University of Southern California Dissertations and Theses

How do securities analysts respond to different types of news events?

(USC Thesis Other)

How do securities analysts respond to different types of news events?

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content HOW DO SECURITIES ANALYSTS RESPOND
TO DIFFERENT TYPES OF NEWS EVENTS?
by
Ying Ying Terry Wang
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)
May 2011
Copyright 2011 Ying Ying Terry Wang
Dedication
To those who believed in me
ii
Acknowledgements
I thank my dissertation committee, Sarah Bonner, Chris Jones, Dave Maber,
Bob Trezevant, and Fernando Zapatero, for their continuous guidance and help-
ful discussions. I am especially grateful to my co-chairs, Sarah Bonner and Bob
Trezevant, for their endless encouragement and unconditional support. I also thank
David Erkens, TK Kim, Yaniv Konchitchki, Suresh Nallareddy, Tatiana Sandino,
Tom Scott, Jesus Sierra, Doug Skinner, Irem Tuna, Mike Welker, Mike Willenborg,
and workshop participants at University of Southern California, Miami University,
Singapore Management University, National University of Singapore, University of
Alberta, Queen's University, London School of Economics and Political Science, and
London Business School for helpful comments and discussions.
iii
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables v
Abstract vi
Chapter 1 Introduction 1
Chapter 2 Literature Review 6
Chapter 3 Data and Sample Selection 16
Chapter 4 Empirical Results 23
Chapter 5 Conclusions 49
References 51
Appendices 57
Appendix 1 List of Sample Firms 57
Appendix 2 Sources of News Articles 58
Appendix 3 Description of News Categories 59
iv
List of Tables
Table 3.1 Summary of Sample Selection Procedures 17
Table 4.1 Descriptive Statistics for Sample Firms 24
Table 4.2 Distribution of News Events 38
Table 4.3 Descriptive Statistics for Analyst Responsiveness 39
Table 4.4 Correlation Matrix for Analyst Responsiveness and News Events 40
Table 4.5 Correlation Matrix for Analyst Responsiveness and Control
Variables 41
Table 4.6 The Logistic Model for = 3, All News Events 42
Table 4.7 The Logistic Model for = 3, Counting Earnings Announce-
ments as the Only News Event on Earnings Announcement
Day 43
Table 4.8 The Logistic Model for = 3, News Events with LargejCARj 44
Table 4.9 The Logistic Model for = 5, All News Events 45
Table 4.10 The Logistic Model for = 5, Counting Earnings Announce-
ment as the Only News Event on Earnings Announcement Day 46
Table 4.11 The Logistic Model for = 5, News Events with LargejCARj 47
Table 4.12 The Discrete-Time Duration Model 48
v
Abstract
Using a unique database hand-collected from approximately 5,000 news articles,
I examine securities analysts' responsiveness to dierent types of news events. I nd
that analysts are most responsive to earnings announcements and less responsive
to management guidance, operations updates, as well as news concerning investing
activities, nancing activities, credit ratings, and analyst activities. It is also in-
teresting to note that analysts are more responsive to management guidance than
they are to operations updates, as well as news concerning investing activities and
nancing activities. These results are robust even if only economically signicant
news events are considered and when dierent measures of analyst responsiveness
are used.
vi
Chapter 1
Introduction
This study examines securities analysts' responsiveness to dierent types of news
events.
1
Analyst responsiveness can be dened as: (i) the likelihood that a securities
analyst will revise her earnings forecast for a given rm following a news event, or
(ii) the time it takes for an analyst to revise her earnings forecast for a given rm
following a news event. In this study, I investigate the dierential likelihood of
analysts' earnings forecast revisions following various types of news events. To the
best of my knowledge, this is the rst study that empirically tests the relation
between analyst responsiveness and types of news events.
I examine analysts' responsiveness to nine types of rm-specic news events:
(i) earnings announcements; (ii) management guidance; (iii) operations updates;
(iv) news concerning investing activities; (v) news concerning nancing activities;
(vi) news concerning credit ratings; (vii) news concerning analyst activities; (viii)
news concerning management issues; and (ix) any other news events that do not
fall under the previous categories. These news events display dierent levels of
information complexity, which I dene as the diculty of translating the eect of a
certain news event into changes in future earnings. According to the judgment and
decision-making literature, information complexity typically has a negative eect on
the quality of people's judgments and decisions (Bonner (2008)) and, in this case,
responsiveness.
1
More precisely, this study examines analysts' responsiveness to dierent types of information
conveyed by news events. Throughout this paper, I use the term news event to refer to information
conveyed by a news event.
1
Specically, information complexity can negatively aect analyst responsiveness
through two channels. First, it can tax an analyst's computational speed constraints
and thus the analyst is simply unable to process complex information in a timely
manner. Second, the analyst may choose not to incorporate complex information
in her forecast because the costs of processing the information outweigh the bene-
ts. The nine types of news events examined in this study carry dierent levels of
information complexity, which can arise from a lack of information precision, the
degree of credibility of the information provider, and/or the number of steps that an
analyst has to go through to incorporate the information in her earnings forecasts.
I therefore argue that these nine types of news events have dierential eects on
analyst responsiveness.
This paper is motivated as follows. First, until very recently, academic research
has primarily focused on examining analysts' use of past earnings (e.g., DeBondt and
Thaler (1990); Abarbanell and Bernard (1992); Ali, Klein, and Rosenfeld (1992))
and management earnings guidance (e.g., Waymire (1986); Hassell, Jennings, and
Lasser (1988)) in forecasting future earnings. However, information available to an-
alysts extends beyond earnings announcements and management earnings guidance
(Bradshaw (2008)) and analysts are responsible for collecting all pertinent publicly
available information about a rm and its business, including, but not limited to,
\nancial statements, research on the rm, industry, product or sector, and public
statements by and interviews of executives of the rm, its customers and suppli-
ers" (Fernandez (2001)). It is therefore important to understand how analysts use
dierent types of information in arriving at their earnings forecasts.
Second, earnings forecasting is a continuous process and requires constant up-
dating as major events occur. As a November 1998 SEC proposal states, analysts
\actively pursue new information, put all of it into context, and act as conduits
in the
ow of information" to investors (Fernandez (2001)). Although this state-
ment suggests that analysts are constantly incorporating new information into their
2
earnings forecasts, academic researchers nd evidence to the contrary (e.g., Stickel
(1989)). One of the more direct approaches to nd out whether analysts incorporate
the most up-to-date information into their forecasts is to investigate whether they
revise their forecasts following certain important news events and how long it takes
them to do so.
Understanding how responsive analysts are to dierent types of news events
is important for several reasons. First, analyst forecasts have long been used as
proxies for market earnings expectations and have been shown to outperform time-
series models as proxies (Fried and Givoly (1985)). Brown, Hagerman, Grin, and
Zmijewski (1987) conclude that such superiority in performance is attributable to
analysts' better use of information at the time-series forecast date and their use
of available information after the time-series forecast date. Learning how analysts
respond to dierent types of news events can provide further insights into analysts'
use of information and help researchers construct better proxies for market earnings
expectations.
Second, responsiveness to news events can serve as an alternative measure of
analyst performance to the traditionally employed forecast accuracy that can only
be measured ex post. One method to evaluate an analyst's performance at any
given point in time, before a proper benchmark (i.e., actual earnings) is available,
is to study whether the analyst has incorporated all the information available at
the time of her forecast, i.e., whether she has responded to all important news
events.
2
On a similar note, analyst forecasts and recommendations are an important
source of information for investors (SRI International (1987)). Prior research has
shown that investors respond to analyst forecast revisions (Stickel (1992); Park
2
This is known as the \process view" of the quality of judgments and decisions. The literature
suggests that the quality of people's judgments and decisions can be evaluated based on either
(i) the extent to which the nal judgment or decision corresponds with some \right nal answer"
(i.e., the performance view) or (ii) the extent to which the judgment or decision-making process
corresponds with some \right process" (i.e., the process view) (Bonner (2008)). In the case of earn-
ings forecasting, the performance view compares an analyst's earnings forecast to actual earnings
to determine its accuracy, while the process view evaluates the analyst's use of information and
forecast models in arriving at her earnings forecast.
3
and Stice (2000); Bonner, Walther, and Young (2003); Clement and Tse (2003);
Gleason and Lee (2003); Bonner, Hugon, and Walther (2007)). For this reason,
when making investment decisions, investors are likely to care about whether an
analyst has incorporated all available information in her forecast. This is especially
important when certain news events are dicult to interpret and investors need to
rely on analysts' professional opinions to facilitate their investment decisions.
In this study, I hand-collect corporate news events from close to 5,000 news arti-
cles about 37 rms in the crude petroleum and natural gas extraction industry (SIC
code 1311) for the period from June 2006 to May 2007. I restrict my sample to the
oil and gas extraction industry for two main reasons. First, Zhang (2008) nds that
analysts covering dierent industries display dierences in responsiveness. Limiting
my sample to one industry provides a more controlled setting and reduces noise.
Second, rms in the oil and gas extraction industry engage in relatively simple op-
erations { the exploration, development, and production of crude petroleum and/or
natural gas { and sell relatively undierentiated and unbranded products. They are,
however, surrounded by complex legal issues, evolving regulatory environments, and
accelerating merger and acquisition (M&A) activities. This presents a simple, yet
rich, context for examining how analysts respond to dierent types of news events.
Consistent with my predictions, I nd that analysts are most responsive to earn-
ings announcements and less responsive to management guidance, operations up-
dates, as well as news concerning investing activities, nancing activities, credit
ratings, and analysts' activities. While no prior prediction is made, it is interesting
to nd that analysts are more responsive to management guidance than they are
to operations updates as well as news concerning investing activities and nancing
activities. These results are robust even if only economically signicant news events
are considered and when dierent measures of analyst responsiveness are used.
4
This paper is only a rst step in studying analysts' decision processes and will be
extended to include various analyst characteristics, rm characteristics, and manage-
ment characteristics in determining analyst responsiveness, as well as to investigate
whether more responsive analysts are also likely to be more accurate. This paper
contributes to the research on analysts' decision processes called for by Schipper
(1991). Additionally, I add to the literature that investigates the eect of informa-
tion complexity on the quality of analysts' outputs.
The rest of the paper is organized as follows. Chapter 2 discusses the related
literature and develops my hypotheses. Section 3 describes the data and the sample
selection procedures. Section 4 presents the results of my study and Section 5
concludes.
5
Chapter 2
Literature Review
Most of the studies that investigate analysts' use of information take an indirect
approach by examining correlations between the information of interest and ana-
lysts' outputs (e.g., earnings forecasts) and draw conclusions on whether the analysts
have fully incorporated such information, i.e., whether the analysts are ecient in
the use of such information. These studies typically regress the information of in-
terest on analysts' forecast errors and if there exists a statistical relation between
the information and analysts' forecast errors, the analysts are said to be inecient
in incorporating that information. Information examined in earlier studies includes
past earnings (DeBondt and Thaler (1990); Ali et al. (1992); Easterwood and Nutt
(1999)), past stock price changes (Lys and Sohn (1990); Abarbanell (1991)), man-
agement earnings guidance (Jennings (1987); Hassell et al. (1988)), and analysts'
own previous forecast errors (Mendenhall (1991); Abarbanell and Bernard (1992);
Elliott, Philbrick, and Wiedman (1995)). Later studies extend the sources of infor-
mation examined to include management discussion and analysis (MD&A) in annual
and quarterly reports (Barron, Kile, and O'Keefe (1999)), corporate presentations
(Francis, Hanna, and Philbrick (1997)), and conference calls (Bowen, Davis, and
Matsumoto (2002)). Several surveys and content analyses also nd that analysts
make use of segmental information and non-nancial information (Previts, Bricker,
Robinson, and Young (1994); Rogers and Grant (1997); Moyes, Saadouni, Jon, and
Williams (2001)).
Studies on analyst responsiveness to any type of news events, on the other hand,
are quite rare. Zhang (2008) nds that analysts have become increasingly responsive
6
to earnings announcements over the period from 1996 and 2002 and that analysts
covering dierent industries display dierences in responsiveness. Charoenrook and
Lewis (2009) nd that analysts are much more likely to revise their earnings fore-
casts following a company-issued announcement after the passage of Regulation Fair
Disclosure. Bagnoli, Levine, and Watts (2005) is the only study that relates analyst
revision activities to types of information; they nd that earnings announcements
tend to trigger greater revision clusters than do management guidance and strategic
disclosures.
In this study, I examine analysts' responsiveness to nine types of news events:
(i) earnings announcements; (ii) management guidance; (iii) operations updates;
(iv) news concerning investing activities; (v) news concerning nancing activities;
(vi) news concerning credit ratings; (vii) news concerning analyst activities; (viii)
news concerning management issues; and (ix) any other news events that do not
fall under the previous categories.
1
These news events display dierent levels of
information complexity, which can arise from a lack of information precision, the
degree of credibility of the information provider, and/or the number of steps that an
analyst has to go through to incorporate the information. The literature concludes
that information complexity typically has a negative eect on the quality of people's
judgments and decisions (Bonner (2008)).
Information complexity can negatively aect analyst responsiveness through two
channels. First, it can tax an analyst's computational speed constraints and thus
the analyst is simply unable to process the information and revise her forecast in
a timely manner. In other words, the likelihood that the analyst will revise her
forecast within a given window following the arrival of new information is lower
when that information is more complex.
1
News events in these categories, such as accounting policy changes, restatements, auditor
changes, SEC inquiries, changes in Dow Jones or S&P index membership, legal issues, environmen-
tal issues, regulations, antitrust, ownership changes, and macro factors, are either insucient in
number or too sketchy to produce an accurate prediction; therefore, no prediction is made for these
categories.
7
Second, information complexity can aect an analyst's assimilation of informa-
tion. From an investor usefulness standpoint, forecast timeliness seems to be at
least as important as forecast accuracy and analysts are rewarded for issuing timely
forecasts (Cooper, Day, and Lewis (2001)). Hence, following the arrival of new in-
formation, an analyst has an incentive to release her updated earnings forecasts as
soon as possible. When faced with a time constraint, it is likely that the analyst will
choose to use a smaller percentage of the available information to reduce her cogni-
tive load so that she can process the information and issue a forecast revision in a
more timely manner. In particular, the analyst is likely to incorporate less complex
information to a greater extent and more complex information to a lesser extent
(Plumlee (2003)). In other words, the analyst is less responsive to more complex
information.
There are several accounting studies that examine the eect of information com-
plexity on the quality of analysts' outputs. Haw, Jung, and Ruland (1994) and
Erwin and Perry (2000) nd that analysts' forecast accuracy decreases after merg-
ers, especially when nancial leverage changes and when there is a low correlation
between the acquirer's and the target's businesses. Similarly, Chaney, Hogan, and
Jeter (1999) nd that analysts' forecast accuracy declines following corporate re-
structurings. Duru and Reeb (2002) and Herrmann, Hope, and Thomas (2008)
nd that international and industry diversication typically have a negative eect
on forecast accuracy. Hirst and Hopkins (1998) and Hirst, Hopkins, and Wahlen
(2004) nd that analysts fail to properly detect and evaluate risks when the infor-
mation needed is not clearly presented. Plumlee (2003) shows that analysts tend
to incorporate less complex information in their forecasts of eective tax rates to
a greater extent than they incorporate more complex information. Most recently,
Lehavy, Li, and Merkley (2009) nd that analysts' forecast accuracy is lower and
their revision response time is longer for rms with less readable corporate 10-K
8
lings. The preceding empirical results are consistent with the ndings in the liter-
ature that information complexity tends to have a negative impact on the quality
of analysts' outputs.
While having a negative eect on the quality of analysts' outputs, information
complexity can also increase the demand for analysts to respond to news events.
When a news event of high complexity occurs, investors are likely to seek analyst'
opinions to facilitate their investment decisions and, as a result, analysts can prot
from selling their outputs to investors. In other words, analysts have an incentive
to respond to news events of high complexity. This is evidenced by the ndings
in Lehavy et al. (2009) that rms with less readable corporate 10-K lings have
higher analyst followings. On the other hand, analysts bear a variety of costs in
processing complex information. First, analysts have to exert greater cognitive eort
when the information is more complex. Second, as discussed previously, information
complexity typically leads to lower output quality, which in turn has a negative
impact on analysts' career outcomes (Mikhail, Walther, and Willis (1999); Hong
and Kubik (2003)). Hence, analysts are likely to respond to news events of high
complexity only when the costs of processing the information do not exceed the
benets. The costs of processing complex information are likely to be lower for
analysts with higher ability and/or more experience, who therefore are more likely
to respond to news events of high complexity.
In arriving at my hypotheses, I refer to the job description of securities analysts
by the Securities Industry Association
2
and assume that securities analysts engage
2
The Securities Industry Association (SIA) was an industry trade association representing se-
curities rms, banks, and asset management companies in the United States, Europe, and Asia.
It later merged with the Bond Markets Association in 2006 to form the Securities Industry and
Financial Markets Association, which, according to Birnbaum (2006), is \the mouthpiece for the
nancial services industry." In a research report dated August 22, 2001, SIA stated: \A securities
analyst, generally employed by a brokerage rm, bank or investment institution, has the princi-
pal task of performing diligent and thorough investigations of specic securities, companies and
industries. The results of these investigations are presented as a research report, which serves as a
basis for making an investment recommendation. Analysts examine all aspects of the current and
prospective nancial condition of certain publicly traded companies. These examinations should
cover all pertinent publicly available information about the company and its businesses. This in-
cludes, but is hardly limited to nancial statements, research on the company, industry, product
9
in fundamental analysis and collect all types of information to ascertain the value
of a rm. Fundamental analysts typically build their own forecast models to esti-
mate a rm's future earnings. These models encompass major facets of the rm's
business(es) and contain various assumptions of the relationships between operating
statistics and earnings.
There are several reasons that earnings announcements should provide an enor-
mous amount of information about past performance and be relatively straightfor-
ward to incorporate into analysts' earnings forecasts. First, earnings announcements
deliver \hard" nancial information that is audited and compiled in accordance with
generally accepted accounting principles and are one of the most important sources
of information for analysts (Stickel (1989); Lys and Sohn (1990)). Second, earnings
announcements typically have an MD&A section in which the management explains
how various factors have aected the rm's performance. This MD&A is considered
as one of the important sources of information to analysts (Barron et al. (1999);
Healy, Hutton, and Palepu (1999)). Third, earnings announcements are usually
accompanied by conference calls, which also increase the information available to
analysts (Bowen et al. (2002)). Taken together, these factors suggest that earnings
announcements provide an enormous amount of information about a rm's past
performance that is relatively straightforward to incorporate into analysts' earnings
forecasts. Therefore, I predict that:
H1: Analysts are likely to be responsive to earnings announcements.
Management guidance is dened here as forward-looking statements issued by
management to provide information about a rm's future nancial and operating
performance, including management's explicit earnings guidance,
3
and is another
important source of information for analysts (Waymire (1986); Jennings (1987);
or sector, and public statements by and interviews of executives of the company, its customers and
suppliers."
3
In my sample, there are very few instances where management provides explicit earnings
guidance. The eect of management earnings guidance on analyst responsiveness is therefore not
discussed separately.
10
Hassell et al. (1988); Miller (2005); Cotter, Tuna, and Wysocki (2006)). Firms in the
oil and gas industry typically issue management guidance that provides production
and reserves forecasts, product pricing estimates, and various cost estimates for
future periods, which can be directly input into analysts' forecast models and be
translated into future earnings forecasts. Although such information appears to be
as straightforward as that contained in earnings announcements, there are at least
two factors the analysts must consider before using such information.
First, the information in management forecasts is not always as precise as that
contained in earnings announcements. For instance, management often provides
range estimates instead of point estimates in their guidance. This lack of precision
increases information complexity, which has a negative eect on the quality of an-
alysts' outputs (Libby, Tan, and Hunton (2006)). Second, Williams (1996) nds
a signicant association between management's prior forecast accuracy and ana-
lysts' subsequent forecast revisions, suggesting that analysts only incorporate man-
agement guidance into their earnings forecasts when they consider such guidance
\useful." Therefore, when incorporating management guidance into their forecasts,
analysts have to take a few extra steps to ascertain the usefulness of the guidance
from management's past forecast accuracy and then to decipher the precision of
the information contained in the guidance. These extra steps increases information
complexity and therefore have a negative impact on the quality of analysts' outputs.
Indeed, Bagnoli et al. (2005) nd that 57% of the revision clusters in their sample
are associated with earnings announcements, while only 11% of the revision clusters
are associated with management guidance. Based on the preceding arguments, I
hypothesize that:
H2: Analysts are likely to be less responsive to management guidance
than they are to earnings announcements.
Management also from time to time provides operations updates to the invest-
ing community to enhance communications between the rm and investors. In the
11
case of the oil and gas industry, these operations updates convey information about
a rm's recent drilling and production activities, reserves replacements, and price
risk management. Such information provides inputs into analysts' forecast models.
However, explanations as to how these new developments are translated into earn-
ings are usually not provided. For instance, if a rm announces that it has recently
completed three wells which are currently producing at 200 barrels of oil per day,
analysts rst will have to ascertain the price of the oil to arrive at revenues, and
then compute costs associated with the production before they can decide on the
impact of this news event on future earnings. These extra steps that the analysts
have to accomplish make the information more complex and given that informa-
tion complexity typically has a negative eect on the quality of analysts' outputs, I
predict that:
H3: Analysts are likely to be less responsive to operations updates than
they are to earnings announcements.
Investing activities are a vital part of a rm's business (Jensen and Ruback
(1983)) and have been shown to be a predictor of future earnings (Wilson (1986);
Ou (1990)). However, the ndings in Abarbanell and Bushee (1997) suggest that
analysts do not incorporate capital expenditure signals in their forecast revisions.
One possible reason is that capital expenditures carry a high level of uncertainty
(Kothari, Laguerre, and Leone (2002)), which makes the process of translating these
expenditures into future earnings highly complex. As suggested by Plumlee (2003),
in the interest of time, analysts may choose not to incorporate such information
in their forecast revisions because the costs of processing the information simply
outweigh the benets.
CAPEX cite (McConnell and Muscarella (1985))
Similarly, M&A transactions have also been shown to have a negative eect on
analysts' forecast accuracy because of the highly complex nature of these transac-
tions. For instance, Haw et al. (1994) and Erwin and Perry (2000) nd that analysts'
12
forecast accuracy declines after mergers, especially when nancial leverage changes
and when there is a low correlation between the acquirer's and the target's busi-
nesses, but the decline in accuracy seems to be temporary and analysts' accuracy
returns to approximately the pre-merger level in four years. Other than the fact that
analysts are likely to take more time to process the impact of M&A transactions
on future earnings because the information is more complex, these ndings also are
consistent with the explanation that analysts spend more time in the short run to
learn strategies to cope with complex information that will ultimately lead to more
accurate forecasts in the long run (Bonner (2008)). In other words, analysts are
likely to be less responsive to complex information in the short run. Considering
the preceding factors, I hypothesize that:
H4: Analysts are likely to be less responsive to news concerning investing
activities than they are to earnings announcements.
Financing is another important corporate decision and includes activities such
as the sale of common stock, preferred stock, and bonds, as well as dividend pay-
outs and share repurchases. Although earlier literature in corporate nance suggests
that management uses dividend payouts to provide information about future cash

ows and earnings (e.g., Linter (1956); Miller and Modigliani (1961); Bhattacharya
(1979)), several later studies nd that dividend payments do not convey informa-
tion beyond what is already re
ected in current earnings (Watts (1973); Gonedes
(1978)) or management forecasts (Penman (1983)). In an attempt to reconcile these
two very dierent ndings, Benartzi, Michaely, and Thaler (1997) show that rms
that increase dividends do not experience subsequent abnormal earnings growth but
are less likely than non-dividend increasing rms to experience future earnings de-
clines, which suggests that management uses dividends to signal that an increase in
concurrent earnings is likely to be \permanent." Similarly, Gustavo and Michaely
(2004) nd that announcements of open-market share repurchase programs are not
followed by an increase in operating performance. Taken together, these ndings
13
seem to suggest that dividends and share repurchases are not good predictors of
future earnings.
On the other hand, Hansen and Crutchley (1990) and Bradshaw, Richardson,
and Sloan (2006) nd that the sale of common and preferred stock and the issuance
of debt are related to future earnings declines. Also, Bradshaw et al. (2006) nd
that analysts tend to issue overoptimistic earnings and long-term growth forecasts
and such overoptimism seems to come from mispricing the values of the stock and
debt with respect to the option pricing theory.
Taken together, the ndings summarized in the two preceding paragraphs sug-
gest that predicting earnings from nancing activities is very dicult. Given the
ndings in Plumlee (2003) that analysts tend not to incorporate more complex in-
formation in their forecasts, I predict that:
H5: Analysts are likely to be less responsive to news concerning nancing
activities than they are to earnings announcements.
Changes in credit ratings, especially downgrades, tend to generate signicant
market reactions. Goh and Ederington (1993) suggest that rating changes due to
changes in the rating agency's evaluation of the rm's nancial prospects provide
new information to investors and thus are likely to generate signicant market re-
actions. Moreover, rating agencies claim to receive inside information unavailable
to securities analysts such as board minutes, prot breakdowns by product, and
new product plans. If securities analysts feel that the change in credit rating likely
re
ects inside information unavailable to them, a change in credit rating could lead
securities analysts to change their earnings forecasts (Ederington and Goh (1998)).
Based on the preceding, I predict that:
H6: Analysts are likely to be less responsive to news concerning credit
ratings than they are to earnings announcements.
14
Analyst revisions are known to cluster (Stickel (1989); Bagnoli et al. (2005)).
One reason is that analysts who revise their forecasts earlier provide additional in-
formation to analysts who revise their forecasts later (Lys and Sohn (1990)). Also,
Li (2006) and Guttman (2008) nd that high-ability analysts revise their forecasts in
a more timely fashion than low-ability analysts, which suggests that low-ability ana-
lysts are likely to follow high-ability analysts' revisions by mimicking their forecasts.
Taken together, these ndings suggest that:
H7: Analysts are likely to be less responsive to news concerning other
analysts' activities than they are to earnings announcements.
In this study, news concerning management issues includes news about changes
of CEO, CFO, other senior management, or board members, as well as news about
management compensation, option exercises, and insider dealings. Although these
news events may generate stock market reactions and media coverage, their eect
on earnings is quite uncertain. Therefore, I hypothesize that:
H8: Analysts are likely to be less responsive to news concerning man-
agement issues than they are to earnings announcements.
15
Chapter 3
Data and Sample Selection
I obtain my data on corporate news events from the Dow Jones Factiva database.
Data on analyst earnings forecasts are obtained from the I/B/E/S Detail History le,
data on stock returns are obtained from the Center for Research in Security Prices
(CRSP), and data on the rms' nancial and segmental information are obtained
from the COMPUSTAT Fundamentals Annual and Segments tables.
Since the news event data must be hand-collected and manually coded, it im-
poses prohibitive costs to obtaining a universal sample of news articles for my anal-
ysis. Given this factor, I choose to limit my sample to one industry and a period
of one year from June 2006 to May 2007. In addition to the practical matter of
hand-collecting data, there is at least one other reason to select only one industry
for this study. In particular, besides having dierences in knowledge and skills,
analysts covering dierent industries also display dierences in responsiveness. For
instance, Zhang (2008) nds that 74% of the analysts covering the business equip-
ment industry revise their earnings forecasts within three trading days of an earnings
announcement, while only 49% of the analysts covering utilities revise their earn-
ings forecasts within three trading days of an earnings announcement. Limiting the
sample to one industry provides a more controlled setting and reduces noise.
I restrict my sample to the crude petroleum and natural gas extraction industry
(SIC code 1311). Firms in this industry engage in relatively simple operations {
the exploration, development, and production of crude petroleum and/or natural
gas { and sell relatively undierentiated and unbranded products. These rms,
however, are surrounded by complex legal issues, evolving regulatory environments,
16
and accelerating merger and acquisition activities. Using the crude petroleum and
natural gas extraction industry presents a simple, yet rich, context for examining
how analysts respond to dierent types of news events.
There are 55 NYSE-traded, US-incorporated rms in the SIC code 1311 during
the sample period. I exclude foreign-incorporated rms and American Depositary
Receipts from my sample as they may have dierent information environments than
US-incorporated rms. Ten of these 55 rms are dropped from the sample because
they are not primarily engaged in oil and gas exploration and production, and one
rm is dropped for having a non-December scal year-end. I further require that
the sample rms (i) are covered by at least two analysts who have issued at least
two earnings forecasts during the sample period and (ii) have at least one year of
trading history prior to the beginning of the sample period. These last two data
requirements reduce the sample size to 38. One of these 38 rms is eventually
dropped from the sample as it does not have sucient news data on Factiva. As
shown in Table 3.1, the sample selection procedure yields a nal sample of 37 NYSE-
traded oil and gas rms, which are covered by 98 analysts, representing a total of
646 analyst-rm pairs. Please refer to Appendix 1 for a list of the sample rms.
Table 3.1
Summary of Sample Selection Procedures
Number
of Firms
NYSE-listed, US-incorporated rms in SIC code 1311 55
Firms not primarily engaged in oil and gas exploration and
production (10)
Firms with non-December scal year-end (1)
Firms with less than two analysts who issue at least two
earnings forecasts during the sample period (2)
Firms with less than one year of trading history (4)
Firms with insucient news data on Factiva (1)
Final sample 37
17
I then use Factiva Intelligent Indexing to search for all news articles related to
these 37 oil and gas rms during the period from June 2006 to May 2007. I limit
the sources of new articles to those originated from the United States, with the
exception of Reuters.
1
These sources include, but are not limited to, Dow Jones
Newswires, Associated Press Newswires, Reuters News, Platts Commodity News,
The Wall Street Journal, The New York Times, and USA Today. Please refer to
Appendix 2 for a complete list of sources of news articles.
2
3.1 News events
Each news article collected from Factiva is coded to correspond to one or more
types of news events. I classify the news events into the following nine categories:
earnings announcements (EARNINGS), which include all nancial information
a rm discloses relating to the release of its annual or quarterly earnings;
management guidance (GUIDANCE), which is dened as forward-looking
statements issued by the management to provide information about a rm's fu-
ture nancial and/or operating performance, including management's explicit
earnings guidance;
operations updates (OPERATIONS), which provide information about a rm's
recent operations such as drilling activities, reserves acquisitions and replace-
ments, production statistics, and price risk management;
news concerning investing activities (INVESTING), which provides informa-
tion about a rm's capital expenditures as well as M&A activities;
1
Reuters is originated from London. However, given that Reuters is one of the world's largest
news services, I felt compelled to include Reuters in the data collection process.
2
To ensure that the news events contained in the news articles are the most current, I randomly
select ve rms and compare their press releases to their SEC Form 8-K \current report" lings. I
nd that all news events reported in the 8-K lings are included in the press releases collected from
Factiva and the dates of the 8-K lings do not precede the dates of the press releases.
18
news concerning nancing activities (FINANCING), which provides informa-
tion about a rm's equity and long-term liabilities nancing;
news concerning credit ratings (RATINGS), which includes changes in and
armations of corporate and bond ratings;
news concerning analyst activities (ANALYSTS), which are related to ana-
lysts' earnings forecast revisions, changes in recommendation, and commen-
taries, as reported in the news articles;
news concerning management issues (MANAGEMENT), such as changes in
management or board members, management compensation, option exercises,
and insider dealings;
other (OTHER), which consists of news events that do not fall under the above
categories.
3
A news article can be coded to fall under more than one category. For instance,
it is common for a rm to announce its quarterly earnings as well as to provide
operations updates for the same quarter and management guidance for the next
quarter in the same press release; this press release is then categorized under earnings
announcements, operations updates, as well as management guidance. I aggregate
the news articles about a given rm by day to obtain the types of news events that
occurred on each day. A news day is dened as a calendar day on which at least one
news event occurs. For instance, if a rm announces an acquisition in a press release
and changes in management in another press release on the same day, the rm is
considered to have only one news day but two news events on that news day. To be
more specic, EARNINGS
j;k
, GUIDANCE
j;k
, OPERATIONS
j;k
, INVESTING
j;k
,
FINANCING
j;k
, RATINGS
j;k
, ANALYSTS
j;k
, MANAGEMENT
j;k
, and OTHER
j;k
equal 1 (and 0 otherwise) when there is a news article related to rm j that falls
under the respective category on news day k.
3
Please refer to Appendix 3 for a detailed description of each category.
19
Since some of the news events may not be economically signicant enough to
elicit analyst response, in order to identify news events that are economically signi-
cant, I calculate the three-day [1; +1] cumulative abnormal return (CAR[1; +1]
j;k
)
centered on news dayk for rmj. A three-day [1; +1] cumulative abnormal return
is used for several reasons. First, some news articles do not have a time stamp and
for those that do have a time stamp, it refers to the local time in the time zone
where the news article is originated. It therefore presents practical diculties in
correctly narrowing down the stock price reaction to within certain trading hours.
Second, having a three-day window allows sucient time for the information to be
re
ected in stock price. CAR[1; +1]
j;k
is estimated from the Fama-French three-
factor model using the CRSP value-weighted market index over 250 trading days
ending 5 days prior to the news day.
3.2 Analyst responsiveness
I use an indicator variable to capture analyst responsiveness. REV()
i;j;k
equals
1 (and 0 otherwise) if analyst i revises her forecast of annual earnings for rm j
for any future years within trading days of news day k. REV()
i;j;k
is dened
to capture the possible delayed eect, if any, of the news events as a news event
may not have an immediate impact on a rm's near-term earnings but may have a
signicant impact on the rm's longer-term earnings. For instance, an acquisition
announced by a rm in November 2006 may not have any impact on the rm's 2006
earnings but is likely to have an impact on the rm's 2007 earnings; as a result,
following this news event, analysts are more likely to revise their forecasts for 2007
than for 2006. Thus, extending REV()
i;j;k
to include analysts' forecast revisions
for any future years allows me to better capture analyst responsiveness, especially
to news events that do not have a near-term eect and/or that occur later in a scal
year.
20
3.3 Control variables
To control for the possible eect of information provided in corporate presenta-
tions (Francis et al. (1997)) and announcements by the Organization of Petroleum
Exporting Countries (OPEC) on analysts' forecast revisions, I also collect the dates
of corporate and conference presentations made by the 37 sample rms from Factiva
and the dates of OPEC announcements from the OPEC website.
4
Similar to the
news event variables discussed in Section 3.1, PRESENTATION
j;k
equals 1 (and 0
otherwise) when rm j has made a presentation on news day k and OPEC
k
equals
1 (and 0 otherwise) when there is an OPEC announcement on news day k. The
three-day [1; +1] cumulative abnormal returns centered on these news days are
calculated in a similar manner to that discussed in Section 3.1.
I also include several rm-level variables to capture the underlying complexity of
the rms. LOSS
j
equals 1 (and 0 otherwise) if rm j suers a loss in the scal year
2006. GEOSEG
j
is the number of geographic segments of rm j reported in 2006.
NONUS
j
is the percentage of rm j's non-US revenue in 2006. I use FOLLOW
j
,
which is the number of analysts following rm j during the period from June 2006
to May 2007, to proxy for rm j's information environment. Finally, to control for
the eect of rm size, I include log(MKTCAP
j
), which is the natural log of market
capitalization of rm j as of December 31, 2006, and SP500
j
, which is an indicator
variable that equals 1 (and 0 otherwise) if rm j is a member of the S&P500 Index
as of December 31, 2006.
Additionally, I use three analyst-level variables to capture an analyst's ability
in responding to news events in a timely manner. IIAA
i
equals 1 (and 0 otherwise)
if analyst i is listed as a member of the Institutional Investor All-America (II AA)
Research Team in 2006. GEXP
i
is the total number of years analyst i has issued
4
http://www.opec.org/opecna/. There are 15 OPEC announcements during the period from
June 2006 to May 2007.
21
forecasts to I/B/E/S, while FEXP
i;j
is the total number of years analysti has issued
forecasts to I/B/E/S on rm j.
22
Chapter 4
Empirical Results
4.1 Descriptive statistics
Descriptive statistics for the sample rms are provided in Table 4.1. The data
indicate a wide range of rm sizes with market capitalization of $1.20 billion at the
lower quartile to $4.82 billion at the upper quartile of the distribution. Mean (me-
dian) market capitalization is $6.04 billion ($2.06 billion). Distributions of several
other measures of rm size, such as total assets, property, plant, and equipment,
total revenue, and net income, also suggest substantial variation in rm size. The
average rm is followed by 17 analysts. Firms in the lower quartile are followed by 9
or fewer analysts, while rms in the upper quartile are followed by 24 or more ana-
lysts. On average, a rm has approximately 134 related news articles and encounters
approximately 40 news days during the sample period.
Distributions of the news events are reported in Table 4.2. Close to 5,000 news
articles are collected for the 37 sample rms, which are translated into 6,250 news
events and 1,470 rm-news days. A majority of the news articles are related to op-
erations updates (27%) and investing activities (27%), while only 19% of the news
articles are related to earnings announcements. Similarly, 27% and 31% of the news
days are attributable to operations updates and investing activities, respectively,
while only 10% of the news days are related to earnings announcements. The aver-
age absolute value of CAR[1; +1] centered on a news day is approximately 2.66%.
On average, analyst activities generate the highestjCAR[1; +1]j, 3.28%, followed
23
Table 4.1
Descriptive Statistics for Sample Firms
Lower Upper Standard
Mean Median Quartile Quartile Deviation
Market capitalization 6,044.57 2,058.77 1,201.08 4,822.45 9,205.21
Total assets 7,417.48 2,139.84 1,199.00 6,971.10 12,363.24
Property, plant, and equipment 6,070.35 1,879.77 1,080.63 6,414.48 9,963.81
Total revenue 2,274.44 761.99 449.55 1,582.05 3,789.54
Net income 669.68 161.57 62.01 597.53 1,187.07
Number of business segments 1.24 1.00 1.00 1.00 0.72
Number of geographical segments 2.00 1.00 1.00 2.00 2.08
Percentage of foreign revenue 11% 0% 0% 18% 21%
Analyst following 17.46 16.00 9.00 24.00 10.01
Number of news articles 130.00 76.00 50.00 94.00 184.47
Number of news days 40.03 32.00 25.00 43.00 29.54
by earnings announcements (3.15%), management guidance (2.91%), operations up-
dates (2.85%), news concerning investing activities (2.79%), and news concerning
credit ratings (2.70%). Other types of news events, such as news concerning nanc-
ing activities and management issues, generate a lower than averagejCAR[1; +1]j.
1
All of the means are greater than the medians, indicating a skewed distribution of
jCAR[1; +1]j.
Table 4.3 provides the descriptive statistics for analyst responsiveness. Panel A
presents the statistics of the time elapsed between a particular type of news event
and analysts' rst forecast revisions immediately following that news event. As
shown in Panel A, it takes an average of approximately 10 trading days for ana-
lysts to revise their earnings forecasts following an earnings announcement. This is
similar to the ndings in Zhang (2008). Additionally, it takes an average of approx-
imately 13, 15, and 16 trading days for analysts to revise their earnings forecasts
following management guidance, operations updates, and news concerning investing
activities, respectively. For other types of news events, it can take up to an average
1
The medians show a qualitatively similar distribution and will not be discussed further.
24
of 20 trading days for analysts to revise their forecasts. The large dierences be-
tween the means and the medians also suggest a skewed distribution, indicating that
some analysts only revise their forecasts after a considerable period of time follow-
ing a news event. Panel B shows the percentage of analysts revising their forecasts
following each type of news events within a revision window. Consistent with the
descriptive statistics presented in Panel A, 64% (71%) of the analysts revise their
earnings forecasts within three (ve) trading days following an earnings announce-
ment and the percentage drops to 49% (55%) for management guidance, 35% (43%)
for operations updates, 33% (41%) for news concerning investing activities, and to
less than 20% (30%) for other types of news events. These descriptive statistics
suggest that, as predicted, analysts are most responsive to earnings announcements
and are less responsive to other types of news events. Also, not unexpectedly, as the
revision window increases from three to ve trading days, the percentage of analysts
revising their forecasts increases.
Table 4.4 provides the correlations between analyst responsiveness and categories
of news events and table 4.5 provides the correlations between analyst responsive-
ness and rm-level and analyst-level control variables. As mentioned previously, it
is common for a rm to announce its quarterly earnings and at the same time pro-
vide operations updates for the current quarter and management guidance for the
next quarter. It is therefore not surprising that EARNINGS is positively and signif-
icantly correlated with OPERATIONS ( =:5864, two-tailed p-value <:0001) and
with GUIDANCE ( =:5413, two-tailed p-value <:0001), and that GUIDANCE is
positively and signicantly correlated with OPERATIONS ( =:5260, two-tailed p-
value<:0001). It is also common for a rm to announce its capital expenditure bud-
get and provide management guidance for the next quarter/year in the same press re-
lease; hence, there is a positive and signicant correlation between INVESTING and
GUIDANCE ( =:4484, two-tailed p-value<:0001). Additionally, a rm sometimes
discloses its actual capital expenditure spending at the same time when it announces
25
its quarterly earnings and provides operations updates. Hence, INVESTING is pos-
itively and signicantly correlated with EARNINGS ( =:3843, two-tailed p-value
< :0001) and OPERATIONS ( = :3826, two-tailed p-value < :0001). All other
types of news events do not exhibit high or signicant correlations with each other.
2
4.2 The logistic model
To investigate the dierential likelihood of analysts' earnings forecast revisions
within a given window following various types of news events, I estimate the following
regression:
log

P (REV()
i;j;k
= 1)
1P (REV()
i;j;k
= 1)

=0 +1EARNINGS
j;k
+2GUIDANCE
j;k
+3OPERATIONS
j;k
+4INVESTING
j;k
+5FINANCING
j;k
+6RATINGS
j;k
+7ANALYSTS
j;k
+8MANAGEMENT
j;k
+9OTHER
j;k
+10PRESENTATION
j;k
+11OPEC
k
+12 log(MKTCAPj )
+13SP500j +14LOSSj +15GEOSEGj
+16NONUSj +17FOLLOWj +18IIAAi
+19GEXPi +20FEXPi;j
(4.1)
Given that more than 60% of analysts revise their earnings forecasts within
three trading days following an earnings announcement (Panel B of Table 4), I set
= 3. To ensure the robustness of the responsiveness measure, I later also estimate
equation (4.1) with = 5. Based on my hypothesis, I expect:

1
>
2
;
3
;
4
;
5
;
6
;
7
;
8
.
2
The correlations between EARNINGS, GUIDANCE, OPERATIONS, and INVESTING are
not expected to create multicollinearity problem in the logistic regression discussed in the next
section. The variance in
ation factors for EARNINGS, GUIDANCE, OPERATIONS, and
INVESTING are all below 2.
26
Since the sample rms are selected to have December scal year-end only, earn-
ings announcements of these rms are likely to be clustered. Additionally, it is
known that analyst revisions also tend to be clustered (Stickel (1989); Bagnoli et al.
(2005)). I therefore calculate the two-way clustered standard errors, by rm and by
analyst, to adjust for the correlation across rms and across analysts.
Column (1) of Table 4.6 reports the results from estimating equation (4.1) for
= 3. The coecients on EARNINGS, GUIDANCE, OPERATIONS, RATINGS,
and ANALYSTS are positive and signicant (p-value<:0001) and the coecient on
FINANCING is negative and signicant (p-value<:0001). As predicted, the coe-
cient on EARNINGS is greater than the coecients on GUIDANCE, OPERATIONS,
FINANCING, RATINGS, and ANALYSTS (at the 5% condence interval). More
specically, the estimated odds of an analyst revising her forecast within three trad-
ing days following an earnings announcement are about 484% higher than that
following other types of news events.
While no prior prediction is made, it is interesting to nd that analysts are
more responsive to management guidance than they are to operations updates and
news concerning nancing activities (at the 5% condence interval) and that they
are least responsive to news concerning nancing activities (at the 5% condence
interval). More specically, the estimated odds of an analyst revising her forecast
within three trading days following management guidance are about 80% higher
than that following other types of news events. On the other hand, the estimated
odds of an analyst revising her forecast within three trading days following news
concerning nancing activities are about 23% lower than that following other types
of news events.
Finally, the coecient on SP500 is negative and signicant (p-value < :05),
suggesting that analysts are less likely to respond to news events of S&P 500 rms.
The coecient on IIAA is positive and signicant (p-value < :05), indicating that
27
II AA ranked analysts are more responsive to news events than non-II AA ranked
analysts.
Since news concerning investing activities consists of news concerning both cap-
ital expenditures and M&A activities (and sometimes negotiations and rumors of
M&A), analysts may respond dierently to dierent types of investing activities.
I therefore separate INVESTING into CAPEX, MERGERACQ, and OTHERINV,
where CAPEX, MERGERACQ, and OTHERINV equal 1 (and 0 otherwise) when
there are news articles related to capital expenditures, conrmed M&A transac-
tions, and other investing activities that do not fall under the previous two cate-
gories, respectively, on a news day. Also, to better understand the eect of debt,
preferred stock, and common stock nancing, I similarly separate FINANCING into
DEBT, PREFERRED, EQUITY, and OTHERFIN, where DEBT, PREFERRED,
EQUITY, and OTHERFIN equal 1 (and 0 otherwise) when there are news arti-
cles related to debt nancing (including loans), preferred stock, common stock, and
other types of nancing activities that do not fall under the previous three cat-
egories, respectively, on a news day. To investigate the eect of news concerning
dierent types of investing activities and nancing activities, I estimate the following
regression with = 3:
28
log

P (REV()
i;j;k
= 1)
1P (REV()
i;j;k
= 1)

=0 +1EARNINGS
j;k
+2GUIDANCE
j;k
+3OPERATIONS
j;k
+4aCAPEX
j;k
+
4b
MERGERACQ
j;k
+4cOTHERINV
j;k
+5aDEBT
j;k
+
5b
PREFERRED
j;k
+5cEQUITY
j;k
+
5d
OTHERFIN
j;k
+6RATINGS
j;k
+7ANALYSTS
j;k
+8MANAGEMENT
j;k
+9OTHER
j;k
+10PRESENTATION
j;k
+11OPEC
k
+12 log(MKTCAPj ) +13SP500j
+14LOSSj +15GEOSEGj +16NONUSj
+17FOLLOWj +18IIAAi +19GEXPi
+20FEXPi;j
(4.2)
The results of estimating equation (4.2) for = 3 are reported in Column (2)
Table 4.6. Similarly, two-way clustered standard errors, by rm and by analyst, are
calculated to adjust for correlation across rms and analysts. Although the coef-
cient on INVESTING is not statistically signicant in estimating equation (4.1),
the coecient on MERGERACQ is now positive and signicant (p-value < :01),
while the coecients on CAPEX and OTHERINV remain statistically insigni-
cant.
3
However, even for conrmed M&A transactions, analysts are still less likely
to revise their forecasts when compared to earnings announcements. On the other
hand, DEBT and EQUITY appear to be driving the results previously found for
FINANCING, i.e., the coecients on DEBT and EQUITY remain negative and
signicant (p-value <:01 and p-value <:05, respectively), while the coecients on
PREFERRED and OTHERFIN are insignicant.
3
This is consistent with the ndings in Abarbanell and Bushee (1997) that analysts do not
incorporate capital expenditure signals in their forecast revisions.
29
The coecients on GEXP and FEXP are positive but statistically insignicant in
estimating both equations (4.1) and (4.2). Mikhail, Walther, and Willis (1997) nd
that analysts improve their forecast accuracy with rm-specic experience, regard-
less of their general and industry experience. I therefore drop GEXP from equation
(4.2) and re-estimate the following regression:
log

P (REV()
i;j;k
= 1)
1P (REV()
i;j;k
= 1)

=0 +1EARNINGS
j;k
+2GUIDANCE
j;k
+3OPERATIONS
j;k
+4aCAPEX
j;k
+
4b
MERGERACQ
j;k
+4cOTHERINV
j;k
+5aDEBT
j;k
+
5b
PREFERRED
j;k
+5cEQUITY
j;k
+
5d
OTHERFIN
j;k
+6RATINGS
j;k
+7ANALYSTS
j;k
+8MANAGEMENT
j;k
+9OTHER
j;k
+10PRESENTATION
j;k
+11OPEC
k
+12 log(MKTCAPj ) +13SP500j
+14LOSSj +15GEOSEGj +16NONUSj
+17FOLLOWj +18IIAAi +20FEXPi;j
(4.3)
Similarly, I calculate two-way clustered standard errors, by rm and by analyst,
to adjust for correlation across rms and analysts. The results of estimating equation
(4.3) for = 3 are reported in Column (3) Table 4.6. While the coecients on
other variables remain approximately the same, the coecient on FEXP becomes
signicant at the p-value <:1 level.
The coecient on MANAGEMENT is, however, not statistically signicant in
estimating equations (4.1), (4.2), or (4.3), suggesting that analysts' forecast revisions
are not attributable to news concerning management issues.
As stated previously in Section ??, information complexity can aect an ana-
lyst's assimilation of information. When faced with a time constraint, it is likely
that the analyst will choose to use a smaller percentage of the information available
30
to reduce her cognitive load so that she can process the information in a more timely
manner. Specically, the analyst is likely to incorporate less complex information
to a greater extent and more complex information to a lesser extent in her fore-
cast revision. Since earnings announcements are often accompanied by other types
of news events, such as management guidance, operations updates, and news con-
cerning investing activities, and earnings announcements are less complex compared
to these other types of news events, it is likely that, in order to be able to revise
her forecast in a more timely manner, the analyst will incorporate in her forecast
revision information conveyed by the the earnings announcement only. If, by any
chance, analysts respond only to earnings announcements and do not respond to any
other types of news events at all, including other types of news events on an earn-
ings announcement day may be erroneously driving the results previously found for
other types of news events such as management guidance and operations updates.
To rule out this possibility, I re-estimate equations (4.1), (4.2), and (4.3) for
= 3, counting earnings announcement as the only news event on an earnings
announcement day and also adjusting standard errors for correlation across rms
and across analysts. The results, presented in Table 4.7, are qualitatively similar to
those presented in Table 4.6. The coecients on all variables remain statistically
signicant, thus ruling out the possibility that the results found previously are driven
by model misspecication.
One may argue that some of the news events are not economically signicant
and analysts respond only to news events that are economically signicant and
the results may not hold once economic signicance is controlled for. In other
words, analysts may respond to all types of news events equally, as long as they are
economically signicant and the results previously found are driven by the fact that
some types of news events are more economically signicant than others. To rule
out this possibility, I re-estimate equations (4.1), (4.2), and (4.3), for = 3, using
only news events that generate ajCAR[1; +1]j of at least 2% on the news day, also
31
adjusting standard errors for correlation across rms and across analysts. I choose
2% since it is close to the median of thejCAR[1; +1]j of all news days (Table 3).
The results, presented in Table 4.8, are qualitatively similar to those presented
in Table 4.6, with the coecient on EARNINGS greater than the coecients on
GUIDANCE, OPERATIONS, MERGERACQ, FINANCING, RATINGS, and ANALYSTS
(at the 5% condence interval). This rules out the possibility that analysts respond
to all economically signicant news events equally and that the results found previ-
ously are driven by failing to control for economic signicance.
To ensure robustness of the analyst responsiveness measure, I repeat all of the
above estimates by using = 5. The results are presented in Tables 4.9, 4.10, and
and are qualitatively similar to those using = 3.
In summary, the results presented above are consistent with ndings in the lit-
erature that information complexity has a negative eect on the quality of analysts'
judgments and decision. Earnings announcements, being the least complex infor-
mation, are most likely to elicit analyst response, while other types of news events,
such as management guidance, operations updates, and news concerning investing
activities are less likely to elicit analyst response as their eects are more dicult
to be translated into changes in future earnings.
4.3 The discrete-time duration model
One may argue that analysts do not respond to individual news events. Instead,
they accumulate information to a point where they feel that there is a signicant
change in the rm's future earnings and they revise their forecasts at that point.
To test this conjecture, I use a discrete-time duration model (Allison (1982)). More
specically, I estimate the following regression:
32
log

P (REVi;j;t = 1)
1P (REVi;j;t = 1)

=
0 +
1EARNINGS(n)i;j;t +
2GUIDANCE(n)i;j;t
+
3OPERATIONS(n)i;j;t +
4INVESTING(n)i;j;t
+
5FINANCING(n)i;j;t +
6RATINGS(n)i;j;t
+
7ANALYSTS(n)i;j;t +
8MANAGEMENT(n)i;j;t
+
9OTHER(n)i;j;t +
10PRESENTATION(n)i;j;t
+
11OPEC(n)i;j;t +
12 log(MKTCAPj )
+
13SP500j +
14LOSSj +
15GEOSEGj
+
16NONUSj +
17FOLLOWj +
18IIAAi
+
19GEXPi +
20FEXPi;j
(4.4)
where REV
i;j;t
is an indicator variable, which equals 1 (and 0 otherwise) if analysti
revises her forecast of annual earnings for rmj for any future years on any trading
dayt, and the independent variables are the cumulative number of news days related
to each respective news categories between analysti's last forecast revision and any
trading day t. This model takes into account the accumulation of information by
counting the number of news events in each news category that has been available
to an analyst since her last forecast revision and determines whether this set of new
information is sucient to trigger the analyst to revise her earnings forecast and
more importantly, cross-sectionally, which types of news events are more likely to
trigger analyst revisions.
To investigate the eect of news concerning dierent types of investing and
nancing activities (as before), I also estimate the following regression:
33
log

P (REVi;j;t = 1)
1P (REVi;j;t = 1)

=
0 +
1EARNINGS(n)i;j;t +
2GUIDANCE(n)i;j;t
+
3OPERATIONS(n)i;j;t +
4aCAPEX(n)i;j;t
+
4b
MERGERACQ(n)i;j;t +
4cOTHERINV(n)i;j;t
+
5aDEBT(n)i;j;t +
5b
PREFERRED(n)i;j;t
+
5cEQUITY(n)i;j;t +
5d
OTHERFIN(n)i;j;t
+
6RATINGS(n)i;j;t +
7ANALYSTS(n)i;j;t
+
8MANAGEMENT(n)i;j;t +
9OTHER(n)i;j;t
+
10PRESENTATION(n)i;j;t +
11OPEC(n)i;j;t
+
12 log(MKTCAPj ) +
13SP500j +
14LOSSj
+
15GEOSEGj +
16NONUSj +
17FOLLOWj
+
18IIAAi +
19GEXPi +
20FEXPi;j
(4.5)
Similarly, to examine the eect of an analyst's general and rm-specic experi-
ence on her responsiveness, I also estimate the following regression:
log

P (REVi;j;t = 1)
1P (REVi;j;t = 1)

=
0 +
1EARNINGS(n)i;j;t +
2GUIDANCE(n)i;j;t
+
3OPERATIONS(n)i;j;t +
4aCAPEX(n)i;j;t
+
4b
MERGERACQ(n)i;j;t +
4cOTHERINV(n)i;j;t
+
5aDEBT(n)i;j;t +
5b
PREFERRED(n)i;j;t
+
5cEQUITY(n)i;j;t +
5d
OTHERFIN(n)i;j;t
+
6RATINGS(n)i;j;t +
7ANALYSTS(n)i;j;t
+
8MANAGEMENT(n)i;j;t +
9OTHER(n)i;j;t
+
10PRESENTATION(n)i;j;t +
11OPEC(n)i;j;t
+
12 log(MKTCAPj ) +
13SP500j +
14LOSSj
+
15GEOSEGj +
16NONUSj +
17FOLLOWj
+
18IIAAi +
20FEXPi;j
(4.6)
The results of estimating equations (4.4), (4.5) and (4.6) are reported in Columns
(1), (2), and (3) of Table 4.12, respectively. Two-way clustered standard errors, by
34
rm and by analyst, are also calculated to adjust for correlation across rms and
across analysts. Similar to the results presented in Section 4.2, the coecients on
EARNINGS(n), GUIDANCE(n), OPERATIONS(n), and ANALYSTS(n) are posi-
tive and signicant (p-value<:05 or lower).
4
These results suggest that while earn-
ings announcements, management guidance, operations updates, as well as news
concerning analyst activities are all likely to trigger analyst forecast revisions, earn-
ings announcements, having the highest coecient, are most likely to trigger analyst
forecast revisions. More specically, one additional news day related to earnings an-
nouncement is associated with at least 153% increase in the predicted odds of analyst
revision, while one additional news day related to management guidance, operations
updates, and news concerning analyst activities only increases the predicted odds
of analyst revision by at least 12%, 6%, and 9%, respectively.
The coecient on SP500 is negative and signicant (p-value<:01) in estimating
equations (4.4), (4.5) and (4.6), consistent with the results presented in Section
4.2 that analysts are less likely to respond to news events of S&P500 rms. The
coecient on IIAA is positive and signicant (p-value < :0001), suggesting that
II AA ranked analysts are more likely to respond to news events than non-II AA
ranked analysts. Additionally, the coecient on FEXP is positive and signicant
(p-value < :01), while the coecient on GEXP is insignicant. This suggests that
analysts with more rm-specic experience are more likely to respond to news events.
In summary, these results are consistent with the results presented in Section
4.2 that information complexity has a negative eect on analyst responsiveness.
4
The coecient onGUIDANCE(n) in estimating equation (4.4) is only signicant at thep-value
<:1 level.
35
4.4 Robustness Tests
4.4.1 Using an aggregate measure of analyst responsiveness
Given the argument in Lehavy et al. (2009) that information complexity creates
demand for analysts to respond to new events, it is possible that my results are
driven by changes in analyst following during my sample period. I therefore use
an aggregate measure of analyst responsiveness in estimating the eect of types of
news events. Specically, REVP()
j;k
equals the percentage of analysts revising
their forecasts of annual earnings for rm j for any future years within trading
days of news day k. I then estimate the following regression using = 3 and 5:
REVP()
j;k
=0 +1EARNINGS
j;k
+2GUIDANCE
j;k
+3OPERATIONS
j;k
+4INVESTING
j;k
+5FINANCING
j;k
+6RATINGS
j;k
+7ANALYSTS
j;k
+8MANAGEMENT
j;k
+9OTHER
j;k
+10PRESENTATION
j;k
+11OPEC
k
+12 log(MKTCAPj )
+13SP500j +14LOSSj +15GEOSEGj +16NONUSj
+17FOLLOWj
(4.7)
The results (not tabulated) are qualitatively similar to those presented in Section
4.2. Therefore, my primary conclusions are robust when an aggregate measure of
analyst responsiveness is used.
4.4.2 Expanding the measure of analyst responsiveness
Since certain news events may not have an impact on earnings (e.g., changes in
credit ratings) but may directly aect an analyst's estimate of a rm's long-term
growth rate, recommendation, or price target estimate, it is possible that my current
responsiveness measure does not fully capture an analyst's response to news events.
Hence, I expand my responsiveness measure to include revisions of an analyst's
36
long-term growth rate forecast, recommendation, and price target. Specically,
REV()
i;j;k
equals 1 (and 0 otherwise) if analysti (i) revises her forecasts of annual
earnings for rmj for any future years, (ii) revises her forecasts of rmj's long-term
growth rate, (iii) changes her recommendation on rm j, or (iv) revises her price
target for rm j within trading days of news day k. I then re-estimate equations
(4.1), (4.2), and (4.2) using = 3 and 5 and nd results qualitatively similar to
those presents in Panels A1 and B1 of Table 6. Thus, my primary conclusions are
robust to dierent measures of analyst responsiveness.
37
Table 4.2
Distribution of News Events
jCAR[1; +1]j Centered on News Day (%)
Number of Number of Lower Upper Standard
Articles News Days Word Count Mean Median Quartile Quartile Deviation
Total number 4,965 1,470 2.77 2.08 0.97 3.82 2.74
Related to:
Earnings announcements 958 (19.3%) 149 (10.1%) 3; 920 3.31 2.41 1.05 4.60 3.00
Management guidance 626 (12.6%) 242 (16.5%) 2; 886 3.05 2.29 1.05 4.05 2.77
Operations updates 1,318 (26.6%) 390 (26.5%) 2; 152 2.99 2.33 0.98 4.00 2.62
Investing activities 1,348 (27.2%) 462 (31.4%) 2; 035 2.73 1.88 0.94 3.82 2.56
Financing activities 566 (11.4%) 276 (18.8%) 896 2.44 1.76 0.96 3.16 2.17
Credit ratings 142 (2.9%) 92 (6.3%) 1; 223 2.67 2.18 1.22 4.07 1.86
Analyst activities 270 (5.4%) 232 (15.8%) 318 3.47 2.50 1.25 4.74 3.06
Management issues 314 (6.3%) 172 (11.7%) 362 2.67 2.24 1.08 3.94 2.05
Other 708 (14.3%) 536 (36.5%) 584 2.77 1.93 0.93 3.74 3.29
38
Table 4.3
Descriptive Statistics for Analyst Responsiveness
Panel A: Number of trading days between a news day and analysts' rst
forecast revisions
Lower Upper Standard
Type of News Event Mean Median Quartile Quartile Deviation
Earnings announcements 11.6 2:0 1.0 11.0 24.3
Management guidance 14.3 3:0 1.0 18.0 24.3
Operations updates 16.7 8:0 1.0 23.0 23.7
Investing activities 18.6 10:0 2.0 25.0 26.1
Financing activities 22.2 15:0 5.0 30.0 25.5
Credit ratings 21.0 12:0 4.0 27.0 25.4
Analyst activities 19.7 13:0 4.0 26.0 25.0
Management issues 21.3 15:0 6.0 30.0 22.7
Other 19.3 12:0 5.0 24.0 23.8
Panel B: Percentage of analysts revising forecasts within trading days of
of a news day
Type of News Event = 3 = 5
Earnings announcements 66.5 69.8
Management guidance 50.7 55.3
Operations updates 38.2 43.4
Investing activities 34.3 40.8
Financing activities 19.4 25.6
Credit ratings 22.3 28.5
Analyst activities 23.0 30.4
Management issues 16.0 24.2
Other 20.0 29.2
39
Table 4.4
Correlation Matrix for Analyst Responsiveness and News Events
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
(1) REV (3) 1:0000
(2) EARNINGS 0:3143 1:0000
(<:0001)
(3) GUIDANCE 0:2246 0:5213 1:0000
(<:0001) (<:0001)
(4) OPERATIONS 0:2133 0:5430 0:4501 1:0000
(<:0001) (<:0001) (<:0001)
(5) INVESTING 0:1404 0:3998 0:4133 0:3060 1:0000
(<:0001) (<:0001) (<:0001) (<:0001)
(6) FINANCING 0:0144 0:0575 0:0179 0:0276 0:0019 1:0000
(0:0207) (<:0001) (0:0040) (<:0001) (0:7633)
(7) RATINGS 0:0065 0:0345 0:0222 0:0685 0:0160 0:1188 1:0000
(0:2937) (<:0001) (0:0004) (<:0001) (0:0102) (<:0001)
(8) ANALYSTS 0:0161 0:0809 0:0959 0:0887 0:1205 0:0769 0:0430 1:0000
(0:0099) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
(9) MANAGEMENT 0:0306 0:0591 0:0512 0:0868 0:0808 0:0565 0:0500 0:0604 1:0000
(<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
(10) OTHER 0:0135 0:1088 0:0539 0:1121 0:1384 0:1019 0:0295 0:0928 0:1014 1:0000
(0:0301) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
(11) PRESENTATION 0:0470 0:0868 0:0353 0:0685 0:0793 0:0721 0:0168 0:0656 0:0654 0:1027 1:0000
(<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (0:0071) (<:0001) (<:0001) (<:0001)
(12) OPEC 0:0818 0:1704 0:1799 0:2792 0:2391 0:1643 0:1034 0:1949 0:1257 0:2687 0:1403 1:0000
(<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
40
Table 4.5
Correlation Matrix for Analyst Responsiveness and Control Variables
(1) (13) (14) (15) (16) (17) (18) (19) (20) (21)
(1) REV (3) 1:0000
(13) log(MKTCAP) 0:0165 1:0000
(0:0079)
(14) SP500 0:0297 0:8552 1:0000
(<:0001) (<:0001)
(15) LOSS 0:0074 0:3925 0:1969 1:0000
(0:2378) (<:0001) (<:0001)
(16) GEOSEG 0:0030 0:5524 0:3886 0:1815 1:0000
(0:6301) (<:0001) (<:0001) (<:0001)
(17) NONUS 0:0040 0:5507 0:4463 0:1380 0:8799 1:0000
(0:5222) (<:0001) (<:0001) (<:0001) (<:0001)
(18) FOLLOW 0:0113 0:9086 0:8025 0:3850 0:5935 0:5332 1:0000
(0:0700) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
(19) IIAA 0:0281 0:2061 0:1629 0:0831 0:1596 0:1446 0:1905 1:0000
(<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
(20) GEXP 0:0231 0:0965 0:0557 0:0271 0:1196 0:1105 0:0950 0:2802 1:0000
(0:0002) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
(21) FEXP 0:0246 0:2005 0:1325 0:0542 0:2663 0:2581 0:1946 0:2814 0:6013 1:0000
(<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001) (<:0001)
41
Table 4.6
The Logistic Model for = 3, All News Events
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 2:0675 <:0001 2:1066 <:0001 2:0936 <:0001
EARNINGS 1:7642 <:0001 5:837 1:8051 <:0001 6:081 1:0854 <:0001 6:082
GUIDANCE 0:5849 <:0001 1:795 0:6212 <:0001 1:861 0:6201 <:0001 1:859
OPERATIONS 0:3082 <:0001 1:361 0:3234 <:0001 1:382 0:3233 <:0001 1:382
INVESTING 0:0112 0:8352 1:011
- CAPEX 0:1195 0:1310 0:887 0:1189 0:1328 0:888
- MERGERACQ 0:2398 0:0015 1:271 0:2396 0:0015 1:271
- OTHERINV 0:0894 0:3020 0:914 0:0898 0:3005 0:914
FINANCING 0:2592 <:0001 0:772
- DEBT 0:3685 0:0002 0:692 0:3687 0:0002 0:692
- PREFERRED 0:1090 0:3956 0:897 0:1094 0:3940 0:896
- EQUITY 0:1848 0:0399 0:831 0:1842 0:0405 0:832
- OTHERFIN 0:1021 0:7275 0:903 0:1024 0:7267 0:903
RATINGS 0:3938 <:0001 1:483 0:4084 <:0001 1:504 0:4081 <:0001 1:504
ANALYSTS 0:4139 <:0001 1:513 0:4247 <:0001 1:529 0:4242 <:0001 1:528
MANAGEMENT 0:0284 0:7212 0:972 0:0129 0:8693 0:987 0:0132 0:8660 0:987
OTHER 0:1728 0:0001 1:189 0:1919 <:0001 1:212 0:1917 <:0001 1:211
PRESENTATION 0:1535 0:0275 0:858 0:1276 0:0676 0:880 0:1274 0:0682 0:880
OPEC 0:0075 0:8885 0:993 0:0098 0:8535 1:010 0:0098 0:8531 1:010
log(MKTCAP) 0:0353 0:5470 1:036 0:0381 0:5184 1:039 0:0379 0:5222 1:039
SP500 0:2629 0:0138 0:769 0:2704 0:0119 0:763 0:2716 0:0117 0:762
LOSS 0:0316 0:8343 0:969 0:0148 0:9220 0:985 0:0141 0:9262 0:986
GEOSEG 0:0213 0:4701 0:979 0:0235 0:4272 0:977 0:0238 0:4221 0:977
NONUS 0:2686 0:3252 1:308 0:3035 0:2704 1:355 0:3010 0:2738 1:351
FOLLOW 0:0032 0:6163 1:003 0:0030 0:6337 1:003 0:0031 0:6278 1:003
IIAA 0:2161 0:0208 1:241 0:2158 0:0212 1:241 0:2240 0:0161 1:251
GEXP 0:0036 0:5569 1:004 0:0036 0:5581 1:004
FEXP 0:0123 0:1951 1:012 0:0121 0:2018 1:012 0:0157 0:0576 1:016
Adjusted R-squared 0:1360 0:1371 0:1370
42
Table 4.7
The Logistic Model for = 3, Counting Earnings Announcements as the Only News Event on Earnings
Announcement Day
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 2:1248 <:0001 2:2642 <:0001 2:2508 <:0001
EARNINGS 2:4057 <:0001 11:086 2:5675 <:0001 13:033 2:5671 <:0001 13:029
GUIDANCE 0:9437 <:0001 2:570 0:9878 <:0001 2:685 0:9866 <:0001 2:682
OPERATIONS 0:2324 <:0001 1:262 0:2566 <:0001 1:293 0:2567 <:0001 1:293
INVESTING 0:0124 0:8338 0:988
- CAPEX 0:1536 0:0503 0:858 0:1532 0:0510 0:858
- MERGERACQ 0:2448 0:0010 1:277 0:2447 0:0010 1:277
- OTHERINV 0:0897 0:2988 0:914 0:0901 0:2970 0:914
FINANCING 0:2928 <:0001 0:746
- DEBT 0:3303 0:0008 0:719 0:3306 0:0008 0:719
- PREFERRED 0:0950 0:4659 0:909 0:0954 0:4641 0:909
- EQUITY 0:2239 0:0137 0:799 0:2233 0:0140 0:800
- OTHERFIN 0:0764 0:7942 0:926 0:0767 0:7935 0:926
RATINGS 0:3617 0:0001 1:436 0:3713 <:0001 1:450 0:3710 <:0001 1:449
ANALYSTS 0:4361 <:0001 1:547 0:4588 <:0001 1:582 0:4583 <:0001 1:581
MANAGEMENT 0:0146 0:8560 0:985 0:0126 0:8739 1:013 0:0124 0:8757 1:013
OTHER 0:1660 0:0004 1:181 0:1954 <:0001 1:216 0:1952 <:0001 1:216
PRESENTATION 0:1621 0:0213 0:850 0:1271 0:0721 0:881 0:1268 0:0728 0:881
OPEC 0:0056 0:9177 1:006 0:0358 0:4964 1:036 0:0359 0:4957 1:037
log(MKTCAP) 0:0434 0:4561 1:044 0:0593 0:3123 1:061 0:0591 0:3162 1:061
SP500 0:2289 0:0311 0:795 0:2526 0:0185 0:777 0:2538 0:0182 0:776
LOSS 0:0029 0:9843 0:997 0:0142 0:9242 1:014 0:0150 0:9210 1:015
GEOSEG 0:0203 0:4890 0:980 0:0177 0:5481 0:982 0:0180 0:5420 0:982
NONUS 0:2208 0:4158 1:247 0:2214 0:4176 1:248 0:2189 0:4223 1:245
FOLLOW 0:0018 0:7708 1:002 0:0005 0:9359 1:001 0:0006 0:9290 1:001
IIAA 0:2172 0:0202 1:243 0:2175 0:0199 1:243 0:2259 0:0150 1:253
GEXP 0:0037 0:5464 1:004 0:0037 0:5460 1:004
FEXP 0:0122 0:1997 1:012 0:0121 0:2021 1:012 0:0158 0:0561 1:016
Adjusted R-squared 0:1392 0:1405 0:1404
43
Table 4.8
The Logistic Model for = 3, News Events with LargejCARj
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 2:1036 <:0001 2:0619 <:0001 2:0367 <:0001
EARNINGS 1:7370 <:0001 5:680 1:7243 <:0001 5:608 1:7242 <:0001 5:608
GUIDANCE 0:6279 <:0001 1:874 0:6475 <:0001 1:911 0:6451 <:0001 1:906
OPERATIONS 0:1533 0:0278 1:166 0:1408 0:0464 1:151 0:1416 0:0454 1:152
INVESTING 0:0902 0:2350 1:094
- CAPEX 0:0414 0:7297 1:042 0:0427 0:7212 1:044
- MERGERACQ 0:3173 0:0015 1:373 0:3181 0:0014 1:375
- OTHERINV 0:3484 0:0189 0:706 0:3500 0:0185 0:705
FINANCING 0:2338 0:0135 0:791
- DEBT 0:5601 0:0004 0:571 0:5596 0:0004 0:571
- PREFERRED 0:0991 0:5683 1:104 0:0991 0:5685 1:104
- EQUITY 0:1214 0:3872 0:886 0:1205 0:3908 0:887
- OTHERFIN 0:2706 0:4884 0:763 0:2721 0:4858 0:762
RATINGS 0:3442 0:0027 1:411 0:3514 0:0023 1:421 0:3495 0:0024 1:418
ANALYSTS 0:4706 <:0001 1:601 0:4600 <:0001 1:584 0:4595 <:0001 1:583
MANAGEMENT 0:3859 0:0008 0:680 0:3911 0:0006 0:676 0:3917 0:0005 0:676
OTHER 0:0919 0:1787 1:096 0:0933 0:1634 1:098 0:0939 0:1607 1:098
PRESENTATION 0:1261 0:2119 0:882 0:1230 0:2263 0:884 0:1224 0:2285 0:885
OPEC 0:1461 0:0646 0:864 0:1534 0:0463 0:858 0:1529 0:0470 0:858
log(MKTCAP) 0:0303 0:6760 1:031 0:0253 0:7331 1:026 0:0249 0:7389 1:025
SP500 0:4361 0:0007 0:647 0:4402 0:0008 0:644 0:4417 0:0008 0:643
LOSS 0:3296 0:0793 1:390 0:3675 0:0493 1:444 0:3687 0:0502 1:446
GEOSEG 0:0682 0:0540 0:934 0:0723 0:0447 0:930 0:0726 0:0429 0:930
NONUS 0:7214 0:0303 2:057 0:7854 0:0219 2:193 0:7811 0:0214 2:184
FOLLOW 0:0140 0:0625 1:014 0:0144 0:0615 1:015 0:0144 0:0607 1:015
IIAA 0:3459 0:0012 1:413 0:3437 0:0013 1:410 0:3601 0:0007 1:433
GEXP 0:0068 0:3105 1:007 0:0066 0:3279 1:007
FEXP 0:0101 0:3340 1:010 0:0107 0:3041 1:011 0:0173 0:0605 1:017
Adjusted R-squared 0:1522 0:1549 0:1547
44
Table 4.9
The Logistic Model for = 5, All News Events
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 1:7151 <:0001 1:8139 <:0001 1:8062 <:0001
EARNINGS 1:6131 <:0001 5:018 1:6984 <:0001 5:465 1:6985 <:0001 5:446
GUIDANCE 0:4740 <:0001 1:606 0:5382 <:0001 1:713 0:5376 <:0001 1:712
OPERATIONS 0:1804 <:0001 1:198 0:2137 <:0001 1:238 0:2137 <:0001 1:238
INVESTING 0:0934 0:0425 1:098
- CAPEX 0:1387 0:0633 0:870 0:1384 0:0638 0:871
- MERGERACQ 0:3289 <:0001 1:389 0:3288 <:0001 1:389
- OTHERINV 0:0236 0:7639 0:977 0:0238 0:7621 0:976
FINANCING 0:2153 0:0001 0:806
- DEBT 0:3522 0:0001 0:703 0:3524 0:0001 0:703
- PREFERRED 0:1435 0:1924 1:154 0:1433 0:1933 1:154
- EQUITY 0:1963 0:0130 0:822 0:1959 0:0132 0:822
- OTHERFIN 0:2614 0:2361 1:299 0:2611 0:2366 1:298
RATINGS 0:2725 0:0011 1:313 0:2948 0:0004 1:343 0:2946 0:0004 1:343
ANALYSTS 0:3507 <:0001 1:420 0:3666 <:0001 1:443 0:3663 <:0001 1:442
MANAGEMENT 0:0660 0:3248 1:068 0:0865 0:1869 1:090 0:0863 0:1879 1:090
OTHER 0:2714 0:0001 1:312 0:2977 <:0001 1:347 0:2976 <:0001 1:347
PRESENTATION 0:2020 0:0012 0:817 0:1656 0:0074 0:847 0:1654 0:0074 0:848
OPEC 0:0460 0:2883 1:047 0:0724 0:0899 1:075 0:0724 0:0899 1:075
log(MKTCAP) 0:0316 0:5856 1:032 0:0427 0:4646 1:044 0:0426 0:4667 1:044
SP500 0:2378 0:0218 0:788 0:2620 0:0126 0:770 0:2626 0:0125 0:769
LOSS 0:0495 0:7064 0:952 0:0346 0:7920 0:966 0:0342 0:7951 0:966
GEOSEG 0:0160 0:5741 0:984 0:0181 0:5251 0:982 0:0183 0:5218 0:982
NONUS 0:0886 0:7320 1:093 0:1299 0:6176 1:139 0:1283 0:6220 1:137
FOLLOW 0:0084 0:1728 1:008 0:0076 0:2216 1:008 0:0076 0:2198 1:008
IIAA 0:1781 0:0613 1:195 0:1767 0:0640 1:193 0:1815 0:0531 1:199
GEXP 0:0021 0:7261 1:002 0:0022 0:7221 1:002
FEXP 0:0140 0:1719 1:014 0:0138 0:1767 1:014 0:0160 0:0868 1:016
Adjusted R-squared 0:1008 0:1026 0:1026
45
Table 4.10
The Logistic Model for = 5, Counting Earnings Announcement as the Only News Event on Earnings
Announcement Day
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 1:7727 <:0001 1:9306 <:0001 1:9225 <:0001
EARNINGS 2:1279 <:0001 8:397 2:2965 <:0001 9:939 2:2963 <:0001 9:937
GUIDANCE 0:7641 <:0001 2:147 0:8434 <:0001 2:324 0:8427 <:0001 2:323
OPERATIONS 0:1206 0:0080 1:128 0:1591 0:0004 1:172 0:1591 0:0004 1:172
INVESTING 0:0815 0:1033 1:085
- CAPEX 0:1728 0:0215 0:841 0:1726 0:0217 0:841
- MERGERACQ 0:3311 <:0001 1:393 0:3310 <:0001 1:392
- OTHERINV 0:0257 0:7438 0:975 0:0259 0:7417 0:974
FINANCING 0:2279 0:0002 0:796
- DEBT 0:3245 0:0004 0:723 0:3247 0:0004 0:723
- PREFERRED 0:1547 0:1636 1:167 0:1544 0:1644 1:167
- EQUITY 0:2260 0:0047 0:798 0:2256 0:0047 0:798
- OTHERFIN 0:2762 0:2116 1:318 0:2760 0:2121 1:318
RATINGS 0:2366 0:0061 1:267 0:2573 0:0030 1:293 0:2571 0:0031 1:293
ANALYSTS 0:3681 <:0001 1:445 0:3896 <:0001 1:476 0:3893 <:0001 1:476
MANAGEMENT 0:0799 0:2366 1:083 0:1045 0:1139 1:110 0:1043 0:1144 1:110
OTHER 0:2654 <:0001 1:304 0:2961 <:0001 1:345 0:2960 <:0001 1:344
PRESENTATION 0:2074 0:0010 0:813 0:1673 0:0074 0:846 0:1671 0:0074 0:846
OPEC 0:0556 0:2051 1:057 0:0876 0:0410 1:092 0:0876 0:0410 1:092
log(MKTCAP) 0:0401 0:4899 1:041 0:0593 0:3092 1:061 0:0592 0:3114 1:061
SP500 0:2190 0:0346 0:803 0:2510 0:0169 0:778 0:2517 0:0166 0:777
LOSS 0:0264 0:8392 0:974 0:0129 0:9210 0:987 0:0125 0:9239 0:988
GEOSEG 0:0157 0:5824 0:984 0:0143 0:6184 0:986 0:0144 0:6146 0:986
NONUS 0:0643 0:8043 1:066 0:0753 0:7726 1:078 0:0737 0:7774 1:076
FOLLOW 0:0073 0:2383 1:007 0:0056 0:3656 1:006 0:0057 0:3627 1:006
IIAA 0:1788 0:0603 1:196 0:1777 0:0622 1:194 0:1827 0:0514 1:200
GEXP 0:0022 0:7171 1:002 0:0022 0:7120 1:002
FEXP 0:0138 0:1779 1:014 0:0138 0:1779 1:014 0:0160 0:0863 1:016
Adjusted R-squared 0:1032 0:1052 0:1052
46
Table 4.11
The Logistic Model for = 5, News Events with LargejCARj
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 1:8310 0:0001 1:8236 0:0002 1:8001 0:0002
EARNINGS 1:5902 <:0001 4:905 1:5970 <:0001 4:938 1:5970 <:0001 4:938
GUIDANCE 0:4569 <:0001 1:579 0:4977 <:0001 1:645 0:4954 <:0001 1:641
OPERATIONS 0:0583 0:3453 1:060 0:0509 0:4187 1:052 0:0515 0:4133 1:053
INVESTING 0:1458 0:0263 1:157
- CAPEX 0:0353 0:7512 1:036 0:0367 0:7416 1:037
- MERGERACQ 0:4056 <:0001 1:500 0:4060 <:0001 1:501
- OTHERINV 0:3494 0:0075 0:705 0:3505 0:0073 0:704
FINANCING 0:1969 0:0223 0:821
- DEBT 0:5390 <:0001 0:583 0:5387 <:0001 0:583
- PREFERRED 0:0536 0:7543 1:055 0:0535 0:7458 1:055
- EQUITY 0:1192 0:3569 0:888 0:1184 0:3598 0:888
- OTHERFIN 0:5004 0:0644 1:649 0:4987 0:0658 1:647
RATINGS 0:2109 0:0444 1:235 0:2289 0:0261 1:257 0:2271 0:0271 1:255
ANALYSTS 0:3998 <:0001 1:492 0:3910 <:0001 1:478 0:3904 <:0001 1:478
MANAGEMENT 0:2206 0:0148 0:802 0:2292 0:0103 0:795 0:2298 0:0101 0:795
OTHER 0:2309 0:0002 1:260 0:2271 0:0002 1:255 0:2275 0:0002 1:255
PRESENTATION 0:1026 0:2532 0:903 0:0964 0:2778 0:908 0:0961 0:2800 0:908
OPEC 0:0461 0:4708 0:955 0:0512 0:4092 0:950 0:0508 0:4127 0:950
log(MKTCAP) 0:0418 0:5534 1:043 0:0431 0:5486 1:044 0:0427 0:5545 1:044
SP500 0:4039 0:0018 0:668 0:3963 0:0027 0:673 0:3976 0:0027 0:672
LOSS 0:3134 0:0474 1:368 0:3440 0:0281 1:411 0:3449 0:0290 1:412
GEOSEG 0:0658 0:0523 0:936 0:0685 0:0459 0:934 0:0688 0:0444 0:934
NONUS 0:4914 0:1053 1:635 0:5490 0:0760 1:732 0:5441 0:0768 1:723
FOLLOW 0:0182 0:0135 1:018 0:0175 0:0205 1:018 0:0175 0:0203 1:018
IIAA 0:2885 0:0110 1:334 0:2871 0:0116 1:333 0:3025 0:0072 1:353
GEXP 0:0064 0:3427 1:006 0:0063 0:3541 1:006
FEXP 0:0067 0:5521 1:007 0:0071 0:5271 1:007 0:0134 0:1948 1:014
Adjusted R-squared 0:1084 0:1124 0:1122
47
Table 4.12
The Discrete-Time Duration Model
(1) (2) (3)
Coecient p-value Odds ratio Coecient p-value Odds ratio Coecient p-value Odds ratio
Intercept 3:9844 <:0001 3:9984 <:0001 4:0036 <:0001
EARNINGS(n) 0:9282 <:0001 2:530 0:9603 <:0001 2:613 0:9606 <:0001 2:613
GUIDANCE(n) 0:1166 0:0542 1:124 0:1330 0:0013 1:142 0:1333 0:0013 1:143
OPERATIONS(n) 0:0592 0:0486 1:061 0:0733 0:0007 1:076 0:0732 0:0007 1:076
INVESTING(n) 0:0073 0:7923 1:007
- CAPEX(n) 0:0456 0:2486 0:955 0:0453 0:2512 0:956
- MERGERACQ(n) 0:0910 0:0070 0:913 0:0912 0:0069 0:913
- OTHERINV(n) 0:0478 0:0733 1:049 0:0479 0:0725 1:049
FINANCING(n) 0:0367 0:3609 0:964
- DEBT(n) 0:0199 0:5700 0:980 0:0199 0:5706 0:980
- PREFERRED(n) 0:0979 0:0698 1:103 0:0979 0:0697 1:103
- EQUITY(n) 0:0703 0:0327 0:932 0:0703 0:0327 0:932
- OTHERFIN(n) 0:0834 0:5011 0:920 0:0837 0:4994 0:920
RATINGS(n) 0:0269 0:6294 1:027 0:0402 0:3382 1:041 0:0400 0:3396 1:041
ANALYSTS(n) 0:0923 0:0128 1:097 0:0834 0:0004 1:087 0:0837 0:0004 1:087
MANAGEMENT(n) 0:0421 0:3014 1:043 0:0292 0:3246 1:030 0:0293 0:3226 1:030
OTHER(n) 0:0016 0:9417 1:002 0:0071 0:6492 1:007 0:0072 0:6474 1:007
PRESENTATION(n) 0:0962 0:0177 0:908 0:1087 <:0001 0:897 0:1088 <:0001 0:897
OPEC(n) 0:1375 0:0001 0:872 0:1327 <:0001 0:876 0:1330 <:0001 0:875
log(MKTCAP) 0:0639 0:2070 1:066 0:0681 0:0844 1:070 0:0682 0:0842 1:071
SP500 0:2340 0:0057 0:791 0:2519 0:0002 0:777 0:2514 0:0002 0:778
LOSS 0:0331 0:7979 0:967 0:0752 0:3921 0:928 0:0754 0:3905 0:927
GEOSEG 0:0258 0:3629 1:026 0:0252 0:1801 1:026 0:0254 0:1753 1:026
NONUS 0:1828 0:5181 0:833 0:1930 0:2539 0:824 0:1922 0:2555 0:825
FOLLOW 0:0000 0:9963 1:000 0:0005 0:9045 1:000 0:0005 0:8979 0:999
IIAA 0:3607 <:0001 1:434 0:3607 <:0001 1:434 0:3568 <:0001 1:429
GEXP 0:0017 0:7277 0:998 0:0016 0:6357 0:998
FEXP 0:0296 0:0004 1:030 0:0293 <:0001 1:030 0:0277 <:0001 1:028
Adjusted R-squared 0:0413 0:0418 0:0418
48
Chapter 5
Conclusions
This study is the rst that empirically test the relation between analyst responsive-
ness and types of news events. Consistent with my predictions, I nd that analysts
are most responsive to earnings announcements and less responsive to management
guidance, operations updates, and investing activities. More specically, the odds
of analysts revising their earnings forecasts following an earnings announcement
within three trading days are 6.9, which is signicantly higher than the odds of
analysts revising their forecasts following management guidance (1.89), operations
updates (1.28), and news concerning investing activities (1.15). On a similar note,
earnings announcements are most likely to trigger analyst revisions; one earnings
announcement is associated with a 141% increase in the predicted odds of analyst
revision. I also nd that analysts are unlikely to revise their forecasts following news
concerning any types of nancing activities but are likely to revise their forecasts
following news concerning credit ratings and other analysts' activities. These results
are robust even if only economically signicant news events are considered.
Analyst responsiveness to news events provides an alternative measure of ana-
lyst performance to the traditionally employed forecast accuracy that can only be
measured ex post. This is known as the \process view" of performance evaluation in
the JDM literature (?). Understanding how analysts respond to dierent types of
news events is also important to both academic researchers (for developing better
proxies for market earnings expectations) and investors (for facilitating investment
decisions).
49
This paper is only a rst step in studying the analyst decision processes and will
be extended to include analyst characteristics and rm characteristics in determining
analyst responsiveness and to investigate whether more responsive analysts are also
likely to be more accurate. This paper contributes to the research on analysts'
decision processes that Schipper (1991) calls for and also provide insights into the
process view of analysts' JDM quality. Additionally, I add to the JDM literature
that investigate the eect of information complexity on JDM quality.
50
References
Abarbanell, J. S. (1991). Do analysts' earnings forecasts incorporate information
in prior stock prices changes. Journal of Accounting & Economics 14 (2), 147{165.
Abarbanell, J. S. and V. L. Bernard (1992). Tests of analysts' overreac-
tion/underreaction to earnings information as an explanation for anomalous stock
price behavior. The Journal of Finance 47 (3), 1181{1207.
Abarbanell, J. S. and B. J. Bushee (1997). Fundamental analysis, future earnings,
and stock prices. Journal of Accounting Research 35 (1), 1{24.
Ali, A., A. Klein, and J. Rosenfeld (1992). Analysts' use of information about
permanent and transitory earnings components in forecasting annual EPS. The
Accounting Review 67 (1), 183{198.
Allison, P. D. (1982). Discrete-time method for the analysis of event histories.
Sociological Methodology 13, 61{98.
Bagnoli, M., S. Levine, and S. G. Watts (2005). Analyst estimation revision clusters
and corporate events, part i. Annals of Finance 1, 245{265.
Barron, O. E., C. O. Kile, and T. B. O'Keefe (1999). MD&A quality as mea-
sured by the SEC and analysts' earnings forecasts. Contemporary Accounting Re-
search 16(1), 75{109.
Benartzi, S., R. Michaely, and R. Thaler (1997). Do changes in dividends signal
the future or the past? The Journal of Finance 52 (3), 1007{1034.
Bhattacharya, S. (1979). Imperfect information, dividend policy, and the bird in
the hand fallacy. The Bell Journal of Economics 10 (1), 259{270.
Birnbaum, J. H. (2006). Merger of Wall Street groups creates a lobbying power-
house. The Washington Post (November 27), D01.
Bonner, S. E. (2008). Judgment and decision making in accounting. Upper Saddle
River, NJ: Pearson Prentice Hall.
Bonner, S. E., A. Hugon, and B. R. Walther (2007). Investor reaction to celebrity
analysts: the case of earnings forecast revisions. Journal of Accounting Re-
search 45(3), 481{513.
51
Bonner, S. E., B. R. Walther, and S. M. Young (2003). Sophisticated-related dif-
ferences in investors' models of the relative accuracy of analysts' forecast revisions.
The Accounting Review 78 (3), 679{706.
Bowen, R. M., A. K. Davis, and D. A. Matsumoto (2002). Do conference calls
aect analysts' forecasts. The Accounting Review 77 (2), 285{316.
Bradshaw, M. T. (2008). Analysts' forecasts: what do we know after decades of
work? Working paper.
Bradshaw, M. T., S. A. Richardson, and R. G. Sloan (2006). The relation between
corporate nancing activities, analysts' forecasts and stock returns. Journal of
Accounting & Economics 42 (1{2), 53{85.
Brown, L. D., R. L. Hagerman, P. A. Grin, and M. E. Zmijewski (1987). Security
analyst superiority relative to univariate time-series models in forecasting quarterly
earnings. Journal of Accounting & Economics 9 (1), 61{87.
Chaney, P. K., C. E. Hogan, and D. C. Jeter (1999). The eect of reporting restruc-
turing charges on analysts' forecast revisions and errors. Journal of Accounting &
Economics 27 (3), 261{284.
Charoenrook, A. and C. M. Lewis (2009). Information, selective disclosure, and
analyst behavior. Financial Management 38 (1), 39{57.
Clement, M. B. and S. Y. Tse (2003). Do investors respond to analysts' forecast
revisions as if forecast accuracy is all that matters? The Accounting Review 78 (1),
227{249.
Cooper, R. A., T. E. Day, and C. M. Lewis (2001). Following the leader: a study
of individual analysts' earnings forecasts. Journal of Financial Economics 61 (3),
383{416.
Cotter, J., I. Tuna, and P. D. Wysocki (2006). Expectations management and beat-
able targets: how do analysts react to explicit earnings guidance? Contemporary
Accounting Research 23 (3), 593{624.
DeBondt, W. F. M. and R. H. Thaler (1990). Do security analysts overreact? The
American Economic Review 80 (2), 52{57.
Duru, A. and D. M. Reeb (2002). International diversication and analysts' forecast
accuracy and bias. The Accounting Review 77 (2), 415{433.
Easterwood, J. C. and S. R. Nutt (1999). Ineciency in analysts' earnings forecasts:
systematic misreaction or systematic optimism. The Journal of Finance 54 (5),
1777{1797.
Ederington, L. H. and J. C. Goh (1998). Bond rating agencies and stock analysts:
who knows what when? Journal of Financial and Quantitative Analysis 33 (4),
569{585.
52
Elliott, J. A., D. R. Philbrick, and C. I. Wiedman (1995). Evidence from archival
data on the relation between security analysts' forecast errors and prior forecast
revisions. Contemporary Accounting Research 1 (2), 919{938.
Erwin, G. R. and S. E. Perry (2000). The eect of foreign diversication on analysts'
prediction errors. International Review of Financial Analysis 9 (2), 121{145.
Fernandez, F. (2001). The roles and responsibilities of securities analysts. SIA
Research Report 2 (7), 3{10.
Francis, J., J. D. Hanna, and D. R. Philbrick (1997). Management communications
with securities analysts. Journal of Accounting & Economics 24 (3), 363{394.
Fried, D. and D. Givoly (1985). Financial analysts' forecasts of earnings: a better
surrogate for market expectations. Journal of Accounting & Economics 4 (2), 85{
107.
Gleason, C. A. and C. M. C. Lee (2003). Analyst forecast revisions and market
price discovery. The Accounting Review 78 (1), 193{225.
Goh, J. C. and L. H. Ederington (1993). Is a bond rating downgrade bad news,
good news, or no news for stockholders? The Journal of Finance 48 (5), 2001{2008.
Gonedes, N. J. (1978). Corporate signaling, external accounting, and capital mar-
ket equilibrium: evidence on dividends, income, and extraordinary items. Journal
of Accounting Research 16 (1), 26{79.
Gustavo, G. and R. Michaely (2004). The information content of share repurchase
programs. The Journal of Finance 59 (2), 651{680.
Guttman, I. (2008). The timing of analysts' earnings forecasts. Working paper.
Hansen, R. S. and C. Crutchley (1990). Corporate earnings and nancings: an
empirical analysis. Journal of Business 63 (3), 347{371.
Hassell, J. M., R. H. Jennings, and D. J. Lasser (1988). Management earnings fore-
casts: their usefulness as a source of rm-specic information to security analysts.
The Journal of Financial Research 11 (4), 303{319.
Haw, I.-M., K. Jung, and W. Ruland (1994). The accuracy of nancial analysts'
forecast after mergers. Journal of Accounting, Auditing & Finance 9 (3), 465{483.
Healy, P. M., A. P. Hutton, and K. G. Palepu (1999). Stock performance and in-
termediation changes surrounding sustained increases in disclosure. Contemporary
Accounting Research 16 (3), 485{520.
Herrmann, D. R., O.-K. Hope, and W. B. Thomas (2008). International diversi-
cation and forecast optimism: The eect of reg fd. Accounting Horizon 22 (2),
179{197.
53
Hirst, D. E. and P. E. Hopkins (1998). Comprehensive income reporting and
analysts' valuation judgments. Journal of Accounting Research 36 (Supplement),
47{75.
Hirst, D. E., P. E. Hopkins, and J. M. Wahlen (2004). Fair values, income mea-
surement, and bank analysts' risk and valuation judgments. The Accounting Re-
view 79(2), 453{472.
Hong, H. and J. D. Kubik (2003). Analyzing the analysts: Career concerns and
biased earnings forecasts. The Journal of Finance 58 (1), 313{351.
Jennings, R. H. (1987). Unsystematic security price movements, management earn-
ings forecasts, and revisions in consensus analyst earnings forecasts. Journal of
Accounting Research 25 (1), 90{110.
Jensen, M. C. and R. S. Ruback (1983). The market for corporate control: the
scientic evidence. Journal of Financial Economics 11 (1{4), 5{50.
Kothari, S. P., T. E. Laguerre, and A. J. Leone (2002). Capitalization versus
expensing: evidence on the uncertainty of future earnings from capital expenditures
versus r&d outlays. Review of Accounting Studies 7 (4), 355{382.
Lehavy, R., F. Li, and K. Merkley (2009). The eect of annual report readability
on analyst following and the properties of their earnings forecasts. Working paper.
Li, L. Y. (2006). Strategic forecast timing: theory and evidence. Working paper.
Libby, R., H.-T. Tan, and J. E. Hunton (2006). Does the form of manage-
ment's earnings guidance aect analysts' earnings forecasts? The Accounting
Review 81(1), 207{225.
Linter, J. (1956). Distribution of incomes of corporations among dividends, retained
earnings, and taxes. The American Economic Review 46 (2), 97{113.
Lys, T. Z. and S. Sohn (1990). The association between revisions of nancial
analysts' earnings forecasts and security-price changes. Journal of Accounting &
Economics 13 (4), 341{363.
McConnell, J. J. and C. J. Muscarella (1985). Corporate capital expenditure de-
cisions and the market value of the rm. Journal of Financial Economics 14 (3),
399{422.
Mendenhall, R. R. (1991). Evidence on the possible underweighting of earnings-
related information. Journal of Accounting Research 29 (1), 170{179.
Mikhail, M. B., B. R. Walther, and R. H. Willis (1997). Do security ana-
lysts improve their performance with experience? Journal of Accounting Re-
search 35(Supplement), 131{157.
Mikhail, M. B., B. R. Walther, and R. H. Willis (1999). Does forecast accuracy
matter to security analysts? The Accounting Review 72 (4), 185{200.
54
Miller, J. S. (2005). Eects of preannouncements on analyst and stock price re-
actions to earnings news. Review of Quantitative Finance and Accounting 24 (3),
251{275.
Miller, M. H. and F. Modigliani (1961). Dividend policy, growth, and the value of
shares. Journal of Business 34 (4), 411{433.
Moyes, G. D., B. Saadouni, S. Jon, and P. A. Williams (2001). A comparison of
factors aecting UK and US analyst forecast revisions. 36(1), 47{63.
Ou, J. A. (1990). The information content of nonearnings accounting numbers as
earnings predictors. Journal of Accounting Research 28 (1), 144{163.
Park, C. W. and E. K. Stice (2000). Analyst forecasting ability and the stock price
reaction to forecast revisions. Review of Accounting Studies 5 (3), 259{272.
Penman, S. H. (1983). The predictive content of earnings forecasts and dividends.
The Journal of Finance 38 (4), 1181{1199.
Plumlee, M. A. (2003). The eect of information complexity on analysts' use of
that information. The Accounting Review 78 (1), 275{296.
Previts, G. J., R. J. Bricker, T. R. Robinson, and S. J. Young (1994). A content
analysis of sell-side nancial analyst company reports. Accounting Horizon 8 (2),
55{70.
Rogers, R. K. and J. Grant (1997). Content analysis of information cited in reports
of sell-side nancial analysts. The Journal of Financial Statement Analysis 3 (1),
17{30.
Schipper, K. (1991). Commentary on analysts' forecasts. Accounting Horizon 5 (4),
105{121.
SRI International (1987). Investor information needs and the annual report. Mor-
ristown, NJ: Financial Executives Research Foundation.
Stickel, S. E. (1989). The timing of and incentives for annual earnings forecasts
near interim earnings announcements. Journal of Accounting & Economics 11 (2{
3), 275{292.
Stickel, S. E. (1992). Reputation and performance among security analysts. The
Journal of Finance 47 (5), 1811{1836.
Watts, R. (1973). The information content of dividends. Journal of Business 46 (2),
191{211.
Waymire, G. (1986). Additional evidence on the accuracy of analyst forecasts before
and after voluntary management earnings forecasts. The Accounting Review 61 (1),
129{142.
55
Williams, P. A. (1996). The relation between a prior earnings forecast by man-
agement and analyst response to a current management forecast. The Accounting
Review 71(1), 103{113.
Wilson, G. P. (1986). The relative information content of accruals and cash
ows:
combined evidence at the earnings announcement and annual report release date.
Journal of Accounting Research 24 (Supplement), 165{200.
Zhang, Y. (2008). Analyst responsiveness and the post-earnings-announcement
drift. Journal of Accounting & Economics 46 (1), 201{215.
56
Appendix 1
List of Sample Firms
Anadarko Petroleum Corporation The Meridian Resource Corporation
Apache Corporation Neweld Exploration Company
Arena Resources, Inc. Noble Energy, Inc.
Berry Petroleum Company Occidental Petroleum Corporation
Bill Barrett Corporation Penn Virginia Corporation
Cabot Oil & Gas Corporation Petrohawk Energy Corporation
Callon Petroleum Company Petroquest Energy, Inc.
Chesapeake Energy Corporation Pioneer Natural Resources Company
Cimarex Energy Co. Plains Exploration & Production Company
Comstock Resources, Inc. Pogo Producing Company
Denbury Resources Inc. Quicksilver Resources Inc.
Devon Energy Corporation Range Resources Corporation
Encore Acquisition Company St. Mary Land & Exploration Company
Energy Partners, Ltd. Stone Energy Corporation
EOG Resources, Inc. Swift Energy Company
Forest Oil Corporation W&T Oshore, Inc.
Goodrich Petroleum Corporation Whiting Petroleum Corporation
Harvest Natural Resources, Inc. XTO Energy Inc.
McMoran Exploration Co.
57
Appendix 2
Sources of News Articles
Press Release Wires: Business Wire
Market Wire
Moody's Investors Service Press Release
PR Newswire
PrimeNewswire
Newswires: Associated Press Newswires
Cox News Service
Dow Jones Newswires, including
- Dow Jones Business News
- Dow Jones Capital Markets Report
- Dow Jones Commodities Service
- Dow Jones Corporate Fillings Alert
- Dow Jones Emerging Markets Report
- Dow Jones Energy Service
- Dow Jones International News
- Dow Jones News Service
Platts Commodity News
Reuters News
States News Service
Major US News and The Atlantic Journal-Constitution
Business Publications: Barron's
The Boston Globe
BusinessWeek
The Denver Post
Forbes
Fortune
The Houston Chronicle
The New York Times
The Times-Picayune
USA Today
The Wall Street Journal
58
Appendix 3
Description of News Categories
Earnings announcements, which include all nancial information a rm dis-
closes relating to the release of annual or quarterly earnings
Management guidance, which is dened as forward-looking statements issued
by the management to provide information about a rm's future nancial and
operating performance, including:
- management's earnings and cash
ow estimates
- production and reserves guidance
- product pricing estimates
- cost estimates
- eective tax rate estimates
Operations updates, which provide information about a rm's recent operations
such as:
- drilling activities and results
- reserves acquisitions and replacements
- production statistics
- price rick management
- production sharing contracts
- leases and bids
News concerning investing activities, which provides information about changes
in a rm's long-term assets, including:
- capital expenditure budget and spending
- mergers, acquisitions, and divestments
- other investing activities such as closing of M&A transactions, as well as
M&A negotiations and plans
59
Appendix 3
Description of News Categories
(continued)
News concerning nancing activities, which provides information about a rm's
equity and long-term liabilities such as:
- dividends
- share repurchases
- issue or redemption of debt securities
- issue of preferred or common stock
- term loan or credit facility arrangements
- contingent interest payments
- debt covenants
News concerning credit ratings, such as:
- changes in bond or corporate ratings
- armations of bond or corporate ratings
- new bond ratings
News concerning analyst activities, which includes:
- earnings forecast revisions
- changes in recommendation and target price
- initiations
- commentary
News concerning management issues, such as:
- changes in management or board members
- corporate governance
- management compensation
- stock compensation plans
- management's exercise or relinquishment of options
- insider dealing
60
Appendix 3
Description of News Categories
(continued)
Other, which includes news events that do not fall under the above categories, for
example:
- accounting policy changes
- restatements
- auditor changes
- SEC inquiries
- exchange listing and registration of securities
- changes in index membership
- legal issues and settlement, including antitrust dispute and approval
- environmental issues
- regulations
- macro factors
61

Abstract (if available)

Abstract Using a unique database hand-collected from approximately 5,000 news articles, I examine securities analysts' responsiveness to different types of news events. I find that analysts are most responsive to earnings announcements and less responsive to management guidance, operations updates, as well as news concerning investing activities, financing activities, credit ratings, and analyst activities. It is also interesting to note that analysts are more responsive to management guidance than they are to operations updates, as well as news concerning investing activities and financing activities. These results are robust even if only economically significant news events are considered and when different measures of analyst responsiveness are used.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays in tail risks

PDF

Essays on interest rate determination in open economies

PDF

Essays on delegated portfolio management under market imperfections

PDF

Essays on the role of entry strategy and quality strategy in market and consumer response

PDF

Nonprofit organizations and the environmental policy outcomes: a systematic inquiry into the role of different types of nonprofits to influence the processes and outcomes of the environmental policy

PDF

Essays on understanding consumer contribution behaviors in the context of crowdfunding

PDF

Defending industrial control systems: an end-to-end approach for managing cyber-physical risk

Asset Metadata

Creator Wang, Ying Ying Terry (author)

Core Title How do securities analysts respond to different types of news events?

School Marshall School of Business

Degree Doctor of Philosophy

Degree Program Business Administration

Publication Date 05/04/2011

Defense Date 05/03/2010

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,securities analysts

Place Name USA (countries)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Bonner, Sarah (committee chair), Trezevant, Robert (committee chair), Jones, Christopher S. (committee member), Maber, David (committee member), Zapatero, Fernando (committee member)

Creator Email gnawyrret@hotmail.com,yingying.wang@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m3886

Unique identifier UC1414111

Identifier etd-Wang-3894 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-612909 (legacy record id),usctheses-m3886 (legacy record id)

Legacy Identifier etd-Wang-3894-1.pdf

Dmrecord 612909

Document Type Dissertation

Rights Wang, Ying Ying Terry

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu