of an earnings release is the
ﬁrst evidence we present consistent with the Market Timing Hypothesis. If ﬁrms are
releasing earnings during the quiet period after periods of higher market volatility, then
managers appear to release earnings information to signal their quality despite the high
market volatility.

In Model (3) we look at the cumulative market return for the ﬁrst ﬁve days that
the IPO trades. We ﬁnd that the ﬁrm is less likely to release earnings information if the
market return is positive, although this variable is not statistically signiﬁcant. We then
test a dummy that takes a value of one if the ﬁrm is a technology stock and zero
otherwise. We use the deﬁnition of a technology stock given in Loughran and Ritter
(2004) Appendix 4. We ﬁnd that technology stocks are more likely to release earnings
during the quiet period. Although it initially might appear that high technology stocks
being more inforrnationally opaque are using the quiet period to release more
information, we shall see later that this may simply be due to higher than usual operating

performance for IPO ﬁrms in the technology industry.

56

We then test the IPO stock price cumulative abnormal return and the excess return
volatility for the ﬁrst ﬁve days the stock trades on the open market to determine if these
stock price movements are inﬂuencing managers to release news during the quiet period.
We deﬁne excess return as the stock return less the equally-weighted stock index.
Models (5) and (6) provide results for the IPO stock price movements and show that
negative stock price movements and high stock volatility imply a higher probability of
releasing earnings information during the quiet period. Note that neither of these
coefﬁcients are statistically signiﬁcant.

We then test a major implication of the Superior Performance Hypothesis: ﬁrms
with higher abnormal earnings will be more likely to release earnings during the quiet
period. We measure abnormal operating performance as the IPO operating income in the
quarter that has not yet been reported divided by the assets of the IPO minus the average
operating income divided by assets for all ﬁrms matched to the IPO based on two digit
SIC code. This proxy for abnormal performance is necessary since there are no analyst
projections to compare the earnings performance to. As such, we use the two-digit SIC
code matched industry average income over assets as our benchmark. There is a problem
with this proxy in that it will only be able to detect a certain type of superior earnings
performance. If the IPO ﬁrm performs better than all other ﬁrms in the same industry,
then this measure of abnormal operating performance will accurately show that the ﬁrm
performs well. However, if all the ﬁrms in the industry of the IPO ﬁrm do well
compared to the rest of the market, then this proxy will not detect abnormal performance.
Note that we lose 538 data points when we use the quarterly accounting information from

CRSP in the regression.

57

Model (8) shows that IPO ﬁrms with better abnormal operating performance are
more likely to release earnings during the quiet period. The statistically signiﬁcant
coefﬁcient on abnormal operating performance implies that an increase in abnormal
operating performance of one standard deviation results in a 0.8% increase in the
probability of releasing earnings during the quiet period. This result supports the
Superior Performance hypothesis in that ﬁrms with better performance are more likely to
release earnings during the quiet period.18

Combining all the explanatory variables into one regression shows that the
explanatory power for market volatility and abnormal operating performance are not
reduced when both are taken into consideration. The results of this regression are in
Model (9). This model shows that both the market volatility and the abnormal operating
performance of the ﬁrm have explanatory power for the release of earnings during the
quiet period. Note, however, that the technology dummy variable is no longer
signiﬁcant. This result implies that the technology dummy is being subsumed by the
abnormal performance variable (compare models (7) and (9)). So we have seen that
ﬁrms are more likely to release earnings if markets have recently been volatile or if the
ﬁrm has performed better. Both of these ﬁndings are consistent with the Market Timing
hypothesis but only the latter is consistent with the Superior Performance hypothesis.
2.5.2 Determinants of the Stock Price Response to Earnings Announcements

We then conduct tests on the determinants of the stock price excess return at the
time of the earnings announcement. Once again, we drop two observations for ﬁrms that

release earnings during the ﬁrst ﬁve days that the IPO is trading. This gives us a sample

 

‘8 This ﬁnding is also consistent with the literature that shows that ﬁrms with better results tend to release
information earlier. See, for instance, Begley and Fischer (1998).

58

of 130 ﬁrms that release earnings during the quiet period. The dependent variable in
Table 2.5 is the cumulative abnormal return (IPO return less the equally-weighted index)
for the IPO ﬁom day -5 to day 5 around the earnings announcement date. We use the
same control variables in Table 2.5 as in Table 2.4: ﬁrm underpricing, the log of the IPO
proceeds, a dummy variable for a high underwriter rank, the offer price revision, and a
dummy variable for venture backing. Note that of these variables, only the underpricing
is statistically signiﬁcant. However, the interpretation of this variable is difﬁcult because
underpricing proxies for two difference things. Higher underpricing implies more
uncertainty for the stock issue. But higher underpricing also implies a higher amount of
demand for the stock.

We then check for a relationship between the announcement day return and the
market volatility the ﬁrst ﬁve days the stock is traded, the market return the ﬁrst ﬁve days
the stock is traded, a technology dummy variable, the IPO cumulative abnormal return
the ﬁrst ﬁve days the stock is traded, and the IPO volatility the ﬁrst ﬁve days the stock is
publicly traded. Of these variables, only the IPO volatility is statistically signiﬁcant in its
relationship with the earnings announcement date. The coefﬁcient on the [PO volatility
implies that an increase in stock volatility of 1% for the ﬁrst ﬁve days the stock trades
will result in an increase in the announcement returns of 1.5%. This result is very
economically signiﬁcant considering the average return on the announcement date is
2.0%.

We then include the abnormal operating performance for the IPO as deﬁned
earlier in the regression. The sign on the coefﬁcient for this variable is positive as would

be expected, but is not statistically signiﬁcant. This result tells us that the market does

59

not react more positively to an earnings release if the earnings release is more positive.
At this point, it is worth making a few comments about regressions (8) and (9) in Table
2.5. First, note that there is a loss of 31 data points once we use the quarterly income
variables for these regressions. This might explain why there is no statistical signiﬁcance
for the variable. It is also possible that our proxy for abnormal operating performance is
not as precise as we would like.

Our results thus far have shown that the probability of releasing earnings is
positively related to how good the earnings are and the market volatility just before the
earnings release. However, the stock price reaction is not related to these things, but is
related to the stock volatility for the ﬁrm releasing earnings. This result implies that a
high operating performance value only allows the manager the opportunity to signal to
the market the ﬁrm’s quality. If the operating performance of the ﬁrm is low, then the
managers are not likely to release earnings. However, the market is not particularly
interested in the operating performance of the IPO ﬁrm. The market seems more
concerned with the volatility of the IPO ﬁrm. Under conditions where the IPO ﬁrm has a
high volatility conditional on releasing earnings during the quiet period, the IPO has a
higher abnormal return. This result is unique in that it implies that the market is reacting
to the signal of high earnings by adjusting its expectations relative to ﬁrm stock volatility,
not relative to the earnings information per se. This result is consistent with the Market
Timing hypothesis in that managers release information in response to periods of high
market volatility and high earnings performance, but the market reacts to the ﬁrm

volatility, not to the earnings information itself.

60

 

Note that another possibility is that ﬁrms with higher volatility are faced with
higher information asymmetry. Thus, by releasing earnings for a high volatility ﬁrm, the
management has effectively reduced the information asymmetry more than for a lower
volatility ﬁrm. In either case, the market seems to be more concerned with the ﬁrm
characteristics than the information contained in the earnings release.

2.5.3 Analyst Coverage and Quiet Period Earnings Release

We are interested to see what the beneﬁt of releasing earnings during the quiet
period might be. Research by Aggarwal, Krigman, and Womack (2002) implies that
insiders might be interested in creating information momentum to keep stock prices high
until they can sell at the lockup expiration. They show that ﬁrms are able to attract more
analyst coverage through information momentum. Managers are not concerned with
underpricing in this model because they consider the stock price at the expiration of the
lock-up, not on the ﬁrst day of trading. Such an argument for information momentum
might also be made for earnings information release. Managers have incentives to
establish track record for early and credible disclosure and may be concerned with the
cost of equity for future stock issuances. Johnson and Marietta (2006) show that 28% of
IPO ﬁrms during this sample period go through seasoned equity offerings within two
years of the IPO date. As such, releasing high quality information quickly and
consistently is important for the ﬁrm to minimize its cost of future equity issuance.

We look at analyst coverage because the literature has shown that ﬁrms beneﬁt
from analyst coverage (see Bradley, Jordan, and Ritter (2003)). Further, analyst coverage
is a public piece of information that is easily quantiﬁable unlike the establishment of an

earnings release track record. We use the I/B/E/S database to track the initiation of

61

 

analyst coverage.19 For each ﬁrm, we examine the number of analyst initiations as well
as the strength of the initiation for the ﬁrst 90 days after the IPO.

Table 2.6 provides summary statistics for the analyst recommendations separated
out by ﬁrms releasing earnings during the quiet period and ﬁrms not releasing earnings
during the quiet period. In Panel A we can see that of the 1,943 ﬁrms not releasing
earnings during the quiet period, 702 or 36.1% of the IPOs have no analyst coverage
within 90 days of the IPO. For ﬁrms releasing earnings during the quiet period, 19.7%
(26 out of 132) do not receive any analyst coverage. Looking through the balance of the
table, it appears that there might be a slightly higher incidence of analyst
recommendation for the earnings period releasing ﬁrms compared to the non-earnings
releasing ﬁrms. However, it should be remembered that these ﬁrms are also larger and
have higher underwriter rankings. These variables are correlated with analyst coverage
(see Cliff and Denis (2004)).

Moving to Table 2.6 Panel B. we can see that there were 2,944 analyst
recommendations made for the 1,943 ﬁrms not releasing earnings during the quiet period.
Of the 2,944 recommendations, 1,538 or 52.2% were strong buy recommendations.
There were 286 analyst recommendations for the 132 ﬁrms releasing earnings during the
quiet period and 55.2% of these were a strong buy (158 out of 286). We use the I/B/E/S
practice of coding a 1 as a strong buy, a 2 as a buy, a 3 as a hold, a 4 as a sell, and a 5 as a
strong sell. Table 2.6. Panel C. shows the average recommendation level aggregated over
each ﬁrm receiving recommendations. We see that on average, ﬁrms not releasing

earnings information during the quiet period have a stock recommendation of 1.53 where

 

'9 We also examine the analyst coverage data from Briefmg.com. This data is more complete, but is only
available starting in 1998. For more information concerning analyst coverage data, see Bradley, Jordan,
and Ritter (2003).

62

ﬁrms releasing earnings have a higher recommendation value of 1.48 (recall, a 1 is a
strong buy, a 2 is a buy, etc.). The difference between these two values is small, but is
statistically signiﬁcant at the 10% level.

We are now interested in determining if underwriter/analysts are more likely to
cover a ﬁrm if the ﬁrm releases earnings during the quiet period. Such coverage could be
coordinated by the underwriter prior to the IPO. The underwriter could trade off
coverage for greater information release. Lang and Lundholm (1996) state that analysts
are more likely to provide analyst coverage for ﬁrms that release more information as this
coverage is less costly to the analyst. Underwriter/analysts make decisions about
covering a ﬁrm based on the information available about the ﬁrm and more information
may make it less costly for the underwriter/analyst to provide coverage. However, it is
also possible that the underwriter/analyst has more private information gleaned from the
underwriting relationship than other analysts and does not want more public disclosure.
For these analysts, it is possible that public disclosure would reduce their informational
advantage.

To test this conjecture, we use a logistic regression with the dependent variable
being a dummy variable that takes a value of one if the underwriter has an analyst that
provides coverage of the ﬁrm within 90 days of the IPO and zero otherwise. This
regression has control variables for the backing of the IPO by a venture capitalist,
whether or not the ﬁrm is listed on the Nasdaq, the ranking of the underwriter, the
number of managing underwriters, the size of the ﬁrm, a dummy for a high technology
company, the performance of the ﬁrm (buy and hold return from the IPO day closing

until three days before the quiet period end), and the underpricing of the ﬁrm. These

63

explanatory variables are chosen based on previous work by Cliff and Denis (2004) and
Bradley, Jordan, and Ritter (2003). Of these explanatory variables, only Nasdaq listing
and ﬁrm size have explanatory power for when an underwriter/analyst provides coverage
for the IPO ﬁrm. Table 2.7 Panel A shows these results.

We then include a dummy variable taking a value of one if the ﬁrm releases
earnings during the quiet period. From this regression, we ﬁnd that underwriters are not
more likely to initiate analyst coverage for the ﬁrm if the ﬁrm releases earnings during
the quiet period (see Model (2) in Table 2.7 Panel A). This result implies that
underwriter/analysts do not use the information environment of the ﬁrm to make a
decision about analyst coverage. It may also be that underwriters decide on providing
analyst coverage long beforethe stock ever goes public.

In Table 2.7 Panel B we use the same model to determine if analysts overall use
the information environment to make decisions about coverage. Only now the dependent
variable is a dummy variable taking a value of one if any analyst (not just an underwriter
afﬁliated analyst) provides analyst coverage and zero otherwise. Note that the
determinants of any analyst coverage shown in Model (1) are slightly different than the
determinants of underwriter afﬁliated analysts. For instance, we now see that venture
backing is a statistically signiﬁcant determinant of analyst coverage. This is likely the
case as non-underwriter analysts are at an informational disadvantage to
underwriter/analysts and may get a beneﬁt from outside certiﬁcation by venture
capitalists. Further, the number of managing underwriters is also an important
determinant of non-underwriter afﬁliated analyst coverage. This result is consistent with

Bradley, Jordan, and Ritter (2003).

64

When we include a dummy variable for the ﬁrm disclosing earnings during the
quiet period, we ﬁnd that unlike the underwriter analysts who are not more likely to
provide coverage after a quiet period earnings release, analysts in general are more likely
to provide coverage after a quiet period earnings release. In fact, an earnings release
results in an increase in the probability of analyst coverage of 9.2%. This result is
economically signiﬁcant considering the unconditional probability of analyst coverage is
42.3%. This result is statistically signiﬁcant at the 5% level.

Thus far we have seen that the probability that a ﬁrm is covered by an underwriter
afﬁliated analyst is not affected by quiet period earnings release. However, the
probability that the universe of analysts cover a ﬁrm is increased when the ﬁrm releases
earnings during the quiet period. This result implies that unafﬁliated analysts are more
likely to provide coverage of the ﬁrm if the ﬁrm releases earnings during the quiet period.
But we are interested in more than if a ﬁrm is covered. Bradley, Jordan, and Ritter
(2003) show that multiple analyst initiations result in subsequently higher returns at the
quiet period expiration. Therefore, the quantity of analyst coverage is important in
addition to whether a ﬁrm is covered or not.

To test if the quantity of analyst coverage is related to quiet period earnings

release, we use a multinomial logistic model.20

The dependent variable is a count
variable taking a value of one if there is one analyst initiation, two if there are two analyst
initiations, three if there are three or more analyst initiations, and zero otherwise. For

explanatory variables, we use a dummy variable taking a value of one if the IPO is

backed by a venture capitalist and zero otherwise, a dummy variable taking a value of

 

2° This regression follows the regression contained in Table IX of Bradley, Jordan and Ritter (2003) except
that we use a different time period and omit two variables that they ﬁnd to be insigniﬁcant: the market
share of the underwriter and the turnover of the stock.

65

one if the IPO is Nasdaq listed and zero otherwise, the number of managers for the IPO
offering”, the market capitalization of the IPO, a technology dummy variable, the quiet
period performance as measured by the stock return from the closing of the IPO offering
day to three days prior to the quiet period expiration, and the underpricing of the IPO.

Regression results for the multinomial logistic model are contained in Table 2.8.
We ﬁnd results consistent with Bradley, Jordan, and Ritte'r (2003) for most of the control
variables. We ﬁnd a positive relationship between the number of analysts and whether or
not a ﬁrm is backed by a venture capitalist. There is a positive relationship with the
number of managing underwriters in the IPO syndicate and the number of analysts
covering the ﬁrm. Further, larger ﬁrms and technology ﬁrms are more likely to attract
analyst coverage. Finally, we ﬁnd that ﬁrms with higher underpricing are less likely to
attract analyst coverage. This last result contradicts earlier ﬁndings reported in Bradley,
Jordan, and Ritter (2003). However, it should be pointed out that their results are for a
later time period and their regressions using the time period from 1996-1998 show a
negative (albeit insigniﬁcant) coefﬁcient for underpricing. We therefore propose that
there has been a regime shift in the relationship between underpricing and analyst
coverage. .

Finally, we test if ﬁrms releasing earnings during the quiet period are able to
attract more analyst coverage. We include a dummy variable for quiet period earnings
release in Model (2) of Table 2.8 and ﬁnd that there is a positive and statistically
signiﬁcant relationship between the release of earnings during the quiet period and the
number of analysts covering the ﬁrm. This result implies that quiet period earnings

release is a good way for managers to attract analyst coverage of their ﬁrm.

 

66

2.6 Conclusions

We have shown that a subset of IPO ﬁrms release earnings during the quiet period
through a press release rather than through the normal lOQ or 8K ﬁling. These ﬁrms
have no history of releasing earnings through press releases so the'ﬁrm clearly has ﬁrll
discretion over when and how to release earnings. We test two hypotheses for why ﬁrms
would release earnings during the quiet period. The Superior Performance Hypothesis
states that ﬁrms releasing earnings during the quiet period will be releasing earnings
news that exceeds the expectations of the market. The Market Timing Hypothesis says
that managers consider ﬁrm operating performance when releasing earnings information,
but also consider other factors such as their own stock performance or the overall market
performance.

We show that consistent with the Superior Performance and the Market Timing
Hypothesis, ﬁrms with better operating performance are more likely to release earnings
during the quiet period. The market volatility also has a substantial ability to predict
future earnings information release, consistent with the Market Timing Hypothesis.
When we look at the earnings release stock returns, we see that returns are related to
previous stock volatility, but not earnings performance. This result implies that the
market responds favorably to an earnings release when the past [PO stock performance
has been highly volatile, but not when the abnormal operating performance is better.
This result contradicts the Superior Performance Hypothesis but is consistent with the

Market Timing Hypothesis.

67

Ultimately, this study is interesting because it considers that managers may time
their information release to beneﬁt the ﬁrm the most. Managers appear to release
earnings information when their ﬁrm performs well and when the overall stock market
has a high volatility. The stock price response of the ﬁrm is related not to the operating
performance of the ﬁrm, but to the previous stock price performance of the ﬁrm. Thus, it
appears that managers release earnings strategically and the market knows that'managers
release earnings strategically. For the individual investor this is important because
investors need to know what earnings information release means. It does not simply
mean that ﬁrms have performed well, therefore, the manager wants to let the market
know. It means that managers have a signal to send which is related to not only their
operating performance, but also to the overall market conditions.

One question of importance is why the market would not value higher abnormal
operating performance through higher abnormal returns. It may be that the ex post proxy
we use for abnormal operating performance is not readily available in the market at the
time of information release. This study back ﬁlls the quarterly performance of the IPO
and the industry average quarterly performance to the time the IPO releases earnings.
But obviously, this information is not yet public at the time of disclosure. Thus, it may
be that the market cannot interpret the information release to know how good it is. But
the market does know how volatile the ﬁrm is. Perhaps the market is only able to
ascertain that the fact the ﬁrm is making a disclosure is good news and the higher the
volatility of the ﬁrm, the better the news is at reducing information asymmetry.

Just as importantly, releasing earnings during the quiet period results in ﬁrms

having more ability to attract analyst coverage. Analyst coverage ultimately serves to

68

increase stock liquidity and reduce the cost of capital. As has been stated previously,
28% of the ﬁrms in this sample go through a subsequent equity offering within two years
so the cost of capital for these ﬁrms is a real concern. In addition, we can see clearly that
early information release unambiguously increases the coverage of ﬁrms by analysts.
This is the result of one of two forces. Either analysis is costly and IPO ﬁrms can lower
this cost by releasing more information, or IPOs ﬁrms can attract the attention of more '

analysts simply be disclosing more information.

69

APPENDIX B.

TABLES OF ESSAY 2.

7O

Table 2.1
IPOs by their Release of Earnings during the Quiet Period

The sample is comprised of 2,075 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998. All REITs, unit offerings, closed-end funds, ADRs, ﬁrms not covered by
CRSP, and IPOs with an offer price below $5 are removed from the sample. We use the Factiva database
to identify whether the IPO ﬁrms release earnings information during the ﬁrst 25 days that the ﬁrm is
publicly traded (the quiet period) prior to releasing the information in a quarterly or annual ﬁling. In the
ﬁnal sample, there are 1,943 ﬁrms that do not release earnings information during the quiet period and 132
ﬁrms that do release earnings information during the quiet period.

 

 

Panel A. Total number of IPOs by release of eam'ﬂs during the quiet period 1
Total Number of IPOs 2,075

IPOs not releasing earnings during the quiet period 1,943

IPOs releasing earnings during the quiet period 132

 

 

Panel B. Classiﬁcation of IPOs by year and earningrelease during the quiet period

 

 

Industry No quiet period release Quiet period release
N Percent N Percent
1993 359 18.47 23 17.42
1994 283 14.56 3 2.27
1995 318 16.36 1 0.76
1996 462 23.77 50 37.88
1997 316 16.31 32 24.24
1998 205 10.55 23 17.42
Total 1,943 100 132 100

 

71

Table 2.] (cont)

 

Panel C. Classiﬁcation of IPOs by industry and earnings release during the quiet period

 

 

Industry No quiet period release Quiet period release
N Percent N Percent
Non-durables 79 4.06 4 3.03
Durables 68 3.50 5 3.79
Oil & Gas 36 1.85 1 0.76
Chemicals 99 5.09 3 2.27
Manufacturing 472 24.33 22 16.67
Telecommunications 85 4.37 10 7.58
Utilities 5 0.26 O 0.00
Wholesale Shops 681 35.03 59 44.70
Financial Industry 194 9.98 1 1 8.33
Other 224 11.52 17 12.88
Total 1,943 100 132 100

 

72

Table 2.2
IPOs Sample Characteristics

The sample is comprised of 2,075 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998. All REITs, unit offerings, closed-end funds, ADRs, ﬁrms not
covered by CRSP, and [PCs with an offer price below $5 are removed from the sample. We use the
F activa database to identify whether the IPO ﬁrms release earnings information during the ﬁrst 25 days
that the ﬁrm is publicly traded (the quiet period) prior to releasing the information in a quarterly or
annual ﬁling. In the ﬁnal sample, there are 1,943 ﬁrms that do not release earnings information during
the quiet period and 132 ﬁrms that do release earnings information during the quiet period.
Underwriter rank is based on the ranking system developed by Carter and Manaster (1990) as updated
by Loughran and Ritter (2004). The ranking is based on the investment bank location in the prospectus
with 9 being the highest ranking and 1 being the lowest ranking. Revision is the offer price minus the
midpoint of the offer range divided by the midpoint of the offer range. Percent venture backed is the
percentage of [PCs that are backed by a venture capitalist at the time of the IPO. Underpricing is the
ﬁrst day closing price minus the offer price divided by the offer price. Percent of ﬁrms sued is the
percent of ﬁrms that have class action suits against them as reported in the Security Class Action Alert
newsletter between January 1990 and August 2002. [PO firm age is the number of years from the
founding of the ﬁrms until the IPO date. Previous operating income is the net income in the quarter
that has not yet been released at the date of the initial public offering. Previous total assets is the total
assets of the IPO ﬁrm in the quarter that has not yet been released at the time of the initial public
offering. *", ”, and " denote that the differences are signiﬁcantly different from zero at the 0.01,
0.05, and 0.10 levels, respectively.

73

Table 2.2

 

Average values for IPOs by whether or not the ﬁrm releases earnings information during the quiet

 

 

period
No Quiet Period Release Quiet Period Release Difference
(P-value)
23:11:): qaprtalrzatron ($ 204 424 219...,
(0.00)
Shares offered (millions) 3.63 5.36 -1.74M**"
(0.00)
Underwriter rank 6.95 7.53 -0.58***
(0.01 )
Offer price ($) 12.07 13.27 -1.20***
(0.00)
Range low ($) 1 1 .21 12.02 -0.81***
(0.00)
Range high ($) 12.98 13.97 -0.94***
(0.00)
Revision -0.3% 1 .08% -1.39%
(0.37)
Percent venture backed 37.8% 35.6% 2.15%
(0.26)
Underpricing 15.7% 21 .5% -5.84%***
(0.01)
Percent of ﬁrms sued 3.65% 3.03% 0.06
(0.15)
IPO ﬁrm age (years) 13.3 13.4 -0.15
(0.46)
:lriqpiflpss) operating income ($ 6.05 7.14 1.09
(0.36)
Total assets ($ millions) 312.4 398.4 86.0
(0.37)

 

74

Table 2.3
Market Adjusted Mean Returns Around Earnings Releases

The sample is comprised of 2,075 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998. All REITs, unit offerings, closed-end funds, ADRs, ﬁrms not covered by
CRSP, and IPOs with an offer price below $5 are removed from the sample. We use the Factiva database
to identify whether the IPO ﬁrms release earnings information during the ﬁrst 25 days that the ﬁrm is
publicly traded (the quiet period) prior to releasing the information in a quarterly or annual ﬁling. In the
ﬁnal sample, there are 1,943 ﬁrms that do not release earnings information druing the quiet period and 132
ﬁrms that do release earnings information during the quiet period. CAR for ﬁrst 25 (5) days of trading is
the return of the IPO less the equally-weighted index for the ﬁrst 25 calendar days (5 trading days) that the
IPO is trading (the quiet period). Volatility of excess returns for ﬁrst 25 (5) days of trading is the standard
deviation for the IPO return less the equally-weighted index for the ﬁrst 25 calendar days (5 trading days)
that the IPO is trading. ***, **, and * denote that the values are signiﬁcantly different from zero at the
0.01, 0.05, and 0.10 levels, respectively.

 

Panel A: Excess returns for IPOs releasing eamirgs during the quiet period

 

 

Event window Mean t—test Median Wilcoxon test
Q-value) (P-value)
AD-l 0.95% 2.21*** 0.15% 1.78*
(0.01) (0.08)
AD 1.10% 2.38*** 0.59% 256*“
(0.01) (0.01)
AD+1 0.11% 0.32 -0.22% -0.26
(0.37) (0.79)
AD-l to AD 2.04% 294*" 1.20% 365*"
(0.00) (0.00)
AD-l to AD+1 2.16% 2.55*** 0.97% 2.72***
(0.01) (0.01)
AD-5 to AD+5 1.40% 1.15 2.13% 1.26
(O. 13) (0.21)

 

Panel B. Excess returns for IPOs for the ﬁrst ﬁve and 25 days by whether or not the ﬁrm releases earnings
information during the quietperiod

 

No quiet period earnings Quiet period earnings Difference
release release (P-value)
CAR for ﬁrst 25 days of trading 0.53% 1.46% -0.93%
(0.25)
CAR for ﬁrst 5 days of trading -0.06% -0.15% -0.08
(0.30)
Volatility of excess returns for 3.60% 3.93% -0.35%*
ﬁrst 25 days of trading (0.07)
Volatility of excess returns for 3.51% 3.85% -0.33%**
ﬁrst 5 days of trading (0.04)

 

75

Table 2.4
Logistic Regression Results with Quiet Period Earnings Release as the Dependent
Variable

The sample is comprised of 2,073 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998 that do not release earnings prior to the sixth day of trading. We use the
Factiva database to identify whether the IPO ﬁrms release earnings information during the ﬁrst 25 days that
the ﬁrm is publicly traded (the quiet period) prior to releasing the information in a quarterly or annual
ﬁling. The dependent variable is a dummy variable taking a value of one if the IPO ﬁrm releases earnings
during the quiet period and zero otherwise. Underpricing is the ﬁrst day closing price minus the offer price
divided by the offer price. Log(proceeds) is the logarithm of the dollar value of IPO proceeds. High
underwriter is a dummy variable taking on a value of one if the underwriter has a ranking of nine and zero
otherwise. Revision is the offer price minus the midpoint of the offer range divided by the offer price.
Venture backed is a dummy variable taking a value of one if the IPO is backed by a venture capitalist and
zero otherwise. Log(1+age) is the logarithm of one plus the number of years from founding date to IPO
date. Market volatility is the ﬁve day standard deviation for the value-weighted index starting at the IPO
date. Market return is the ﬁve day return for the value-weighted index starting at the IPO date.
Technology dummy is a dummy variable taking a value of one if the IPO is in the high technology sector.
1P0 CAR is the IPO return less the equally weighted index for the ﬁrst ﬁve days the stock is publicly
traded. 1P0 volatility is the standard deviation for the IPO return less the equally weighted index for the
ﬁrst ﬁve days the IPO is publicly traded. Abnormal operating performance is the IPO operating income in
the quarter that has not yet been reported at the time of the IPO divided by the assets of the IPO minus the
average operating income divided by assets for all ﬁrms matched to the IPO based on two digit SIC code.
Each regression contains dummy variables for the IPO year. White heteroskedasticity consistent standard
errors clustered by year are below the regression coefficients. ***, **, and * denote statistical signiﬁcance
at the 0.01, 0.05, and 0.10 levels, respectively.

76

Table 2.4

 

 

Variable (1) (2) (3) (4) (5)
Underpricing 0558*" 0484*" 0569*" 0533*" 0554*“
(0.150) (0.131) (0.159) (0.168) (0.132)
Log(proceeds) 0296*” 0283*" 0300*” 0312*" 0296*"
(0.082) (0.092) (0.080) (0.081) (0.080)
High Underwriter 0.057 0.048 0.062 0.057 0.056
(0.263) (0.268) (0.270) (0.270) (0.271)
Revision -0.359 -O.283 -0.387 -0.563 -0.357
(0.579) (0.566) (0.541) (0.730) (0.590)
Venture backed 0.042 0.023 0.043 -0.020 0.042
(0.324) (0.334) (0.328) (0.293) (0.323)
Market volatility 29.557“
(15.357)
Market return —8.503
(9.186)
Technology dummy 0486*
(0.298)
IPO CAR -0.731
(5.670)
IPO volatility
Abnormal operating
Performance
Sample Size 2,073 2,073 2,073 2,073 2,073
Pseudo R2 0.0877 0.1008 0.0906 0.0917 0.0877

 

77

Table 2.4 (cont)

 

 

Variable (6) (7) (8) (9)
Underpricing 0.438" 0.371 * 0457* 0.242
(0.239) (0.206) (0.256) (0.292)
Log(proceeds) 0310*” 0311*“ 0.325*** 0318*“
(0.072) (0.083) (0.059) (0.081)
High Underwriter 0.062 0.049 -0.097 -0.119
(0.269) (0.289) (0.199) (0.220)
Revision -0.389 -0.518 -0.727 -0.703
(0.605) (0.693) (0.757) (0.810)
Venture backed 0.020 -0.057 0.048 -0.010
(0.303) (0.293) (0.411) (0.385)
Market volatility 28.258“ 29.946"
(14.420) (15.565)
Market return ~2.744 4.091
(5.366) (4.598)
Technology dummy 0.490” 0.262
(0.232) (0.180)
IPO CAR -2.040 -4.891
(4.091) (4.072)
IPO volatility 3.627 2.583 2.726
(4.177) (3.180) (3.210)
Abnormal operating 0.576“ 0.599"
Performance (0.289) (0.273)
Sample Size 2,073 2,073 1,535 1,535
Pseudo R2 0.0887 0.1058 0.0900 0.1043

 

78

Table 2.5
Multivariate Regression of Earnings Announcement Day Returns for IPO ﬁrms
Releasing Earnings during the Quiet Period

The sample is comprised of 130 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998 that release earnings information during the ﬁrst 25 days that the f'um
publicly trades, but not during the ﬁrst ﬁve days the ﬁrm trades. The dependent variable is the sum of the
IPOs return less the equally weighted index from ﬁve days before the earnings announcement to ﬁve days
after the earnings announcement. Underpricing is the ﬁrst day closing price minus the offer price divided
by the offer price. Log(proceeds) is the logarithm of the dollar value of [PO proceeds. High underwriter
is a dummy variable taking on a value of one if the underwriter has a ranking of nine and zero otherwise.
Revision is the offer price minus the midpoint of the offer range divided by the offer price. Venture backed
is a dummy variable taking a value of one if the IPO is backed by a venture capitalist and zero otherwise.
Log(1+age) is the logarithm of one plus the number of years from founding date to IPO date. Market
volatility is the ﬁve day standard deviation for the value-weighted index starting at the [PO date. Market
return is the ﬁve day return for the value—weighted index starting at the IPO date. Technology dummy is a
dummy variable taking a value of one if the IPO is in the high technology sector. 1P0 CAR is the [PO
return less the equally weighted index for the ﬁrst ﬁve days the stock is publicly traded. [P0 volatility is
the standard deviation for the IPO return less the equally weighted index for the ﬁrst ﬁve days the IPO is
publicly traded. Abnormal operating performance is the IPO operating income in the quarter that has not
yet been reported at the time of the IPO divided by the assets of the IPO minus the average operating
income divided by assets for all ﬁrms matched to the IPO based on two digit SIC code. Each regression
contains dummy variables for the IPO year. White heteroskedasticity consistent standard errors clustered
by year are below the regression coefﬁcients. ***, **, and * denote statistical signiﬁcance at the 0.01,
0.05, and 0.10 levels, respectively.

79

Table 2.5

 

 

Variable Q) (2) (3) (4) (5)
Underpricing -0.062** 0061 ** -0.065** -0.062** -0.056*
(0.017) (0.017) (0.018) (0.019) (0.025)
Log(proceeds) 0.009 0.009 0.01 1 0.007 0.008
(0.006) (0.007) (0.008) (0.007) (0.007)
High Underwriter -0.023 -0.022 -0.023 -0.020 -0.018
(0.024) (0.025) (0.022) (0.023) (0.019)
Revision 0.059 0.061 0.062 0.066 0.058
(0.065) (0.061) (0.065) (0.074) (0.062)
Venture backed 0.029 0.028 0.028 0.032 0.026
(0.025) (0.024) (0.025) (0.021) (0.028)
Market volatility -0.291
(0.863)
Market return 0.672
(0.523)
Technology dummy -0.027
(0.021)
IPO CAR 0.843
(1.075)
IPO volatility
Abnormal Operating
Performance
Sample Size 130 130 130 130 130
Adjusted R2 0.0210 0.0134 0.0222 0.0188 0.0225

 

80

Table 2.5 (cont)

 

 

Variable (6) (7) (8) (9)
Underpricing -0.110*** -0.115*"‘ -0.064*** -0.127**
(0.022) (0.037) (0.014) (0.034)
Log(proceeds) 0.020“ 0.021“ 0.006 0.018”
(0.006) (0.009) (0.009) (0.005)
High Underwriter -0.033 -0.028 -0.017 -0.032
(0.019) (0.018) (0.035) (0.034)
Revision 0.076 0.093 0.051 0.102
(0.059) (0.061) (0.125) (0.120)
Venture backed 0.019 0.022 0.036 0.036
(0.027) (0.024) (0.023) (0.021)
Market volatility -0.640 -0.418
(0.319) (0.509)
Market return 0.547 0.785
(0.492) (0.397)
Technology dummy -0.041* —0.057*
(0.018) (0.025)
IPO CAR 0.097 -0.836
(0.768) (0.872)
IPO volatility 1.499*** 1.593" 1.546“
(0.304) (0.477) (0.658)
Abnormal operating 0.027 0.048
Performance (0.042) (0.093)
Sample Size 130 130 99 99
Adjusted R2 0.0685 0.0596 0.0225 0.0380

 

81

   

Table 2.6
Summary Statistics for Analyst Coverage

The sample is comprised of 2,075 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998. We use the Factiva database to identify whether the IPO ﬁrms release
earnings information during the ﬁrst 25 days that the ﬁrm is publicly traded (the quiet period) prior to
releasing the information in a quarterly or annual ﬁling. We then use the IBES database to determine the
analyst coverage for each ﬁrm for the ﬁrst 90 days that the IPO is publicly traded. Panel A contains
information about the overall distribution of initiations by analysts. There are 1,347 [PCs that receive
analyst initiation and 728 that do not. There are a total of 3,230 analyst recommendations in the sample.
Panel B. gives detailed information about the number of analysts giving strong buy, buy, etc ratings for the
stocks. Panel C. shows the average rating for ﬁrms that do not release earnings during the quiet period and
ﬁrms that do release earnings during the quiet period. Consistent with IBES, we code a strong buy as 1, a
buy as 2, a hold as 3, a sell as 4, and a strong sell as 5. ***, **, and * denote that the differences are
signiﬁcantly different from zero at the 0.01, 0.05, and 0.10 levels, respectively.

 

Panel A. Number Of analysts initiating coverage

 

 

Number of analysts NO quiet period release Quiet period release
N Percent N Percent

0 702 36.1 26 19.7

1 286 14.7 14 10.6

2 51 1 26.3 35 26.5

3 260 13.4 35 26.5

4 l 18 6.1 16 12.1

5 40 2.0 4 3.0

>5 26 1.3 2 1.5

Total 1,943 100 132 100

 

 

Panel B. AnalLst recommendations

 

 

 

Recommendation NO quiet period release Quiet period release
N Percent N Percent
Strong Buy 1,538 52.2 158 55.2
Buy 1,255 42.6 120 42.0
Hold 141 4.8 8 2.8
Sell 2 0.1 0 0
Strong Sell 8 0.3 0 0
Total 2,944 100 286 100
82

Table 2.6 (cont)

 

Panel C. Analyst recommendations

 

 

No quiet period release Quiet period release Difference
(P-value)
Average reconnnendation level 1.53 1.48 0.06*
conditional on analyst coverage (0.06)
N 2,944 286

 

83

Table 2.7
Logistic Regression Results with Analyst Coverage as the Dependent Variable

The sample is comprised of 2,075 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998. In Panel A. the dependent variable is a dummy variable that equals one
if the underwriter provides analyst coverage within the ﬁrst 90 days of trading for the stock and zero
otherwise. In Panel B. the dependent variable is a dummy variable that equals one if any analyst provides
coverage within the ﬁrst 90 days of trading for the stock and zero otherwise. Venture backed is a dummy
variable taking a value of one if the IPO is backed by a venture capitalist and zero otherwise. Nasdaq listed
is a dummy variable taking a value Of one if the issue is NASDAQ listed and zero otherwise. High rank
underwriter takes a value of one if the underwriter has a ranking of 9 and zero otherwise. Number of
managing underwriters is a count variable for the number of managing tunderwriters in the IPO.
Technology dummy is a dummy variable taking a value of one if the IPO is in the high technology sector.
Quiet period performance is the return of the stock from the closing price on the IPO day until three days
prior to the end Of the quiet period. Underpricing is the ﬁrst day closing price minus the Offer price divided
by the Offer price. Quiet period release is a dummy variable taking a value of one if the IPO released
earnings information during the quiet period and zero Otherwise. Each regression contains dummy
variables for the IPO year. White heteroskedasticity consistent standard errors clustered by year are below
the regression coefﬁcients. ***, **, and * denote statistical signiﬁcance at the 0.01, 0.05, and 0.10 levels,
respectively.

 

Panel A. Dependent variable is underwriter analyst coverage

 

 

Variable (l) (2)
Venture backed 0.021 0.022
(0.150) (0.151)
Nasdaq listed 0.771*** 0.769***
(0.295) (0.295)
High rank underwriter -0.297 -0.298
(0.267) (0.268)
Number Of managing underwriters 0.026 0.026
(0.026) (0.026)
Log market capitalization 0.221" 0.220“
(0.112) (0.113)
Technology dummy 0.141 0.139
(0.164) (0.162)
Quiet period performance 0.096 0.097
(0.435) (0.439)
Underpricing -0.663 -0.662
(0.404) (0.403)
Quiet period release 0.059
(0.102)
N 2,075 2,075
Pseudo R2 0.0232 0.0232

 

84

Table 2.7 (cont)

 

Panel B. Dependent variable is analyst coverage

 

 

Variable (l) (2)
Venture backed 0.366*** 0369*”
(0.067) (0.067)
Nasdaq listed 0.173 0.168
(0.190) (0.196)
High rank underwriter -0.145 -0.150
(0.191) (0.186)
Number of managing underwriters 0.063“ 0062*
(0.038) (0.037)
Log market capitalization 0.565*** 0.561 *"
(0.085) (0.084)
Technology dummy —0.107 -0.120
(0.087) (0.088)
Quiet period performance -0.343 -0.341
(0.275) (0.253)
Underpricing -0.498 -0507
(0.371) (0.337)
Quiet period release 0.407"
(0.188)
N 2,075 2,075
Pseudo R2 0.1394 0.1408

 

85

Table 2.8
Multinomial Logistic Regression Results with Analyst Coverage as the Dependent
Variable

The sample is comprised of 2,075 IPOs reported in the Securities Data Corporation (SDC) New Issues
database between 1993 and 1998. The dependent variable is a multilevel variable that equals one if
coverage is initiated by one analyst, two if coverage is initiated by two analysts, three if coverage is
initiated by more than two analysts, and zero otherwise. Venture backed is a dummy variable taking a
value of one if the IPO is backed by a venture capitalist and zero otherwise. Nasdaq listed is a dummy
variable taking a value of one if the issue is NASDAQ listed and zero otherwise. High rank underwriter
takes a value of one if the underwriter has a ranking of 9 and zero otherwise. Number of managing
underwriters is a count variable for the number of managing underwriters in the IPO. Technology dummy
is a dummy variable taking a value of one if the IPO is in the high technology sector. Quiet period
performance is the return of the stock from the closing price on the IPO day until three days prior to the end
of the quiet period. Underpricing is the ﬁrst day closing price minus the offer price divided by the offer
price. Quiet period release is a dummy variable taking a value of one if the IPO released earnings
information during the quiet period and zero otherwise. Each regression contains dummy variables for the
IPO year. White heteroskedasticity consistent standard errors clustered by year are below the regression
coefficients. ***, **, and * denote statistical signiﬁcance at the 0.01, 0.05, and 0.10 levels, respectively.

 

 

Variable (l) (2)
Venture backed 0.374*** 0383*"
(0.081) (0.084)
Nasdaq listed 0.155 0.143
(0.116) (0.126)
High rank underwriter -0.065 -0.073
(0.107) (0.112)
Number Of managing underwriters 0083*" 0081*“
(0.017) (0.016)
Log market capitalization 0.619*** 0.613***
(0.052) (0.052)
Technology dummy 0.106M 0.070
(0.048) (0.063)
Quiet period performance -0.107 -0.088
(0.342) (0.277)
Underpricing -0.525** -0.540**
(0.253) (0.213)
Quiet period release 0815*”
(0.203)
N 2,075 2,075
Pseudo R2 0.1040 0.1045

 

ESSAY 3. THE EFFECT OF NEWS ON VOLATILITY: A STUDY OF IPOS

3.1 Introduction

We analyze the hypothesis that a greater number Of news announcements
contributes to greater return volatility. More news presents investors with a larger
information set, and the processing Of this information set can increase volatility.
Campbell, Lettau, Malkiel, and Xu (2001, hereafter CLMX) were the ﬁrst to show that
average ﬁrm-level volatility more than doubled over the last four decades while market
volatility remained constant. A subsequent paper by Wei and Zhang (2003) provides
evidence that the upward trend in idiosyncratic volatility is strongest for newly listed
ﬁrms. We ﬁnd that the rising trend in volatility for initial public Offerings (IPOs) is
accompanied by a similarly increasing trend in news citations. For example, keeping the
number Of news sources constant, the average IPO in its ﬁrst ﬁve years Of trading had 23
news citations in the 19705, 72 in the 19803, and 202 in the 19905. In contrast, ﬁrms
matched to the IPOs based on size and bOOk-tO-market had an average Of 23 citations in
the 19708, 57 in the 1980s, and 77 in the 19908.

We conﬁrm an increase in idiosyncratic IPO volatility over the period from 1973
through 2003. The increase in volatility is over two times as large for [PCs as for
matching ﬁrms. Next, we form a hand-collected database Of 129,737 public news
disclosure announcements for IPOs and a set Of matching ﬁrms. Analyzing our news
database, we provide information on the distribution Of news-day returns for both

samples. The number Of news citations per IPO has risen sharply over the last three

87

decades, and the reaction to news has become signiﬁcantly larger for IPOs than for
matching ﬁrms. Our ﬁnal contribution presents evidence that, for both [PCs and
matching ﬁrms, the increased idiosyncratic volatility over time is signiﬁcantly related to
an increase in news in recent decades. We conclude that the reason IPOs show the
strongest increase in idiosyncratic volatility over time is driven at least in part by the fact
that they have more news.

There are many reasons to study not only total volatility, but idiosyncratic
volatility as well. CLMX (2001) argue that effective diversiﬁcation depends on
knowledge Of the idiosyncratic volatility of ﬁrms in the portfolio. They also point out
that arbitrageurs who attempt to exploit mispricing in stocks face risks related to
ﬁrm-speciﬁc volatility rather than total market volatility. Our study is the ﬁrst to focus
exclusively on the idiosyncratic volatility Of IP03. We believe this is an important area
Of research, as IPOs appear to show the strongest increase in idiosyncratic volatility over
time. Since IP03 are relatively young ﬁrms without a history Of publicly available
information, the effect of news on IPO volatility could be larger than that for other
(non-IPO) ﬁrms. IPOs are also particularly interesting due to the high level Of apparent
mispricing for these ﬁrms in the marketplace. Research concerning the limitations to
arbitrage, such as short sales constraints, make the study of IPO idiosyncratic risk
particularly important [see Schleifer and Vishny (1997) and Ljungqvist, Nanda, and
Singh (2003)].

First, we conﬁrm the rise in idiosyncratic IPO volatility across decades. For IPOs
issued in the 19703, the average monthly idiosyncratic standard deviation Of returns over

the ﬁrst ﬁve years Of trading is 12%. For IP03 issued in the 19903, the standard

88

 

deviation rises to 21%. A matching ﬁrm sample shows a rise in idiosyncratic standard
deviation from 14% to 18% over the same period, an increase less than half that of IPOs.

To our knowledge, there are only a few papers that try to explain the rise in
idiosyncratic volatility documented by CLMX (2001). While not directly testing these
theories, CLMX (2001) do postulate that such factors as corporate governance,
companies going public earlier in their life cycle, and ﬁnancial innovation may all
contribute to increased ﬁrm volatility. Wei and Zhang (2003) present evidence that
corporate earnings and retum-on-equity (ROE) have become more volatile over time,
especially for newly listed ﬁrms. This is consistent with the ﬁndings Of Fama and French
(2003) that in recent years, newly listed ﬁrms have had lower proﬁts and higher growth
rates than in the past. Wei and Zhang (2003) also provide evidence Of an inverse link
between ROE and idiosyncratic volatility. Xu and Malkiel (2003) show a positive
relation between institutional ownership and ﬁrm volatility, and Pastor and Veronesi
(2003) examine the link between return volatility and ﬁrm proﬁtability. Whenever
possible, we control in our own empirical tests for the variables these studies have found
to impact idiosyncratic volatility.

While the focus of our paper is on idiosyncratic ﬁrm news and volatility, there are
several theoretical and empirical studies Of market news and its effect on market
volatility. For instance, Cutler, Poterba, and Summers (1989) ﬁnd that macroeconomic
news can explain approximately one-third Of the variation in Standard and Poor’s
Composite Index returns. Mitchell and Mulherin (1994) ﬁnd a direct relation between
aggregate market volatility and the occurrence of Dow Jones news stories. Klibanoff,

Lamont, and Wizman (1998) show that closed-end country funds exhibit higher return

89

volatility during weeks with increased country-speciﬁc news. Only Roll (1988) tracks
public, ﬁrm-speciﬁc news events in his test Of the CAPM and APT pricing models.
However, his results indicate that news is not driving a substantial portion of ﬁrm
volatility.

While the literature has demonstrated a link between market news and market
volatility, empirical evidence has not documented a positive relationship between ﬁrm
news and ﬁrm volatility. As discussed in CLMX (2001), cash ﬂow news is less
correlated across ﬁrms implying that idiosyncratic volatility is more likely to be caused
by cash ﬂow news and market volatility is more likely to be caused by discount rate
news. Therefore, the news we examine in this study is likely cash ﬂow as Opposed to
discount rate news [see Vuolteenaho (2002) for a further discussion Of cash ﬂow and
discount rate news]. In a constant discount rate framework, improved information about
future cash ﬂows will result in a decrease in ﬁrm return volatility. This is the result Of
news arriving sooner, when these uncertain cash ﬂows are discounted at a higher rate.
Thus, it is not clear that an increase in ﬁrm-level news should result in increased ﬁrm
volatility. Our results are unique in that they deﬁnitively show that the ﬁrm-speciﬁc
news we document results in an increase in ﬁrm level volatility.

Since our goal is to understand changes in ﬁrm-speciﬁc volatility, we focus on
ﬁrm-level news as measured by the number Of cites in the Factiva® database of newswire
and newspaper publications. With this method, we avoid choosing speciﬁc
announcements that ex-post have been shown tO inﬂuence returns. Recent papers use the
number of ﬁrm news cites in either Factiva or Lexis/Nexus as a measure of news ﬂow,

although they do not study the link between news and volatility [See Chan(2003),

90

Demers and Lewellen (2003), and Reese (2003)]. For each year, we randomly select
10% of the ﬁrms that go through IP03. We collect the number Of Factiva cites for the
IPO subsample and a matching ﬁrm sample over the ﬁrst ﬁve years that the IPO trades.
Across decades, the number of cites over the ﬁrst ﬁve years of trading increases for both
IPOs and the matching ﬁrm sample. The number of news cites is approximately equal
for IPOs and matching ﬁrms in the 19703, but in the 19903 the IPOs have almost three
times as many news citations as the matching ﬁrms.

Next we turn to investor reaction on news days. We build on prior research in the
news area in that not only dO we calculate the number Of news announcements for each
IPO, we also record every date that a news report is published in the ﬁrst ﬁve years Of
trading. This allows us to track investor reaction to news announcements over time, and
we believe we are the ﬁrst to create such a comprehensive database Of news affecting
IPOs. The reaction to news is more pronounced for IPO ﬁrms compared to matching
ﬁrms. In keeping with Chan (2003) and Engle and Ng (1993) we deﬁne “good” news
days as days with F activa cites and positive abnormal daily returns. The average return
on good news days increases signiﬁcantly for IPOs from 3% in the 19703 to 5% in the
19903. The average return on bad news days decreases for [PCs from -2% in the 19703
to -4% in the 19903. The matching ﬁrm reaction on good and bad news days remains
relatively ﬂat across decades. The standard deviation of returns on news days also
[increases over decades, with IPOs again showing a larger increase than for matching
ﬁrms.

We show in a regression framework that the number Of news hits each month has

a signiﬁcant impact on monthly IPO and matching ﬁrm idiosyncratic volatility. An

91

increase Of just one extra news day each month leads to an additional 0.6% in monthly
standard deviation Of returns for IPO and matching ﬁrms, or 2.1% per year. With
average monthly volatility in the range of 17% to 19% for [PCs and matching ﬁrms, a
0.6% addition to monthly volatility from just one extra news day is also economically
signiﬁcant. These regressions tell us IPOs show more volatility than non-IPOs in part
because they have more news. In each regression, we make sure to control for several
variables that are known contributors to future volatility such as size and lagged
volatility. We also ﬁnd that the increase in idiosyncratic volatility for both IPOs and
matching ﬁrms is Granger-caused by the increase in news.

Finally, we look at exogenous events that would cause an increase in the amount
of news that ﬁrms release. Utilizing exogenous events affecting news allows us to
deﬁnitively determine causality for the news-volatility relationship. We document that
these exogenous shocks to news correspond to an increase in both news and volatility.
Such a link allows us tO rule out the cases Of volatility driving news or a third common
factor driving both news and volatility.

The remainder of the paper is organized as follows. Section I describes our IPO
ﬁrm sample and the technique for ﬁnding matching ﬁrms. Section II docmnents the
increase in idiosyncratic [PO volatility over time. Section IH examines the relation
between news ﬂow and volatility for both IPO and matching ﬁrms. Section IV
investigates the robustness of our F activa sample. We include a discussion of exogenous
shocks to news in Section V. In the paper’s last section, we Offer conclusions and

implications Of our ﬁndings.

92

3.2 IPO and Matching Firm Sample

Our data source for IPOs from 1973 through 1998 is the Securities Data Company
(SDC) new issues database. Stock return information is collected from the Center for
Research in Security Prices (CRSP). Firms must meet criteria as follows. The IPOs must
have an Offer price of at least $5 per share. Unit offerings, closed-end frmds, Real Estate
Investment Trusts (REITs), partnerships, non-U.S. Operating companies (as deﬁned by
CRSP), utilities (three-digit SIC codes 491-494), and American Depository Receipts
(ADRs) are excluded from the sample universe. Our ﬁnal sample includes 5,955 IPOs.

TO form a matching sample for the IPO ﬁrms, we select a methodology that will
match on sample ﬁrm characteristics while avoiding look-ahead biases and/or
contamination problems. Matching ﬁrms must be listed on CRSP for at least ﬁve years.
Each year from 1972 to 1998, we form size quintiles according to the June market
capitalizations of ﬁrms listed on the New York Stock Exchange (NYSE). On a yearly
basis, bOOk-tO-market (B/M) quintiles using NYSE, Amex, and Nasdaq ﬁrms are then
formed inside each size quintile. The book values are taken from Compustat as Of
December Of the preceding year.

For every IPO in our sample, we select a matching ﬁrm with the closest size and
BM values in the (July to June) year of the IPO. If no B/M value is available for the new
issue ﬁrm, we choose the matching ﬁrm on the basis Of size only.22 For IPO ﬁrms
without a reported book value in Compustat, we use the post-Offering book value if it is

available from SDC.

 

22 Due to missing book values, 836 (of 5,955) lPOs are matched only on the basis of size.

93

A matching ﬁrm may be chosen only once every year. The matching ﬁrm’s
volatility is calculated over the same time period as the particular IPO ﬁrm. If the
matching ﬁrm does not have the full post-measurement period of returns, another
matching ﬁrm is spliced in to continue where the ﬁrst matching ﬁrm’s returns stop.
About 71% of the IPO sample needs only a single matching ﬁrm, while about 25% needs

two matching ﬁrms.

3.3 IPO volatility

We calculate the monthly idiosyncratic volatility of IPOs and their matching ﬁrms
over the ﬁrst ﬁve years that the [PCs publicly trade. Each month, we calculate the
volatilities for all ﬁrms that are within ﬁve years Of their IPO date. Separately, we also
calculate volatilities Of ﬁrms that are matched with each IPO. We follow CLMX (2001)
for our calculation procedures. We start measuring volatilities in the ﬁrst full month after
the Offering date. Using daily returns from CRSP, we calculate the monthly idiosyncratic
volatility Of a ﬁrm as the sum Of the daily squared differences between the ﬁrm returns

and the value-weighted index returns. Thus, monthly ﬁrm idiosyncratic volatility is:

 

T
2
Umonthly : J2 (Rﬁrm _ RVWindex) (1)

i=1
where T is the number of days in the month, Ram, is the one-day ﬁrm return and RVWindex
is the one-day value-weighted index return. The volatility average is then calculated two
ways: equal weighting and value weighting. The equally weighted volatilities are

averaged across all months and all ﬁrms. The value-weighted volatilities are found by

94

value weighting across all sample ﬁrms within each month and then averaging equally
across all months. We also calculate the equally weighted (EW) and value-weighted
(VW) monthly index volatilities by summing the squared daily returns over the month.

Thus, the index volatility becomes:

 

T
amonthly : Z (Rindex )2 (2)
(:1

where T is the number Of days in the month and Rindex is the one-day return for either the
value-weighted or equally weighted CRSP stock index.

Table 3.1 contains the monthly idiosyncratic volatility results for the IPOs, the
matching ﬁrms, the EW index, and the VW index. This table shows that there has been a
dramatic increase in idiosyncratic volatility for IPO ﬁrms over the last few decades,
rising from 12% in the 19703 to 21% in the 19903 for the equal-weighted measure.
Although ﬁrms matched to the [PCs based on book-tO-market and size also show an
increase in volatility, the difference is much smaller (only rising from 14% tO 18%). The
value-weighted volatilities are lower than the equal weighted measures, but the increase
for IPOs is still almost three times as large as that for matching ﬁrms. Consistent with
CLMX (2001), the value-weighted and equally weighted indices are essentially ﬂat
across decades. Figure 3.1 shows the idiosyncratic ﬁrm volatility over time for IPOs and
matching ﬁrms.

Also contained in Table 3.1 is the pooled daily return kurtosis for the IPOs, the

23

matching ﬁrms, the VW index, and the EW index. The reported kurtosis ﬁgures are

 

23 The VW index and EW index values do not contain returns for October l4-October 26, 1987. With all
daily returns for the 19803 decade, the VW index kurtosis goes from 6.4 to 48.8 and the EW index kurtosis
goes from 11.1 to 40.7. The daily returns around the date of October 19, 1987 have a limited affect on the
kurtosis Of the IPOs and matching ﬁrms.

95

ﬁ'om equally weighted returns across all ﬁrms and all months. Note that like volatility,
the kurtosis Of daily returns for IPOs and matching ﬁrms is monotonically increasing
across decades. For IPOs, the kurtosis goes from 45 in the 19703 to 125 in the 19803 to
360 in the 19903. The fourth moment of returns is important because it allows us to
discriminate between the same news being broken into smaller pieces and a
fundamentally different kind Of news. The monotonically increasing kurtosis values
given in Table 3.1 contradict the idea that the same amount Of news is released across

decades but the news in the 19903 is released in smaller pieces.

3.4 News Flow and Volatility

3. 4.1 News measurement

We wish to study the effect Of news on the returns and volatility Of ﬁrms, and thus
we require a method to track news ﬂows. We use the F activa database of newspaper,
newswire, and periodical publications from Dow Jones & Reuters to ﬁnd when
information becomes available to investors. The overall spirit Of our news collection
using Factiva which we describe below is similar to Chan (2003), Demers and Lewellen
(2003), and Reese (2003).

Rather than searching over all sources available in the Factiva database, we
choose newspaper and newswire sources we consider most poignant to the vast majority
of investors. We also wish to keep our number of news sources as constant as possible
across decades. Our selection criteria are as follows: the top six US. newspapers by

circulation with Factiva listings going back to at least 1987, the two newswires with

96

longest data availability, the top two Canadian newspapers by circulation and the top
British business newspaper by circulation. Table 3.2 contains the news sources we utilize
along with their date Of availability in the Factiva database.

In our collection process, we record a dummy variable that takes on a value of one
if the citation comes from the New York Times Abstracts, two if the source is The Wall
Street Journal or the Dow Jones Newswire, and zero otherwise. These are the US. news
sources with the longest data availability. This allows us to later conduct robustness
checks using news sources available over the whole time period versus our extended
database where some sources are not available until the early to mid-19803.

Our IPO sample consists of 5,955 IP03 from 1973 tO 1998. Collecting the date Of
every IPO news article over the ﬁrst ﬁve years Of trading is very labor intensive. TO
make the process more manageable, we randomly select 10% of the ﬁrms that went
through IP03 in each year from 1973 through 1998 along with their matching ﬁrms.24
When multiple matching ﬁrms are used over the ﬁve—year period, the Factiva search is
spliced the same way the returns would be spliced in for the new matching ﬁrm. This
provides us with data for a total Of 609 IP03 and 818 matching ﬁrms. We have a very
large sample Of daily news retums consisting of 88,577 total cites for IPO ﬁrms and
41,160 total cites for matching ﬁrms. 25

For our search in Factiva, we use the company name as given by the CRSP

database. For IP03, the search is done from the IPO date until ﬁve years after the IPO

 

2‘ We see no signiﬁcant differences in monthly volatility for our Factiva sample of IPOs versus the IP03
for which we do not collect news. Our IPO news sample has a monthly volatility of 19.0%, the same as the
IPO sample with no news collected. For the corresponding matching ﬁrms, we calculate volatilities of
16.2% for those ﬁrms matched to the news IP03, and 16.9% for the ﬁrms matched to the IP03 without
news collected.

25 These results include redundant hits, i.e., citations on the same day in multiple news sources. Some ﬁrms
had as many as 36 hits in various publications on a single day.

97

 

date plus the next full calendar month or until the stock is delisted. For matching ﬁrms,
the search is performed until another matching ﬁrm is chosen, until the IPO ﬁrm is
delisted, or until ﬁve years after the IPO date plus one calendar month. For every citation
in Factiva, the date Of the citation and a ﬂag for its source is recorded.

As an example of the data we collect, Bear Creek Corporation has 19 days with
hits from its IPO in 1976 through our ﬁve-year measurement period ending in 1981.
Roadhouse Grill has 55 days with hits ﬁom its IPO in 1996 until 2001. In the 19703,
only 1.5% of trading days in the ﬁrst ﬁve years have news hits; this percentage increases
to 4.4% in the 19803 and 8.4% in the 19903.

Table 3.3 contains the results of our data collection. The table contains the
number of citations (including multiple hits on each day) and the number of news days
(with multiple hits purged from the sample).26 The percentage Of news days with
multiple news citations goes from 6% in the 19703 to 18% in the 19803 to 46% in the
19903. From this point on in the paper, when we refer to news citations, we are using this
term synonymously with days on which there is news released about a ﬁrm. Paring our
sample down in this way reduces our data set to 53,166 news citations for the IPOs and
30,500 news citatiOns for the matching ﬁrms.

Several comments can be made about news citations for [PCs and matching ﬁrms
based on Table 3.3. First, it is clear that both in terms Of raw number Of hits and in terms
Of non-redundant citations, the IP03 have increasing news coverage across decades.

IP03 go from 19 news days to 105 news days over ﬁve years in the 19703 and the 19903,

 

26 Three ﬁrms account for 14.4% of the total number of hits. All were ﬁrms that went public in the late
19903. Amazon.com has 5176 hits, CIENA has 4467 hits, and Verisign has 3116 hits. Excluding the hits
for these three companies, the 1990-1998 mean hits per ﬁrm is 166.9 and the 1973-1998 mean hits per ﬁrm
is 125.5.

98

respectively. This trend is only marginally shared by the matching ﬁrms. It is also clear
that the IP03 are receiving more news coverage in all time periods when compared tO
ﬁrms matched on size and book-to-market. For instance, in the 19903, IP03 have 105
news days over ﬁve years whereas matching ﬁrms have 52 news days.

Once news dates are generated for all the ﬁrms in our sample, returns on the news
dates and nO-news dates are Obtained from the CRSP database.27 Following Chan (2003)
and Engle and Ng (1993) we then categorize returns as good or bad according to whether
the daily return was positive or negative on the news day. We also separate returns
across decades. The results Of this tabulation are given in Table 3.4 Panel A.

The table shows the dramatic change in news day return distributions, particularly
for IPOs. Going from the decade of the 19703 to the 19903, we see a change in mean IPO
news day returns from 0.15% to 0.39%. The returns on nO-news days show no clear
trend for either IP03 or matching ﬁrms. Separating out good news from bad news, the
average returns on good news days are increasing while the average returns on bad news
days are decreasing. A piece of good IPO news in the 19703 produces a 2.7% daily
return while that return grows tO 4.9% for a piece of good news in the 19903. Similarly,
bad news returns start at -2.3% for IP03 in the 19703 and drops to -4.2% in the 19903.28
The daily good news returns and bad news returns for matching ﬁrms remain relatively

ﬂat across time.

 

27 We do not include the returns for the stock on the ﬁrst day of trading.

28 It is possible that a news item could be reported by the Associated Press newswire aﬁer 4:00 pm. on one
day, and the same story reported in a print newspaper the next day. In our methodology, we would record
this as two news days instead of one. However, this type of event would actually bias against our results.
We would record the stock return associated with the Associated Press newswire as a valid “news day”
return. The market would not yet have been able to process this information though, given a 4:00 pm.
close. Additionally, one would need to argue that this type of event has a time trend — i.e.) it occurs more
frequently in the 19903 than in the 19803.

99

We also see in Table 3.4 that the standard deviation of daily returns increases
across decades. This is true for both news and non-news days. For example, the standard
deviation Of IPO daily returns increases from 4.1% in the 19703 to 8.1% in the 19903.
The matching ﬁrms have relatively stable values across the 19703 and 19903, with a
lower standard deviation in the 19803.

It is also important to note that there is no increase in the percentage Of bad news.
Although the number Of events is increasing substantially over time, the news spread is
consistently about half good news and half bad news. Thus, it is clear that the increase in
volatility is not coming as a result Of more bad news being released in recent decades.

Panel B of Table 3.4 includes an analysis showing the signiﬁcance Of the changes
in news day returns and volatilities. TO control for the fact that both the return means and
standard deviations are changing across decades, we use a regression framework to test
for statistically signiﬁcant differences across decades. The base year is the 19803
allowing us to test for changes from the 19703 to the 19803 and from the 19803 to the
19903. We regress daily returns onto dummy variables for time, whether or not the frrrn
is an IPO, and the requisite cross dummy variables. The regression equation for
regression (I) in the table is as follows:

Return, = by + b; 19703 Dum + b2 19903 Dum + b3 IPO Dum+
b4 IPO Dum * 19703 Dum +b5 IPO Dum * 19903 Dum + e, (3)
All t-statistics are heteroskedasticity-consistent and are given in parentheses.

These results are shown in the ﬁrst row Of Table 3.4 Panel B. Note that the news

day returns are approximately 0.3% across all time periods with no difference between

IP03 and matching ﬁrms. We ﬁnd no signiﬁcant difference between 19703 news day

100

returns and 1980 news day returns but we ﬁnd that news day returns became more
positive (by 0.2%) in the 19903 for all ﬁrms. The fact that an IPO was issued in any
particular decade has no signiﬁcance for the news day returns.

Once coefﬁcients from this ﬁrst regression are available, we can use them to
calculate a set of squared residuals. These squared residuals are then regressed onto the

same variables per the equation below used in regression (II).

0’, = b0 + 6, 1970s Dum + b219903 Dum + b3 1P0 Dum +

b4 IPO Dum * 19703 Dum +b5 IPO Dum * 19905 Dum +e, (4)

The second regression allows us to test which variables are driving increases in
volatility on news days. The volatility going from the 19703 to the 19803 does not
change signiﬁcantly. However, moving from the 19803 to the 19903 causes a doubling Of
the volatility on news days. The fact that news is about an IPO versus a matching ﬁrm
also implies a higher volatility.

An interesting ﬁnding from regression (11) deals with the cross effects dummy for
the 19703 Dum * IPO Dum. The fact that this particular coefﬁcient is negative and
statistically signiﬁcant indicates that going from the 19703 to the 19803, IP03 became
much more volatile on news days. This is consistent with the proposition by Fama and
French (2003) and CLMX (2001) that IP03 are going public earlier in their lifecycles
resulting in higher volatility. If cash ﬂows are pushed out farther into the future for ﬁrms
listing in the 19803 compared to ﬁrms listing in the 19703, then for equivalent news

releases, the news day volatility will be much higher in the 19803. However, regression

101

(II) also shows clearly that there is no change in news day volatility for IP03 when
moving from the 19803 to the 19903. These results are consistent with ﬁrms going public
earlier in their lifecycles in the 19803 and 19903 when compared to the 19703, but
inconsistent with a shift from the 19803 to the 19903. The positive and highly signiﬁcant
19903 Dum coefﬁcient implies that all ﬁrms were becoming more volatile in the 19903,
not only IP03. Thus we see that while IP03 going public earlier in their lifecycle might
have caused some of the increase in volatility going from the 19703 to the 19803, any
volatility increase thereafter appears to be affecting both IP03 and non-lPOs.

Figure 3.2 shows the news day return distributions of IP03 for the 19703, 19803,
and 19903. The distributions appear to be a mean preserving spread across decades; the
increase in volatility is easily observable from this ﬁgure.

3.4.2 Estimating the relationship between news and idiosyncratic volatility

Now that we have documented the rise in news citations and the rise in volatility
over the last few decades, we turn our focus to estimating the relationship between
publicly reported news and idiosyncratic volatility. First, we identify variables known to
inﬂuence volatility. We control for lagged volatility, because volatility is known to be an
autocorrelated series. Based on arguments in Duffee (1995) that risk and return are
contemporaneously correlated, we also include a monthly return variable. Wei and
Zhang (2003) and Pastor and Veronesi (2003) provide evidence that ﬁrm proﬁtability or
return on equity (ROE) has a signiﬁcant inﬂuence on volatility. In addition to controlling
for ROE, we also include variables for leverage, size, and book-to-market (B/M). In

theory, more highly levered ﬁrms should have a higher probability of ﬁnancial distress,

102

which could lead to more volatile returns. Size and book-to-market have been identiﬁed
by F ama and French (1992) as signiﬁcant factors affecting expected returns.

Fama and French (2003) conjecture that an increasing dispersion in proﬁtability
and growth rates for IPOs may explain the increase in idiosyncratic volatility documented
by CLMX (2001). Their proxy for proﬁtability is return on assets, and their proxy for
growth is change in assets. Most IP03 have little debt, so we believe our use of ROE
would pick up proﬁtability in essentially the same manner as return on assets. Also,
many IPOs have few tangible assets. As such, using a variable such as B/M may be a
better proxy for growth than change in assets.

Since our focus is on IPO volatility, we have also included a control variable that
may be important for IP03. We use the technology and Internet industry classiﬁcations
from Appendix 4 of Loughran and Ritter (2004) to form a tech dummy variable. This
variable takes on a value one (zero otherwise) if the IPO or matching ﬁrm is classiﬁed as
a tech or intemet stock at the time of its IPO.

All accounting data are pulled from the annual Compustat database. The previous
ﬁscal year’s ﬁnancial statements are matched with the current month’s return, allowing
for a six-month window at the end of each ﬁrm’s ﬁscal year. This ensures that all data
are publicly available to investors.

The variables are formed as follows. We continue to use the monthly measure of
idiosyncratic volatility (St. Devi) formed from daily returns as in CLMX (2001). We use
the equal weighted measure in order to explore cross-sectional implications that would
not be possible with the value-weighted measure. We also cumulate daily excess returns

to form a monthly retum measure (Retumt). The monthly market capitalization variable

103

(Mkt. Cap“) uses the previous month’s market value (price times shares outstanding).
We cumulate the number of days with news citations each month to form a monthly hits
variable (Hitst). Leverage is deﬁned as the ratio of last ﬁscal year’s long-term debt to last
ﬁscal year’s total assets (LevH). ROE is the prior year’s net income divided by the prior
year’s book value (ROEH). Book-to-market uses the prior year’s book value and divides
by the prior month’s market value (B/MH).29

Table 3.5 provides summary statistics of the control variables for both IP03 and
matching ﬁrms. These are monthly averages across the ﬁrst ﬁve years that the IPO
trades. Panel A shows that the size of the IP03 ($307 million) is slightly larger than that
of matching ﬁrms ($291 million). Our size and book-to-market matching ﬁrm technique
matches ﬁrms only once — at the date of the IPO. Since the table shows an average of our
monthly variables across the ﬁrst ﬁve years the IPO trades, it is possible that our size and
book-to-market variables will differ over that period. IPOs show a book-to-market ratio
of 0.67 with matching ﬁrms tilted slightly more towards a value-type ratio at 1.18. More
IPOs come from the tech industry than do the matching ﬁrms (14% versus 8%). Finally,
the leverage ratios are approximately the same for each sample at about 16%.

Panel B of Table 3.5 shows correlations among the variables we use in our
regression analysis. Note that Hits] and St. Dev] are signiﬁcantly positively correlated
with a correlation coefﬁcient of 0.06. Consistent with Chan (2003), the relationship
between ﬁrm size and the number of news hits is positive and highly signiﬁcant at 0.32.30

Lagged standard deviation, St. Dev“, is very highly correlated with the current standard

 

29 The precise Compustat variables used are: data 60 for book value, data 6 for total assets, data 9 for
long-term debt, and data 172 for net income per share.
30 Chan (2003) reports a correlation of 0.37.

104

deviation with a coefﬁcient of 0.65. Lagged size is signiﬁcantly negatively correlated
with St. Dev,.

Figure 3.3 documents the simple bivariate relationship between news and
volatility. We combine IP03 and matching ﬁrms into groups based on the average
number of days with hits per month. For IP03, the average number of days with hits per
month is 1.4. For matching ﬁnns this number drops to 0.84. We see a strong upward
trend in monthly standard deviation as the number of days with hits increases.

The correlation matrix and Figure 3.3 provide information on bivariate relations,
but in order to control for the effect of all our variables on volatility, we move to a
regression framework. We conduct a pooled regression using all months of data and all

ﬁrms from 1973 to 2003 in the following format:

St. Dev, = b0 + b, Hits, + b2 IPO Dum + b 3 Hits, * IPO Dum + b., Ln (Mkt. Cap),-, +
b5 St. Dev,-, + b6 Return, + byROE,-1 + by Lev,-, + b9 Ln(b/m),-1 + big Tech Dum +

b1119808Dum + bu 19903 Dum + e, (5)

We take the natural logarithm of both size and B/M to help reduce the effect of
heteroskedasticity. We include both IPO and matching ﬁrms in the same regression and
use an IPO dummy variable to capture differences between the two samples. We include
an interaction term that will help us determine the differential effects of news hits on
volatility for IPOs versus matching ﬁrms.

Columns two and three of Table 3.6 provide our regression results using all

months of data ﬁom 1973 to 2003. Note the signiﬁcantly positive coefﬁcient on the IPO

105

dummy variable of 0.96 (t-statistic of 8.47). This indicates that IPOs have an additional
0.96% contribution to monthly idiosyncratic volatility over matching ﬁrms, even after
controlling for the other factors in the regression. The Hits, variable is signiﬁcant with a
coefﬁcient of 0.6 (t-statistic of 13.04). An increase of one news day per month adds
0.6% to monthly idiosyncratic standard deviation for both IPO and matching ﬁrms, or
approximately 2.1% per year. With the average monthly volatility range between 17%
and 19% for matching ﬁrms and IP03, an additional 0.6% contribution to volatility from
just one extra news day per month is economically signiﬁcant as well. The Hits, * IPO
Dum variable is not signiﬁcantly different from zero. Thus, it appears that news items
affect both IPO and non-IPO volatility in a similar manner. The evidence reveals that the
increase in idiosyncratic volatility can be partially explained by the fact that there has
simply been an increase in news over the last few decades. [PCs are more volatile than
matching ﬁrms in part because they have experienced greater levels of news in recent
years.

The time dummies are signiﬁcant for both the 19803 and the 19903, indicating
there are unknown factors inﬂuencing idiosyncratic volatility over time other than the
factors for which we currently control. The other variables in the regression show the
signs that we would expect. Larger ﬁrms have lower monthly volatility, and lagged
volatility has a signiﬁcantly positive effect on current monthly volatility. Leverage does
not signiﬁcantly impact idiosyncratic volatility. While the impact of R0E,-1 is
statistically signiﬁcant (t-statistic of -3.79), it does not appear to be economically

signiﬁcant with a coefﬁcient of -0.00. Our results in this respect are inconsistent with

106

 

Wei and Zhang (2003) and Pastor and Veronesi (2003). Firms in the tech industry do
have a signiﬁcantly higher monthly volatility than those in non—tech industries.

As a robustness check, we split the time periods into pre- and post-1990 and run
the same regressions (but without time dummies). The results are also reported in Table
3.6. The IPO dummy is again positive and statistically signiﬁcant, indicating higher
average idiosyncratic volatility for IP03. Note that the hits variable has a positive and
signiﬁcant impact on volatility for both IP03 and matching ﬁrms in both regressions. No
other qualitative differences emerge from these regressions versus the results from the
entire time period.

We also use the natural logarithm of the number of hits [as in Mitchell and
Mulherin (1994) and Reese (2003)] and see no difference in results (not tabulated).
Finally, we winsorize ROE at the 1St and 99th percentile and also book-to-market at the
99th percentile. This action produces no change in inference (not tabulated).

If we make the assumption that return distributions are independent across days,
we can calculate the expected monthly volatility (as measured by variance) as the sum of
the daily volatilities within a month. We know that the average number of news days
within a month for our sample is one. Thus, taking the variance values for news and
non-news days for 1973-1998 from Table 3.4 and the fact that on average we have 20
trading days per month and one news day, we can calculate the monthly expected

volatility for the average ﬁrm as:

 

Monthly Std = [zl‘fmsmz + ijwssmﬁcw, (6)

no news

The average monthly IPO volatility calculated in this manner from 1973 to 1998 is

22.8%. Note that this value is substantially above the average value of 19.2% given for

107

monthly volatility given in Table 3.1. This is symptomatic of the fact that return
distributions are not independent across days. If we wish to determine the effect of
increasing the number of hits per month from 1 to 2, we can duplicate the calculations
and ﬁnd that the monthly volatility goes from 22.8% to 23.6%. This increase of 0.8% per
month corresponds well to the 0.6% reported in Table 3.6.
3.4.3 Granger Causality

To this point, we have identiﬁed a contemporaneous relationship between the
number of days with news citations each month and the corresponding idiosyncratic ﬁrm
volatility. This tells us nothing about causality, however. It may be the case that news
publications tend to write more about volatile stocks than less volatile stocks. Similar to
CLMX (2001) and Xu and Malkiel (2003), we conduct Granger causality tests to
determine the direction of causality between monthly hits and monthly idiosyncratic
volatility. First, we form an equally-weighted time series average every month from
1973 to 2003 of IPO idiosyncratic volatility and corresponding monthly news hits.31 The
data are linearly detrended as in CLMX (2001). The average monthly volatilities for
each month are regressed on a monthly time trend variable. We calculate the difference
between the predicted volatilities from the linear model and the actual monthly
volatilities. These differences are then used in the Granger causality tests. A similar
detrending method is used for the hits variable.

The regressions for Granger causality are of the form:

St. Dev, = a0 + a, St. Dev,-1 + ...+ a,- St. Dev,.j + b,Hits,-, + + ijits,.j + en, (7)

 

3' There are 360 total months from January 1973 to December 2003. Due to the fact that some months do
not have any hit observations, the total months used in the Granger causality tests is less than 360. We
require at least 50 observations in a given month.

108

Hits, = co + c, St. Dev,-, + ...+ 0;, St. Dev,_,, + d, Hits,-, + + (7,, Hits,-,, + e2,, (8)

The idea of the test in equation seven is to determine if the hits variable provides
information about ﬁrture idiosyncratic volatility above and beyond the information
contained in past volatility. The appropriate number of lags j or k in equations seven and
eight are chosen through the Akaike criterion and are reported in parentheses in Table
3.7. We also provide p-values regarding the null hypothesis that monthly hits have no
explanatory power for monthly volatility (and vice versa). The number of Observations
used in each regression will depend on the number of lags chosen by the Akaike criterion.
More lags will reduce the number of observations available. The number of observations
for each regression is also provided in Table 3.7.

Panel A of Table 3.7 shows that with a p-value of 0.00 we can reject the
hypothesis that lagged hits have no explanatory power for idiosyncratic volatility at the
1% level. However, we cannot reject the hypothesis (p-value of 0.20) that monthly
volatility has no explanatory power for monthly news hits. The evidence is consistent
with our claim that the number of news citations explains at least part of the increase in
idiosyncratic volatility over time. As the number of news citations concerning IP03
increases, so does their idiosyncratic volatility.

We repeat the Granger causality tests for the matching ﬁrms in Panel B of Table
3.7. The results are similar to those for IP03. With a p-value of 0.00, we can reject the
hypothesis that news hits have no explanatory power for idiosyncratic volatility. We

cannot reject the hypothesis (p-value of 0.46) that volatility has no explanatory power for

109

monthly news hits. In sum, for both the IPO and matching ﬁrms, it appears the monthly
news hit variable has signiﬁcant explanatory power for monthly idiosyncratic volatility.
We ﬁnd no evidence of the reverse hypothesis that monthly volatility actually drives

monthly news.

3.5 Identifying Exogenous Shocks to News

We provide evidence that an increase in news announcements in recent years
contributes to the increase in idiosyncratic volatility documented by CLMX (2001).
From our Granger causality tests, we learn that the increase in news is driving the
increase in volatility, and not vice versa. Ideally, we would like to obtain a measurable
exogenous shock to news releases to help reinforce the ﬁndings of this paper. The
question is whether a shock to news increases the number of news announcements
released by ﬁrms over our time period as well as the volatility over the same time period.
Recall that we document an increase in IPO news items per ﬁrm ﬁom an average of 72 in
the 19803 to 202 in the 19903 (matching ﬁrms rise from 57 to 77 over the same time
period). Firms themselves could choose to release more news at any time. Our goal in
this section is to discuss four entities that may give ﬁrms incentives to release more news.
Groups that have the power to affect corporate news releases are the stock markets, the
Financial Accounting Standards Board (F ASB), the Securities and Exchange
Commission (SEC), and the court system. Another consideration is actions by our

sources themselves, e. g. newspapers and news wire services.

110

 

We start by considering the effect of stock market regulations on news. There is
the possibility that individual exchanges might alter disclosure rules over our sample
period. The exchanges provide disclosure rules to be followed by stocks they list, but the
rules are rather vague. For instance, the New York Stock Exchange has an immediate
release policy.

“Any release of information that could reasonably be expected to

have an impact on the market for a company’s securities should be given

to the wire services and the press ‘For Immediate Release.’...To insure

adequate coverage, releases requiring immediate publicity should be given

to Dow Jones & Company, Inc., Reuters Economic Services and
Bloomberg Business News”32
New York Stock Exchange ofﬁcials tell us that the language of the disclosure guidelines
above has not changed in at least 20 years. The exchange has so broadly deﬁned news as
“anything that could reasonably be expected to have an impact on. . .ﬁrrn securities,” that
the news release decision seems to be entirely in the hands of the individual ﬁrm. As
such, we rule out changes in stock market disclosure regulation as providing possible
exogenous shocks to news between the 19703 and 19903.

It is also possible that FASB may impose new mandatory disclosure rules on
ﬁrms. If these new rules require ﬁrms to provide new information not previously
mandated under former guidelines, then ﬁrms may feel compelled to release this new

information to the press as well. However, our reading of FASB changes over the years

shows that these changes tend to be highly industry speciﬁc and would apply to very few

 

32 NYSE web page: http://lcm.nyse.com

lll

ﬁrms in general. We did not uncover any major F ASB changes over the last few decades
that would cause an increase in news releases for a wide variety of ﬁrms.

The SEC creates regulation targeting mandatory ﬁrm disclosure in the quarterly
and annual reports ﬁrms must ﬁle with the SEC. Federal securities laws primarily target
the remediation of information asymmetries. It has been argued that mandatory
disclosure to the SEC should impact the amount of news released in other channels as
well. For example, an SEC report issued by the Sommers Committee in the mid-19703
argued that much of the information contained in mandatory disclosure reports was
redundant and,

“...at best conﬁrmed information that earlier had been disseminated by

other media. But the report concluded that the mandatory system still played a

pivotal role in ensuring the initial corporate disclosure through whatever means

and in ensuring the accuracy of the data disclosed.” (Seligrnan 1995)

This report conﬁrms that the media is releasing pertinent and valuable information to
investors, often ahead of the ofﬁcial documents ﬁrms ﬁle with the SEC.

Before 1972, the SEC did not allow any reports to contain predictive data or ﬁrm
speculation about upcoming ﬁnancial trends. By the late 19703, the SEC reversed its
position and encouraged this type of “soft” analysis in ﬁrms’ ﬁled reports. The SEC
mandated management projections only in reference to the ﬁrm’s liquidity, capital
resources, and income in 1982. In 1989, however, “. . .the SEC now regarded disclosure
of predictive...information as mandatory in a wide variety of circumstances.” (Seligrnan
1995) This decision in 1989 affected not only ﬁrms’ ﬁled annual reports, but also the

registration statements for initial public offerings. It is possible that this change in SEC

112

 

regulation in the late 19803 could lead to the increase in ﬁrm press releases we see in the
19903. If ﬁrms are mandated to include more management projections in their ofﬁcial
SEC ﬁlings, it is very possible that these projections ﬁnd their way into the news media
as well.33

We also consider court action occurring within our sample period. A 1988
Supreme Court decision in Basic Inc. v. Levinson made it easier to engage in large class
action suits. People could join in the suits without having speciﬁcally relied on
misleading statements themselves. The damage assessments from such suits skyrocketed
as a result (Seligrnan 1995). This type of threat might have made ﬁrms more likely to
disclose more information to the media and also to disclose news earlier. The idea that
ﬁrms disclose news as soon as possible in order to avoid litigation has also been proposed
by Healy and Palepu (2001). We feel it is possible that the large increase in news in the
19903 could be related to a ﬁrm’s desire to avoid potential lawsuits.

Our ﬁnal consideration is a broadening of news coverage by our selected
newspaper and newswire sources. If a news source broadens its coverage to include
more ﬁrms, this could translate to the increase in news we document. Additionally, if a

news source actually expands the scope or type of ﬁrm news covered, this could

contribute to the increase in news announcements per ﬁrm. For our purposes, it does not

 

33 Unfortunately, some major SEC regulation changes that might have an affect on news releases occur
outside or right at the very end of our sample. For instance the Secruities Exchange Act of 1934 would
certainly have an impact on the way that ﬁrms release pertinent news. Regulation Fair Disclosure went
into effect in 2000 and would also inﬂuence press releases. Regulation FD is the source of some
controversy though as to whether the regulation results in more information release or less. Bailey, Li,
Mao, and Zhong, (2004) state that the amount of news released after Regulation Fair disclosure has
increased. On the contrary, Venkatraman, Thompson, and Eleswarapu (2004) state, “...that information
ﬂow around mandatory announcements has decreased” since Regulation Fair Disclosure has gone into
effect. The Sarbanes-Oxley Act of 2002 could also have an impact on the disclosure of information by
ﬁrms. However, this act is predominantly related to the auditing and ofﬁcial disclosure statements of a
ﬁrm as opposed to the release of voluntary news updates.

113

matter which way coverage is broadened. Mitchell and Mulherin (1994) state that Dow
Jones broadened their news coverage in 1989, although they do not state whether this
larger coverage derives from more ﬁrms or a larger scope of news covered. Dow Jones is
our primary news source, accounting for 82% of total news announcements between the
Dow Jones newswire and the Wall Street J oumal.

It is very possible that more than one of the discussed shocks contributes to the
increase in news announcements per ﬁrm between the 19703 and 19903. The three events
we feel are most likely to increase news disclosure — court action, SEC mandate for
management projections, and increasing coverage from Dow Jones — all occur in the
same two-year period from 1988 to 1989. We do not have a way to differentiate among
these three competing possibilities, but we do test for signiﬁcant changes in news and
volatility pre- and post this two-year period.

We look at ﬁrm news releases before and after 1988 and 1989, the two-year
period we wish to examine for shocks to news hits. We calculate the average monthly
number of news hits before and after the shock period and ﬁnd that IP03 from October
1984 to September 198734 go from 0.85 hits per month to an average of 1.16 hits per
month for January 1990 to December 1992. This difference is statistically signiﬁcant at
the 5% level. We then detrend this data for the IP03 by regressing each month from the
sample period onto a time trend and subtracting out the trend for each month. Once the
trend has been removed from the data, we see that the IP03 go from 0.85 hits per month
before 1988 to 1.11 hits per month after 1989. Once again, this result is statistically

signiﬁcant at the 5% level. We then examine a combination of the matching ﬁrms and

 

3‘ We exclude data from October 1987 as we expect the market crash in this month could cause spurious
results for both news and volatility.

114

 

the IP03, ﬁnding similar results to those seen by the [PCs These results are contained in
Table 3.8.

We have established that the news shocks we are concerned about have caused an
increase in news. However, we need to show that they are also responsible for a requisite
increase in volatility. To this end, we examine the monthly volatility for the IPOs. We
look at volatility from October 1984 through September 1987 ﬁnding an average monthly
volatility of 15.7%. From January 1990 through December 1992 we ﬁnd an average
monthly volatility of 20.5%. These ﬁgures are signiﬁcantly different at the 5% level.
Detrending the data for the volatility in the later period, we ﬁnd the volatility to be
16.4%. This ﬁgure is also signiﬁcamly different from 15.7% at the 5% level. As seen in
Table 3.8, the results for the pooled matching and IPO ﬁrms are similar to the IPO results
for monthly volatility.

In summary, we see that there are statistically signiﬁcant changes to news before and
after the 1988-1989 time period. We propose that these changes to news are related to at
least one of the exogenous events discussed above. We also see that the volatility of
ﬁrms increases substantially over the same time period, even once the data has been
detrended. We conclude that exogenous shocks to news result in increases to both news

and volatility.

3.6 Robustness Tests

3. 6. 1 Sample Evaluation
In order to ensure that a reasonable number of the Factiva dates collected are

related to the companies in our search, we evaluate the quality of our data. We select

115

 

three ﬁrms from each decade and read every Factiva article for these companies. The
ﬁrms are selected based on having a generic name, which might be subject to data
collection errors.35

The resulting articles are split into three categories: directly related articles,
indirectly related articles, and unrelated articles. Directly related articles contain news
that is about the ﬁrm’s proﬁts, sales, management, or any other event that would have a
direct impact on the ﬁrm. For directly related articles, we do not discriminate on the
basis of timeliness. In other words, it is often clear that directly related news is quite
stale. Indirectly related articles include news about direct competitors, for instance, but
the Factiva search ﬁrm is also mentioned in the article. Indirectly related news also
includes situations where an article cites an industry expert and the employer of the
expert was the F activa search ﬁrm. Unrelated articles have nothing to do with the ﬁrm in
the F activa search. Table 3.9 tabulates results for the sample.

The only time there are completely extraneous hits is in the 19903, making this
decade subject to the worst potential data problems (based on our sampling). Therefore,
ﬁrrther statistical tests involve only the 19903 decade (it is presumed that earlier decades
will be no worse than the 19903). T 0 make our results even more conservative, we have
assumed that neither indirect hits nor unrelated news hits provide poignant news for the
ﬁrm. This results in 19 fallacious articles out of the 295 articles from the 19903. Using a
binomial distribution for a sample size of 250, we can determine with 95% certainty the
limits of the distribution as 3.4% to 9.7% non-direct news. Our conclusion based on this

result is that our database is relatively representative of the news being released by the

 

35 The companies selected are Superior Services Inc, Community Care Services Inc, International Network
Services, Avantek, Bear Creek Corp, Vapor Corp, Laser Photonics, Sigma Research, and Crawford Energy.

116

frrrns. Essentially, we can say with 95% certainty that at least 90% of our articles are
related to the F activa search ﬁrms.
3. 6.2 Restricting Sources to the New York Times and Dow Jones publications

Due to the fact that several of our news sources are not available over the whole
sample period, we check to conﬁrm that our results are robust to this data problem. We
ﬁnd that for IP03 2% of our data points come frOm the New York Times, which is
available over the whole sample period while 80% of our data points come from the Dow
Jones Newswire or the Wall Street Journal. The remaining 17% of our data come from
other sources listed in Table 3.2.36 Upon replication of tests using just the New York

Times citations, we ﬁnd results similar to those with the full data set.

3.6 Conclusions

We have shown that idiosyncratic volatility has increased over the last few
decades for both IP03 and a size and book-to-market matching sample of ﬁrms. The
news day return distributions change over time for both samples. The investor reaction to
good news is larger for IPO ﬁrms in the 19903 than in the 19703 and bad news day
returns are becoming more negative across time. The standard deviation of news day
returns has increased for both samples, with IPO ﬁrms showing a sharper rise in standard
deviation across decades. We conclude with cross-sectional and Granger causality

regressions that show a statistically signiﬁcant link between idiosyncratic volatility and

 

3‘6 For matching ﬁrms, the results are similar with 3% coming from the New York Times, 85% coming from
the Wall Street Journal or Dow Jones Newswire, and 12% coming from other sources.

117

public news. An increase of one news day per month leads to an additional 0.6% in
idiosyncratic monthly standard deviation for IPO ﬁrms.

While our results are consistent with news driving idiosyncratic ﬁrm volatility for
[PCs and matching ﬁrms, the mechanism of this driving force is not clear. As pointed
out by CLMX (2001), an increase in cash ﬂow news should result in decreasing return
volatility. There are several potential explanations for why news is driving increased
ﬁrm volatility. We have controlled for many of the ﬁrm effects discussed in the past
literature so we consider only additional theories here.

First, it is possible that the type of news is fundamentally changing over time.
News from the 19703 might be fundamentally different in nature and scope than news in
the 19903. This View is consistent with the SEC change requiring disclosure of predictive
information as well as historical information. However, if this is the case, the predictive
information is not providing any beneﬁt for investors, but is negatively impacting value
through an increase in idiosyncratic volatility. The increase in news could also be the
result of news being broken into smaller pieces and released more consistently across
time. As previously stated though, this result would imply a decrease in return kurtosis,
not the increase demonstrated in Table 3.1 [see Shiller (1981)]. It is also possible that the
news is becoming less informative compared to past years, although it is not clear why
this would be the case. In addition, it may be that news is fundamentally the same, but
the investor reaction to news is changing. This could manifest itself in over-reaction and
under-reaction phenomena, which would tend to increase volatility.

Finally, it is possible that news is driving certain trading behaviors, which are

causing the increase in volatility. If the visibility of the ﬁrm is increasing due to news

118

 

ﬂows, the addition of new investors with divergent opinions about the ﬁrm could cause
an increase in volatility.37 It is also likely that an increase in the amount of public news
about a ﬁrm would lower the costs of trading that stock}8 Such a decrease in trading
costs would lead to increased transactions and thus, an increase in the volatility of the
stock [Jones, Kaul, and Lipson (1994)]. Thus, increased news ﬂows could lead to
changes in trading behavior for investors resulting in higher ﬁrm volatility. We certainly
do not propose that any one of these effects is exclusive of any other.

Xu and Malkiel (2003) propose that increased volatility is related to increased
trading activities by ﬁnancial institutions. In an interesting bridge between Xu and
Malkiel (2003) and our work, Falkenstein (1996) shows that mutual funds avoid stocks
with little information, “as measured by the number of major newspaper articles...”
written about the ﬁrm. In addition, Diamond and Verrecchia (1991) and Kim and
Verrechia (1994) argue that voluntary disclosure will lead to increased stock liquidity and
higher institutional ownership. If mutual funds are investing in ﬁrms with more news
releases, then it may be that news releases are the driving force behind increases in
idiosyncratic volatility. Thus, it is not clear that our results do not actually explain why
institutional ownership appears to be driving volatility as in Xu and Malkiel (2003).

CLMX (2001) were the ﬁrst to document an increase in idiosyncratic ﬁrm
volatility over the last few decades. We focus on a subsample -- IP03 and size and

book-to-market matched ﬁrms -- of the ﬁrms originally studied. For these ﬁrms, we

 

37 Miller (1977) makes the case that increasing the number of investors informed about the ﬁrm could raise
the price of the stock. However, with more investors trading on beliefs of over or under-pricing, it is also
conceivable that such an increase in investor trading would lead to higher ﬁrm volatility.

33 Copeland and Galai (1983) show that increases in public information about a ﬁrm should result in a
decrease in bid-ask spread. Other decreases in trading costs could come in the form of lower research
costs.

119

provide evidence that the rise in volatility is signiﬁcantly linked to a rise in public news
in recent decades. Given our ﬁndings hold for both IPOs and matching ﬁrms, we believe
more research on the relation between news and volatility across all types of ﬁrms would

be ﬁ'uitful.

120

APPENDIX 3.

TABLES OF ESSAY 3.

121

 

Table 3.1

Mean Monthly Standard Deviation
in the First Five Trading Years

The sample consists of all IP03 from 1973 to 1998 where the IPO offer price is at least $5. Stock and
index return information is collected from the CRSP daily database. Matching the IP03 with ﬁrms of
similar size and BM ratio creates the matching ﬁrm sample. The matching ﬁrms are companies with at
least ﬁve years on the CRSP database. Monthly volatility is calculated by summing the squared differences
in daily returns between the individual stocks and the value-weighted index over the month. For the
equal-weighted (EW) calculations, the monthly volatilities are averaged equally across all ﬁrms and all
months relevant to the given time period. For the value-weighted (VW) calculations, the volatilities are
value weighted by market capitalization within each month. The value-weighted volatilities are then
averaged equally across all months relevant to the given time period. Kurtosis data is for pooled daily
returns for each decade.

 

 

Mean EW Mean VW
Standard Standard
Deviation (%) Deviation (%) Kurtosis

 

 

 

 

 

IP03 issued in

1973-1979 12.4 11.6 44.6
1980-1989 16.7 11.9 125.0
1990-1998 20.9 14.7 359.7
1973-1998 19.2 12.9

Matching Firms

1973-1979 14.4 10.5 25.4
1980-1989 15.4 9.8 219.0
1990-1998 17.8 11.6 259.3
1973-1998 16.8 10.7

Value Weighted

Index.

1973-1979 3.7 4.6
1980-1989 3.8 6.4
1990-1998 4.2 7.2
1973-1998 4.0

Equal Weighted

129%

1973-1979 2.9 8.0
1980-1989 2.5 11.1
1990-1998 3.1 9.9
1973-1998 2.8

 

122

Table 3.2

Factiva News Sources

The Factiva database consists of newspaper, newswire, and periodical publications from Dow Jones &
Reuters. Our selection criteria are as follows: the top six US. newspapers by circulation with Factiva
listings going back to at least 1987, the two newswires with longest data availability, the top two Canadian
newspapers by circulation, and the top British business newspaper by circulation.

 

 

Source Publication Type Coverage Dates
Associated Press Newswire 12/1/85-present
Chicago Tribune Newspaper 1/ 1/ 85 -present
Dow Jones Newswires Newswire 6/13/79-present
Financial Times International Newspaper 1/5/80-present

The Globe and Mail
Los Angeles Times
New York Times
Toronto Star

International Newspaper
Newspaper

Abstracts from NYT articles
International Newspaper

1 l/14/77-present
1/1/85-present
1/1/69-present
1/ 1/86-present

USA Today Newspaper 1/4/87-present
The Wall Street Journal Newspaper 6/13/79-present
The Washington Post Newspaper 1/1/84-present

 

123

 

Table 3.3
Newspaper and N ewswire Citations in the First Five Trading Years

For each IPO or matching ﬁrm, the company name as given by the CRSP database is entered into Factiva.
The search is done from the IPO date until ﬁve years after the IPO date plus the next ﬁrll calendar month or
until the IPO is delisted. For matching ﬁrms, the search is performed until another matching ﬁrm is
chosen, until the IPO ﬁrm is delisted, or until ﬁve years after the IPO date plus one calendar month. The
number of news citations are summed over the ﬁrst ﬁve years of trading, then averaged over the given time
period. For News Days per Firm, the days with cites are summed over the ﬁrst ﬁve years of trading and
then averaged across all ﬁrms in the given time period.

 

Number of IP03 Mean News Hits Mean News Days

 

 

 

in the sample per Firm per F irm

IP03 issued in

1973-1979 19 23.0 18.8
1980-1989 237 71.5 55.9
1990-1998 353 201.6 105.3
1973—1998 609 145.6 87.3
Matching Firms

1973-1979 22.9 20.8
1980-1989 57.0 49.2
1990-1998 77.1 52.2
1973-1998 67.6 50.1

 

124

 

 

 

 

 

 

 

 

 

 

sane. an? OS. 38 8.8 m: OS. 2 .4 :3 8o snowman
sea. 83: :8 m2 m3 mos Nam- was. one 36 03:52
.53 was: 8.4 an sea was. 32. man «no :5 32.82
.303 as 8.4 as :s Rs 80 an :8 sod 32-32
math MCEBNE
sea. Eda m3 5.» a: 43. eon- 8.4 one So mania
.309. an? :3 a: so.” 83 8.4- as. one :3 32.82
seas and NZ. New Mae 3.4 e3- an ode as 33.32
31.8 5 83 :2 24 Ba 42- 03 26 3o 32.32
E UDSmmm won:
mama $502 2502 @502 $502 $502 9502 9502 @502 9502
am a. =< am 88 :3. oz 30 88 =< oz
eeosm g 4.63656 8:360 g meaom baa
me 03:52 93:85 502 $08G 502
053.0% 030 Z ”< 055
.382 20

now made mmm can .032 05 08 08E 5mm .353 05 5m manna 3 ma 00% 05 no.“ 050, 0383 2E. .830 £5 E 009:2: Sc 03 0832
REE Evie On: .55 a pow 00383 «>303 0a 5 8:36 a mm? 005 moi? no 930 00 35.00 03 view @502 .0003 00:30?
-03? ammo 08 $088 03 $5500 $15 003300 ammo 05 no 030% 0>c ~82 S 53, 00:89:00 03 08.5 wcﬁoumﬁ 05. .0383.
SE waEBmE 05 000008 038 gm we“ 09m 3:86 we 05.6 55 005 0% @3032 00.38% .330 ammo 05 Sec 0080:00
mm non—~86; E80“ x005 28 06on .mm 030— «a a 00% 0&0 On: 05 20:3 woa 8 9% 80¢ 00% .3 $2 mo 806:8 0388 25.

when 0302-52 was mB0z :0 2:50”— .«e 55325 ocean—Sm ES 2:30”— 803m
Wm 03.;

125

 

 

84.0-: 83-: 30: 8:: se.: 5.0::

 

 

08.0- 08.0- :80 80.0 30.0 80.0 N00328:: :0
5.3 80.0-: 80.0-: :00: and: 5.0:
:0. 80- 00.0- ~00 2.0 40.0 .828 €
5 3 s 3 5 .0
as: 0:: as: 0:: as: 0:: as: wood: as: 83: noses: oases:
x 8m. 0000. 0: 8:0 030. :0.000:w0~.

 

0020?: ”m .80:

80:0.0:00-b.0.:00.:00.00:0:0: 0.0 00005000: :. :0>.w 00.:0.:0:0-: 0: ... On. :0
0. :8: 0:: :. 30.8050 0:00: 0:0 :0 0:.0> 0 :0 000.0: 0.:0.:0> >82: 0:. 0:... 008:0: 30.8.0 0. 8.8:: 0000. 0:... 6379:.
w:.::.: 0080000 0:. 0:: :. 00.3050 0:00: 0:0 :0 0:.0> 0 000.0: >888: 03.0. 0:... 8.2.: 0:: 80:0 80:0: 030: 0 8.0800 000:0:00
03:00... 0:: :0.:3 :0 m0: 0 00 008:0: 0. A0: 030: < .93 00.00080: :0: e :0.000:w0: 80:: .0800: 00:013.. 0:: .80 e :0.000:w0:
:0: :80: 0.: 030: 0:: 0. 0.:0.:0> 80.80000 0:... 30:80.: ..me 0:: :0 0:00.: 0>: :000. :0 ::.3 0088800 0:0 08:: 8280::
0: H .0.:0: 0808-90.00: .80 00.0 :0 00:80:: 88: 5.3 w:0_0 M30. 0: m3. 80:: :00» :000 0.3:: 803 :0:: 00:. 0:: :0 $0.
:0 0:00:00 29:00 0:... .00.:0.:0> 0800 0:: :0 00000:w0: 0:0 00:0:60 0:0 5 :0.000:w0: 80:: 0.0.6.00: 0:: .00.000:w0: .8800 0:: :.

.o + :3: .82 . as: 0:: 3+ :5: .33: .. :30. 0.: .0 +35 0:: s + as: .08: s + as: 33: s + s n .558

088-0080 80.80850

0:: 0:0 .9: :0 003 8:: 0:: :0: :0 050:? 0.0000: :0: 00.:0.:0> >880: 0::0 0000000: 0:0 0:80: 0:: 0020:. .< .050: 80:: 0880:

:. 03:80:"...0 :0: 0.00:0 0: :00:-: 0.950 0 08:8 :0: 0.: 03 .08.: :0>0 0w:0:0 :00:0::. :0 0030.0.» ::0 :0 000:0.:0> 0:0 0:008 085
0.3: 0307. :0 0:0—30¢ 0000:6— 0w0:0>< :0 80.000801

083360 4.0 2:00

126

Table 3.5
Summary Statistics for [PCs and Matching Firms

The sample consists of all IPOs and matching ﬁrms from 1973 to 1998 where the IPO offer price is at least
$5 and complete data on all of the variables is available. Stock return information is collected ﬁ'om the
CRSP daily database. Accounting variables come from Compustat. Matching the IPOs with ﬁrms of
similar size and BM ratio creates the matching ﬁrm sample. The matching ﬁrms are companies with at

least ﬁve years on the CRSP database. All variables are calculated on a ﬁrm-level basis. St. Dev t is the

monthly standard deviation of daily returns. Hits. is the monthly number of days the ﬁrm is cited in our
Factiva database. The IPO dummy takes a value of one (zero otherwise) if the ﬁrm conducted an IPO

within the last ﬁve years. Mkt. Cap t..1is last month’s market value, and Retum is the monthly excess

return. ROE.-. is the trailing year’s earnings divided by trailing year book value. Lev“ is the trailing
year’s long-term debt divided by trailing year total assets. The Tech dummy takes on a value of one (zero

otherwise) if the ﬁrm belonged to the technology or Internet industry. Ln(b/m) t-1is calculated as the

natural logarithm of last year’s book value divided by the previous month’s market capitalization. P-values
are given in brackets.

Panel A: Means Across First Five Trading Years

 

 

Item IPOs Matching
Market Cap. $307 million $291 million
% Tech 13.9% 7.5%
B/M 0.67 1.18
Leverage 15.5% 16.6%

 

127

 

Table 3.5 (cont)

Panel B: Monthly Correlations

 

 

IPO StDev StDev Tech Ln Lever- Ln
Dum Return. Dum (MkCa m
HiISt t t-l p) aget-l (bl)
t-l
t-l
mo
Dum
. 0.14
H‘tSt [0.00]
0.07 0.06
StDcvt [0.00] [0.00]
0.07 0.01 0.65
StDth-l [0.00] [0.00] [0.00]
.000 0.03 0.34 0.16
m [0.04] [0.00] [0.00] [000]
Tech 0.10 0.03 0.06 0.06 0.01
Dum [0.00] [0.00] [0.00] [0.00] [0.00]
mm 0.02 0.32 -0.34 -0.34 -0.07 0.01
Cap)“ [000] [0.00] [0.00] [0.00] [0.00] [0.00]
Lever-age -0.03 0.02 -0.05 -0.06 -0.02 -0.09 0.01
H [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
Ln -0.13 -0.14 0.11 0.13 0.05 -0.08 -0.39 0.00
Wm)“ [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.62]
0.00 -0.00 -0.01 -0.01 -0.00 0.01 0.01 0.02 -0.03
ROEt-l [0.08] [0.46] [0.00] [0.00] [0.14] [0.00] [0.00] [0.00] [0.00L

 

128

 

a + as: 0009. S: + S: .000: :0 + as: 00...: 2.: + 2:50:53 + 2.00.0 +
2.0.0:... + 5.00:... + 2.0: :0 00 + .800 0:0: =0 .0 + as: 0.: .. .00: s + 55 0.: .0 + 5.0: .0 + 2: u 0...: :0

80:00:00 b0:00:0:00:0:0: 0:0 000.0023: ..<
8800:5800 :0::08 0.:808 00030:: 0:: .3 0003... 0:.0> :00: 0.000.. :00. :0 85:000. .230: 0:: 00 030.0200 0. 2.8.8:... .8030.
.0830. :0 30.05.00: 0:: 0: 00w:0.0: 8:: 0:: :. 30.8050 0:00. 0:0 :0 0:.0> 0 :0 00:0: 88:: :00... 0:... 0.88:: 0:. 0:: 0:: .3 .0:0:
0:: :58: 2:: 8:003: as: 808.20: 2: 0808 0:00:00 0: 0050 8%: 2:. 002-002 00.50 0258.. 0:: 20:: .3020?
0:00. 0:0 :0 0:.0> 0 00:0: 08:50 0000. 0:... $800 .0:0: :00» 850:: .3 00.03.: :00: 83.000. 0.30.. 0:58: 0:: 0. 3.00.. .080.» :00:
:8: 050:: :0 3.3: 0:88 0:8: 05:05 2:: 0: :00: .832 :58: 2:: 0: :.: ea”: 20 .8? 0.00.: 0.58: :8: :o 8.0502

.830: 0. ...: 300 0.2.0.. 0:00.. 0.0.: :00. 0:: 85.3 on: :0 00:00:08 :8: 0:: :. 60.8050 0:00: 0:0 :0 00.0.» 0 00:0: 8:80.: 0:. 0:...

.8038 2.8: :8 a 09: 0: E: 20 0:00 :o :35: 8:8: 2:: 0: .30: 2:32 :0:. :o 80060 0:00:00 508. 2:: 0: 200g
80:00:00 0:... 0.00: .0838: 0 :0 02030.00 0:0 00.:0.:0> .2. 000080.: n.00.“: 0:: :0 0:00.. 0.»: :08. :0 8.3 00.000800 0:0 0:::
055.0: 2F 2:83 E0 00:20: 2:: 8:85 2:0: 2:0 08 on: 008:0 :o 0:5: 5:3 00.: 2:: 00:20:): 032:3 m: 83000, 2::
:o =0 8 :00 22:08 08 00 :08: a m: 8:: :0:: 0.: 2:: 00:3 0a: 3 08: as: 2:0 055:0: 05 00:: =0 :o 00:88 2:05 2:0

0030.00.» .2800 .80 0:... 0.507. :0 :0::87. £500.). 0:: :0 3.8.8:. 9.0.805 5:80.». :0 003000.800.
9m 030...

129

 

 

 

 

 

$0.00 0.0.: :20 020000-:

80.00 08:: 0.30 08.0.5000

00.00 00.: $.00 3.: 080 8.: .0850:

- -- -- -- 8.00 8.0 0a: 800:

-- -- -- -- 0:: 3.0 0a: 0000:

02 6.0 Rd :0 00:. 00.0 0a: :80

E

00.0- :00- :0..- 2.0- 8.0- 00.0- .830:

. . . . . . 3 >00
00 :- R o- 00 : R o :0 o- 8 a.

ﬁn.
00.0.- 8.0- 00.:- 80- 0:0- 80- mo:
3
00.: 00.: 00.0: :0:: 8.2 03.: 833.
E .
8.: 00.00 00.00 8.00 00.0. 03.0 5: :0
:
I“ Q 0

0.00. 00.0- 00.00- 00..- :30- 3..- . 00 00>: 5

00.: 2.0 00.0- 00.0- 00.0 00.0 0s: on: .00.:

0:. 0:0 00.: 00.. 2.0 00.0 05: 0::

00.0 00.0 00.: 00.0 3.2 8.0 .00:

0.:0.:0:0-: A00: 80.05000 0.:0.:0:0-: AS. 80.05000 0.:0.:0:0-: Aﬁb 80.05000.
0000-000: 000:-m~0:-:-.% _. 3: - 002-08:

 

PM 0.:0...

130

Table 3.7
Granger Causality

AmhﬁmnmofSOhitsobservationsismquimdforeachmonthlyobservation Thisrequirement
corresponds to months from June 1983 to Februaly 2002. Stock retum information is collected from the
CRSPdailydatabase. MatchingtheIPOswithﬁnnsofsimihrsizeandB/Mmﬁocreatesthematching
ﬁrmsample. ThematchingﬁnnsareoompanieswithatleastﬁveywrsontheCRSPdatabase. SLDevds
themomhlystandarddeviationofdailymtums. lﬁmisthemonthlynumberofdaystheﬁrmiscitedinmn
Factiva database. The table reports the p-values of Granger-causality VAR tests. The null hypothesis is
thatlagslthroughkofthevaﬁableineolumnldonothelppredictthevaluesofthevaﬁablesineolumnsZ
and 3, aftercontrolling for other variables in the VAR The number of lags kapptopriate for each VARis
chosenmmnghtheAkaikemiwrionandisgiveninWbelowmep-valm. BothStDev.andI-lits.
amlinearlyden‘endedhefomeonductingtheVARs 'I'hemnnberofobservationsNusedineachmgreadon
isalsoreported

 

 

 

 

 

 

 

Panel A: IPOs
811(11)th Hits.
0.20
Std. Dev“ - (9)
N = 216
0.00
Hitst-l (2) "
N = 223
Panel B: Matching Firms
Std. Devt Hita
0.46
Std. Dev” - (6)
N = 219
0.00
Hits” (2) -

N=223

131

 

 

Table 3.8
Changes in Monthly News Hits and Volatility,
Pre- and Post-Exogenous News Shocks

Thesample consists oflO% oftheIPOsthatwentpubliceachyearfrom 1973 to 1998 alongwithﬁrms
matchedonsizeandbook-to-marketratio. 'l‘hematchingﬁrmsarecompanieswithatleastﬁveyearson
theCRSPdatabase. Factivawasutilizedtoﬁndeverynewsreportfortheﬁrstﬁveyearsastockwas
publiclytraded Themeanmonthlynewsdaysperﬁrmistheaveragenumberofhitspooledacmssﬁrms
and time. Volatilities are pooled across ﬁrms and over time as well. The third column gives the number of
ﬁrms that are within ﬁve years of going public in the sample period. T-stats for diﬂ‘erences between the
two time periods (pm-l988/89 and post 1988/89) are contained in parentheses. Detrended data uses the
monthly values from hits per ﬁrm or monthly volatilities from Jarmary 1980-December 1990 and regresses
themonatimetrend Thecoeﬁ'rcientforthetimetrendprovides theincrease inhits(orvolatility)per
month This coefﬁcient is then multiplied by the number of months separating the two time periods and
subtracted from the second data point. There are three exogenous events that occur in 1988 and 1989 that
could affect ﬁrm release of information. First, in 1989 the Securities and Exchange Commission mandated
managerial release of predictive content in ﬁrm statements Second, the Supreme Court Decision in Basic
Inc vs. Levinson occurred in 1988. This court decision made it more likely that ﬁrms would release timely
informationto decrease theirlawsuitrisks. lastly, theDow Jones newswire servicebroadened its coverage
in 1989. Thisincreaseincovemgewmddinevitablyresuhmanmcreaseinnewsabmdﬁrms

 

Mean Detrended Mean Detrended
Monthly Monthly Monthly Monthly

 

 

News Number of News News Standard Standard

Years Firms Days/Firm Days/Firm Dev. (%) Dev. (%)
Oct. 1984

IPOs to 114 0.85 0.85 15.70 15.70
Sep.l987

““1390 88 1.16 1.11 20.50 16.42

Dec. 1992 (8.58) (7.26) (14.41) (2.15)
Oct. 1984

ﬁgsﬁd to 209 0.86 0.86 13.39 13.39
c “3 Sep.1987

“at?” 159 1.06 1.02 18.12 14.28

Dec. 1992 (8.61) (6.63) (17.98) (3.37)

 

132

Table 3.9
Articles Related to the Factiva Search Firms

Asampleofthrceﬁrmsistakenﬁ'omeachdceadeandallthenewsarticlesfoundinFactivaforthcseﬁrms
areread Articlesarecategorizcdintodirectnews, indirectnews, andunrelatednews. Basedonabinomial
distribution with sample size of 250, the range of non-related news is conservatively estimated at
3.4% - 9.7% with 95% certainty.

Panel A: News Content

 

 

 

 

 

Direct News Indirect News Unrelated News Total

All 484 24 3 51 1

Firms .

94.7% 4.7% 0.6% 100%

1990s 276 16 3 295

Firms
93.6% 5.4% 1.0% 100%
Panel B: Estimation of Unrelated News
Max Unrelated Min
News Unrelated

News
N=250 Upper limit = 9.7% Lower Limit = 3.4%

 

133

 

APPENDIX A.

DATA CONSTRUCTION

134

 

Data Construction for Essay 3

There are two possible data errors to consider in our F activa collection process.
First, we may omit dates where news is actually released, and second, non-news dates
may erroneously be recorded as news dates. These errors may cause two potential biases
in our procedure.

If news days have the same return distribution as non-news days, then any data
collection error will not have an effect on the results. If the distributions are the same,
then the sorting of news days and non-news days can be at random and will still show no
difference between the distributions. Of course, this is the trivial case.

If news days have more volatile returns than non-news days, this actually biases
against our ﬁnding a difference. For instance, if news days are omitted (potential error
one), they will cause an increase in the non-news day volatility decreasing the difference
between the two groups. If non-news days are included in the news day returns (potential
error two), they will have the propensity to lower news day return volatility. Thus, it can
be seen that any errors in the database will be biased against the hypothesis that news

days have higher volatilities than non-news days.

135

APPENDIX B

FIGURES

136

 

1.6

0
’ o

0.8 ~ I
I

 

Average Percentage of IPO Shares Held

 

 

 

 

 

 

0.6 .
0.4 4 O UW
I Non-UW
0.2 .
o . . . ,
0 2 4 6 8

Quarters from IPO

Figure 1.1 Underwriter and Non-Underwriter IPO Holdings. We examine underwriters
that are classiﬁed in the SDC IPO database as lead underwriters between 1993 and 1998. These
underwriters must also have an asset management division that ﬁles quarterly 13f statements with the SEC.
For each IPO, we refer to “underwriters” as those investment bank/asset managers that underwrite the IPO
and hold shares in a given quarter. We refer to “non-underwriters” as those investment bank/asset
managers that do not underwrite the given IPO, but do hold shares in the given quarter. Measurement of
the ﬁrst eight quarters of trading begins at the ﬁrst calendar quarter after the IPO date. The percentage of
shares held is deﬁned as the number of shares stated in the quarterly 13f ﬁlings (obtained from Thomson
Financial) divided by the number of shares outstanding (obtained from CRSP).

137

2.50%
2.00% -

N?
1.50% - - /
1.00%

0.50% /
t /
0.00% \

-0. 50% v

 

 

 

 

 

 

 

 

 

Figure 2.1 Cumulative market-adjusted returns: Returns around the Earnings

Announcement made during the quiet period. This ﬁgure plots cumulative market-adjusted
returns for the period prior to and alter the release of earnings news during the quiet period. The average
abnormal return for the 132 ﬁrms that release earnings information during the quiet period from the day
before to the day of the announcement is 2.04%.

138

 

 

24~

.3 .3 N N
C) m C N

Monthly Standard Deviation (%)

.3
.b

12 4

 

 

 

+ IPOs
-I— Matching

 

 

 

10

Figure 3.1. Monthly volatility over time for IPOs issued from 1973-1998. This ﬁgure
shows the average monthly volatility for the IPOs in our sample averaged over ﬁve year increments. The
graph also shows the average monthly volatility for ﬁrms matched to the IPOs based on size and book-to-
market. Volatility is calculated using the CLMX (2001) method. Volatility over each ﬁve-year block is the

1973-1978 1979-1983

1

I

1984-1988 1989-1993 1994-1998

Year IPO Issued

equally weighted average monthly volatility for each ﬁrm

139

 

Newsday Return Distributions Across Decades

 

0.50 *

/\
0'45 / \ —— 1970s

0-40 -——-1980s
,- - -1990s

 

 

 

 

 

 

 

0.35
0.30
0.25
0.20 -
0.15 ,
0.10 .
0.05 -~ ~ \1

0.00 r 1 I 1
-0.05 -0.025 0 0.025 0.05 0.075

Newsday Return

 

 

 

 

 

Probability of Return

 

 

 

 

Figure 3.2. News Day return probability distribution across decades for sampled
IPO ﬁrms. The sample consists of 10% of the IPOs that go public each year. The fu'm sample size is
19, 237, and 353 across the decades of the 19705, 1980s, and 19908, respectively. Factiva is utilized to
search for every day on which news is released for each ﬁrm in our sample. The CRSP database is then
used to obtain the news day returns for each time news is released about the ﬁrm. The total number of
news day returns for each decade is 357, 13,129, and 37,289.

140

 

 

26

25*

24 ‘ 305

23.

22.

21~

Monthly Standard Deviation (%)

20*

19~

 

 

 

 

0 1t03 4t06 7t09 10to12 13andup
Number of Days with Hits per Month

Figure 3.3. Relation Between Number of News Days per Month and Monthly

Volatility. The sample consists of 10% of the IPOs that go public each year from 1973 to 1998 as well
as ﬁrms matched to the IPOs based on size and book-to-market. The number of days with news (obtained
from Factiva) is summed over each month for both IPOs and matching ﬁrms. Volatility is calculated using
the CLMX (2001) method. The number of observations for each grouping is listed above the
corresponding data node.

141

BIBLIOGRAPHY

142

BIBLIOGRAPHY

Aggarwal, Rajesh K., Laurie Krigman, and Kent L. Womack, 2002, Strategic IPO
underpricing, information momentum, and lockup expiration selling, Journal of Finance
66, 105-137.

Agrawal, Anup and Mark A. Chen, 2004, Analyst conﬂicts and research quality,
University of Maryland working paper.

Bailey, Warren, Haitao Li, Connie X. Mac, Rui Zhong, 2004, Regulation fair disclosure
and earnings information: Market, analyst, and corporate responses, Forthcoming Journal
of Finance.

Begley, J ., and P. E. Fischer, 1998, Is there information in an earnings announcement
delay? Review of Accounting Studies 3, 347-363.

Ben Dor, Arik, 2003, The performance of initial public offerings and the cross section of
institutional ownership, working paper, Northwestern University.

Ber, Hedva, Yeshay Yafeh, and Oved Yosha, 2001, Conﬂict of interest in universal
banking: bank lending, stock underwriting, and fund management, Journal of Monetary
Economics 47, 189-218.

Bradley, Daniel J., , Bradford D. Jordan, and Jay R. Ritter, 2003, The quiet period goes
out with a bang, Journal of Finance 58, 1-36.

Campbell, J. Y., Martin Lettau, Burton G. Malkiel, and Yexiao Xu, 2001, Have
individual stocks become more volatile? An empirical exploration of idiosyncratic risk,
Journal of Finance 56, 1-43.

Carhart, Mark M., 1997, On persistence in mutual fund performance, Journal of Finance
52, 57-82.

Carter, Richard B., and Steven Manaster, 1990, Initial public offerings and underwriter
reputation, Journal of Finance 45, 1045-1067.

Chambers, A. E., and S. H. Penman, 1984, Timeliness of reporting and stock price
reaction to earnings announcements, Journal of Accounting Research 22, 21-47.

Chan, Wesley S., 2003. Stock price reaction to news and no-news: Drift and reversal after
headlines, Journal of Financial Economics 70, 223-260.

Cliff, Michael T., and David J. Denis, 2004, Do initial public offering ﬁrms purchase
analyst coverage with underpricing? Journal of Finance 59, 2871-2901.

143

Copeland, Thomas E., Dan Galai, 1983, Information effects on the bid-ask spread,
Journal of Finance 38, 1457-1469.

Core, John E., 2001, A review of the empirical disclosure literature: discussion, Journal
of Accounting and Economics 31, 441-456.

Cutler, David M., James M. Poterba, and Lawrence H. Summers, 1989, What moves
stock prices? Journal of Portfolio Management 15, 4-12.

Demers, Elizabeth A., and Katharina Lewellen, 2003. The marketing role of IPOs:
Evidence from the intemet stocks. Journal of Financial Economics 68, 413-437.

Diamond, D., and R. Verecchia, 1991, Disclosure, liquidity, and the cost of capital,
Journal of Finance 46, 1325-1355.

Duffee, Gregory R., 1995, Stock returns and volatility a ﬁrm level analysis, Journal of
Financial Economics 37, 399-420.

Dugar, A., and S. Nathan, 1995, The effect of investment banking relationships on
ﬁnancial analysts’ earnings forecasts and investment recommendations, Contemporary
Accounting Research 12, 131-160.

Ellis, K., R. Michaely and M. O'Hara (2000). When the underwriter is the market maker:
An examination of trading in the IPO aftermarket. Journal of Finance 55, 1039-1074.

Engle, Robert F. and Victor K. Ng, 1993, Measuring and testing the impact of news on
volatility, Journal of Finance 48, 1749-1798.

F ama, Eugene, and Kenneth French, 1992, The cross section of expected returns, Journal
of Finance 47, 427-465.

F ama, Eugene, and Kenneth French, 2003, New Lists: Fundamentals and survival rates,
Forthcoming, Journal of Financial Economics.

Falkenstein, Eric G, 1996, Preferences for stock characteristics as revealed by mutual
fund portfolio holdings, Journal of Finance 51, 111-135.

Field, Laura, and Michelle Lowry, 2005, Institutional versus individual investment in
IPOs: The importance of ﬁrm fundamentals, working paper, Penn State University.

Healy, Paul M., and Palepu, Krishna, 2001, Information asymmetry, corporate disclosure,
and the capital markets: A review of the empirical disclosure literature. Journal of
Accounting & Economics 31, 405-440.

Healey, Paul M. and Krishna G. Palepu, 2001, A review of the empirical disclosure
literature, Journal of Accounting and Economics 31 405-440.

144

 

Hoberg, Gerard, and H. Nejat Seyhun, 2005, Do underwriters collaborate with venture
capitalists in IPOs? Implications and evidence, working paper, University of Maryland.

Johnson, William C. and Jennifer Marietta-Westberg, 2006, Universal banking, asset
management, and stock underwriting, working paper, Michigan State University.

Jones, Charles M., Gautam Kaul, and Marc L. Lipson, 1994, Transactions, volume, and
volatility, Review of Financial Studies 7, 631-651.

Kim, 0., and R. Verecchia, 1994, Market liquidity and volume around earnings
announcements, Journal of Accounting & Economics 17, 41-68.

Klibanoff, Peter, Owen Lamont, and Thierry A. Wizman, 1998, Investor reaction to
salient news in closed-end country funds, Journal of Finance 53, 673-699.

Krigman, Laurie, Wayne H. Shaw, and Kent L. Womack, 1999, The persistence of IPO
mispricing and the predictive power of ﬂipping, Journal of Finance 54, 1015—1044

Lang, M., and R. Lundholm, 1996, Corporate disclosure policy and analyst behavior, The
Accounting Review 71, 467-493.

Leland, Hayne E., and David H. Pyle, 1977, Information asymmetries, ﬁnancial structure
and ﬁnancial intermediation, Journal of Finance 32, 371-3 87.

Lin, Hsiou-Wei and Maureen McNichols, 1998, Underwriting relationships and analysts’
earnings forecasts and investment recommendations, Journal of Accounting and
Economics 25, 101-127.

Ljungqvist, Alexander, Felicia Marston, and William J. Wilhelm, 2005, Competing for
securities underwriting mandates: Banking relationships and analyst recommendations,
Journal of Finance, forthcoming.

Ljungqvist, Alexander, Vikram Nanda, and Rajdeep Singh, 2003, Hot markets, investor
sentiment, and IPO pricing, Working paper, NYU Stern School of Business.

Ljungqvist, Alexander, and William Wilhehn Jr., 2005, Does prospect theory explain IPO
market behavior?, Journal of Finance 60, 1759-1790.

Loughran, Tim, and Jay Ritter, 2004, Why has IPO underpricing changed over time?,
Financial Management 33, 5-37.

Lyon, John, Brad Barber, and Chih-Ling Tsai, 1999, Improved methods for tests of long-
run abnormal stock returns, Journal of Finance 54, 165-201.

145

Michaely, Roni and Womack, Kent, 1999, Conﬂict of interest and the credibility of
underwriter analyst recommendations, Review of Financial Studies 12, 653-686.

Miller, Edward M., 1977, Risk, uncertainty, and divergence of opinion, Journal of
Finance 32,1151-1168.

Mitchell, Mark L., and J. Harold Mulherin, 1994, The impact of public information on
the stock market, Journal of Finance 49, 923-950.

Myers, Stewart C., and Nicolas S. Majluf, 1984, Corporate ﬁnancing and investment
decisions when ﬁrms have information that investors do not have, Journal of Financial
Economics 13, 187-221.

Pastor, Lubos and Pietro Veronesi, 2003, Stock valuation and learning about proﬁtability,
Journal of Finance 58, 1749-1789.

Petersen, Mitchell, 2005, Estimating Standard Errors in Finance Panel Data Sets:
Comparing Approaches, working paper, Northwestern University.

Puri, Manju, 1999, Commercial banks as underwriters: implications for the going public
process, Journal of Financial Economics 54, 133-163.

Puri, Manju, 1996, Commercial banks in investment banking conﬂict of interest or
certiﬁcation role?, Journal of Financial Economics 40, 373-401.

Reese, William A., 2003. [PO underpricing, trading volume, and investor interest.
Working paper, Tulane University.

Ritter, Jay R., and Donghang Zhang, 2005, Afﬁliated mutual funds and the allocation of
initial public offerings, working paper, University of South Carolina

Roll, Richard, 1988, R2, Journal of Finance 43, 541-566.

Schleifer, Andrei and Robert W. Vishny, 1997, The limits of arbitrage, Journal of
Finance 52, 35-55.

Schrand, Catherine and Robert E. Verrecchia, 2005, Information disclosure and adverse
selection explanations of IPO underpricing, Working paper, University of Pennsylvania.

Seligman, Joel, 1995, The Transformation of Wall Street, Northeastern University Press.

Shiller, Robert J ., 1981, Do stock price move too much to be justiﬁed by subsequent
changes in dividends?, American Economic Review 71, 421-436.

146

Venkatraman, Kumar, Rex Thompson, and Venkat Eleswarapu, 2004, The impact of
regulation fair disclosure: Trading costs and information asymmetry, Forthcoming,
Journal of Financial and Quantitative Analysis.

Vuolteenaho, Tuomo, 2002, What drives ﬁrm-level stock retums?, Journal of Finance
57, 233-264.

Wei, Steven X., Chu Zhang, 2004, Why did individual stocks become more volatile?
Forthcoming, Journal of Business.

Wermers, Russ, 1999, Mutual fund herding and the impact on stock prices, Journal of
Finance 54, 581-622.

Xu, Yexiao, and Burton G. Malkiel, 2003, Investigating the behavior of idiosyncratic
volatility, Journal of Business 76, 613-644.

147