University of Southern California Dissertations and Theses

Effects of redundancy in media coverage on nonprofessional investors' earnings forecasts

(USC Thesis Other)

Effects of redundancy in media coverage on nonprofessional investors' earnings forecasts

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content NOTE TO USERS
This reproduction is the best copy available.
®
UMI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
EFFECTS OF REDUNDANCY IN MEDIA COVERAGE ON
NONPROFESSIONAL INVESTORS’ EARNINGS FORECASTS
Copyright 2004
by
J. Artur Hugon
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTFIERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)
August 2004
Artur Hugon
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 3180777
Copyright 2004 by
Hugon, J. Artur
All rights reserved.
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
®
UMI
UMI Microform 3180777
Copyright 2005 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ii
ACKNOWLEDGEMENTS
I appreciate the assistance of my committee members: Randy Beatty,
Sarah Bonner, Jim Manegold, Norman Miller, and Beverly Walther. I am
especially grateful for the comments and support of my chair, Sarah Bonner. I
would also like to thank the following persons for comments and assistance:
Larry Brown, Clara Chen, Joel Demski, Brenda Flores, Chrislynn Freed, Frank
Hodge, Abbe Hugon, Cecil Jackson, Kathryn Kadous, Tom Kida, Lisa Koonce,
Hai Lu, Molly Mercer, William Messier, Lay Khim Ong, Kristy Towry, Shiing-
wu Wang, Susan Young and seminar participants at University of Southern
California, University of Massachusetts, University of Florida, Emory University,
and Georgia State University. I appreciate the financial support of the Leventhal
School of Accounting, Marshall School of Business, and the Deloitte & Touche
Foundation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
iii
TABLE OF CONTENTS
Acknowledgements ii
List of Tables iv
List of Figures v
Abstract vi
Chapter 1: Introduction 1
Chapter 2: Theory and Hypotheses 14
Chapter 3: Method 63
Chapter 4: Data Analysis and Results 80
Chapter 5: Conclusion 120
Bibliography 125
Appendix A: Explanation of the Sobel Test 135
Appendix B: Libby Boxes: Overview of the Experimental Design 137
Appendix C: Example of the Case Materials 138
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
iv
LIST OF TABLES
Table 1: Summary of Asymmetrical Redundancy Studies 24
Table 2: Summary of Symmetrical Redundancy Studies 38
Table 3: Summary of Redundancy Effect Hypotheses 46
Table 4: Summary of Credibility Effect Studies 58
Table 5: Dimensions of Redundancy Treatments 72
Table 6: Redundancy Comparisons 87
Table 7: Redundancy Comparisons Scaled by Initial Forecast 91
Table 8: Mediational Analysis of Credibility 99
Table 9: Mediational Analysis of Credibility Scaled by Initial Forecast 101
Table 10: Mediational Analysis of Credibility- Asymmetrical 107
Comparison: Redundant (0) and Non-Redundant (3) Conditions
Table 11: Mediational Analysis o f Credibility- Asymmetrical 110
Comparison: Redundant (0) and Non-Redundant (4) Conditions
Table 12: Mediational Analysis of Credibility- Symmetrical Comparison 113
Redundant (0) and Non-Redundant (1) Conditions
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF FIGURES
Figure 1: Example of Redundancy Comparisons
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
vi
ABSTRACT
Recent empirical research finds that the financial media influences market
participants in general. This study seeks to understand how redundancy, an
institutional feature of the financial media, influences investors’ judgments in
particular. Specifically, I conduct a between-participants experiment to examine
the impact of redundant financial press articles on nonprofessional investors’
earnings forecasts. The influence of redundancy is evaluated by contrasting
investors’ earnings forecasts in a redundant press article condition to those of
investors in non-redundant press article conditions. Results indicate that investors
in the redundant condition overweight the information provided by their press
article sets when making an earnings forecast. Moreover, a formal process
investigation provides support for the notion that this effect is due to a perception
that redundancy is positively associated with credibility. There are two potential
alternative explanations for the results: redundant news articles convey an
information signal to investors, and redundant news articles increase the salience
of the information content. I do not find support for these alternative
explanations. In sum, the results suggest that investors confronted with a stream
of repetitive financial news may overweight this information when engaging in
investment decision-making.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
CHAPTER 1: INTRODUCTION
The introductory chapter o f this thesis first discusses the research
questions addressed in this study. Then, the general importance of these research
questions is presented through a discussion of the significance of the financial
media, why redundancy is an inherent feature of the financial media, the rise of
nonprofessional investors in the market place, and intersection of the financial
media and nonprofessional investors. Next, the academic importance of the
research questions is discussed in the context of design challenges in previous
research and the process explanation explored in this study. Finally, a summary
of the study results and implications is presented.
RESEARCH QUESTIONS
This thesis reports experimental results concerning the influence of
redundancy in financial press articles on nonprofessional investors’ judgments
about companies’ earnings. Specifically, I compare investors’ earnings forecasts
in conditions with redundant financial press articles to those of investors in
conditions with non-redundant press articles. Across both redundant and non-
redundant conditions, the press articles are comprised of information items
pertaining to the future financial performance of a company. The purpose of these
comparisons is to examine the following research questions: (1) Does mere
redundancy in press articles affect investors’ financial judgments, and, if so, (2)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2
What is the psychological process that explains the observed effect on investors’
financial judgments?
IMPORTANCE OF FINANCIAL MEDIA
In recent years, there has been an expansion of the financial press
(Madrick 2000; Schuster 2003) and a corresponding growth in empirical
research supporting the notion that the financial press influences a variety of
market participants (Mutchler et al. 1997; Busse and Green 2002; DuCharme et
al. 2002; Joe 2003; Miller 2003; Gadarowski 2004; Bonner et al. 2004). As one
example of this work, Bonner et al. (2004) show that investors react to the
amount of press coverage associated with a particular sell side analyst. That is,
analysts who receive greater amounts o f financial press coverage attract a greater
reaction to their earnings forecast revisions. Further, these results hold after
controlling for the accuracy of the forecast revisions. This recent body of
empirical evidence showing an influence of the financial press on market
participants supports anecdotal evidence that has long suggested that the media is
an important factor in the formation of investors’ judgments and decisions
(Shiller 2001). The intuition behind the significance of the financial media is
straightforward- important market events can happen only if a large number of
people share a similar set of beliefs and there is a dissemination mechanism in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
place to facilitate the communication of financial news that underlies belief
sharing (Shiller 2001). 1
WHY REDUNDANCY IS A KEY DIMENSION OF THE
MEDIA ENVIRONMENT
The media plays an important role in society as an information intermediary
(Bagdikian 2000; Islam 2002; Stromberg 2002; Djankov et al. 2002). In order to
survive, the media industry depends on three primary functional features:
regulation by the government, sponsorship from advertisers, and subscriptions
from consumers (Islam 2002). I argue that the two functional features of media
dependence on revenues from advertisers and consumers result in a proliferation
of redundant or overlapping media coverage. That is, revenues from advertisers
and consumers are ultimately a function of consumer preferences. Due to the fact
that advertisers cater to specific subsets of consumers, media coverage is largely
a function of these consumers’ preferences.2 Servicing this readership’s
preferences necessarily results in a large body of media outlets reporting on a
small body of salient news. The end result is a proliferation of stories that are
either redundant or highly redundant in information content and sources.
1 The agency problems of the financial media taking on the primary role of information dissemination
are not addressed in this study. For a discussion of these problems, see Herman 2002.
2 A salient example o f this phenomenon was the cancellation of the television show “Gunsmoke”.
Despite the fact that the program had high ratings, it was cancelled because the show’s audience was
too old and too rural to satisfy advertisers (Bamouw 1978).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
Moreover, contributing to the redundancy that results from these functional
features, are several institutional characteristics of the media: industry
consolidation, the advent of dissemination technology, and a focus on returns to
scale.
There is strong trend towards consolidation in the media resulting in both
private and government concentrations of media resources (Herman and
McChesney 1997; Bagdikian 2000; Djankov et al. 2002). This trend has resulted
in the increasing homogeneity of news reporting and practices because
international media conglomerates such as CNN and News Corporation are
responsible for shaping and reporting an increasingly greater proportion of the
news (Herman 2001). Therefore, increasing redundancy in media coverage, i.e.,
focusing on the same subset of highly salient stories, stems from managers at
these media conglomerates who seek to reach similar, lucrative audiences (with
similar stories) and satisfy a related group of advertisers (Herman 2001).
The advent of dissemination technology, making the transfer o f news both
timely and efficient, has eased the duplication of news reporting and practices.
For example, when there is an economic event in the marketplace such as the
release of an earnings announcement there are variety of ways the financial news
is instantaneously disseminated. Reporters may register for news updates directly
through corporate websites, receive press wires directly or through tracking
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5
services, or have direct contact via email or conference calls with media relations
personnel organized by companies’ business units. The resulting redundant media
coverage then is due to the fact that there are fixed number of salient news events
that are disseminated instantaneously to a large number of media outlets. Thus,
the advent of dissemination technology provides news reporting organizations,
both large and small, with the ability to focus cost effectively on the similar
subset of the current news events that are deemed to provide value for their
target readership and satisfy advertisers.
A focus on returns to scale in the media industry has important implications
for the level of redundancy in media coverage (Stromberg 2002). As an example,
researching stories, editing, and compiling a newspaper is expensive, however,
the cost of printing additional newspapers is minimal (Stromberg 2002).
Likewise, for large news agencies, who are often the only agencies with sufficient
resources to cover many important stories, researching and compiling stories is
expensive; however, the cost involved in selling these stories to a multitude of
outlets is relatively small. This cost structure creates an incentive for the few
large news agencies to sell stories to as many outlets as possible and an additional
incentive to produce highly salient stories with broad readership appeal.
In sum, I argue that the effects of redundant media coverage are important to
understand because redundancy is the natural by-product of the functional
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
features of sponsorship from advertisers and subscriptions from consumers.
Further contributing to the prevalence of redundant media coverage is the trend
towards consolidation in the media, which diminishes independence and
correspondingly leads to homogeneity of news practices and reporting. In
addition, the advent of dissemination technology, i.e., the relatively inexpensive,
instantaneous dissemination of news from a small group of news agencies to a
large body of media outlets, and the industry focus on returns to scale further
contribute to the redundancy resulting from the functional features.
IMPORTANCE OF NONPROFESSIONAL INVESTORS
Examining nonprofessional investors’ judgments is important due to
the integral role that individual investors assume in the market today. Individual
investor participation in the stock market increased rapidly during the 1990s
(Cummings 2002). The most recent comprehensive investment data, from the
1998 Survey of Consumer Finances (conducted through the Federal Reserve
Board) indicates that 84 million individual investors own shares of stock. An
increasing trend in individual stock ownership is apparent when the 1998 data is
compared to that from earlier years. The number of individual investors
increased by approximately 32 million between 1989 and 1998, and by 15 million
between 1995 and 1998 (NYSE 2000). The primary forms of individual stock
ownership are direct ownership in publicly traded companies, equity mutual
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7
funds, self-directed retirement plans, and defined contribution pension plans.
When the overlap is accounted for, a total of 84 million shareowners hold stock
through at least one of these channels, and three million hold stock through all
four channels (NYSE 2000). It is interesting to note that all forms of stock
ownership increased during the late 1990s, with stock ownership through self
directed retirement accounts and equity mutual funds being the two largest
contributors to the growth of share ownership (NYSE 2000).
Both empirical and theoretical studies indicate that nonprofessional
investors, even those who act irrationally, may have an impact on the stock
market that is not easily arbitraged away by professional investors (Thaler 1993;
Shleifer 2000; Libby et al. 2002; Bonner 2005).3 Even with the rather tenuous
assumption that professional investors do not fall prey to the same biases as
nonprofessional investors (Bonner 2005), it is not likely that nonprofessionals can
be driven from the market through arbitrage.4 This is due to the fact that arbitrage
is normally limited due to the lack of perfect or even good substitutes for
securities. Additionally, when suitable substitutes do exist, arbitrage is
3 For a counter-interpretation of the evidence, see Rubinstein (2001).
4 See the reviews of Belkaoui (1984), Maines (1990), and Libby and Tan (1999) in Chapter 2 for
evidence in support of the idea that even professionals may react inappropriately to redundant
information.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
inherently risky because of the delay in prices converging to fundamental values
(Shleifer 2000; Brav and Heaton 2002).
INTERSECTION OF THE FINANCIAL MEDIA AND
NONPROFESSIONAL INVESTORS
Due to functional features and institutional trends in the media, a
significant part of financial press coverage is overlapping or redundant. For
example, a recent search for Microsoft earnings information in Factiva5 yielded
1,018 perfectly redundant articles in a set of 21,595 total articles. It is likely that
there are many additional highly redundant articles as well. Because of the
recency of this intersection between nonprofessional investor interest in the stock
market and proliferation of financial media coverage, the effects of redundant
press coverage on nonprofessional investors are not well understood. These
effects are important to understand because there may be negative implications
for the decision-making of nonprofessional investors who inappropriately weight
redundant press coverage (Bonner 2005). For example, chasing recently
successful companies is a common investment mistake o f nonprofessional
investors. A recent, successful firm is more likely to be in the news and
correspondingly is more likely to have repeated news stories touting the
5 Factiva, a Dow Jones & Reuters Company, is a searchable database containing approximately 9,000
sources including newspapers, magazines, press wires, television, radio and Internet sources.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9
company’s success. Therefore, the mistake of ignoring the concept of “regression
to the mean” and purchasing stocks at their peak is likely exacerbated by
redundancy of media coverage. Another common example of a nonprofessional
investor mistake is under diversification. If there is a recent trend in the media of
covering oil company stocks, for example, then the high level of press coverage
and corresponding redundant coverage may lead investors to focus too narrowly
on investment sectors. These negative implications, i.e., “bandwagon” investing
and under diversification, have a direct impact on the economic welfare of the
nonprofessional investors and also may have an indirect impact on market
welfare in general due to the limitations of the arbitrage mechanism.
DESIGN ISSUES IN PREVIOUS RESEARCH
Previous studies examining redundancy effects in both psychology and
business may confound the concept of redundancy with either the amount of
information or the amount of data. The approach adopted in one group of studies
that examines redundancy is to employ a design in which redundant content is
added to baseline information. In this setup, the baseline information is
considered the non-redundant condition and either an explicit or implicit
comparison is made to the full information set, which is considered the redundant
condition. These types of comparisons are problematic due to the fact that they
confound redundancy with the amount of data (see Chapter 2 for additional
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
detail). Therefore, when there is an effect of redundancy, it can be attributed
plausibly to either variation in data or variation in redundancy. A second
approach examines redundancy by adding both redundant and non-redundant
content to baseline information. In this design, the baseline information plus the
non-redundant content is considered the non-redundant condition, and a
comparison is made to the baseline information plus the redundant content, which
is considered the redundant set. This approach is also problematic because it
confounds redundancy with the amount of information. In this case, variation in
the dependent variable can be attributed to either variation in redundancy or
variation in amount of information. The present study implements a unique
design that focuses more precisely on the effects of redundancy.
PROCESS EXPLANATION FOR REDUNDANCY EFFECTS
Extant research typically does not investigate process explanations for
redundancy effects. This study extends the examination of redundancy by
formally investigating such an explanation. To examine a process explanation,
research demonstrating a redundancy effect is integrated with research in
psychology and marketing that suggests that perceived credibility mediates the
observed variation in investor judgment and decision-making. Previous research
has not linked these two literatures in a formal manner. Investigation of process
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
factors is an important first step in remediation of any negative observed effects
of redundancy on judgment quality (Bonner 1999).
RESULTS OF THE STUDY
An experiment is conducted to examine the effects of redundancy in
financial press articles on investors’ earnings forecasts, and also to study whether
a perception that redundancy is positively associated with information credibility
explains these effects. In the study, nonprofessional investors, proxied by
graduate business students, are randomly assigned to treatments that vary the
level of redundancy. The design employs two distinct types of comparisons
between the redundant and non-redundant conditions6 to control for the natural
confounds that arise when attempting to vary the level of redundancy.7 To
measure the effects of redundancy, I use the difference between a baseline
earnings forecast that investors make after reading company background and
financial statement information and a second earnings forecast that investors
make after reading either a redundant or non-redundant set of press articles. It is
predicted that investors in the redundant press article condition will overweight
6 I refer to the conditions as either “redundant” or “non-redundant” throughout the paper for
expositional ease; however, in operationalizing these concepts, 1 do not achieve exact representations
of redundancy and non-redundancy. Rather, I approximate these definitions as closely as possible
given the constraints inherent in an experimental design. Therefore, my conditions are most accurately
interpreted as “relatively more redundant” and “relatively less redundant” conditions. See the
discussion on this topic in Chapter 3.
7 See Chapter 3 for additional detail as to the design.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
their information sets when forming earnings forecasts.8 Moreover, it is predicted
that investors’ responses to the redundant information will be mediated by a
perception that redundancy is positively associated with information credibility.
That is, investors in the redundant conditions will perceive that their information
sets are more credible than will investors in the non-redundant conditions. In
addition to the primary mediational analysis, theoretically plausible alternative
explanations are evaluated. Specifically, I evaluate whether either the potential
increased salience of redundant information, or the perception of editorial
signaling associated with redundant press articles mediates earnings forecast
changes. Consistent with expectations, the results indicate that investors in the
redundant press article condition overweight their information sets. Furthermore,
the mediational results are consistent with the investors’ responses being due to
enhanced credibility.
SUMMARY
This chapter presented the research questions proposed and investigated in
this study. In addition, the chapter addressed the general importance of these
research questions, i.e., the exploration of the intersection of two dynamic factors
in the market today: the financial media and nonprofessional investors. The
academic importance of the chapter focused on two key contributions:
8 See Chapter 3 for a discussion of the term “overweight”.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
opportunities for design improvements from previous research on information
redundancy and the formal investigation of a process explanation for the
redundancy effect. Lastly, a summary of the results was presented.
The remainder of the thesis is organized as follows. Chapter 2 discusses
prior literature and hypothesis development. Chapter 3 outlines the experimental
method and procedures. Chapter 4 discusses the statistical analyses and the
results. Chapter 5 summarizes the dissertation findings, discusses the limitations
of the study, and provides future research directions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
CHAPTER 2: THEORY AND HYPOTHESES
OVERVIEW
The theory and hypotheses chapter of this thesis begins by discussing the
factors leading to a proliferation of redundant financial press coverage, why
redundant sources of information are problematic for judgment and decision
making in general, and why redundant financial press coverage is problematic for
investors’ judgment and decision making in particular. Next, prior research
investigating the effect of redundancy on judgment and decision-making in
psychology and business is reviewed and hypotheses pertinent to the present
study are presented. Finally, research in psychology documenting a potential
process mechanism underlying the redundancy effect is reviewed and hypotheses
linking perceived credibility, the process mechanism, to investors’ earnings
forecast changes are generated.
REDUNDANT MEDIA COVERAGE: THE POTENTIAL NEGATIVE
IMPACT ON JUDGMENT AND DECISION MAKING
Factors Underlying Redundant Media Coverage
As mentioned above, there has been a boom in financial media
coverage in recent years. Currently, there are approximately 10,739 newspapers,
13,876 periodicals, 240 financial magazines, 112 financial television shows, and
165 financial newsletters in the United States alone (Gale Group 2001, Bacon’s
Information Inc. 2003). Due to the aforementioned functional features and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
institutional trends within the media industry, a non-trivial portion of this press
coverage is redundant.
Why Redundancy is Problematic for Judgment and Decision-Making
Ceteris paribus, redundancy in information sources is problematic due
to positive correlation in the sources’ deviations. When a composite estimate, i.e.,
an estimate comprised of multiple sources, of a criterion differs from the
criterion itself, there exists a deviation. In the case of redundant sources, this
deviation can be parsed into bias (systematic) and error (non-systematic) terms.
In the case of non-redundant sources, the deviation is comprised of only non-
systematic error. Therefore, in the case of two sources that produce a composite
estimate, the deviation terms are pairwise correlated in the redundant case and
pairwise uncorrelated in the non-redundant case. Assuming the accuracy of the
composite estimate is determined by its variance, the accuracy will be increasing
in the correlation of the deviation terms. That is, the two sources that are
redundant and possess correlated deviation terms will result in a less accurate
composite estimate. Thus, two sources that are non-redundant result in an
estimate that is greater in accuracy relative to the two sources that are redundant.
Identical reasoning applies to the more specific case of redundant
financial press coverage. That is, when the same source of information is used
multiple times to produce multiple financial press articles, there will be positively
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
correlated deviations among these articles. Therefore, when accuracy is
controlled across sources (as is the case in the current study), statistical theory
implies that investors should place less weight on information derived from
redundant sources than on information derived from non-redundant (independent)
sources.9 However, care should be taken to note that the primary point is not that
redundant sources are devoid of value for decision- making. In fact, Einhorn et
al. (1979) discuss several potential benefits of redundancy including limited
information search, facilitation of cue intersubstitutability, and alleviation of cue
unreliability. Rather, the primary point is that, ceteris paribus, redundant
financial press articles are less “informative” than non-redundant articles due to
the problem of positively correlated deviations. Therefore, if investors prefer
redundant articles to non-redundant articles when making a decision, they are
making an error in judgment, i.e., over-weighting the redundant information. This
error in judgment may have direct negative implications for the economic welfare
of individual investors. As previously mentioned, over-weighting redundant
media stories may result in a “bandwagon” effect in investment purchase
decisions as well as contribute to under diversified investment portfolios. In
addition, these investment errors on the part of nonprofessional investors have
9 Of course, there is the potential belief by investors that the repetition of news events in the financial
press represents an information signal (Spence 1973). See Chapter 4 for discussion and analysis
related to this possibility.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
implications for the market in general due to the limitations of the arbitrage
mechanism.
Next, I review prior research on the effects of redundancy on judgment
and decision-making in a variety of contexts. I organize my review of this
literature around the two different approaches utilized in prior studies for
investigating the effects of redundancy.
EFFECTS OF REDUNDANCY ON INVESTORS ’ JUDGMENTS:
ASYMMETRICAL COMPARISONS
Introduction to Asymmetrical Redundancy Comparisons
An approach employed by a number of studies examining redundancy is a
design in which redundant content is added to baseline information. In this setup,
the baseline information is considered the non-redundant condition and either an
explicit or implicit comparison is made to the full information set, which is
considered the redundant condition. I term this type of comparison asymmetrical
redundancy comparison, hereafter (AR), because there is a greater quantity of
data in the redundant condition than in the non-redundant condition. AR
comparisons are problematic due to the fact that they confound redundancy with
the quantity of data. (See Figure 1 for an example of an AR comparison and a SR
comparison- explained in the following section.)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
A Note on the Distinction between Information and Data in this Study
In the context of this study, information is defined as data that is relevant
in forming an expectation of future earnings. Therefore, if a decision-maker
receives a new item of information regarding the positive future expectations of
earnings for a company, the decision-maker should revise the expectation of
earnings in a manner corresponding to the direction (positive in this study) of the
information. Conversely, if a decision-maker receives a new item of data
regarding the positive future expectation of earnings for a company, the decision
maker should not the expectation of earnings since the “data” contains no new
information.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
FIGURE 1
EXAMPLE OF REDUNDANCY COMPARISONS
Asymmetrical Comparison
Example Comparison between Redundant Condition (0) and Non-
Redundant Condition (4)
Redu n d a n t Non-redundant
al = fl (ild2? I3 4 4) al = fl ObL, I3 4 4)
&2~ fiihA&hM)
where each a represents a financial press article and
each / represents an independent item of information.
Symmetrical Comparison
Example Comparison between Redundant Condition (0) and Non-
Redundant Condition (1)
Redundant
a i = fi (11,12, 13,14)
a 2 ~ f2 ( 11,12, 13,14)
Non-redundant
ai = fi (11,12, 13,14)
a 2 = {2 ( 15, 16,17,18)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
Review of Prior Literature Employing Asymmetrical Redundancy
Comparisons
Research in psychology provides support for the idea that redundancy
positively influences perceptions of information content.1 0 Research in finance
and accounting using AR comparisons is largely consistent with these findings
from psychology.
In a financial setting, Ho and Michaely (1988) examine redundancy in an
archival study. Specifically, they study the impact of newspaper stock
commentaries on stock returns. The authors select 29 pessimistic newspaper
commentaries published between 1982 and 1984 in Barron’ s and The Wall Street
Journal. This design is AR due to the fact that the authors select newspaper
commentaries that are intended to be redundant with previously available
financial information. Therefore, they are implicitly comparing the market
reaction to an information set that contains a baseline of publicly available
financial information (non-redundant condition) and the market reaction to a set
that contains the publicly available financial information plus “repackaged”
newspaper commentaries (redundant condition). The authors analyze daily stock
returns over an event period consisting of five trading days before and five
1 0 Note that research in psychology has not focused specifically on AR comparisons. Rather, the
research has centered on SR comparisons (reviewed in the following section) and process
explanations for the mechanism underlying the credibility effect. However, these results have
relevance for how participants may react to redundancy in the AR comparison.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
trading days after the date of the article publication. The results indicate that the
publication of these redundant newspaper commentaries is associated with
significant abnormal returns on the event date as well as the day prior to
publication and the second day after the event. These findings suggest that
investors react to financial press coverage that is redundant with previously
released financial information. However, an important caveat to their results is
that the findings appear to be driven by small companies, suggesting a possible
confound between information and redundancy.
Belkaoui (1984) uses an experiment and examines the effects of redundant
financial ratios on loan officers’ predictions of bankruptcy. In this study, thirty
bank loan officers predict which of ten actual companies will declare bankruptcy
and state their confidence in these predictions based on the availability of either
diagnostic (non-redundant) or redundant financial ratio sets. This comparison is
AR because a comparison is made between investors viewing a set of six
financial ratios (non-redundant condition) and another group of investors viewing
the same six ratios plus three redundant financial ratios (redundant condition).
The additional three financial ratios are considered redundant because a
classification model based on adding these ratios to the baseline of six ratios does
not improve upon the original model. The loan officers in the non-redundant
condition receive a consecutive three-year set of six financial ratios for each of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ten firms. These ratios are net income/total assets, cash/total assets, current
assets/total assets, current assets/current liabilities, net assets/current assets and
total liabilities/owners’ equity. Loan officers in the redundant condition receive
the same ratios listed above in addition to the following ratios: total
liabilities/total assets, cash/current liabilities, and current assets/sales. The results
indicate that redundant information did improve the loan officer’s bankruptcy
predictions; however, an important caveat to these results is the fact that this
study employs an AR comparison. In particular, the participants in the redundant
condition in this setting may have benefited from a reduction in measurement
error, whereas the participants in the non-redundant condition were not provided
with the corresponding benefit from a reduction in measurement error and bias
(through presentation of additional non-redundant cues). In addition to the
primary results, the judgments of the participants in the redundant condition did
not conform to statistical theory in the sense that they displayed greater
overconfidence in their bankruptcy predictions in the redundant information
conditions.
Joe (2003) also uses an experiment to examine the effects of redundancy
in press articles. Specifically, she examines why auditors are more likely to issue
a going-concem opinion when their audit client has negative press coverage that
is redundant with information already contained in the audit workpapers. The
permission of the copyright owner. Further reproduction prohibited without permission.
purpose of her study is to investigate whether there is a strategic or cognitive
explanation behind these results that appear in previous archival work (Mutchler
et al. 1997; Frost 1991). That is, previous archival work documents such a
reaction and attributes it to an economic or strategic reason, i.e., auditors’
reactions to redundant press coverage as an economic or strategic reason in that
auditors are acting in a conservative manner in their opinion choice due to an
increased risk of litigation brought on by the press coverage. The Joe (2003)
design is AR due to the fact that she makes a comparison between a baseline of
the client’s loan default information (non-redundant condition) and the loan
default information plus a fictitious Wall Street Journal article that repeats the
loan default facts (redundant condition). Her findings confirm the archival results
that auditors do modify their opinions conditional on press coverage; however,
her findings support the idea that the reason is cognitive rather strategic.
Specifically, she hypothesizes that redundant press coverage increases the
salience of the company’s debt problems, which in turn increases the auditors’
perceptions of the probability of client bankruptcy and leads to a greater
propensity to modify the audit opinion.
permission of the copyright owner. Further reproduction prohibited without permission.
24
TABLE 1
SUMMARY OF ASYMMETRICAL
REDUNDANCY STUDIES
Study Participants Method How is
Redundancy
Operationalized?
Did
participants’
reactions to
redundancy
conform to
theory?
Ho and
Michaely
(1988)
Market
(Archival
study)
Market reaction to
pessimistic
newspaper
commentaries
published between
1982 and 1984 in
B arron’ s and The
Wall Street
Journal.
Press articles are
redundant with
publicly available
information.
No. There are
significant
abnormal
returns on the
event date as
well as the day
before and
two days
afterwards.
Beikaoui
(1984)
30 bank loan
officers
Participants use
financial ratios to
predict which o f
10 actual
companies declare
bankruptcy and
state confidence in
these predictions.
The addition o f
financial ratios
(that do not
increase
predictive value)
to a baseline o f
financial ratios.
No.
Redundancy
increases
overconfidenc
e in
predictions.
J o e (2003) 90 in-charge
auditors
Auditors decide
whether to issue a
going-concem
opinion for an
audit client with
negative press
coverage that is
redundant with
information
already in the
audit work papers.
A Wall Street
Journal article
that repeats the
loan default facts
already contained
in the audit work
papers.
No.
Redundant
press coverage
increases
auditors’
perceptions of
the probability
o f client
bankruptcy.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
As summarized in Table 1 above, the findings from research on
redundancy in AR settings indicate that decision makers do not perceive
redundant information in a manner that conforms to statistical theory; however,
there are at least two reasons why redundant information may not influence
investors in the AR comparison in my study. First, because redundant financial
press coverage adds no new information, there is little logical basis for the
influence of redundant information on beliefs (Bacon 1979; Begg et al. 1992).
Therefore, in the current study setting where the redundant press articles are
juxtaposed (as opposed to a time lag between presentations in previous studies),
investors may infer that the second, redundant article adds no new information.
Second, central to research in accounting and finance is the idea that investors in
financial markets interpret new information efficiently (Fama 1970). Although
the investors in this study are nonprofessionals, the setting is significantly less
complex than an external market setting; therefore, again, these investors may
discount for redundancy in the relatively unambiguous redundant information
conditions.
The research reviewed above provides evidence that redundancy
influences judgments and decisions in AR comparisons;1 1 despite the fact that
1 1 As previously mentioned, the AR studies may confound the number of data items with redundancy,
therefore, the implications of these studies for the current study should be viewed only in terms of this
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
redundancy adds no new information and the current setting is relatively
transparent, I believe that nonprofessional investors in the current study will act
similarly to participants in the reviewed research, and, therefore, conjecture that
redundancy operationalized asymmetrically in this study will influence their
earnings forecasts. However, it is crucial to note that this is only one form of
comparison and that this representation necessarily confounds the quantity of data
with redundancy. Therefore, even if an effect is demonstrated in the AR
comparison, this result is not sufficient to support the idea that redundancy per se
influences investors. Therefore, I also review studies that employ a second form
of comparison between non-redundant and redundant content- the symmetrical
redundancy comparison, which controls for the quantity of data.
EFFECTS OF REDUNDANCY ON NONPROFESSIONAL INVESTORS’
JUDGMENTS: SYMMETRICAL COMPARISONS
Introduction to Symmetrical Redundancy Comparisons
Symmetrical comparison studies examine redundancy by adding both
redundant and non-redundant content to baseline information. In this design, the
baseline information plus the non-redundant content is considered the non-
redundant condition and a comparison is made to the baseline information plus
the redundant content, which is considered the redundant set. I term this
study’s overall design. Please see the discussion surrounding Table 3 for additional detail as to the
overall design investigating an effect of redundancy on investors’ judgment and decision-making.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
representation a symmetrical redundancy comparison, hereafter (SR), because
there is an equal quantity of data in the non-redundant and redundant conditions.
Review of Prior Literature Employing Symmetrical Redundancy
Comparisons
Research in psychology provides support for the idea that redundancy will
influence participants’ judgment and decision-making in an SR setting (Goethals
and Nelson 1973; Kahneman and Tversky 1973).1 2 Kahneman and Tversky
(1973) examine the effect of information redundancy on judgment and decision
making in a prediction task. In their study, participants predict grade point
averages based on the results of two pairs of aptitude tests. Participants are
informed that the first pair of aptitude tests (creative thinking and symbolic
ability) are highly correlated, the second pair of tests (mental flexibility and
systematic reasoning) are not correlated, and all four individual tests are equally
predictive of grade point average. The results indicate that the participants prefer,
i.e., have more confidence in, the correlated information set even though the set
has inferior predictive ability relative to that of the uncorrelated set.1 3 These
1 2 Goethals and Nelson (1973) and Kahneman and Tversky (1973) represent distinct departures from
earlier literature that addresses redundancy from a learning perspective rather than a decision-rnaking
perspective (see, for example, the multiple-cue probability learning literature: Naylor and Schenck
1968; Knowles et al. 1971; Lindell and Stewart 1974; Brehmer 1974; Armelius and Armelius 1974;
and Schmitt and Dudycha 1975).
1 3 Note that Kahneman and Tversky do not report the results of the predictions. That is, they do not
report the results o f participants’ predictions as compared to either a relative or objective accuracy
criterion.
permission of the copyright owner. Further reproduction prohibited without permission.
28
results provide evidence that people engaged in a prediction task may have
difficulty properly weighting information cues due to an improper consideration
of information redundancy, and that this may lead to an overestimation of the
value of redundant information.
Goethals and Nelson (1973) conduct an experiment to investigate
differential reliance on similar and dissimilar others conditional on whether a
belief or a value is at issue. The similarity of others is manipulated by providing
the participants with two profiles: their own and that of the other evaluator. These
profiles indicate their ratings on a test designed to measure their style of judging
personal characteristics. The participants are informed that they can use these
ratings to determine how similar their own style is relative to the other evaluator.
In the redundant condition, the profile sheets are almost identical and in the non-
redundant condition, the profile sheets are very divergent. This study represents
an interesting comparison of the implications of Festinger (1954) and Kelley
(1967). Festinger’s (1954) social comparison theory suggests that people prefer to
evaluate their opinions by comparing them with those of people similar to
themselves. This notion lends support to the idea that, in some situations, people
may prefer redundant opinions. In contrast to this view, an implication of
Kelley’s (1967) attribution theory is that people may realize that similar others
who agree share the same bias and that this results in less accurate opinion
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
evaluations. Goethals and Nelson conjecture that when the issue is a question of
belief, i.e., a potentially verifiable assertion, Kelley’s view is more relevant, but
when a value is at stake, i.e., an assertion about the goodness or badness of an
entity of state of affairs, a person’s interest rests in preference for what is
affectively positive and Festinger’s view is more relevant. To test their
hypotheses, the authors investigate whether, when a belief is in question, an
individual’s confidence in the decision will be increased by agreement from a
similar or dissimilar other but, when a value is at issue, whether a similar other
provides more social support and causes a greater increase in confidence. In the
study, participants judge either the relative academic success of two students
(belief) or which of the two they like more (value). The participants are given an
evaluation of the students by another participant who they are led to believe is
similar or dissimilar to themselves in terms of style of judging other people. The
results support the authors’ hypotheses that when a belief is at issue agreement
from a dissimilar other increases confidence and when a value is at issue
agreement from a similar other increases confidence. Thus, the authors
demonstrate mixed effects for the influence of redundancy that are conditional on
the whether a belief or value is at stake. The context of the current study is most
similar to the belief treatment in Goethals and Nelson. That is, forecasting
earnings per share is a potentially verifiable outcome that, in accordance with the
permission of the copyright owner. Further reproduction prohibited without permission.
30
Goethals and Nelson results, would result in an appreciation for the relative
superiority of non-redundant information.
Research in finance and accounting also provides mixed evidence for the
influence of redundancy when operationalized using the SR comparison. Davis
et al. (1994) examine the effects of redundancy in an experiment where
participants forecast earnings per share for actual companies (using a
computerized news retrieval system). Their design is SR because they have a
baseline of graphical financial information and then add either a management
disclosure derived from the baseline information (redundant) or one that is
independent of the baseline information (non-redundant). The results indicate that
investors have significantly more confidence in the non-redundant information
than the redundant information; however, forecast accuracy is diminished in both
the redundant and non-redundant conditions relative to the baseline condition.
Care should be taken in interpreting these results due to the fact that when
constructing the redundant and non-redundant conditions the content added was
not the same in both conditions. That is, the content added to form the non-
redundant condition was associated with aggregate business results, e.g.,
“Results will be the best in our history”, while the content added to form the
redundant condition pertained to financial facts derived from the data, e.g.,
“Earnings increased 112% from the 1st quarter to the 3rd quarter.” This
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
construction is important due to the fact that perceived relevance needs to be
controlled for across conditions when evaluating redundancy effects on
confidence. Furthermore, it is not obvious that the specific items of information
added to the baseline condition in order to construct the non-redundant condition
are actually predictive of earnings per share in general or earnings per share for
the companies forecasted in this study in particular. In sum, the authors provide a
degree of evidence that investors appropriately weight the redundancy in
information in attributing confidence; however, the results for judgment and
decision making need to be interpreted in light of the design problems.
Gonzalez (1994) employs a stock portfolio reallocation task where
participants make a series of portfolio reallocations on the basis of financial
reports. The investors view the responses of another person who uses either the
same information as the participant (redundant information) or different but
equally predictive information (non-redundant information). This design is SR
because redundant and non-redundant content are added to baseline information
creating the redundant and non-redundant conditions, respectively. The
dependent measure is an index of the degree of conformity to the other investor’s
decision. The first set of results show that investors do not appreciate the relative
benefit o f non-redundant information and, furthermore, investors claim that they
are not influenced by other investors’ even though their actual responses differed
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
from those recorded in the control condition. Therefore, this set of results
provides evidence that investors overweight redundant information and also are
unaware of this error. In the second set of results, the procedure focuses on
investors’ attention to other investors’ evaluations rather than decisions. Results
indicate that investors do consider the redundancy in other investors’ evaluations.
The results show that investors are sensitive to the redundant information content
of another's responses when observing how the investor evaluates their
information, but are not sensitive to redundancy when observing how the investor
acts on their information. The lack of sensitivity to redundancy when participants
only observe another’s actions may be due to the fact that participants are unable
to evaluate the relevance of the methods employed by the other participants.
Similar to Goethals and Nelson (1973), this study provides mixed evidence for
the effect of redundancy conditional on whether the investors see other investors’
methods or merely their final decisions.
In an archival setting, Mutchler et al. (1997) investigate the use of
contrary information and mitigating factors for auditors’ going-concern
modification decisions for companies prior to bankruptcy. As part of this
investigation, the authors examine the impact of negative Wall Street Journal
coverage on the audit opinion, after controlling for the possibility of bankruptcy.
The news measures are based on news items referenced in the Wall Street
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
Journal Index during the period beginning two years before the audit-report date
for the last audit opinion issued prior to the bankruptcy and ending with the
bankruptcy date. A comprehensive categorization of news items relevant to the
companies’ financial health are classified into one of 82 predetermined
categories. These categories include, for example, financing and credit (e.g.,
securities offerings), corporate structure (e.g. mergers), bankruptcy (e.g.
subsidiary filing for Chapter 11), management (e.g. new CEO, CEO resigns), etc.
(Mutchler et al. 1997, p. 301). This study is an example of an SR comparison
because the authors include extreme negative news events that are both redundant
and non-redundant with the client’s footnote disclosure of default events. The
results show that negative news items concerning default events that are
redundant with information already contained in the footnotes are associated with
a modification in the audit opinion. Thus, the auditors do not appear to properly
evaluate the redundant information contained in the news items.
An example of an implicit test of redundant information is that by Libby
and Tan (1999).1 4 In this study, the authors investigate analysts’ reactions to
qualitative warnings of adverse earnings. The authors conjecture that an
inconsistency arises between analysts’ reactions to earnings warnings and
analysts’ perceptions of their reactions. This inconsistency exists due to the fact
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
that analysts revise their earnings forecasts sequentially in response to both the
warning and the earnings announcement; however, their responses to
retrospective questions about the effects of warnings are the result of a
simultaneous response to both signals (Libby and Tan 1999, p. 416). In the study,
twenty-eight financial analysts predict future years’ earnings in one of three
conditions that vary as to the presence of an earnings warning. In a no warning
condition, analysts receive an earnings announcement but no warning and then
make an adjusted earnings forecast. In the sequential warning condition, analysts
receive the warning first, adjust their earnings forecast and then receive the
earnings announcement and adjust the earnings forecast a second time. Finally, in
the simultaneous warning condition analysts receive both the warning and the
actual earnings announcement simultaneously and make a single adjustment in
response to both signals. The results indicate that analysts in the sequential
warning condition forecast lower future earnings than analysts in either the
simultaneous warning or the no warning condition. Since the information in the
warning is redundant with the information in the actual earnings announcement,
the analysts should not have a greater reaction to the sequential and simultaneous
warning conditions than the no warning condition. Thus, the results in this study
1 4 In this context, “implicit” connotes that the authors did not explicitly set out to study redundant or
correlated information events.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
provide evidence that investment professionals may not properly weight
redundant financial information.
Maines (1990) examines the effects of information redundancy in a
financial forecasting setting using both between-participants and within-
participants designs. In her first experiment, Maines uses a statistical measure of
redundancy, i.e., correlation of forecast errors, 120 MBA student participants, and
confidence intervals as the response mode. The experimental materials provides
participants with ten forecasts for total sales prepared by sales representatives.
Participants receive one of three possible types of information about the
redundancy of the sales forecasts: correlation among the forecast errors is 80%,
correlation among forecast errors is 10%, or a baseline case where participants
receive no information regarding the correlation among forecast errors.
Participants provide a best estimate of sales, an optimistic estimate of sales, and a
pessimistic estimate of sales, where the optimistic estimate and the pessimistic
estimate serve as bounds for the confidence interval. The results of the between-
participants design indicate that the participants do not appreciate the positively
correlated forecast errors. In the within-participants design, however, the results
indicate that participants incorporate the effects of redundancy.
In her second experiment, Maines uses the data on past forecasts and
actual outcomes, 120 MBA student participants, and confidence intervals as the
permission of the copyright owner. Further reproduction prohibited without permission.
response mode. Participants receive three analysts’ forecasts of earnings per
share for a fictional company. In addition, the case contains 24 quarters of past
forecast data that varies as to redundancy. The results indicate that participants in
neither the between-participants nor the within-participants designs incorporate
forecast redundancy into their judgments. In her third experiment, Maines uses
descriptions of forecasters’ use of similar or dissimilar information and
forecasting methods, 47 financial analyst participants, and probability judgments
as the response mode. Participants are told that they recently had prepared an
annual earnings forecast for a corporation and are 60% confident that actual
earnings per share will be within plus or minus $0.10 of their forecast. Their
supervisor asks them to find another analyst’s earnings per share forecast for this
company in order to provide a better assessment of their forecast’s accuracy.
Redundancy is operationalized by the similarity between methods and
information used by the subject and the other analyst to determine their forecasts.
The between-participants results (based on thirty-two participants) does not show
that analysts properly incorporate redundancy into their judgments; however, the
fifteen participants in the within-participants design version do incorporate
redundancy into their probability judgments. Overall, Maines’ results indicate
that participants recognize the negative effect of information redundancy when
they are able to compare the redundant condition to a less redundant condition;
permission of the copyright owner. Further reproduction prohibited without permission.
however, they are unable to recognize this effect when a comparison group is not
made salient. An important consideration in interpreting these results is Maines’
use of a forecasting environment. In this environment, participants may have
been susceptible to an inference of redundancy implying greater accuracy due to
the fact that high redundancy in forecasts normally is indicative of more accurate
forecasts (Lipe 1990). This is an important distinction because perceptions of
accuracy may interact with actual accuracy manipulations and redundancy
manipulations.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
TABLE 2
SUMMARY OF SYMMETRICAL REDUNDANCY
STUDIES
Study Participants Method How Redundancy is
Operationalized
Did participants’
perceptions o f
redundancy
conform to
theory?
Kahneman
and
Tversky
(1973)
131
college
students
Participants
attempt to
predict grade
point
averages
based on
aptitude tests.
One pair o f aptitude
tests are highly
correlated (redundant),
and the second pair o f
tests are not correlated
(non-redundant).
No. Participants
prefer the
correlated
information set
even though the
set has inferior
predictive ability.
Goethals
and Nelson
(1973)
34 high
school
graduates
Participants
judge either
the relative
academic
success o f
two students
(belief) or
which o f the
two they like
more (value).
Participants are given
an evaluation o f the
students by another
participant who is
similar or dissimilar to
themselves.
Mixed.
Participants
incorporate
redundant
information
properly when a
belief was at issue,
but not when a
value was at issue.
Davis et al.
(1994)
22 MBA
students
Using a
computerized
news
retrieval
system,
participants
forecast
earnings per
share for
actual
companies.
Participants receive a
baseline o f graphical
financial information
plus either a
management disclosure
derived from the
baseline information
(redundant) or one that
is independent o f the
baseline information
(non-redundant).
Yes. The results
indicate that
investors have
significantly more
confidence in the
non-redundant
information than
the redundant
information.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
TABLE 2 (Continued)
SUMMARY OF SYMMETRICAL REDUNDANCY
STUDIES
Gonzalez
(1994)
107
college
students
Participants
complete a
stock portfolio
reallocation
task after
reviewing
responses o f
other (fictional)
participants.
Participants view
the responses of
another who uses
either the same
information as the
participant
(redundant) or
different but
equally predictive
information (non-
redundant).
Mixed. Participants
properly weight
redundant
information when
viewing another’s
evaluations but do
not properly weight
redundancy when
viewing another’s
actions.
Mutchler et
al. (1997)
Market
(Archival
study)
The authors
investigate the
impact o f
negative Wall
Street Journal
coverage on the
audit opinion
after
controlling for
the possibility
o f bankruptcy.
They include
extreme negative
news events that
are both redundant
and non-redundant
with the client’s
footnote disclosure
o f default events.
No. They find that
negative news items
concerning default
events that are
redundant with
information already
contained in the
footnotes are
associated with a
modification in the
audit opinion
Libby and
Tan (1999)
28
financial
analysts
Financial
analysts in
earnings
warning
conditions
make adjusted
earnings
forecasts.
Earnings warning
information is
repeated since the
warning
information is
contained in the
final earnings
announcement.
No. Results indicate
that analysts in the
sequential warnings
condition forecast
lower future earnings
relative to the
remaining conditions.
Maines
(1990)
Exp. 1
120
MBA
students
Participants
forecast sales
in both
between and
within-
participants
conditions.
Statistical
correlation among
the forecast errors
o f the sales
forecasts provided
to the participants.
Mixed. The between-
participants design
does not incorporate
redundancy and the e
within-participants
design does.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
TABLE 2 (Continued)
SUMMARY OF SYMMETRICAL REDUNDANCY
STUDIES
Maines
(1990)
Exp. 2
120 MBA
students
Participants
determine a
consensus EPS
forecast and a
confidence
interval.
There was an
implicit
correlation
among
analysts’ past
earnings per
share forecasts.
No. In both the between-
participants and within-
participants designs,
participants did not
properly weight the
redundant information.
Maines
(1990)
Exp. 3
47
financial
analysts
Participants are
asked to form a
consensus
earnings forecast
and rate the
probability that it
falls within a
confidence
interval.
Redundancy
was
characterized
by the
similarity
between
methods and
information
used by the
subject and the
second analyst
to determine
their forecasts.
Mixed. In the between-
participants design,
participants did not
properly incorporate
redundancy, and in the
within-participants
design they did.
Overall, the research summarized in Table 2 above provides mixed
evidence that redundant information operationalized using an SR comparison
influences judgment and decision-making. Because the current study controls for
the perceived accuracy and relevance of the information across conditions
(factors contributing to a reaction to redundancy in the reviewed studies), I
conjecture that redundant information will positively influence investors’
judgments in the SR comparison. It is important to note that, while this form of
comparison corrects the problem of confounding redundancy and quantity of data
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
that occurs in the AR studies, it creates a new confound between redundancy and
the quantity of information. Thus, the results from an SR comparison are not
sufficient to support the idea that investors inappropriately weight redundant
information. The next section outlines an approach that combines AR and SR
comparisons into an experimental design to account for the quantity of data and
the quantity of information confounds for redundancy effects. It is the premise of
this study that this comprehensive design in conjunction with a process
explanation provides sufficient evidence to conclude that investors overweight
redundant financial press information.
APPROACH FOR STUDYING THE EFFECTS OF REDUNDANCY ON
NONPROFESSIONAL INVESTORS’ EARNINGS FORECASTS
Introduction to the Approach for Studying the Effects of Redundancy
This section explicates the approach taken to study the effects of
redundant financial press information on investors’ earnings forecasts while
accounting for the confounds inherent in the AR and SR comparison formats. As
reviewed in the previous sections, extant research creates either an AR
comparison by adding redundant content to a baseline case and considering the
baseline case as the non-redundant condition or, alternatively, creates an SR
comparison by adding redundant content to baseline information, creating a
redundant condition, and adding non-redundant content to the baseline
information, creating the non-redundant condition. The former approach
permission of the copyright owner. Further reproduction prohibited without permission.
42
confounds the quantity of data with redundancy and the latter approach
confounds the quantity of information with redundancy.
The Approach for Studying the Effects of Redundancy
The approach taken in this study to account for the naturally arising
confounds is to develop a set of hypotheses related to both AR and SR
comparisons. Consistent with the discussion above pertaining to the AR
comparison literature, the first hypothesis predicts that investors’ mean earnings
forecast change will be larger in the redundant information conditions than it will
be in the non-redundant conditions.
HI a: Investors’ earnings forecast changes in the redundant condition will
be greater than those of investors in the asymmetrical non-
redundant condition.
As previously stated, the AR comparison allows both quantity of data and
redundancy to vary; therefore, quantity of data is controlled for by also
operationalizing redundancy using an SR comparison. As discussed above in the
SR comparison literature review, I expect to find that redundancy will affect
investors’ judgments when redundancy is operationalized using an SR
comparison format. Although the SR design corrects for the quantity of data
confound in the AR design, the SR design creates a new problem by confounding
the quantity of information with redundancy. That is, there will be a greater
amount of information in the non-redundant condition than there will be in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
redundant information condition. If participants’ earnings forecast changes in the
redundant condition are larger than those of investors in the SR non-redundant
conditions, then the results demonstrate a strong response to redundancy due to
the fact that a higher level of redundancy is preferred over a higher level of
information content. However, the more likely case is that of offsetting
preferences for redundancy and information content. Specifically, if investors in
the redundant conditions react to the high level of redundancy, and the investors
in the non-redundant condition react to the high level of information content,
there will be no significant difference between the investors’ earnings forecast
changes in an SR comparison.
To complicate matters further, a non-significant difference could
alternatively be attributed to participants whose reactions are neither a function of
redundancy nor information quantity.1 5 Therefore, the demonstration of a reaction
to redundancy in an SR requires an additional, “information control,” hypothesis.
In this “information control” hypothesis, redundancy is controlled and
information content is allowed to vary. In this fashion, I can verify that
participants’ earnings forecast changes vary conditional on information content.
This verification of an investor reaction to higher information content then allows
1 5 There are a variety of reasons this may be true. For example, participants might not attend to the
case materials, expend sufficient effort, or recognize “greater” information content in a between-
participants setting.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
an inference of an influence for redundancy within the SR comparison. In sum,
to examine whether redundancy affects investors’ judgments in the SR
comparison, the following hypotheses are made:
H lb: Investors’ earnings forecast changes in the redundant conditions
will not be significantly different than those of investors in the
symmetrical non-redundant conditions.
H lc: Investors’ earnings forecast changes in the symmetrical non-
redundant condition will be larger than those of investors in the
non-redundant information control conditions.
The purpose of H lb is to investigate whether there is a non-significant,
difference between investors’ earnings forecast changes in the redundant and
non-redundant financial press article treatments, and the purpose of H lc is to
serve as an information control hypothesis. That is, H lc is meant to verily that
investors’ earnings forecast changes vary conditional on information content.
SUMMARY OF HYPOTHESES INVESTIGATING THE EFFECTS OF
REDUNDANCY ON NONPROFESSIONAL INVESTORS’ JUDGMENTS
As Table 3 illustrates, the current study attempts to provide evidence for an
effect of redundant financial press information on investors’ judgments by
employing a set of three hypotheses. The first hypothesis tests an AR comparison
of the redundant and non-redundant conditions. This comparison is in line with
previous studies that have manipulated redundancy but allowed the amount of
data to vaiy between the redundant and non-redundant conditions. As previously
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
mentioned, this design creates a confound between redundancy and the amount of
data, thus rendering the results difficult to interpret. Therefore, this study also
employs a second hypothesis testing an SR comparison of the redundant and non-
redundant conditions. Again, as previously described, this comparison equalizes
the data across conditions, thus remedying the confound in the AR comparison;
however, it creates a new confound between redundancy and the amount of
information content. Thus, an information control hypothesis is necessary to
control for redundancy while evaluating whether investors’ earnings forecasts are
affected by information content. Each individual hypothesis and the related
findings within this set are flawed and not capable of providing evidence for the
contention that redundant financial press information influences investors;
however, the set o f three hypotheses taken together does provide an important
first step to establishing the influence of redundancy. In the next section, the
process explanation is explained and hypotheses are developed. Evidence of a
process explanation coupled with evidence for an effect using the multiple tests
within the set of three hypotheses provides strong support for an influence of
redundant financial information on investors.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
TABLE 3
SUMMARY OF REDUNDANCY EFFECT
HYPOTHESES
Type o f
Comparison
HI Prior Research Manipulates
Red.
Varies
Info.
Varies
Data
Asymmetrical H la Ho & Michaely 1988
Belkaoui 1984
Joe 2003
V V
Symmeterical Hlb Kahneman & Tversky 1973
Goethals & Nelson 1973
Davis etal. 1994
Gonzalez 1994
Mutchler et al. 1997
Libby and Tan 1999
Maines 1990
V V
Information
Control
H lc
V
PERCEIVED CREDIBILITY AS A PROCESS EXPLANATION FOR THE
EFFECTS OF REDUNDANCY
Introduction to Perceived Credibility as a Process Explanation
Thus far, the focus of the study has been on the effects of redundancy on
investors’ judgments after taking into account the quantity of data and quantity of
information confounds. Demonstrating an effect for redundancy in both the AR
and SR comparisons is necessary but not sufficient for proving an influence for
information redundancy per se. That is, it is the premise of this study is that only
a demonstrated effect in both the AR and SR comparisons in conjunction with a
valid process explanation provides sufficient evidence to support the idea that
investors are inappropriately influenced by information redundancy. The purpose
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of this section then is to review prior literature and provide hypotheses
concerning such a process explanation for investors’ judgments.
Review of Prior Literature on the Credibility Effect
Introduction to Review of Prior Literature on the Credibility Effect
There are several goals of this literature section. First, a review is
conducted of prior research that demonstrates a credibility effect stemming from
repeated or redundant information. This part of the overall review is performed
to establish the existence of a coherent stream of research that supports the
proposed process explanation. Second, prior research that investigates the process
underlying the credibility effect is reviewed. To be clear, I propose the
credibility effect as a process explanation to explain variation in investors’
earnings forecast changes in this study; however, to gain a deeper understanding
of this link, the process mechanism underlying the credibility effect is also
explored. In other words, the second order process explanation, while not
formally investigated in this study, serves an important role in understanding how
the hypothesized model functions. Finally, research from both psychology and
business is examined. While the study of the credibility effect originates in
psychology, it is important to review how these results have translated into the
more restricted business domain as well.
permission of the copyright owner. Further reproduction prohibited without permission.
Review of Prior Literature on the Credibility Effect
As discussed in Gigerenzer (1984, p. 186), the origin of the idea that
redundant information is positively associated with perceived credibility can be
traced to Locke’s An Essay Concerning Human Understanding (1700) and to
Hartley’s Observations on Man (1749). In step with this observation, modern
empirical research in psychology shows that redundant information is perceived
as being more credible than non-redundant information in a variety of contexts.
Hasher et al. (1977) is one of the first studies to suggest that repetition of
information produces a credibility effect. In this study, participants rate the
validity of each of sixty statements that pertain to various areas of knowledge
including history, government, current affairs, sports, etc. While the statements
are objective in the sense that they can be labeled as factually true or factually
false, the validity of the statements are plausibly ambiguous to the participants.
For example, one statement used in the study is: “The total population of
Greenland is about 50,000” (Hasher et al. 1977, p. 107). Participants provide
three sets of ratings spaced over two-week intervals with the second and third set
of statements comprised of both repeated and non-repeated statements. Their
results indicate that for both the factually true and factually false statements that
are repeated, there is a significant increase in the validity of judgments of the
repeated statements and no increase in validity for the non-repeated statements.
permission of the copyright owner. Further reproduction prohibited without permission.
49
This result persists for the participants that are able to discriminate between the
factually true and false statements. The authors do not explore the precise role of
frequency in cognition and memory in this study, but their results tie to studies
that show a relationship between frequency of exposure to a stimulus and positive
affect for the stimulus (Zajonc 1968; Smith and Dorfman 1975; Stang 1975).
As an alternative to the frequency explanation for the credibility effect
offered by Hasher et al. (1977), Bacon (1979) puts forth a recognition explanation
for the mediational process and compares this to several alternative theoretical
processes.1 6 Specifically, Bacon (1979) suggests that frequency per se does not
mediate credibility, but rather it is the perceived recognition of statements
brought on by the fact that the statements were repeated. This chain of reasoning
attempts to integrate the social psychology research on credibility (Asch 1940,
1948; Hovland et al. 1953) with the Hasher et al. (1977) research on repetition. In
Bacon’s first experiment, participants rate sentences for truth on two occasions
that are three weeks apart. The statements are divided evenly between true and
false statements, and one half of the statements rated on the second occasion are
repetitions of the ones presented on the first occasion. The second experiment is
1 6 That is, Bacon is investigating the process mechanism underlying the credibility effect, rather than
the mechanism underlying the redundancy effect. In this study, it is important to understand the
process behind the credibility effect in order to inform the examination of the credibility effect itself
as a potential process mechanism.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
meant to determine if the influence of recognition decisions on truth ratings is
conditional on experimenter-provided information and whether the results
generalize to contradictions of the original statements. In accordance with the
hypotheses, the results indicate that the repetition effect for validity judgments
depends on the participants’ detection of the fact that a statement is repeated.
That is, statements that are perceived as being repeated are thought to be more
valid than statements that are not perceived as being repeated, regardless of the
actual repetition of the statements. In addition, Bacon (1979) finds a credibility
effect for repeated statements even when the participants are informed that the
statements are repeated. Thus, the author’s conclusion is that the repetition effect
is not really a repetition effect per se, but rather it is a recognition effect that
extends to the situation where the repetition is transparent.
Schwartz (1982) extends previous work on repetition by addressing
questions about the conditions under which the credibility effect occurs.
Interpreted in combination with results of Hasher et al. (1977) and Bacon (1979),
results of her first experiment indicate that repetition of statements increases the
perceived credibility of the statements regardless of whether the statements are
presented weeks apart or minutes apart. Results of the second experiment
indicate that the credibility effect does not depend on the composition of the list
of statements. That is, the credibility effect persists whether the list is comprised
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
of both the repeated and non-repeated statements or only the repeated statements,
thus ruling out the possibility that the effect depends on a mixed list contrast
effect.
In a novel study, Gigerenzer (1984) evaluates the external validity of the
credibility effect. Rather than using college students, participants are selected
randomly from a telephone listing of adults living in Schwaging, Munich. To
further extend external validity, participants are tested in their homes rather than
in a controlled laboratory environment. The results indicate that both the mean
results and effect size for the credibility effect are in conformity with previous
“controlled” experimental research.
Begg et al. (1985) extends the memory-based recognition explanations in
Bacon (1979) by reasoning that if rated truth depends on the match between
tested information and remembered information, then statements with familiar
topics should appear more credible (even when the original statements do not
contain supporting facts). The intuition behind the familiarity idea is that
repetition increases the familiarity with the contents of the statement. With
Bacon’s recognition effect, a participant would consciously recognize the
repetition of the statement when evaluating its credibility; however, Begg et al.
argue that the familiarity mechanism is more subtle in that participants could
merely perceive that a statement was familiar without explicitly recognizing the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
repetition. The familiarity mechanism underlying the credibility effect is related
to the mere exposure effect where the positive affect is increasing in frequency of
encounters with an unobjectionable stimulus (Zajonc 1968; Kunst-Wilson and
Zajonc 1980).1 7 Consistent with the authors’ predictions, statements like “the
extended right arm of the Statue of Liberty is 42 feet long” are rated as more
probably true if participants had previously studied “Statue of Liberty”, or even if
the participants merely had been asked, “Do you have any idea how long the
Statue of Liberty has been in New York?”. Thus, the authors extend support for
the familiarity explanation of the credibility effect by finding evidence consistent
with the idea that the credibility effect is dependent on the relationship between
the tested statement information and the cognitive information made available by
the tested information.
Arkes et al. (1989) evaluate the generality of the relation between
familiarity1 8 and judged validity. In contrast to prior research using factual
information of uncertain truth value, Arkes et al. (1989) utilize opinion
statements and statements that are initially judged as either true or false. In this
study, participants review two 108-statement lists one week apart. The second list
1 7 However, conditions in which the mere exposure-liking relationship does not hold have also been
recognized (Saegert and Jellison 1970; Smith and Dorfman 1975; Cacioppo and Petty 1979).
1 8 Research shows that this familiarity effect is separable from the influence of recall (Jacoby and
Dallas 1981; Begg etal. 1985).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
53
is comprised of repeated statements from the first list plus non-repeated
statements. Results from the first experiment indicate that both initially perceived
true and false statements are rated as more valid than the non-repeated
statements. Results of a second experiment indicate that the credibility effect of
repetition does not occur for participants that are unfamiliar to participants. That
is, the credibility effect only arises when participants perceive themselves as
knowledgeable in a particular domain. The authors conclude that familiarity is the
mechanism underlying the credibility effect.
Begg and Armour (1991) study the influence of biasing comments that
accompany repeated statements on the credibility effect. Further the authors
propose that the effect of repetition on credibility is memory based. That is,
repeated statements “ring more true” due to the consistency between the facts
they express and the facts the participants retrieve from memory. The authors
propose that biasing comments affect the likelihood that facts are learned.
Research on memory supports this notion with examples where the prior
occurrence of an item in an experiment influences performance on tasks (Jacoby
and Dallas 1981; Tulving et al. 1982; Tulving 1983). The authors provide
examples of biasing comments that include hedges, e.g., perhaps, by and large,
technically speaking; emphases, e.g., I’ve told you a thousand times before, etc.;
and estimates of plausibility, e.g., few would believe, everybody knows, you have
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
often heard (Begg and Armour 1991, p. 197). Participants are presented with
plausible but unfamiliar facts accompanied by comments that bias participants to
believe or doubt the facts, and the authors propose that the negative biases will
reduce the probability that presented facts are learned, relative to affirming
comments. Consistent with the authors’ hypotheses, the results support the idea
of an effect on credibility for statements that are accompanied by biasing phrases.
Begg et al. (1992) study the cognitive processes that influence the ratings
of probable truth. Their contention is that source recollection and statement
familiarity are independent influences on rated truth because recollecting and
using source information requires intent whereas the feeling of familiarity that
occurs while processing occurs unintentionally (Jacoby et al. 1989; Jacoby and
Kelley 1987, 1990, 1991). This result of unintentional processing has been
labeled the “sleeper effect” and occurs if an argument from a discredited source
has a greater delayed than immediate effect on attitudes (Greenwald et al. 1986;
Pratkanis et al. 1988). That is, this effect occurs due to the fact that the link
between memory for the message and memory the source is lost over time
(Gruder et al. 1978). Thus, the familiarity of the message is influential because it
is no longer associated with the source that would discredit the statement if the
message-source link had been remembered. To investigate this idea, the authors
develop an approach where they dissociate the intentional influence of source
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
recollection from the unintentional influence of statement familiarity. In their
study, statements that vary on validity are paired with differentially credible
sources. In this setup, old true statements should be rated as true due to source
recollection and statement familiarity; however, old false statements should be
rated as false if sources are recollected which leaves the unintentional influence
of familiarity as their only basis for being rated true. The results support their
hypotheses, and the authors conclude that rated truth is influenced independently
by both unintentional familiarity and intentional memory.
As mentioned in the introductory paragraph, the first and majority of
research on the credibility effect emanates from psychology; however, since the
current study is in a business environment, it is also important to examine
research beyond this environment. The majority of academic research on the
credibility effect in business lies in the field of marketing.
In marketing, the perspective of research on the credibility effect has
centered on the idea of low involvement learning. That is, for many purchases, a
deliberate decision process never occurs because consumers are not sufficiently
motivated to think deeply about routine purchase decisions (Krugman 1965;
Olshavsky and Granbois 1979; Hoyer 1984; Sujan 1985; Alba and Marmorstein
1987; Hoyer and Brown 1990). In line with this stream of research, Hawkins and
Hoch (1992) examine how participants’ level of involvement during initial
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
exposure to consumer oriented statements influences the credibility of the
statements. The authors find that participants rate statements as more true when
they are previously exposed to the statements, i.e., consistent with the findings in
the psychology literature previously reviewed that simple repetition increases
subsequent truth ratings. In line with a familiarity explanation of the credibility
effect, the authors find that the level of processing involvement moderates the
effect of repetition. Specifically, low involvement non-evaluative processing
leads to a larger truth effect.
Following on their previous study and earlier work on repetition induced
belief in advertising claims (Law et al. 1998), Hawkins et al. (2001) investigate
the effect of repetition induced credibility effect for advertising claims that are
hierarchically related. That is, advertising claims that consist of a super-ordinate
general benefit claim and multiple subordinate feature claims. The authors find
that beliefs in feature claims increase monotonically with the number of
exposures although at a diminishing marginal rate. The authors find no evidence
of horizontal spillover, i.e., repetition induced increased in credibility from one
subordinate feature claim to another; however, they do find vertical spillover, i.e.,
repetition induced increases in credibility from individual subordinate feature
claims to the super-ordinate general benefit. The results of a dual mediation
analysis suggest that the vertical spillover effect is due to both an increase in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
familiarity of the general benefit and a greater belief in the set of subordinate
feature claims.
Also in marketing, Roggeveen and Johar (2002) test competing
explanations of familiarity and perceived source variability for the credibility
effect. The intuition behind the idea of source variability is that if one recognizes
seeing a statement multiple times but cannot recollect the source, the statements
may inadvertently be attributed to different sources. Thus, the perception that
multiple independent sources endorse a statement increases the credibility of a
statement relative to a new statement that is attributed to a single source. Overall,
the results indicate that familiarity rather than source variability is the
underlying mechanism for the credibility effect.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
TABLE 4
SUMMARY OF CREDIBILITY EFFECT STUDIES
Study Participants Task Contribution to Study o f
Credibility Effect
Hasher et al.
(1977)
40 students Participants rate the
validity o f repeated and
non-repeated
statements on three
occasions that are
spaced 2 weeks apart.
First study to find that
participants perceive
repeated statements as
greater in credibility than
non-repeated statements.
Bacon (1979) 98 students in
experiment 1
and 64
students in
experiment 2
Participants rate the
validity o f repeated and
non-repeated
statements on two
occasions that are
spaced 3 weeks apart..
Finds evidence that the
credibility effect may be
due to a recognition effect
rather than a repetition
effect suggested in Hasher
et al. (1977)
Schwartz (1982) 72 students Participants rate the
validity and familiarity
o f repeated and non-
repeated statements
minutes apart.
Shows that credibility
effect exists for short time
lag (minutes apart) and
that the credibility effect is
not the artifact o f a mixed-
list presentation.
Gigerenzer
(1984)
109 residents
o f Schwabing,
Munich.
Participants rate the
validity o f repeated and
non-repeated
statements.
Lends external validity to
the credibility effect
findings.
Begg et al.
(1985)
30-125
students across
4 experiments
Participants rate the
validity o f statements
that vary as to the level
o f factual details.
Furthers support for the
familiarity explanation.
Arkes et al.
(1989)
93 students Participants review two
108-statement lists one
week apart.
Extends the credibility
effect to opinion
statements and finds
further support for the
familiarity mechanism.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
TABLE 4 (Continued)
SUMMARY OF CREDIBILITY EFFECT STUDIES
Begg et al.
(1992)
74-195 college
students
Participants rate the
truth o f statements that
vary on validity and are
paired with
differentially credible
sources.
Findings indicate that
rated truth is influenced
independently by both
familiarity and memory.
Hawkins and
H och(1992)
96
undergraduate
and graduate
students
Participants rate the
validity o f consumer-
oriented statements
during two sessions.
Find support for a
familiarity explanation o f
the credibility effect in a
consumer oriented context.
Hawkins et al.
(2001)
50
undergraduate
students
Participants evaluated
20 sets o f 5 product
related claims: 1
general claim and 4
specific claims.
Finding o f vertical
spillover effects from the
subordinate feature claims
to the superordinate
general benefit claims.
Roggeveen and
Johar(2002)
80
undergraduate
students
Participants rate the
validity o f statements
that varied on repetition
and number o f sources.
Findings indicate that
familiarity rather than
source variability is the
underlying mechanism for
the credibility effect.
Begg and
Armour (1991)
60-145
students
Participants rate the
validity o f statements
that are accompanied
by biasing phrases.
Presents evidence that
biasing facts
accompanying statements
influence the credibility
effect. This contribution
also adds to the memory-
based explanation o f the
credibility effect.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
60
In summary, the research on the credibility effect in Table 4 demonstrates
a consistent pattern of results in both psychology and business that supports the
idea that perceived credibility mediates the effects of redundancy on judgment
and decision-making. Further, the results from this literature seem to point to
familiarity as the process explanation for the credibility effect.
Recall that the redundant information conditions are formed through the
repetition of news items in two different press articles in an identical fashion
across comparisons. That is, in the AR and SR designs different non-redundant
conditions are utilized to construct the relevant comparisons.1 9 Therefore, the
research above provides support for an inference of greater credibility of
information for the investors in the redundant information conditions relative to
the investors in both the AR and SR non-redundant conditions. However, it is
important to note that investors in the non-redundant SR condition have an
information advantage over the investors in the redundant condition. Thus,
investors in the SR non-redundant conditions may perceive their information sets
as high in credibility due to the fact that they possess multiple, non-redundant
sources of information.
Given the review above, I propose that investors in the redundant
conditions will perceive their press article sets as higher in credibility than will
1 9 Please see Chapter 3 for additional detail as to the construction of the AR and SR comparisons.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
61
investors in the non-redundant conditions. Furthermore, I conjecture that
investors will weight the press article sets in congruence with their perceived
credibility when making earnings forecasts.2 0 Therefore, I make the following
hypotheses:
H2a: Investors’ perceptions of information credibility in the redundant
conditions will be greater than those of investors in the non-
redundant conditions.
H2b: Perceived credibility of information will mediate the impact of
redundancy on investors’ earnings forecast changes.
SUMMARY
The first third of this chapter presented the factors leading to a
proliferation of redundant financial press coverage, why redundant sources of
information are problematic for judgment and decision making in general, and
why redundant financial press coverage is problematic for investors’ judgment
and decision making in particular. The second third of this chapter reviewed
research documenting the effect of redundancy on judgment and decision-making
in psychology and business, and hypotheses pertinent to the first research
question were presented. The last third of this chapter reviewed research
concerning a potential process mechanism underlying the redundancy effect and
2 0 Since I conjecture that the same differential perceptions of reliability exist across redundant and
non-redundant condition, a distinction between asymmetrical and symmetrical comparisons is not
made in the hypotheses.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
hypotheses relevant to the second research question were generated. The next
chapter discusses the experimental method employed in this study to examine the
proposed hypotheses.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
CHAPTER 3: METHOD
OVERVIEW
The method chapter of this thesis first discusses the characteristics of the
participants who completed the experimental materials. Next, the design and
procedures are presented with an emphasis on the details of the individual
treatments and the construction of the AR and SR comparisons. In addition, notes
are presented that discuss the following topics pertinent to the design and
procedures: the operationalization of the redundancy construct; the choice of the
dependent variable, earnings forecast change; and the design aspects that
facilitate interpretation of investors’ problematic judgment and decision making
as “over-weighting” of redundant information.
PARTICIPANTS
Participants were recruited via e-mail and signed up for one of two
possible experimental sessions.2 1 For completing the case materials, participants
were paid a fixed rate of $20 and enrolled in a lottery that paid two $100 prizes.
One hundred and thirteen graduate business students participated in the
experiment (43% male and 57% female), averaging 29.5 minutes to complete the
2 1 The response rate to the email solicitation was 76%.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
materials.2 2 The results of participants’ self-reported answers to investing and
financial media questions indicate the following: 60% have invested in common
stock,2 3 the average participant made 2.18 stock trades during the previous 12
months, 95% plan to invest in common stock at some time in the future, 99%
have read a business press article, the average participant read 95 articles during
the previous year, and 95% have previously analyzed the financial statements of a
company. Finally, participants reported that they had taken 11.8 accounting or
finance courses and worked full time for 1.4 years.2 4
DESIGN AND PROCEDURES
Participants began by listening to an overview of the case directions and
materials. Then the participants read specific directions and an overview of the
case that was contained in the case materials. Next, they read company
background information for the fictional, medium-sized, medical supply company
2 2 Two participants failed to follow directions and were excluded from the analysis. Thus, there were
111 usable responses.
2 3 To investigate whether the participants that invested in common stock differ from the participants
that did not invest in common stock, a common stock purchase independent variable is included in an
ANOVA of earnings forecast changes (the dependent variable). The results indicate that the common
stock purchase variable is not significant overall (F=1.883, p=0.17), in the condition (0) versus (1)
comparison (F=0.011, p=0.917), the condition (0) versus conditions (3) and (4) comparison (F=0.036,
p=0.85), and condition (2) versus conditions (3) and (4) comparison (F=0.032, p=0.86). Furthermore,
the interactions between the common stock purchase variable and redundancy are not significant.
These results are consistent for the second formulation of the dependent variable, scaled earnings
forecast change.
2 4 The amount of course work is likely overstated due to the fact that some courses in the program are
half-semester courses.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
referred to as Principal Health, Inc. The company background information
informed the participants concerning the types of products and services the
company provided (e.g., drug-delivery technologies) as well as their customers
(e.g., health care manufacturers) and a brief description of the employees (e.g.,
5,000 employees including 400 pharmacists). The provision of company
background information was deemed important to provide a context for the
presentation of the financial statement information. For example, it was thought
likely that the financial statement information would be interpreted in a different
light depending on whether the company was an Internet firm, a
telecommunications company, a retailer, etc. Therefore, the intention of the
company background information was to portray a relatively stable and neutral
company and industry that would not interact with any of the information in the
case.2 5
The company financial information consisted of three years of financial
statements. The intent behind the construction of the financial information was to
provide a mildly positive set of financial statements. The reasoning was that a set
2 5 In the context of this study, a “neutral company and industry” implies a company whose financial
performance is neither abnormally poor, e.g., on the verge of liquidation, nor abnormally good, .e.g.,
earnings growth of 300%. The phrase “neutral company and industry” also implies an industry that
does not have the potential for abnormal volatility, for example, Internet firms, bio-technology firms,
telecommunications firms, etc. The primary goal is to avoid an unanticipated interaction between
company and/or industry characteristics and the redundancy treatments.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
66
of financial statement containing negative earnings might indicate liquidity
problems or participants might even infer bankruptcy. Therefore, to prevent
inadvertent reactions by participants the financial information was generally
positive in nature. Due to the fact that a strong positive trend across all three
years may also imply undesired information to participants, the financial
statements contained positive information in all three years, but the first and third
year had larger earnings than the second (middle) year. To present as complete a
set o f financial information as possible, all three primary financial statements
were provided to the participants, i.e., the income statement, balance sheet, and
statement of cash flows. There is additional financial information that could have
been provided such as the 10-K; however, due to time constraints, a reasonable
balance between the amount o f financial information and the complexity of the
task was sought by using only the three primary financial statements.
After their review of the company background and financial statements,
the participants made a baseline annual earnings forecast for the company. That
is, participants had annual financial information in the case materials for years
2000-2003 and then they were asked to provide a baseline earnings forecast for
the year 2004. Participants then received one of five different financial press
information sets that varied as to redundancy. To increase realism, the press
articles contained in the information sets were constructed from a review of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
articles in the business press.2 6 The five press article treatments consisted of the
following: Redundant (0), Non-Redundant (1), Redundant (2), Non-Redundant
(3), and Non-Redundant (4)- see Table 5 below.
The Redundant (0) treatment, hereafter, R(0) treatment, consisted of two
redundant press articles that each contained four items of information.2 7 Each of
these items of information was represented by a paragraph of information in the
press articles. Therefore, each item consisted of several sentences pertaining to
financial events that suggested a positive expectation of future earnings. For
example, specific items pertained to the successful launch of a product line, a
successful takeover of another company, and the recent positive performance of a
business segment. In R(0), the two press articles contained the same items of
information; however, the order of the items of information was different to lend
realism to the two articles. That is, two authors writing two independent articles
may write about similar news items regarding a company; however, it is unlikely
that they would address the items in an identical order within their articles. In
summary, the R(0) treatment contained two articles each with four items of
information that were identical with respect to information content. The
2 6 See Appendix C for examples of the press article sets.
2 7 Please see Table 5 at the end of this section for an overview of the dimensions of the five
redundancy treatments.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
68
participants in R(0) viewed one of two possible sequences of articles and one of
four possible combinations of articles and information items. The sequences of
article presentation and article-information items were randomized across
participants in R(0) so that the results of the R(0) condition represented an
aggregate of the counterbalanced representations of the article materials. The
purpose of R(0) was to serve as the primary redundant information treatment for
comparative purposes with the non-redundant information treatments explained
below.
The Non-Redundant (1) treatment, hereafter, NR(1) treatment, consisted
of two non-redundant press articles that each contained four items of information.
As in R(0), these items of information contained several sentences pertaining to
recent business activities suggesting future positive earnings for the company.
However, contrary to R(0), the two press articles contained items of information
that were relatively independent with respect to each other. In summary, the
NR(1) treatment contained two articles each with four items of information that
were unique with respect to information content. The participants in NR(l)
viewed one of two possible sequences of articles and one of two possible
combinations of articles and information items. The sequences of article
presentation and article-information items were randomized across participants in
NR(1) so that the results of NR(1) condition represent an aggregate of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
counterbalanced representations of the article materials. The primary purpose of
the NR(1) is to compare the participants’ responses in this treatment to the
participants’ responses in the R(0) treatment. This treatment allows for control of
the amount o f data while it allows the amount of information to vary. That is, the
amount o f data is held constant across these two treatments (data items are equal
to eight in both treatments), while the amount of information is allowed to vary
(the amount of information items is equal to four in the R(0) and equal to eight in
NR(1).
The Redundant (2) treatment, hereafter, R(2) treatment, consisted of two
redundant press articles that each contained two items of information. In R(2), the
two press articles contained the same items of information; however, the order of
the items of information were rotated. That is, the two paragraphs of information
were presented in a different order in the two articles. In summary, the R(2)
treatment contained two articles each with two items of information that were
identical with respect to information content. The participants in R(2) viewed one
of two possible sequences of articles, one of two possible combinations of articles
and information items, and four of eight possible information items. The
sequences of article presentation, article-information items, and information items
were randomized across participants in R(2) so that the results of the R(2)
condition represent an aggregate of the counterbalanced representations of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
article materials. The purpose of R(2) is twofold. First, utilizing a second
redundant press article treatment allows replication of results from the primary
redundant treatment comparisons. Second, utilizing a redundancy treatment with
one half the information content allows me to conduct sensitivity analyses with
the additional non-redundant treatments explained below.
The Non-Redundant (3) treatment, hereafter, NR(3) treatment, consisted
of two non-redundant press articles that each contained two items of information.
Consistent with the previous treatments, these paragraphs of information
contained two to four sentences of information pertaining to the performance of
the company. In NR(3), each of the two press articles contained different items
and order of items of information. In summary, the NR(3) treatment contained
two articles each with two items of information that were unique with respect to
information content. The participants in NR(3) viewed one of two possible
sequences of articles and one of two possible combinations of articles and
information items. The sequences of article presentation and article-information
items were randomized across participants in NR(3) so that the results of NR(3)
condition represent an aggregate of the counterbalanced representations of the
article materials. The primary purpose of NR(3) was to create a non-redundant
condition to compare with R(0). In contrast to the R(0) and NR(1) comparison,
this comparison holds the amount of information content equal while the number
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
of data items is allowed to vary. That is, the number of information items in both
R(0) and NR(3) are equal to four, while the number of data items is equal to eight
in the R(0) treatment and four in the NR(3) treatment.
The Non-Redundant (4) treatment, hereafter, NR(4) treatment, consisted
of one non-redundant press article that contained four items of information. The
participants in NR(4) viewed one of two possible articles, one of two possible
authors, and four of eight possible items of information. The authors, articles and
items of information were randomized across participants in NR(4) so that the
results of NR(4) condition represent an aggregate of the counterbalanced
representations of the article materials. The primary purpose of NR(4) was to
create a non-redundant condition with the same number of items per article as the
R(0) treatment. Unlike the R(0) versus NR(3) comparison, however, this
comparison holds the number of items per article constant while the number of
articles is different from that of the R(0) condition.
The counterbalancing of both press articles and information content across
treatments allowed for control of the participants’ perceived accuracy and
strength of information items since the same information appeared in both the
redundant and non-redundant information conditions.2 8
2 8 I limit the experimental materials to two press articles due to time constraints and also due to the
fact that it is not reasonable for decision makers to actively process all financial communications
(Greenwald and Leavitt 1984; Payne et al. 1993; Hawkins et al. 2001). Also, decision makers may
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
TABLE 5
DIMENSIONS OF REDUNDANCY TREATMENTS
Redundant Non- Redundant Non- Non-
Redundant Redundant Redundant
Information Set:
Number o f articles
(0)
2
(1)
2
(2)
2
(3)
2
(4)
1
Number of data
items per article 4 4 2 2 4
Number of data
items total 8 8 4 4 4
Number of
information items
total
4 8 2 4 4
Number of articles: The number of financial press articles contained in the treatment.
Number of data items per article: An “item” of information can be defined narrowly or
broadly. Here, I refer to an item of information broadly, for example, the launch of a
product line, a takeover of another company, or the performance of a business segment.
Number of data items total: The number of data items in the treatment
Number of information items total: The number of independent data items in the treatment.
The Operationalization of the Redundancy Construct
Throughout this study, reference is made to the experimental treatments as
either “redundant” or “non-redundant” for expositional ease; however, as the
literature review above indicates, exact representations of redundancy and non
redundancy are not achieved with the current experimental materials. For
become tired or aggravated with the volume of information content which could diminish the
persuasiveness (Belch 1982; Cacioppo and Petty 1979) and cause floor effects.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
example, although the redundant press article treatments R(0) and R(2) contain
the same financial information pertaining to the future performance of the firm,
the articles within these sets differ on several dimensions. These dimensions
include the following: the sequence of the articles, i.e., one article necessarily
precedes the other article in the materials; authors; name and format of journals;
arrangement of paragraphs; and sentence construction within the paragraphs. The
claim cannot be made that these articles are 100% redundant. Therefore, two
important questions need to be addressed: (1) Why do these differences exist, and
(2) Could these differences bias the findings?
The different sequence of the articles is controlled for through a
counterbalancing of the presentation o f the articles. Therefore, when the
aggregate mean earnings forecast change is evaluated, it will contain both types
of sequences. The choice of different authors, journals, paragraph arrangement,
and wording is made due to the artificiality of having two duplicate articles
juxtaposed in the case materials. That is, to construct a 100% redundant treatment
would necessitate that the exact article appears twice in the case materials. This
creates the specific problem that participants may simply ignore the second
article. In summary, superficial differences are integrated within the press article
treatments to limit the artificiality of the experimental materials.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
The second question raised above addresses whether these superficial
differences that lower the level of redundancy in the redundant press article sets
actually bias the findings. It is the contention of this study that rather than bias the
findings, it is likely that the differences strengthen the findings. Take for
example, a hypothetical experiment where redundancy is conceptualized on a
continuous scale from 0% redundancy to 100% redundancy, and the endpoints of
this scale are proxied exactly, i.e., a manipulation with 0% and 100% redundant
materials. If an effect for redundancy is shown within this context, then an effect
has been shown for only the strong form conditions of redundancy. These forms
of redundancy are essentially outlier conditions in the external world of financial
information. It is argued here that a much stronger case for a redundancy effect
can be made when the comparisons are made closer to the midpoint of the scale.
For example, if the current case materials are evaluated as, say, 80% redundant
for the redundant condition and 20% redundant for the non-redundant condition
and an effect is shown, there may be greater confidence in extrapolating these
results to a comparison of 100% and 0% redundancy, than in interpolating from
100% and 0% redundancy.
Dependent Variable- Earnings Forecast Changes
Nonprofessional investors engage in a variety of judgment and decisions.
In this study, I measure the effects of redundancy through the difference between
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
a baseline earnings forecast that investors make after reading company
background and financial statement information and a second earnings forecast
that investors make after evaluating either a redundant or non-redundant set of
press articles. Thus, earnings forecast change and a scaled version of the earnings
forecast change are the primary dependent variables. Other choices for dependent
variables would include dichotomous variables representing the choice to
purchase or sell stock, and the continuous variables of the number of shares to
purchase or sell. It may be argued that the act of purchasing or selling shares of
stock is a more common one for nonprofessional investors than the act of
forecasting earnings.
Given the nature of this experiment, the selected dependent variable
should meet two conditions. First, for external validity purposes, at least some
nonprofessional investors must engage in this activity currently or in the future.
Second, for internal validity purposes, the dependent variable must produce
consistent results conditional on the given experimental treatment. To further
explain, the decision to purchase or sell stock does often is not made consistently
across investors due to variation in investor strategy. That is, investors typically
employ a variety of investment strategies that can yield different actions given the
same set of information. For example, some investors employ a contrarian
strategy. Therefore, if they receive negative news regarding an equity security
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
they may purchase that security rather than sell the security. Conversely, given
the same information set, a momentum investor may decide to sell the security
rather than purchase the security. Therefore, when strategies are heterogeneous,
one cannot attribute variation in the dependent variable due to the intended
manipulation because it may be due to the variation in strategies employed by
investors.2 9 Therefore, earnings forecasts are employed as the dependent variable
because investors, on average, do formally or informally engage in these
activities, albeit to a lesser extent than purchase or sale activities, and the
dependent variable results are less noisy. That is, when investors receive positive
information concerning the future performance of the company this generally
translates into a prediction of a positive increase in future earnings for the
company.
A Note on the Term “Over-weighting”
This study does not employ an objective accuracy criterion for the
dependent variable. If an objective accuracy criterion were employed, e.g., an
actual company’s earnings, the problems would be numerous. For example, the
information provided to the participants in the study is by necessity a subset of
2 9 The implications of variation of investing strategies are twofold for evaluating the experimental
results. First, it may be the case that a strategy interacts with the redundancy treatments, creating an
unintended interaction effect. Second, if there are no interactions and the investing strategies offset,
i.e., there are an equal number of contrarian and momentum investors in the cells, then there is
potential for a great deal of noise in the results.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
77
the actual information that would be required to form an expectation of the future
earnings of the company. Therefore, in the case of an actual company, the
significant variables for earnings prediction may be external to the information
provided in the case materials (e.g., Davis et al. 1994). This could result in non-
systematic responses from participants. In this scenario, results would be difficult
to interpret due to the inherent noise in across participants’ decisions. One
potential remedy for this is to use a theoretically-based subset of information
coupled with numerous participant forecasts. This strategy may result in more
consistent results in the aggregate; however, numerous forecasts require either a
corresponding increase in time required for the experiment or a corresponding
decrease in the size of the subset of financial information provided to the
participants. In the latter case, the experimenter may return to the first scenario
where the subset of information provided to the participant is too small relative to
the information required and the resulting dependent variables are
correspondingly noisy.
In this study, the choice is made to avoid the problems associated with an
objective dependent variable and to interpret the results through comparisons
between the redundant and non-redundant conditions. Although nothing can be
determined from the earnings forecast themselves, generally it can be determined
that if information items are equally positive with regard to forming expectations
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
concerning future earnings (an assumption that I guarantee in the aggregate due
to counterbalancing of information items across conditions), more positive
information items should be preferred to fewer positive information items. For
example, in the case of a comparison between R(0), which has four independent
items of positive information, and NR(1), which has eight independent items of
positive information, NR(1) should be preferred to R(0), ceteris paribus.
Investigation of Hypotheses through Treatment Comparisons
To investigate hypothesis Flla, i.e., whether investors’ earnings forecast
changes in the redundant conditions are greater than those of investors in the AR
non-redundant conditions, I compare the earnings forecast changes of investors in
the R(0) condition to those o f investors in the NR(3) and (4) conditions. In the
AR comparison, I utilize two non-redundant conditions to control for both the
number of press articles and data items per press article. To examine hypothesis
H lb, i.e., whether investors’ earnings forecast changes in redundant condition are
not significantly different than those of investors in the SR non-redundant
condition, I compare the earnings forecast changes of investors in the R(0)
condition to those of investors in the NR(1) condition.3 0 In addition, the SR
comparison analysis is replicated by comparing the R(2) condition to NR
conditions (3) and (4). Finally, to investigate hypothesis H lc, i.e., whether
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
investors’ earnings forecast changes in the non-redundant comparison condition
are larger than those of investors in the non-redundant information control
conditions, I compare the earnings forecast changes of investors in the NR(1)
condition to those of investors in NR(3) and (4) conditions.
To collect information to verily the redundancy manipulations, measure
the proposed process explanation, and rule out alternative process explanations,
participants completed post-experimental questions. Please see Appendix B for
an overview of the design.
SUMMARY
The method chapter of this thesis first discussed the characteristics of the
participants who completed the experimental materials. Then, the design and
procedures were presented with an emphasis on the details of the redundant and
non-redundant treatments and the construction of the AR and SR comparisons.
Also, notes were presented that discussed the operationalization of the
redundancy construct, the choice of dependent variable, and use of the term
“over-weighting”. The next chapter presents the results.
3 0 The symmetrical comparison analysis is replicated by comparing redundant condition (2) to non-
redundant conditions (3) and (4).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4: DATA ANALYSIS AND RESULTS
OVERVIEW
The data analysis and results chapter of this thesis first discusses the
manipulation and other checks designed to ensure proper manipulation of
redundancy and perception of the experimental materials. Then, a discussion of
the analysis and results for the planned comparisons evaluating the set of three
hypotheses that comprise the first hypothesis is presented. This first set of results
pertain to documenting an effect for redundancy on investors’ judgment and
decision-making. Finally, a discussion of the analyses and results for both
regression-based and structural equation modeling approaches for evaluating the
mediational hypotheses is presented. These mediational results are necessary to
establish perceived credibility as the process mechanism underlying the effect for
redundancy on investors’ judgment and decision-making.
MANIPULATION AND OTHER CHECKS
Numerous analyses are conducted to ensure proper manipulation of
redundancy and perception of the experimental materials. Unless otherwise
noted, all post-experimental questions referred to in this section utilize an 11-
point scale with the endpoints labeled Strongly Disagree (-5) and Strongly Agree
(+5) and the midpoint labeled Neither Agree nor Disagree (0). To verify the
redundancy manipulation, participants were asked two questions aimed at
permission of the copyright owner. Further reproduction prohibited without permission.
81
measuring their perception of the level of redundancy in the press article
information set. Specifically, the questions asked if the information contained in
the second press article was “redundant” or “very similar” with the information
contained in the first press article. The participants’ pattern of responses indicate
that the questions measured the same latent construct (Cronbach’s alpha =.84);
therefore, the two sets of responses are averaged into a single measure of
redundancy.3 1 The results indicate that participants in the redundant conditions
view their press article sets as significantly more redundant than the participants
in the non-redundant conditions view their sets.3 2
Related to the participants’ perceptions of redundancy, I investigate the
amount of independent information the participants perceive in the sets of press
articles. The question informs participants that what constitutes an item of
information can be defined quite narrowly or more broadly, provides examples of
both narrow and broad types of information, and states that the broader definition
3 1 The alpha coefficient ranges in value from 0 to 1 with higher scores implying greater internal
consistency. Nunnally (1978) has indicated 0.70 to be an acceptable reliability coefficient but lower
thresholds sometimes are used in the literature. I use confirmatory factor analysis to corroborate the
Cronbach’s alpha results for all latent constructs in this study.
3 2 The R(0) condition was judged significantly more redundant than the NR(1) condition (p=.007) and
NR(3) condition (p=.024), and the second redundant condition, R(2), was judged significantly more
redundant that the NR(1) condition (p=.001) and the NR(3) condition (p=.003). There was not a
significant difference between the two redundant conditions (p= 0.999). The non-redundant condition
(4) was excluded from analysis because it contains only a single article. Due to a significant Levene’s
statistic (p=.015), 1 used a multiple comparison test that does not assume equal variances, Tamhane’s
T2. The results do not change if I use Dunnet’s T3, Games-Howell, or Dunnetf s C.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
is assumed for the purposes of the question. If participants are able to perceive
the relative level of redundancy in the materials, then they should be able to
distinguish the approximate number of independent information items in the
materials as well. In addition, this form of the redundancy manipulation check
has the advantage of including the single press article treatment, non-redundant
condition (4). These judgments can be compared on a relative basis to the actual
“number of information items total” line in Table 5. The mean amount of
perceived information in the redundancy conditions was as follows: R(0) M=4.5,
NR(1) M=7.1, R(2) M=2.6, NR(3) M=3.9, and NR(4) M=4.8. For the participants
to have properly perceived the information content in their conditions, the
following three relations should be observed: (a) the information content of
NR(1) should be greater than conditions R(0), R(2), NR(3), and NR(4); (b) there
should be no significant difference between conditions R(0), NR(3), or NR(4);
and (c) the information content of R(2) should be less than conditions R(0),
NR(3), and NR(4). A series of three contrasts without the assumption of equal
variances was analyzed for the aforementioned comparisons with the following
results: (a) t=4.13, p=.000, (b) t=.30, p=.766, (c) t=4.78, p=000.3 3 Therefore,
3 3 Contrast analysis is employed in the Manipulation and Other Checks, Effects of Redundancy on
Nonprofessional Investors’ Earnings Forecasts, and Perceived Credibility as a Process Explanation
for the Effects of Redundancy sections of Chapter 4. Contrasts are planned comparisons of two or
more groups. The hypothesized results of the comparisons are expressed as lambda weights and can
take any value as long as the total lambda weights sum to zero. Since contrasts are based on a single
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
the results of the manipulation check confirm that the participants did in fact
properly perceive the information content for their treatments.
To investigate if participants understand the effects of redundancy in
general, responses to a set of three questions measuring the conceptual
understanding of redundancy are evaluated.3 4 As an example, the first question
informs participants that a friend has purchased a home pregnancy test and
received a positive result. The question then asks the participants to help the
friend choose between purchasing the same brand again or, alternatively,
purchasing an equally accurate different brand (that uses a slightly different
process to determine pregnancy). The second and third questions are similar in
nature to the first question but use the alternative contexts of choosing an
investment advisor and researching the purchase of a digital camera. The results
indicate that participants answer appropriately, i.e., recognize the impact of
positive error correlation, for the pregnancy test (M=2.61, t=l 1.67, p=.000),
degree of freedom, the sum of squares is identical with the mean square and can be divided by the
mean square error to yield and Ftest for the contrast. Since the F tests employed to test contrasts have
a single degree of freedom in the numerator, the t test can be easily calculated by taking the square
root of the F test. The following coding is used for the planned comparisons: (a) lambda weights of
(+1, -.250, -.250, -.250, -.250) to test if the information content of NR(1) is greater than conditions
R(0), R(2), NR(3), and NR(4)); (b) lambda weights of (+1, -.50, -.50) to test if there is a non
significant difference between conditions R(0) and NR(3) or NR(4); and (c) lambda weights of (+1, -
.333, -.333, -.333) to test if the information content of R(2) is less than conditions R(0), NR(3), and
NR(4).
3 4 The alpha coefficient does not provide support for aggregation of these questions; therefore, the
questions are analyzed separately.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
investment advisor (M=2.16, t=9.79, p=.000), and digital camera scenarios
(M=2.17, t=10.34, p=.000). The company background information and the
financial statements are intended to be relatively neutral. To verify this,
participants are asked two questions concerning the company background and
financial statement information valence. Specifically, these questions ask for
investors’ perceptions concerning the degree of positive or negative information
in the company background and financial statement information. The
participants’ pattern of responses indicate that the questions measure the same
construct (Cronbach’s alpha =.86); therefore, the responses to the two questions
are averaged into a single measure. Results indicate that participants perceive the
company background information and financial statements to be significantly
positive (M=1.94, t=13.4, p=.000).
The financial press article sets are intended to be positive in nature and
realistic. To verify the valence and realism of the press articles, participants are
asked two questions about perceived valence and three questions about perceived
realism of the press articles. The questions regarding the valence of the press
articles ask participants the degree of negative or positive information contained
in the press articles. The questions regarding realism ask participants to compare
the articles in the information sets to typical articles that one might find in the
financial press. The participants’ pattern of responses indicates that the questions
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
\
appropriately measure the latent constructs, valence (Cronbach’s alpha =.68) and
realism (Cronbach’s alpha= .68); therefore, the two questions for each construct
are averaged into single measures. Results indicate that the press articles are
perceived to be both positive (M = 3.37, t = 30.5, p=.000) and realistic (M=1.26,
t=8.2, p= 000).
Additional checks concerning the effect of the experimental session and
participants’ gender on the dependent variable were conducted. Each variable
was included in an ANOVA for the change in earnings forecasts variable. The
results indicated that there were no significant effects caused by either the
experimental session or gender variable.3 5
EFFECTS OF REDUNDANCY ON NONPROFESSIONAL INVESTORS’
EARNINGS FORECASTS
The first hypothesis, HI a, predicts that investors’ earnings forecast changes
in the redundant condition will be greater than those of investors in the AR non-
redundant information sets. To investigate this hypothesis, I compare the
dependent variable, earnings forecast change, in the R(0) condition, to earnings
forecast changes in the NR(3) and (4) conditions.3 6 The results indicate that the
3 5 These comparisons are made separately for conditions R(0) and NR(1), conditions R(0) and NR(3),
and conditions R(0) and NR(4). In the R(0) and NR(1) comparison, the session results are F=.073
(p=,832) and gender results are F=. 191 (p=.738). In the R(0) and NR(3) comparison, the session
results are F=.039 (p=.876) and gender results are F=4.028 (p=.294). In the R(0) and NR(4)
comparison, the session results are F=.071 (p=.834) and the gender results are F=5.026 (p=.267).
3 6 The lambda weights for the planned comparison are (+1, -.50, -.50).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
average earnings forecast for participants assigned to the R(0) condition is
significantly greater than that of participants assigned to NR(3) and (4) conditions
(t=4.70, p=.000, see Table 6). In this AR comparison, both the level of
redundancy and the amount of data vary. Therefore, two possible explanations for
the results are that the participants’ reactions are a function of the level of
redundancy or that the participants’ reactions are a function of the quantity of
data.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
TABLE 6
REDUNDANCY COMPARISONS
PANEL A: Earnings Forecast Changes- Means (Standard Deviations) and
Sample Sizes by Condition
Redundant
(0)
Non-Redundant
(1)
Redundant
(2)
Non-Redundant
(3)
Non-
Redundant
(4)
Overal 1
94.5 98.5 59.9 36.7 39.0 66.3
(52.8) (45.4) (37.4) (52.8) (43.4) (53.2)
23 23 21 22 22 111
PANEL B: Tests of Hypothesis 1
Comparison Hypotheses Contrast d f
/-statistic
P-
vaiue
Asymmetrical H la
Mean earnings forecast change
in the R(0) condition will be
greater than in the NR(3) and
(4) conditions.
106 4.70 .000
Symmetrical Hlb
Mean earnings forecast change
in the R(0) condition will not
be significantly different than
the NR(1) condition.
106 -0.30 .773
Symmetrical Hlb
Mean earnings forecast change
in the R (2) condition will not
be significantly different than
the NR (3) and (4) conditions.
106 1.77 .080
Information
Control
Hie
Mean earnings forecast change
intheNR(l) condition will be
greater than the NR(3) and (4)
conditions.
106 5.03 .000
This table shows the descriptive statistics and results of hypothesis tests to determine if redundant
financial press coverage significantly influences nonprofessional investors’ judgments. Panel A shows
the means, standard deviations, and sample sizes by experimental condition. Panel B shows the
planned comparisons for Hypotheses la, lb, and Ic.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
88
To address the alternative explanation, the variation in quantity o f data is
controlled for by examining the second hypothesis, H lb. This hypothesis predicts
that investors’ earnings forecast changes in the redundant condition will not be
significantly different than those of investors in the SR non-redundant condition.
To investigate this hypothesis, the R(0) condition is compared to the NR(1)
condition and, to replicate the first comparison, the R(2) condition is compared to
NR (3) and (4) conditions.3 7 The results of the first comparison show that there is
no significant difference between earnings forecast changes in the R(0) condition
and the NR(1) condition (t=-.30, p=.773, see Table 6). The results of the second
comparison show a marginally significant difference when comparing the R(2)
condition to NR(3) and (4) conditions (t=l .77, p=.080, see Table 6); however,
this difference is in a direction consistent with the hypothesis. That is, the
hypothesis predicts that participants will react to both redundancy and quantity of
information and that these preferences will offset. Thus, the results indicate a
marginal preference for redundancy over information content.
Because the SR comparison non-redundant conditions hold an informational
advantage over their respective redundant conditions, two possible explanations
for non-significant differences are that the participants’ judgments are a function
3 7 The lambda weights for the first planned comparison are (+1, -1) and for the second planned
comparison are (+1, -.50, -.50).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
89
of both the level of redundancy and amount of information, which offset each
other, or that the participants’ judgments are neither a function of the level of
redundancy nor information.
To rule out the alternative explanation, I examine the information control
hypothesis, H ie, which posits that investors’ earnings forecast changes in the
NR(1) condition will be greater than those of investors in NR(3) and (4)
conditions.3 8 The purpose of this comparison is to verify that participants are
reacting to the level of information content in their respective experimental
conditions. The comparison conditions vary as to information content with the
NR condition (1) possessing approximately twice the information as the NR(3)
and (4) conditions. If participants’ judgments are a function of information
content, the results should indicate that earnings forecast changes in the NR
condition (1) are significantly greater than those of the NR(3) and (4) conditions.
The contrast results support this hypothesis (t=5.03, p=.000, see Table 6).
An alternative explanation for the results thus far is that the investors who
forecast larger initial earnings forecast may in turn make larger earnings forecast
changes. That is, there is a danger that larger forecasters may be either driving the
results or biasing the results. To investigate this potential problem, the earnings
3 8 The lambda weights for the planned comparison are (+1, -.50, -.50).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
90
forecast changes are scaled by the investors’ initial forecasts. Below, Table 7
presents the scaled results.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
91
TABLE 7
REDUNDANCY COMPARISONS SCALED BY INITIAL FORECAST
PANEL A: Scaled Earnings Forecast Changes- Means (Standard
Deviations) and Sample Sizes by Condition
Redundant
(0)
Non-Redundant
(1)
Redundant
(2)
Non-Redundant
(3)
Non-
Redundant
(4)
Overall
0.36 0.34 0.21 0.13 0.14 0.24
(0.31)
(.22) (0.24) (0.18) (0.18) (0.25)
23 23 21 22 22 111
PANEL B: Tests o f Hypothesis 1
Comparison Hypotheses Contrast df t-
statistic
P-
value
Asymmetrical Hla
Mean earnings forecast
change in the R(0)
condition will be greater
than in the NR(3) and (4)
conditions.
106 3.74 .000
Symmetrical Hlb
Mean earnings forecast
change in the R(0)
condition will not be
significantly different
than the NR(1)
condition.
106 0.26 .797
Symmetrical Hlb
Mean earnings forecast
change in the R (2)
condition will not be
significantly different
than the NR (3) and (4)
conditions.
106 1.26 .209
Information Control Hlc
Mean earnings forecast
change in the NR(1)
condition will be greater
than the NR(3) and (4)
conditions.
106 3.74 .000
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
The analysis employing the scaled dependent variable indicates the
following. In the first hypothesis, HI a, the R(0) condition is compared to the
NR(3) and (4) conditions.3 9 The results are consistent with the raw results
presented in Table 6 and indicate that the average scaled earnings forecast for
participants assigned to R(0) is significantly greater than that of participants
assigned to NR (3) and (4) conditions (t=3.74, p=.000, see Table 7). I address the
scaled version o f the second hypothesis, H lb, by comparing the R(0) condition to
the NR(1) condition and, to replicate the first comparison, the second redundant
condition R(2) is compared to non-redundant conditions NR(3) and (4).4 0
Consistent with the raw results, the results of the first comparison show that there
is no significant difference between the scaled earnings forecast changes in the
R(0) condition and the NR(1) condition (t=0.26, p=.797, see Table 7). The
results of the second comparison show no significant difference when comparing
the second redundant condition, R(2) to non-redundant conditions, NR(3) and (4)
(t=l .26, p=.209).4 1 The hypothesis predicts that participants will react to both
redundancy and quantity of information and that these preferences will offset. In
3 9 The lambda weights for the planned comparison are (+1, -.50, -.50).
4 0 The lambda weights for the first planned comparison are (+1, -1) and for the second planned
comparison are (+1, -.50, -.50).
4 1 The lambda weights for the planned comparison are (+1, -.50, -.50).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
93
contrast to the raw dependent variable results, the scaled dependent variable
results indicate no preference for redundancy over information content. Finally, if
participants’ judgments are a function of information content, the results should
indicate that earnings forecast changes in the NR(1) condition are significantly
greater than those of NR(3) and (4) conditions.4 2 The contrast results support this
hypothesis (t=3.74, p=.000).4 3
In summary, the results of the planned comparisons employing both raw
and scaled dependent variables of earnings forecast changes indicate that
investors’ earnings forecast changes in the redundant condition are larger than
those of investors in the AR non-redundant condition. In addition, investors’
earnings forecast changes in the redundant conditions are not significantly
different than those of investors in the SR non-redundant conditions. Finally,
investors’ earnings forecast changes in the SR non-redundant condition are
larger than those of investors in the non-redundant information control
4 2 The lambda weights for the planned comparison are (+1, -.50, -.50).
4 3 An alternative strategy to incorporate the idea of relative earnings forecast changes utilizes an
analysis with the second earnings forecast as the dependent variable and the first earnings forecast as
a covariate. The results consist of the following: in the asymmetrical comparison, HI a, the mean
earnings forecast change in R(0) is greater than those of the NR(3) and NR(4) conditions (t = 4.02, p =
.000); in the symmetrical comparison, Hlb, the mean earnings forecast change in R(0) is not
significantly different from that of the NR( 1) condition (t = .29, p=.769); in the second symmetrical
comparison, Hlb (replication), the mean earnings forecast change in R(2) is marginally greater than
those of the NR(3) and NR(4) conditions (t = 1.845, p = .068); in the information control comparison,
Hlc, the mean earnings forecast change in NR(1) is greater than those of the NR(3) and NR(4)
conditions (t = 4.31 ,p = .000). In sum, these results are consistent with both the raw and scaled results.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
94
conditions. Taken together, these results provide support for the hypothesized
effect of redundancy in press articles on nonprofessional investors’ judgments.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
PERCEIVED CREDIBILITY AS A PROCESS EXPLANATION FOR THE
EFFECTS OF REDUNDANCY
Data Analysis and Results: Hypothesis 2a
Hypothesis 2a predicts that investors’ perceptions of credibility in the
redundant conditions will be greater than those of investors’ in the non-redundant
conditions. To test this hypothesis, the responses to four post-experimental
questions aimed at measuring credibility are analyzed. Specifically, these
questions have participants rate the information concerning Principal Health, Inc.
contained in the press articles for credibility, reliability, dependability, and
consistency. The participants’ pattern of responses indicates that the questions
measure the same latent construct (Cronbach’s alpha =.87); therefore, the four
sets of responses are averaged into a single measure. Next, a series of planned
comparisons are employed to examine if investors’ perceived credibility of the
information set in the R(0) condition is significantly greater than that of investors
in the NR(3) and (4) conditions; if perceived credibility of the information set in
the R(0) condition is greater than that of the NR(1) condition; and if perceived
credibility in the R(2) condition is greater than that of the NR(3) and (4)
conditions. 4 4 The results from the first (t=-3.87, p=.000, see Table 8, Panel B),
second (t--2.43, p=.017, see Table 8, Panel B), and third planned comparisons
4 4 The lambda weights for these planned comparison are (+1, -.50, -.50), (+1, -1), and (+1, -.50, -.50),
respectively.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
(t=2.58, p=.011, see Table 8, Panel B) support the hypothesis that investors in the
redundant conditions perceive their information as more credible than do the
investors in the non-redundant conditions.
Data Analysis and Results: Hypothesis 2b
Hypothesis 2b predicts that perceived credibility of the redundant press
article sets will mediate the effect of redundancy on investors’ earnings forecast
changes. Baron and Kenny (1986) and Judd and Kenny (1981) suggest that the
following findings establish mediation: the independent variable is significantly
associated with the dependent variable; the independent variable is significantly
associated with the mediator; in the presence of the independent variable, the
mediator is significantly associated with the dependent variable; and in the
presence of the mediator, the effect of the independent variable on the dependent
variable is zero. While satisfaction of the last condition would demonstrate
complete mediation, most mediational analyses demonstrate partial mediation by
showing a reduction in the significance of the independent variable rather than a
complete elimination of significance. To satisfy the conditions proposed by
Baron and Kenny (1986) and Judd and Kenny (1981), both regression-based and
structural equation modeling approaches are used to evaluate the hypothesized
process explanation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
Mediational Analysis: Regression-based and Structural Equation Modeling
Approaches
If the mediational model is comprised of only measured variables the
analysis approach suggested is generally regression-based, and if the model is
comprised of latent variables the analysis approach suggested is normally
structural equation modeling (Kenny 2003). The model in the current study is
comprised of both measured variables and latent variables; therefore, both
regression-based and SEM approaches are employed. The convergence of results
from these two approaches, which make different assumptions regarding the data,
strengthens support for the hypothesized mediational model.
Regression-Based Approach to Mediational Analysis
Introduction to the Regression-Based Approach
While the mediational conditions suggested by Baron and Kenny (1986)
and Judd and Kenny (1981) can be useful to informally judge whether mediation
is occurring, a number of statistical tests have been developed to more formally
capture compliance with these conditions (MacKinnon and Dwyer 1993;
MacKinnon et al. 1995). Following these formalized tests, the Sobel test and
relevant variations on the Sobel test statistic are utilized as a first step to
investigate mediation in this study.4 5 The first set of tests examines the R(0)
4 5 Please see Appendix A for test details. Further details can be found in Baron and Kenny (1986),
Sobel (1982), Goodman (1960), and MacKinnon et al. (1995).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
condition and the NR(3) and (4) conditions. I then evaluate the R(0) condition
and the NR(1) condition. Finally, to replicate the SR comparison, I examine the
R(2) condition and NR(3) and (4) conditions.
Results from the Regression-Based Approach
The first set of results for the AR comparison indicates that credibility
significantly mediates the relation between redundancy and the change in
earnings forecast for the R(0) condition and NR(3) (z=2.18, p=.03), and (4)
conditions (z=2.50, p=.01, see Table 8, Panel C).4 6 The second set of results
pertaining to the SR comparison indicates that credibility significantly mediates
the relation between redundancy and the change in earnings forecast for the R(0)
condition and the NR(1) condition (z= 2.14, p=.03, see Table 8, Panel C). The
last set of results indicates marginally significant support for the mediational
hypothesis between the R(2) condition and NR(3) condition (z=1.71, p=.08, see
Table 8, Panel C) and significant support for R(2) condition and NR(4) condition
(z=2.01, p=.04, see Table 8, Panel C).
4 6 The reported z-value are calculated after Sobel (1982). The reported p-value is drawn from the unit
normal distribution under the assumption of a two-tailed Z-test of the hypothesis that the mediated
effect equals zero in the population. Results for alternative variations of the Sobel test, the Goodman I
and Goodman II tests, are reported in Table 8, Panel C.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
TABLE 8
MEDIATIONAL ANALYSIS OF CREDIBILITY
PANEL A: Perceived Credibility of Financial Press Sets- Means (Standard
Deviations) and Sample Sizes by Condition
Redundant
(0)
Non-Redundant
(1)
Redundant
(2)
Non-Redundant
(3)
Non-Redundant
(4)
Overall
2.91 1.95 2.49 1.61 1.53 2.10
(1.35) (1.40) (1.22) (1.20) (1.53) (1.43)
23 23 21 22 22 111
PANEL B: Tests of Hypothesis 2a
Comparison Contrast
df ^-statistic
p-value
(2-tailed)
Asymmetrical
Mean perceived credibility in the R(0)
condition will be greater than in the NR
(3) and (4) conditions.
106 3.87 .000
Symmetrical
Mean perceived credibility in the R(0)
condition will be greater than in the NR
(1) condition.
106 2.43 .017
Symmetrical
Mean perceived credibility in the R(2)
condition will be greater than the NR(3)
and (4) conditions. 106 2.58 .011
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
TABLE 8 (Continued)
MEDIATIONAL ANALYSIS OF CREDIBILITY
PANEL C: Tests of Hypothesis 2b
Comparison Conditions
4 h >
Sobel
z-value
(/j-value)
2-tailed
Goodman 1
z-value
(p-value)
2-tailed
Goodman 1 1
z-value
(p-value)
2-tailed
Asymmetrical (0) (3)
-3.41 2.84 2.18 2.13 2.24
(-03) (-03) (-03)
(4)
-3.22 3.98 2.50 2.46 2.55
(.01) (.01) (.01)
Symmetrical (0) (D
-2.38 4.91 2.14 2.10 2.18
(.03) (.04) (.03)
Symmetrical (2) (3)
-2.38 2.66 1.77 1.71 1.85
(.08) (.09) (.06)
(4) -2.28 4.26 2.01 1.97 2.05
(.04) (.05) (-04)
This table shows the descriptive statistics and results of hypothesis tests to determine if investors’
perceive that redundancy is positively associated with information credibility and if perceived
credibility mediates nonprofessional investors’ earnings forecasts changes. Panel A shows the means,
standard deviations, and sample sizes of perceived credibility of financial press article sets by
experimental condition. Perceived credibility is measured by asking participants their level of
agreement with questions pertaining to the press articles’ level of credibility, reliability, dependability,
and consistency. Their responses are measured on an 1 1 -point scale with the endpoints labeled
Strongly Disagree (-5) and Strongly Agree (+5) and the midpoint labeled Neither Agree nor Disagree
(0). Panel B shows the planned contrasts for Hypothesis 2a. Panel C displays the results of the
mediational analysis consisting of the following: ta , which is the t-statistic of the regression coefficient
for the association between independent variable and the mediator in each condition; tb , which is the t-
statistic of the coefficient for the association between the mediator and the dependent variable in the
presence of the independent variable; Sobel test z-value; Goodman (1) test z-value; and the Goodman
(II) test z-value. The p-values are drawn from the unit normal distribution under the assumption of a
two-tailed Z-test of the hypothesis that the mediated effect equals zero in the population. All
regression models were checked for outliers and homoscedasticity, independence, and normality of
residuals.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
101
Again, an alternative explanation for the mediational results above is that
the investors who forecast larger initial earnings forecast may in turn make larger
earnings forecast changes. To investigate this potential explanation, I scale the
earnings forecast changes by the initial forecasts. The scaled tables are presented
in Table 9.
TABLE 9
MEDIATIONAL ANALYSIS OF CREDIBILITY SCALED BY INITIAL
FORECAST
Scaled Tests of Hypothesis 2b
Comparison Conditions
^ a 4
Sobel
z-value
(/?-value)
2-tailed
Goodman I
z-value
(p-value)
2-tailed
Goodman I I
z-value
(p-value)
2-tailed
Asymmetrical (0) (3)
-3.41 2.72 2.13 2.07 2.18
(.03) (.04) (.03)
(4)
-3.22 3.51 2.37 2.32 2.42
(.02) (.02) (.02)
Symmetrical (0) (1)
-2.38 4.28 2.14 2.04 2.12
(.03) (.04) (.03)
Symmetrical
(2) (3)
-2.38 1.80 1.77 1.36 1.52
(.08) (.17) (.13)
(4)
-2.28 2.87 2.01 1.72 1.86
(.04) (.09) (.06)
This table displays the results of the mediational analysis consisting of the following: t„ which is the
t-statistic of the regression coefficient for the association between independent variable and the
mediator in each condition; tb , which is the t-statistic of the coefficient for the association between the
mediator and the scaled dependent variable in the presence of the independent variable; Sobel test z-
value; Goodman (I) test z-value; and the Goodman (II) test z-value. The p-values are drawn from the
unit normal distribution under the assumption of a two-tailed Z-test of the hypothesis that the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
The first set of scaled tests examines the R(0) condition and the NR(3) and
(4) conditions. Then the R(0) condition and the NR(1) conditions are evaluated.
Finally, to replicate the SR comparison, the R(2) condition and the NR(3) and (4)
conditions are examined.
The first set of scaled results for the AR comparison indicates that
credibility significantly mediates the relation between redundancy and the change
in earnings forecast for the R(0) condition and the NR(3) condition (z=2.13,
p=.03), and the NR(4) condition (z=2.37, p=.02, see Table 9). The second set of
results pertaining to the SR comparison indicates that credibility significantly
mediates the relation between redundancy and the change in earnings forecast for
the R(0) condition and the NR(1) condition (z= 2.14, p=03). The last set of
results indicates marginally significant support for the mediational hypothesis
between the R(2) condition and the NR(3) condition (z=1.77, p=.08) and
significant support for R(2) condition and the NR(4) condition (z=2.01, p=.04).
In summary, the significant mediational test statistics across comparisons
using both scaled and raw versions of the dependent variable indicate that
redundancy, the independent variable, is significantly related to credibility, the
hypothesized mediator, and that credibility while in the presence of redundancy is
significantly related to the earnings forecast change. Therefore, the results
mediated effect equals zero in the population. All regression models were checked for outliers and
homoscedasticity, independence, and normality of residuals.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
presented in this section support the hypothesis that perceived credibility of
redundant information mediates investors’ earnings forecast changes.
Structural Equation Modeling Approach to Mediational Analysis
Introduction to the Structural Equation Modeling Approach
Structural equation modeling (SEM) consists of both a measurement
model and structural model. The measurement model is evaluated through
confirmatory factor analysis, while the structural model is conducted through
path analysis with latent variables. One difference between SEM and the
regression-based methods used above is SEM’s use of maximum likelihood
estimation (MLE). MLE produces estimates based on maximizing the probability
that the observed covariances are drawn from a population assumed to be the
same as that reflected in the coefficient estimates. Unlike the regression based
approach, MLE does not assume uncorrelated error terms.
In this study, the primary advantages of SEM compared to the regression-
based methods include the use of confirmatory factor analysis for the latent
variables, the ability to explicitly model the mediator, and the ability to
investigate a partial mediational model. In contrast to the Sobel test statistics
used in the regression-based methods, SEM uses goodness-of-fit tests to
determine if the pattern of variances and covariances in the data is consistent with
the structural model; however, these goodness-of-fit tests do not verify if the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
104
paths within the model are significant. The individual paths must be examined by
verifying if the coefficients are significantly different from zero and in the
hypothesized direction.
Two types of goodness-of-fit tests are recommended to assess model fit:
those that assess absolute fit and those that assess comparative fit (Tanaka 1993).
The assessment of absolute fit focuses on the ability of the model to reproduce
the actual covariance matrix implied by the parameters, while the assessment of
comparative fit focuses on comparing two or more models (Kelloway 1998). To
assess absolute fit, the chi-square test and the standardized root mean square
residual are employed. The standardized root mean square residual is employed
to supplement the more commonly used chi-square due to sample size concerns
associated with the chi-square test. To assess the comparative fit of the model, the
normed fit index is employed.
As mentioned above, the chi-square is a most commonly used goodness-
of-fit test in SEM. In the obverse of traditional hypothesis testing, the chi-square
value should not be significant if there is a good model fit. That is, the chi-square
may be viewed as a "badness of fit" measure where a significant test statistic
indicates that the hypothesized model is significantly different from the observed
covariance matrix (Kelloway 1998). The standardized root mean square residual,
(SRMR) is also a measure of absolute fit; however, it is less dependent on sample
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
size. SRMR is the average difference between the predicted and observed
variances and covariances in the model, based on standardized residuals. The
standardized residuals are the fitted residuals divided by the standard error of the
residual. In the case of SRMR, a relatively small test statisctic indicates a better
model fit. The normed fit index, (NFI) is a measure of comparative fit. That is, it
compares the covariance structure implied by the model parameters to the
independence or null model. The NFI ranges from 0 to 1 with values greater than
0.90 indicating a good fit. The NFI indicates the percentage improvement in fit
over the baseline null model (Bentler and Bonnet 1980). Therefore, an NFI of
0.90 indicates a model that is a 90% superior fit to the null model.
Results from the Structural Equation Modeling Approach
The first set of results for the AR comparison, R(0) and NR(3), indicates
that the chi-square test statistic does not reject the null of a lack of fit between
the covariance structure and the covariance structure implied by the model
parameters (% 8 2 =11.4, p=0.18, see Table 10). The SRMR test statistic is less than
0.05 further indicating a strong model fit (SRMR = 0.04, see Table 10), and the
NFI is greater than 0.90 also indicating a strong model fit (NFI =0.92, see Table
10). Thus, the goodness-of-fit statistics are consistent across tests in indicating a
strong model fit.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
106
An evaluation of the path coefficients indicates that the coefficient for the
relation between the independent variable redundancy and the mediator
credibility is significant in the hypothesized direction (-0.47, t=-2.67, p=0.00, see
Table 10). The negative coefficient indicates that the redundant financial press
articles are perceived as being greater in credibility than the non-redundant press
articles. The coefficient for the relation between the mediator credibility and the
dependent variable earnings forecast changes is significant in the hypothesized
direction (0.42, t = 2.43, p = 0.01, see Table 10). The positive coefficient
indicates that the greater credibility of the redundant press articles leads to
increased earnings forecast changes. Finally, the coefficient for the relation
between the independent variable redundancy and the dependent variable
earnings forecast change is significant (-0.29, t= -2.10, p=0.02, see Table 10).
The significant coefficient indicates that this model is a partial mediational model
rather than a full mediational model. That is, there is likely an unmeasured
process mechanism or mechanisms in addition to credibility that are also
responsible for the variation in earnings forecast changes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
107
T A B L E 10
MEDIATIONAL ANALYSIS OF CREDIBILITY
ASYMMETRICAL COMPARISON
REDUNDANT (0) AND NON-REDUNDANT (3) CONDITIONS
PANEL A: Hypothesized Model- Path coefficients, t-statistics, and p-values
Independent
(Exogenous)
-0.47
*42 = -2.67
p = 0.00
Mediator
(Endogenous)
0.42
* 4 2 = 2.43
p = 0.01
Dependent
(Endogenous)
Redundancy in
press coverage
Information
Credibility
Investors’
earnings
forecast
changes
w W
-0.29
*42 = -2.10
p = 0.02
t
PANEL B: Hypothesized Model- Goodness of Fit Tests
Fit Tests
Comparison
1 %
SRMR NFI
Asymmetrical (0) (3) 11.4, p= 0.18 0.04 0.92
This table displays the results of the structural equation modeling mediational analysis for the
asymmetrical comparison between R(0) and NR(3) conditions consisting of the following: path
coefficients; t, which is the t-statistic of the regression coefficient for each link in the model; p, which
is the p-value for the associated test statistic; and goodness of fit statistics. The fit statistics include the
following: % s2 (chi-square statistic), SRMR (standardized root mean squared residual statistic) and
NFI (normed fit index test statistic).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
108
The second set of results for the AR comparison, R(0) and NR(4),
indicates that the chi-square test statistic marginally rejects the null of a lack of fit
between the covariance structure and the covariance structure implied by the
model parameters (yx 2 = 15.7, p=0.05, see Table 11). The SRMR test statistic is
equal to 0.05 indicating a good model fit (SRMR = 0.05, see Table 11), and the
NFI is equal to 0.90 also indicating a good model fit (NFI =0.90, see Table 11).
Although not as strong as the first set of results regarding model fit, the
goodness-of-fit statistics taken together for this comparison indicate a relatively
good fit.
An evaluation of the path coefficients indicates that the coefficient for the
relation between the independent variable redundancy and the mediator
credibility is significant in the hypothesized direction (-0.49, t=-2.98, p=0.00, see
Table 11). Again, the negative coefficient indicates that the redundant financial
press articles are perceived as being greater in credibility than the non-redundant
press articles. In addition, the coefficient for the relation between the mediator
credibility and the dependent variable earnings forecast changes is significant in
the hypothesized direction (0.54, t = 3.30, p = 0.00). The positive coefficient
indicates that the greater credibility of the redundant press articles leads to
increased earnings forecast changes. Finally, the coefficient for the relation
between the independent variable redundancy and the dependent variable
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
earnings forecast change is significant (-0.24, fi=-1.86, p=0.03). As in the first set
of results, the significant coefficient indicates support for credibility as a partial
mediator rather than the sole mediator.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
110
TABLE 11
MEDIATIONAL ANALYSIS OF CREDIBILITY
ASYMMETRICAL COMPARISON
REDUNDANT (0) AND NON-REDUNDANT (4) CONDITIONS
PANEL A: Hypothesized Model- Path coefficients, t-statistics, and p-values
Independent
-0.49
^ 4 2 = -2.98 Mediator
(Endogenous)
0.54
^ 4 2 = 3.30
p = 0.00
Dependent
(Endogenous)
Investors’
Redundancy in
press coverage
Information
Credibility
earnings
forecast
changes
w w
-0.24
tn2= "L86
p = 0.03
PANEL B: Hypothesized Model- Goodness of Fit Tests
Fit Tests
Comparison SRMR NFI
Asymmetrical (0) (4) 15.7, p = 0.05 0.05 0.90
This table displays the results of the structural equation modeling mediational analysis for the
asymmetrical comparison between R(0) and NR(4) conditions consisting of the following: path
coefficients; t, which is the t-statistic of the regression coefficient for each link in the model; p, which
is the p-value for the associated test statistic; and goodness of fit statistics. The fit statistics include the
following: ys2 (chi-square statistic), SRMR (standardized root mean squared residual statistic) and
NFI (normed fit index test statistic).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I l l
The results for the SR comparison, R(0) and NR(1), indicates that the chi-
square test statistic does not reject the null of a lack of fit between the covariance
structure and the covariance structure implied by the model parameters (% 8 2 =
13.0, p = 0.11, see Table 12). The SRMR test statistic is equal to 0.05 further
indicating a good model fit (SRMR = 0.05, see Table 12), and the NFI is greater
than 0.90 also indicating a strong model fit (NFI =0.91, see Table 12). Thus, the
goodness-of-fit statistics are consistent across tests in indicating a strong model
fit.
An evaluation of the path coefficients indicates that the coefficient for the
relation between the independent variable redundancy and the mediator
credibility is significant in the hypothesized direction (-0.34, t= -1.99, p=0.03, see
Table 12). Again, the negative coefficient indicates that the redundant financial
press articles are perceived as being greater in credibility than the non-redundant
press articles. The coefficient for the relation between the mediator credibility
and the dependent variable earnings forecast changes is significant in the
hypothesized direction (0.68, t = 3.13, p = 0.00, see Table 12). Again, the positive
coefficient indicates that the greater credibility of the redundant press articles
leads to increased earnings forecast changes. Finally, the co e fficien t for the
relation between the independent variable redundancy and the dependent variable
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
112
earnings forecast change is significant (0.27, t = 2.12, p = 0.02, see Table 12),
indicating support for a partial mediational model.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
113
TABLE 12
MEDIATIONAL ANALYSIS OF CREDIBILITY
SYMMETRICAL COMPARISON
REDUNDANT (0) AND NON-REDUNDANT (1) CONDITIONS
PANEL A: Hypothesized Model- Path coefficients, t-statistics, and p-values
Independent
-0.34
f4 2 =-1-99 Mediator
(Endogenous)
0.68
*42 = 3.13
p = 0.00
Dependent
(Endogenous)
Investors’
Redundancy in
press coverage
Information
Credibility
earnings
forecast
changes
w w
0.27
*42 = 2.12
p = 0.02
PANEL B: Hypothesized Model- Goodness of Fit Tests
Fit Tests
Comparison SRMR NFI
Symmetrical (0) (1) 13.0, p - 0.11 0.05 0.91
This table displays the results of the structural equation modeling mediational analysis for the
symmetrical comparison between R(0) and NR(1) conditions consisting of the following: path
coefficients; t, which is the t-statistic of the regression coefficient for each link in the model; p, which
is the p-value for the associated test statistic; and goodness of fit statistics. The fit statistics include the
following: % $ 2 (chi-square statistic), SRMR (standardized root mean squared residual statistic) and
NFI (normed fit index test statistic).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
114
In summary, the SEM results indicate that the hypothesized model is a
good fit for the actual model implied by the parameters. Furthermore, the
individual path coefficients are significant and in the hypothesized direction. The
results lend support to the hypothesis that redundancy, the independent variable,
is significantly related to credibility, the hypothesized mediator, and that
credibility is significantly related to the earnings forecast change.
To summarize the results, due to the fact that the present model is
comprised of both measured variables and latent variables, both regression-based
and SEM approaches are used to investigate the mediational hypothesis. The
results from these two approaches converge and support the hypothesis that
perceived credibility of redundant information mediates the effects of redundancy
on investors’ earnings forecast changes.
ALTERNATIVE PROCESS EXPLANATIONS
To this point, the results of the study indicate that redundancy
significantly influences investors’ earnings forecasts and perceived credibility of
redundant press articles is a viable process explanation. The analysis thus far,
however, cannot rule out the alternative explanations that investors’ responses
could result from either the increased salience of repeated information or a belief
by investors that the repetition of news events in the financial press articles
represents an information signal.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
115
Joe (2003) finds that auditors exposed to both financial information and
press coverage concerning a company’s debt status are more likely to judge a
company to fail than auditors who view only the company’s financial
information. This finding is attributed to evidence that redundant press coverage
increases the salience of the company’s debt problems.4 7 To examine if salience
is an explanation for my results, I construct a salience variable and submit this
variable to a regression-based mediational analysis. The salience variable is based
on participants’ responses to multiple choice recall questions concerning their
financial press article information sets.4 8 Due to variation in information content
across and within conditions, there is variation in both the quantity and content of
the recall questions. Therefore, I standardize and adjust the recall scores for
relative difficulty to obtain comparable salience measures across conditions.4 9
4 7 Auditors’ attention is measured using recall of influential cues in their going concern judgments.
Using this measure, Joe (2003) characterizes cues that auditors attend to more as “salient” cues. In
this context, the author suggests salience as the variable that mediates the relation between
redundancy and the dependent variable, going concern judgments. This is distinct from the
psychology literature referenced in Chapter 2 that is concerned with the variables, e.g., familiarity,
that mediate the relation between redundancy and the credibility effect.
4 8 As an example, one of the multiple-choice questions was: “According to the information provided,
a substantial impact on the financial statements should follow from the recent acquisition of
____________ .” Then four company names were provided in a multiple-choice format, one of which
was the correct answer.
491 standardize the recall scores across recall sets by converting the scores to a percentage. I adjust for
the relative difficulty of the recall question sets by equalizing the mean percentage recall scores across
recall sets. Note that these adjustments do not equalize the recall means across redundancy conditions.
That is, as a result of the counterbalancing of information content, each redundancy condition is
comprised of multiple different information sets and their corresponding recall question sets. The
results do not change when I use the unadjusted recall scores in the analysis.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
The results of the mediational analysis indicate that salience does not
significantly mediate the relation between redundancy and the change in earnings
forecasts for the R(0) condition and the NR(1) condition (z= .60, p=.57),5 0 NR(3)
condition (z=1.35, p=.18), and the NR(4) condition (z=45, p=.65). In addition,
results are not significant for the R(2) and NR(3) condition (z= 1.204, p=.23) or
NR(4) condition (z=.024, p=.98). An important caveat to this exploration is that
people are sometimes influenced by events that they fail to remember on explicit
tests (Jacoby 1991; Richardson-Klavehn and Bjork 1988; Roediger 1990;
Schacter 1987).
The second alternative explanation is the potential belief by investors that
the repetition of news events in the financial press represents an information
signal. A variety of market and quasi-market phenomena can be usefully viewed
in a signaling context (Spence 1973). Although there is not a generally accepted
definition of exactly what constitutes signaling, there are at least three elements
normally involved: information asymmetry between at least two parties, an
attempt by the more informed to communicate to the less informed party
indirectly rather than directly, and either an explicit or implicit benefit for the
more informed party to make a credible communication (Bannock et al. 1998;
5 0 These z-values are calculated after Sobel (1982). The results do not change with the use of the
Goodman I or Goodman I I tests.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
117
Dixit and Skeath 1999; Dutta 1999; Rutherford 2002). It is reasonable to assume
that investors on average believe in the existence of private financial information.
It also may be reasonable for the average investor to perceive that the editor or
author of a financial publication has access to some degree of private
information. If true, then investors may view the repetition of a financial news
items as a form of signal about the expectations of a firm’s future prospects. That
is, these investors may perceive information content in the mere choice by an
editor to publish or an author to write financial press articles.
To examine this signaling explanation, responses to four post-
experimental questions aimed at measuring the participants’ perceptions of
signaling within the financial press article set are analyzed. Specifically, the
questions address whether participants believe that the authors or editors know
more information than what appears in the articles and attempt to convey this
information to readers through the act of choosing certain companies to feature.5 1
The participants’ pattern of responses indicate that the questions measure the
same latent construct (Cronbach’s alpha =.74); therefore, the responses to the
four questions are averaged into a single measure. The results of the mediational
analysis conducted utilizing the constructed signaling measure indicate that
5 1 Investors’ perceptions of information signaling are investigated for both the specific financial press
articles utilized in this study and for press articles in general. The reported results are based on
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
118
signaling does not significantly mediate the relation between redundancy and the
change in earnings forecasts for the R(0) and NR(1) condition (z - .51, p=.61),
NR(3) condition (z=.28, p=.78), or NR(4) condition (z=.40, p=.69). In addition,
mediational results are not significant for the R(2) and NR(3) condition (z=.17,
p=.86) or NR(4) condition (z=.00, p=.99).
In summary, the insignificant mediational test statistics across
comparisons indicate that neither salience nor signaling provides an adequate
alternative process explanation for how redundant information influences
investors’ earnings forecast changes. However, recall that significant mediational
tests across comparisons and types of analyses do converge to support perceived
credibility as suitable process explanation for how redundancy affects investors’
judgment and decision-making.
SUMMARY
The data analysis and results chapter of this thesis first discussed the
checks designed to ensure the proper manipulation of redundancy and perception
of the experimental materials. Then, a discussion of the analysis and results for a
series of planned comparisons evaluating the set of three hypotheses that
comprise the first hypothesis was presented. Both the raw and scaled versions of
these results documented an effect for redundancy on investors’ judgment and
responses to the financial press articles utilized in this study. The results do not change if the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
decision-making. Next, a discussion of the analysis and results for both
regression-based and structural equation modeling approaches for evaluating the
mediational hypothesis was presented. The convergence of these two approaches
provided strong evidence necessary to establish perceived credibility as the
process mechanism underlying the effect for redundancy on investors’ judgment
and decision-making. Finally, the analysis results did not support the alternative
explanations o f increased salience of repeated information or a possible belief by
investors that the repetition of news events indicates an information signal.
responses to press articles in general are analyzed.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 5: CONCLUSION
SUMMARY OF RESULTS
The impact of redundant financial press coverage on nonprofessional
investors’ earnings forecasts is examined by conducting a between-participants
experiment in which investors examine sets of financial press articles that vary as
to redundancy. The purpose of this experiment was to examine whether mere
redundancy in press articles affected investors’ financial judgments, and, if so,
the psychological process underlying the observed effect on investors’ financial
judgments. The unique design of this study accounted for the naturally arising
confounds inherent in comparisons between redundant and non-redundant
information in prior research. The results provided support for the idea that
redundant financial press coverage can significantly influence investors’
judgments. In addition, a perception that redundancy is positively associated with
credibility appears to have explained these investors’ judgments. My results
suggest that investors confronted with a consistent stream of repetitive financial
news may fail to properly weight this information when engaging in investment
decision-making. An improper weighting of information may lead to a negative
impact on these nonprofessional investors’ investment returns. This negative
impact may manifest itself in the form of investors’ focusing on “newsworthy”
companies that are covered more intensely by the media, rather than companies
permission of the copyright owner. Further reproduction prohibited without permission.
with strong fundamentals. Another negative implication of improper weighting of
media stories is that it may lead to under diversification on the part of
nonprofessional investors. That is, if there is a recent trend in the media of
repetitious coverage of a particular business sector, e.g., oil companies and
suppliers, then the level of press coverage may lead investors to focus too
narrowly on their stock purchases. The negative implications stemming from
“bandwagon” investing and under diversification have a direct impact on the
economic welfare of the nonprofessional investors and also may have an indirect
impact on market welfare in general due to the limitations of the arbitrage
mechanism.
STUDY LIMITATIONS AND FUTURE RESEARCH SUGGESTIONS
There are several limitations to this study and possibilities for future
research directions. First, information redundancy is referred to and studied in a
dichotomous fashion although in reality it is continuous concept. This may have
implications for the interpretation o f the effects of redundancy on judgment. For
example, in constructing the financial press articles, the same four items of
information are used in the redundant set of articles. In reality, it is likely that
several of the topics contained in the press articles would be similar, but not all of
the topics. Also, in reality, it is likely that the author may add additional
information or insights into an article beyond what is relayed in press releases.
permission of the copyright owner. Further reproduction prohibited without permission.
Therefore, realistic depictions of redundancy are generally less precise than what
was constructed in the case materials. Future research might employ multiple
levels of redundancy or “fuzzier” conceptions of redundancy to address these
issues and raise the level of external validity.
Second, the participants in this study are nonprofessional investors in a
fixed incentive experimental setting. Professional investors and/or
nonprofessional investors in a market setting may not be affected by redundancy
in financial press articles. Thus, future research might investigate the impact of
redundancy on the judgments of professional investors or examine redundancy
effects on nonprofessional investors in a laboratoiy market setting.
Third, the nonprofessional investors in this study are not representative of
all nonprofessional investors. In particular, the participants may have a higher
level of education in accounting and finance than typical nonprofessional
investors. In addition, these nonprofessional investors may represent a more
homogenous sample of investors than the population of nonprofessional
investors. That is, the participants in this study are all graduate business students,
whereas the population of nonprofessional investors is represented by a
heterogeneous group of people consisting of, for example, janitors, doctors,
teachers, farmers, etc. Future research might study a more heterogeneous group
of nonprofessional investors that would allow greater generalization of results.
permission of the copyright owner. Further reproduction prohibited without permission.
Fourth, the information sets in this study are comprised of both qualitative
and quantitative information. I cannot rule out the possibility that it is only one
form of redundant information that is responsible for the results. That is, it may
be that investors discount for redundant quantitative information, e.g., earnings
per share, while they do not fully discount for qualitative statements, e.g., “the
performance of the company is strong”. Future research might attempt to
investigate and disentangle varying effects of redundancy for these types of
information.
Fifth, the investors in this study are engaged in a passive search for
information as opposed to an active search for information. That is, investors are
presented with the financial press articles and then required to make an earnings
forecast. In an active search for information, an investor might be asked to select
an article from a group of several redundant and non-redundant financial press
articles. In this scenario, inferences regarding redundancy might be made based
on the selection choices, rather than a decision variable such as earnings forecasts
or stock purchases. Future research might employ active search scenarios or
manipulate passive and active search scenarios. The results of the current study
cannot generalize to active search scenarios.
Finally, this study lacks an examination of remediational strategies. In
other words, given that investors misweight redundant information, how might
permission of the copyright owner. Further reproduction prohibited without permission.
their judgments be improved? Examples o f strategies used in previous judgment
and decision making research include education, training, and decision aids (see
Bonner 1996). Future research might pit theoretically-based remedies against one
another to evaluate which improves investors’ judgments to a greater extent.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
125
BIBLIOGRAPHY
Alba, J. and H. Marmorstein. 1987. The effects of frequency knowledge on
consumer decision making. Journal o f Consumer Research 14: 14-25.
Arkes. H., L. Boehm, and G. Xu. 1991. Determinants of judged validity. Journal
o f Experimental Social Psychology 27: 576-605.
Arkes, H., C. Hackett, and L. Boehm. 1989. The generality of the relation
between familiarity and judged validity. Journal o f Behavioral Decision
Making 2: 81-94.
Armelius, B. and K. Armelius. 1974. The use of redundancy in multiple-cue
judgments: Data from a suppressor-variable task. American Journal o f
Psychology 87: 385-392.
Asch, S. 1940. Studies in the principles of judgments and attitudes: II.
Determination of judgments by group and ego standards. Journal o f Social
Psychology 12: 433-465.
Asch, S. 1948. The doctrine of suggestion, prestige and initiation in social
psychology. Psychological Review 55: 250-276.
Bacon, F. 1979. Credibility of repeated statements: Memory for trivia. Journal o f
Experimental Psychology: Human Learning and Memory 5: 241-252.
Bagdikian, B. 2000. The Media Monopoly, 6th edition. Boston, MA: Beacon
Press.
Bannock, G., R. Baxter, and E. Davis. 1998. Dictionary o f economics. London,
England: Penguin Books.
Bamouw, E. 1978. The Sponsor. New York, NY: Oxford University Press.
Baron, R. and D. Kenny. 1986. The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal o f Personality and Social Psychology 51:
1173-1182.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
126
Begg, I. 1974. Estimation of word frequency in continuous and discrete tasks.
Journal o f Experimental Psychology 102: 1046-1052.
Begg, I., A. Anas, and S. Farinacci. 1992. Disassociation of processes in belief:
Source recollection, statement familiarity, and the illusion of truth. Journal o f
Experimental Psychology General 121: 446-458.
Begg, I. and V. Armour. 1991. Repetition and the ring of truth: Biasing
comments. Canadian Journal o f Behavioural Science 23: 195-213.
Begg, I., V. Armour, and T. Kerr. 1985. On believing what we remember.
Canadian Journal o f Behavioral Science 17: 199-214.
Belch, G. 1982. The effects of television commercial repetition on cognitive
response and message acceptance. Journal o f Consumer Research 9: 56-65.
Belkaoui, A. 1984. The effects of diagnostic and redundant information on loan
officers’ predictions. Accounting and Business Research 14: 249-256.
Bentler, P. and D. Bonnet. 1980. Significance tests and goodness of fit in the
analysis of covariance structures. Psychological Bulletin 88: 588-606.
Boehm, L. 1994. The validity effect: A search for mediating variables. Journal o f
Personality and Social Psychology 20: 285-293.
Bonner, S. 1999. Judgment and decision-making research in accounting.
Accounting Horizons 13: 385-398.
Bonner, S. 2005. Judgment and Decision-Making in Accounting. Upper Saddle
River, NJ: Prentice-Hall, Forthcoming.
Bonner, S., R. Libby, and M. Nelson. 1996. Using Decision Aids to Improve
Auditors’ Conditional Probability Judgments. The Accounting Review :221-
240.
Bonner, S., B. Walther, and J. Hugon. 2004. Investor overreaction to
financial media coverage: The case offinancial analysts ’forecast revisions.
Working Paper, University o f Southern California and Northwestern
University.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
127
Brav, A. and J. Heaton. 2002. Competing theories of financial anomalies. The
Review o f Financial Studies 15: 575-606.
Brehmer, B. 1974. The effect of cue intercorrelation on interpersonal learning of
probabilistic inference tasks. Organizational Behavior and Human
Performance 12: 397-412.
Busse, J. and T. Green. 2002. Market efficiency in real time. Journal o f Financial
Economics 65: 415-437
Cacioppo, J. and R. Petty. 1979. Effects o f message repetition and position on
cognitive response, recall, and persuasion. Journal o f Personality and Social
Psychology 37: 97-109.
Cummings, J. 2002. Small investors now a big bloc market slump could influence
how emergent group votes. Wall Street Journal. A4. September 27.
Davis, F. D., G. L. Lohse, and J. E. Kotterman. 1994. Harmful effects of
seemingly helpful information on forecasts of stock earnings. Journal o f
Economic Psychology 15: 253-267.
Dixit, A. and S. Skeath. 1999. Games o f strategy. New York, NY: W. W. Norton
and Company.
Djankov, S., C. McLiesh, T. Nenova, and A. Shleifer. 2002. Media ownership
and prosperity. The Right to Tell-The Role o f Mass Media in Economic
Development. The World Bank Institute: 141-165.
DuCharme, L, S. Rajgopal, and S. Sefcik. 2002. Why was internet IPO
underpricing so severe? Working paper, University of Washington
Dutta, P. K. 1999. Strategies and games. Cambridge, MA: The MIT Press
Einhom, H. J., D. N. Kleinmuntz, and B. Kleinmuntz. 1979. Linear regression
and process-tracing models of judgment. Psychological Review 86: 465-485.
Fama, E. 1970. Efficient capital markets: A review of theory and empirical work.
Journal o f Finance 25: 383-417.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
128
Festinger, L. A. 1954. A theory of social comparison processes. Human
Relations 7: 117-140.
Gadarowski, C. 2004. Financial press coverage and expected stock returns.
Working paper, Cornell University
Gale Directory of Publications and Broadcast Media. 2001. Detroit, MI: Gale
Research
Goethals, G. R. and R. E. Nelson. 1973. Similarity in the influence process: The
belief-value distinction. Journal o f Personality and Social Psychology 25:
117-122.
Gonzalez, R. 1994. When words speak louder than actions: Another’s evaluation
can appear more diagnostic than their decisions. Organizational Behavior and
Human Decision Processes 58: 214-245.
Goodman, L. A. 1960. On the exact variance of products. Journal o f the
American Statistical Association 55: 708-713.
Greenwald, A. and C. Leavitt. 1984. Audience involvement in advertising: Four
levels. Journal o f Consumer Research 11: 581-592.
Greenwald, A., A. Pratkanis, A. Leippe, and M. Baumgardner. 1986. Under what
conditions does theory obstruct research progress? Psychological Review 93:
216-229.
Grader, C., T. Cook, K. Hennigan, B. Flay, C. Alessis, and J. Halamaj. 1978.
Empirical tests of the absolute sleeper effect predicted from the discounting
cue hypothesis. Journal o f Personality and Social Psychology 36: 1061 -1074.
Hartley, D. 1749. Observations on man, his frame, his duty and his expectations
(Vol. 1). London: Richardson.
Hasher, L., D. Goldstein, and T. Toppino. 1977. Frequency and the conference of
referential validity. Journal o f Verbal Learning and Verbal Behavior 16: 107-
112.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
129
Hawkins, S. and S. Hoch. 1992. Low-involvement learning: Memory without
evaluation. Journal o f Consumer Research 19: 212-225.
Hawkins, S., S. Hoch, and J. Levy. 2001. Low involvement learning: Repetition
and coherence in familiarity and belief. Journal o f Consumer Psychology 11:
1- 11.
Herman, E. 2002. The media and the markets in the United States. The Right to
Tell-The Role o f Mass Media in Economic Development. The World Bank
Institute: 61-81.
Herman, E. and R. McChesney. 1997. The Global Media. London, England:
Cassel.
Ho, T. and R. Michaely. 1988. Information quality and market efficiency.
The Journal o f Financial and Quantitative Analysis 23: 53-70.
Hovland, C., J. Janis, and H. Kelley. 1953. Communication and persuasion.
New Haven, CT: Yale University Press.
Howell, W. 1973. Storage of events and event frequencies: A comparison of two
paradigms in memory. Journal o f Experimental Psychology 98: 260-263.
Hoyer, W. 1984. An examination of consumer decision making for a common
repeat purchase product. Journal o f Consumer Research 11: 822-829.
Hoyer, W. and S. Brown. 1990. Effects of brand awareness on choice for a
common, repeat-purchase product. Journal o f Consumer Research 17: 141-
148.
Islam, R. 2002. Into the looking glass: What the media tell and why. The Right to
Tell-The Role o f Mass Media in Economic Development. The World Bank
Institute: 1-23.
Jacoby, L. 1991. A process dissociation framework: Separating automatic from
intentional influences of memory. Journal o f Memory and Language 30: 513-
541.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
130
Jacoby, L. and M. Dallas. 1981. On the relationship between autobiographical
memory and perceptual learning. Journal o f Experimental Psychology
General: 110: 306-340.
Jacoby, L. and C. Kelley. 1987. Unconscious influences of memory for a prior
event. Personality and Social Psychology Bulletin 13: 314-336.
Jacoby, L. and C. Kelley. 1990. An episodic view of motivation: Unconscious
influences of memory. In E. T. Higgins and R. M. Sorrentino (Eds.),
Handbook o f motivation and cognition: Foundations o f Social Behavior. (Vol.
2, pp. 451-481). New York, NY: Guilford Press.
Jacoby, L. and C. Kelley. 1991. Unconscious influences of memory:
Dissociations and automaticity. In D. Milner and M. Rugg (Eds.),
Neuropsychology o f Unconsciousness (pp. 201-233). San Diego, CA:
Academic Press.
Jacoby, L., V. Woloshyn, and C. Kelley. 1989. Becoming famous without being
recognized: Unconscious influences of memory provided by dividing
attention. Journal o f Experimental Psychology: General 118: 115-125.
Joe, J. 2003. Why press coverage influences the audit opinion. Journal o f
Accounting Research 41:109-133
Judd, C. and D. Kenny. 1981. Process analysis: Estimating mediation in
treatment evaluations. Evaluation Review 5: 602-619.
Kahneman, D. and A. Tversky. 1973. On the psychology of prediction.
Psychological Review 80: 237-251.
Kelley, H. 1967. Attribution theory in social psychology. Nebraska
Symposium on Motivation 15: 192-238.
Kelloway, K. 1998. Using Lisrel fo r Structural Equation Modeling: A
Researcher’ s Guide. Thousand Oaks, CA: Sage Publications.
Knowles, B. A., K. R. Hammond, T. R. Stewart, andD. A. Summers. 1971.
Positive and negative redundancy in multiple cue probability tasks. Journal o f
Experimental Psychology 90: 158-159.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
131
Krugman, H. 1965. The impact of television advertising: Learning without
involvement. Public Opinion Quarterly 29: 349-356.
Kunst-Wilson, W. and R. Zajonc. 1980. Affective discrimination of stimuli that
cannot be recognized. Science 207: 557-558.
Law, S., S. Hawkins, and F. Craik. 1998. Repetition-induced belief in the elderly:
Rehabilitating age-related memory deficits. Journal o f Consumer Research
25:91-107.
Libby, R., R. Bloomfield, and M. W. Nelson. 2002. Experimental research in
financial accounting. Accounting, Organizations and Society 27: 775-810.
Libby, R. and H. T. Tan. 1999. Analysts reactions to warnings of negative
earnings surprises. Journal o f Accounting Research 37: 415-436.
Lindell, M. and T. Stewart. 1974. The effects of redundancy in multiple-cue
probability learning. American Journal o f Psychology 87: 393-398.
Lipe, M. 1990. Discussion of the effect of forecast redundancy on judgments
of a consensus forecast’s expected accuracy. Journal o f Accounting Research
28: 48-54.
Locke, J. 1700. An essay concerning human understanding (Bk. 4), 4th Edition.
London, England: Churchill.
MacKinnon, D. P. and J. H. Dwyer. 1993. Estimating mediated effects in
Prevention studies. Evaluation Review 17: 144-158.
MacKinnon, D. P., G. Warsi, and J. H. Dwyer. 1995. A simulation study of
mediated effect measures. Multivariate Behavioral Research 30: 41-62.
Madrick, J. 2000.The Influence of the Financial Media over International
Economic Policy. International Capital Markets and the Future o f Economic
Policy No. 16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
132
Maines, L. A. 1990. The effect of forecast redundancy on judgments of a
consensus forecast’s expected accuracy. Journal o f Accounting Research 28:
29-47.
Miller, G. S. 2003. The press as a watchdog for accounting fraud. Working paper,
Harvard University.
Mutchler, J. F., W. Hopwood, and J. M. McKeown. 1997. The influence of
contrary information and mitigating factors on audit opinion decisions on
bankrupt companies. Journal o f Accounting Research 35: 295-310.
Naylor, J. C. and E. A. Schenck. 1968. The influence of cue redundancy upon the
human inference process for tasks of varying degrees of predictability.
Organizational Behavior and Human Performance 3: 47-61.
New York Stock Exchange. 2000. Shareowner ship 2000. New York Stock
Exchange, Inc. New York, NY: NYSE.
Nunnaly, J. 1978. Psychometric theory. New York, NY: McGraw-Hill.
Olshavsky, R. and D. Granbois 1979. Consumer decision making: Fact or fiction?
Journal o f Consumer Research 6: 93-100.
Payne, J., J. Bettman, and E. Johnson. 1993. The adaptive decision maker. New
York, NY: Cambridge University Press.
Pratkanis, A., R. Greenwald, M. Leippe, and M. Baumgardner. 1988. In search of
reliable persuasion effect: III. The sleeper effect is dead. Long live the sleeper
effect. Journal o f Personality and Social Psychology 54: 203-218.
Richardson-Klavehn, A. and R. Bjork. 1988. Measure of memory. Annual Review
o f Psychology 39. 475-543.
Roediger, H. 1990. Implicit memory: Retention without remembering. American
P s y c h o lo g is t 45: 1043-1056.
Roggeveen, A. and G. Johar. 2002. Perceived source variability versus
familiarity: Testing competing explanations for the truth effect. Journal o f
Consumer Psychology 12: 81-91.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
133
Rubinstein, M. 2001. Rational Markets: Yes or No? The Affirmative Case.
Financial Analysts Journal 57: 15-29.
Rutherford, D. 2002. Routledge dictionary o f economics. New York, NY:
Routledge.
Saegert, S. and J. Jellison. 1970. Effects of initial level of response competition
and frequency of exposure on liking and exploratory behavior. Journal o f
Personality and Social Psychology 16: 553-558.
Schacter, D. 1987. Implicit memory: History and current status. Journal o f
Experimental Psychology: Learning, Memory and Cognition 13: 501-518.
Schmitt, N. and A. L. Dudycha. 1975. Positive and negative cue redundancy in
multiple cue probability learning. Memory & Cognition 3: 78-84.
Schuster, T. 2003. Fifty-fifty. Stock recommendations and stock prices. Effects
and benefits of investment advice in the business media. Working paper,
Leipzig University.
Schwartz, M. 1982. Repetition and rated truth value of statements. American
Journal o f Psychology 95: 393-407.
Shiller, R. J. 2001. Bubbles, human judgment, and expert opinion. Yale Cowles
Foundation Discussion Paper No. 1303
Shleifer, A. 2000. Inefficient Markets: An Introduction to Behavioral Finance.
New York, NY: Oxford Press.
Smith, G. and D. Dorfman. 1975. The effect of stimulus uncertainty on the
relationship between frequency of exposure and liking. Journal o f Personality
and Social Psychology 31: 150-155.
Sobel, M. 1982. Asymptotic intervals for indirect effects in structural equations
models. In S. Leinhart (Ed.), Sociological methodology (pp.290-312).
San Francisco, CA: Jossey-Bass.
Soli, J. 1999. Intuitive theories of information: Beliefs about the value of
redundancy. Cognitive Psychology 38: 317-346.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
134
Spence, M. 1973. Job market signaling. Quarterly Journal o f Economics 87: 355-
374.
Stang. D. J. 1975. Effects of “mere exposure” on learning and affect. Journal o f
Personality and Social Psychology 31: 7-12.
Stromberg, D. 2002. Distributing news and political influence. The Right to Tell-
The Role o f Mass Media in Economic Development. The World Bank Institute:
95-106.
Sujan, M. 1985. Consumer knowledge: Effects on evaluation strategies mediating
consumer judgments. Journal o f Consumer Research 12: 31-46.
Tanaka, J. 1993. Multifaceted conceptions of fit in structural equation models. In
K. A. Bollen and J. S. Long (Eds.), Testing Structural Equation Models (pp.
10-39). Newbury Park, CA: Sage Publications.
Thaler, R. 1993. Advances in Behavioral Finance. New York, NY: Russell Sage
Tulving, E. 1983. Elements o f Episodic Memory. New York, NY: Oxford
University Press.
Tulving, E., D. Schacter, and H. Stark. 1982. Priming effects in word-fragment
completion are independent of recognition memory. Journal o f Experimental
Psychology: Learning, Memory, and Cognition 8: 336-342.
Underwood, B. J. 1971. Recognition memory. In H. H. Kendler and J. T. Spence
(Eds.), Essays in neo behaviorism (pp. 313-335). New York, NY: Appleton-
Century Crofts.
Zajonc, R. B. 1968. Attitudinal effects of mere exposure. Journal o f Personality
and Social Psychology 9: 1-27.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
135
APPENDIX A
EXPLANATION OF THE SOBEL TEST
Illustration of Mediation
Earnings forecast A
(D ependent V ariable)
Credibility
(M ediator)
Redundancy
(Independent Variable)
The path coefficients are: a, b, and c. The standard errors of the path coefficients
are in parentheses.
To conduct the mediatonal test, I examine a regression model with
redundancy, the independent variable, predicting credibility, the mediator, which
provides the path coefficient, a, and the standard error of the path coefficient, sa.
Next, I examine a regression model with redundancy and credibility predicting
the earnings forecast change, the dependent variable, which provides the path
coefficient, b, and the standard error of the path coefficient, sb. Finally, I calculate
the critical ratio as a test of whether the indirect effect of the independent variable
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
136
on the dependent variable via the mediator is significantly different from zero.5 2
The p-values are drawn from the unit normal distribution under the assumption of
a two-tailed z-test of the hypothesis that the mediated effect equals zero in the
population. There are three versions of the Sobel test for mediation.5 3
Sobel test equation:
z-value = a*b/SQRT(b2*sa2 + a2*sb2)
Goodman (I) test equation:
z-value = a*b/SQRT(b2*sa2 + a2*sb2 + sa2*sb2)
Goodman (II) test equation:
z-value = a*b/SQRT(b2*sa2 + a2*sb2 - sa2*sb2)
5 2 Details can be found in Baron and Kenny (1986), Sobel (1982), Goodman (1960), and MacKinnon,
Warsi, and Dwyer (1995).
5 3 Test equations are from MacKinnon and Dwyer (1994) and MacKinnon, Warsi, and Dwyer (1995).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
137
APPENDIX B
LIBBY BOXES: OVERVIEW OF EXPERIMENTAL DESIGN
Conceptual
Operational
Independent Mediator Dependent
Redundancy
in Press Coverage -
Information
Credibility
—►
Investors’
Earnings
Forecasts
1
r
I
Press Article Sets:
2 Redundant and
3 Non-Redundant
—► 4 Questions
Measuring
Perceived
Credibility
-
Change from
Baseline to
Forecast
Made after
Press Article
Treatment
t
T
Controls
Measured and/or Randomized:
Information salience
Signaling in media
Individual differences (e.g. ability, experience, etc.)
Other factors (e.g. room, time, etc.)______________
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
APPENDIX C
EXAMPLE OF THE CASE MATERIALS
Redundant Condition (0)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
139
General Instructions
Thank you for your participation in this study! You will be compensated $20 for
your effort and, additionally, everyone who has a fully completed questionnaire will
be entered in a lottery that will pay a cash prize of $100.
The purpose of this study is to investigate how people use information to make
investment decisions. It should take you approximately 25 minutes to complete this
study.
In this study, you will be asked to assume the role of an investor evaluating
information about Principal Health Corporation (PHC). After reading through the
relevant materials concerning PHC, you will be asked for your opinion about several
investment issues regarding the company.
The information contained in the materials on the following pages is not intended to
include all the potential information that would normally be available to you for an
investment decision. Rather, this information should be viewed as a representative
subset of the information that is informative enough for the required questions to be
answered. Therefore, for the purposes of this study, base your answers only on the
information provided and do not make any assumptions concerning information that
is not presented.
If you have any questions during this study, please do not hesitate to raise your hand
and someone will assist you.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
140
Part 1 Directions:
As a potential investor, you have obtained the following background
information and financial statem en ts from the Principal Health, Inc. 2002
annual report (year-end 2003 results a re not yet available). P le a se read
through the information and then answ er the questions that follow regarding
your opinion about potential 2003 year-end financial results.________________
Principal Health, Inc.
COMPANY BACKGROUND INFORMATION
Principal Health, Inc. (PHC) is a provider of products and services supporting
the health-care industry. PHC develops, manufactures, packages and markets
products for patient care; develops drug-delivery technologies; distributes
pharmaceuticals, medical-surgical and laboratory supplies; and offers consulting and
other services. PHC serves manufacturers of pharmaceuticals and medical products,
as well as those who care for patients.
Serving Health-Care Manufacturers:
PHC works in partnership with pharmaceutical and biotechnology companies
to develop custom manufacturing solutions and packaging for their drugs. The
Company’s distribution services include warehousing and distribution of drugs,
medical-surgical and laboratory products to hospitals, retail pharmacies, surgery
centers, physician practices and other points of care, as well as specialized inventory
management and logistics services.
Serving Providers of Patient Care:
For hospitals and other health-care providers, PHC develops and
manufactures a broad range of medical and surgical products, as well as automated
supply and pharmaceutical dispensing systems. PHC is one of the largest distributors
of pharmaceuticals and other medical products, and provides systems to help
customers streamline purchasing and inventory management. PHC’s operations and
clinical improvement services include unique consulting and information systems to
help health-care providers reduce supply costs, improve operational efficiencies and
enhance clinical outcomes.
Employees of PHC Health:
PHC employs 5,000 people including 400 pharmacists and 50 clinical,
productivity and logistics consultants. PHC has approximately 900 sales and service
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
141
associates serving hospitals and other clinical-care providers, retail pharmacies, and
pharmaceutical manufacturers.
Principal Health, Inc.5 4
Income Statements-Fiscal Year Ended December 31
Period Ending: 2002 2001 2000
Net Sales $10,254 $10,053 $9,856
Cost of Sales
-6.704 -6.623 -6.542
Gross Profit 3,550 3,431 3,314
Operating Expenses:
Selling Expenses -1,329 -1,454 -1,082
General and Administrative Expenses -1.501 -1.522 -1.541
Total Operating Expenses -2.830 .-2,976 -2.623
Operating Income 721 455 691
Total Other Income & Expenses (Net) 477 467 458
Earnings Before Interest And Taxes 1,197 922 1,149
Interest Expense -709 -745 -780
Income Before Tax 488 177 369
Income Tax Expense (35% tax rate) -171 d62 -129
Net Income $317
$115
$240
5 4 Note: All amounts in the financial statements are in millions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Balance Sheets-
As of December 31
Period Ending
Current Assets
2002 2001 2000
Cash And Cash Equivalents
$810 $776 $800
Short Term Investments
651 632 652
Net Receivables
1,688 1,578 1,562
Inventory
4,578 4,120 4,325
Other Current Assets
1.362 1.273 1.312
Total Current Assets
9,089 8,379 8,651
Property Plant And Equipment
6,780 7,354 9,254
Goodwill
562 605 650
Other Assets
2.516 2.352 2.198
Total Assets
$18,948 $18,689 $20,753
Current Liabilities
Accounts Payable
$3,158 $2,209 $2,145
Short Term And Current Long Term Debt 906 880 854
Total Current Liabilities
4,064 3,089 2,999
Long Term Debt 7,753 9,341 11,254
Other Liabilities 3,125 3,467 4,177
Total Liabilities $14,942 $15,897 $18,430
Stock Holders Equity
Common Stock $500 $500 $500
Capital Surplus
1,750 1,650 1,500
Retained Earnings 1,756 642 323
Total Stockholder Equity 4.006 2.792 2.323
Total Liabilities & Stockholder Equity
$18 948 $18,689 $20,753
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Statements of Cash Flows- December 31 _
Period Ending: 2002..2001 2000
Net Income
$317 $115 $240
Cash Flow Operating Activities
Depreciation 568 564 452
Other Adjustments To Net Income -1,452 -900 -452
Change In Accounts Receivables -110 -16 -92
Change In Inventories -458 205 -289
Change In Liabilities 975 90 84
Change In Other Current Accounts 89 =39 79
Cash Flow From Operating Activities -$71
m
$22
Cash Flow Investing Activities
Purchase of Property & Equipment -$540 -$254 -$1,858
Purchase of Investments -124 -452 -230
Other Cash Flow From Investing Activities 895 435 1.595
Cash Flow From Investing Activities $231 -$271 -$493
Cash Flow Financing Activities
Additions to Long-Term Debt $895 $450 $654
Reductions of Long-Term Debt -665 -634 -100
Repurchase of Stock -1,006 305 -125
Other Cash Flows from Financing Activities 650 107 139
Cash Flow From Financing Activities -$126 $228 $568
Change In Cash And Cash Equivalents $34 =$24 $97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
144
Part 1 Directions (continued):
You are free to look back at the company background information and the financial
statem ents when answering the following questions about potential results in 2003.
1. Considering the company background and financial statement information presented
concerning Principal Health, Inc. (PHC), what do you estimate the 2003 Net Income will
be?5 5
$ ; .
W rite in a num ber
2. Please use the scale below to indicate how confident you are o f your estimate o f the
2003 Net Income in (1) above. (Place an X on the point o f the scale that corresponds to
your answer.)
0 1 2 3 4 5 67 8 9 10
0% 50% 100%
C onfident C onfident C onfident
5 5 For questions that involve a calculation, you may either approximate or use a calculator. If you
would like to use a calculator and do not have one, please raise your hand and a calculator will be
provided to you.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
145
Part 2
Read the information on this page and the
following pages only after you have read and
completed Part 1.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
146
CORPORATE R ESEA R CH B Y TREVO R S.
STANFORD
PRINCIPAL
HEALTH: THE
PROGNOSIS IS
GOOD!
SEVERAL REASONS W HY PHC SHOULD
BE A STRONG PERFORMING COMPANY
IN 2003
erving manufacturers of
S
pharmaceuticals and medical
products, Principal Health is
gaining strength in the health
care industry. Information
gleaned from current press releases
indicates several reasons why Principal
Health should be able to deliver value
to shareholders.
R reason 1: Recently, PHC
announced the completion of a
successful offer for the outstanding
common stock of Taylor, Inc. Taylor,
Inc. is a provider of integrated medical
education to the health care industry.
Taylor’ s educational focus is well suited
for PHC’ s current strategy of moving
into the medical education market.
Services and products that Taylor, Inc.
provides include content development,
strategic consultation, logistics
management, accredited and non
Health Industry News * September 2003
accredited program development, and a
variety of related Internet-based
solutions. The highly profitable offerings
of Taylor Inc. include: sales force
education meeting support, accredited
and non-accredited symposia, large-
scale video conferences, patient
services, web conferencing, and
publication planning.
R
eason 2: A substantial positive
impact on the financial statements
should follow from the recent acquisition
of Harris International Corp., a leader in
retail pharmacy services. "With
PHC’ s management structure, the
Harris International pharmacy services
divisions will be well positioned
strategically to achieve long-term
earnings growth," stated Phil D. Genis,
chief financial officer at PHC. Harris
International possesses many attributes
that complement PHC’ s business
strategy such as: a unique customer
mix, high productivity, and a rich variety
of cross-selling opportunities. These
important attributes should add mightily
to PHC’ s bottom line in the near future.
R
eason 3: PHC reports steady
performance in the Pharmaceutical
Distribution and Provider Services
Segment (PDPS), which comprises half
the Company’ s operating earnings. The
strong financial performance in the
Segment is due to record quarterly
highs in revenues, operating earnings
and return on sales. Disciplined
expense control and a strong customer
mix also contributed to the PDPS’ s
operating performance. Thus far in the
fiscal year, total revenues have
increased 27 percent, caused by
increased growth in the pharmaceutical
distribution and specialty distribution
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
businesses. Increasing revenue growth,
continued vendor margin programs, and
improved productivity, including the
realized benefits from the acquisition of
Emma W est Industries, were the main
drivers of the strong financial
performance.
R
eason 4: The Consulting Services
segment, representing one-fifth of
the Company’ s operating earnings,
segment is also producing another
exceptional year. Strong demand for
patient safety applications and the
supply automation product line is a
central reason behind the performance.
Increased demand for automation
products has driven revenues up
twenty-four percent over the first two
quarters. In addition, demand has been
increasing for bedside technology
products such as PatientPlus, an
integrated medical information system.
Relative to last year, gross margins
have risen substantially as a result of
product mix and manufacturing
efficiencies. The Consulting Services
Segment’ s strong margin performance
and productivity have driven operating
earnings up relative to 2002.
Trevor S. Stanford, CFA,
CPA, is a New York-
based money manager
contributing to Health
Industry News.
Health Industry News * September 2003
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
148
September
2003
Mike R.
Kerstetter
CFA, CPA, is
a Chicago-
based
money
manager
contributing
to Medical
Industry
Report.
Medical Industry Report
Good News for Principal Health
Shareholders!
Principal Health, Inc. (PHC), a provider o f products and
services supporting the health care industry, is gaining strength in
the health care industry. According to company press releases,
there are several recent developments that bode well for Principal
Health shareholders.
PHC recently announced the successful acquisition o f
Harris International Corp., a leader in retail pharmacy services that
should add substantial positive financial statement benefits. "With
PHC’s management structure, the Harris International pharmacy
services divisions will be well positioned strategically to achieve
long-term earnings growth," stated Phil D. Genis, chief financial
officer at PHC. Indeed, Harris possesses a strong blend o f unique
customer mix, productivity, and cross-selling opportunities that
should contribute heavily to PHC’s future earnings growth.
The tender offer for the outstanding shares o f common
stock o f Taylor, Inc. was successful. Founded in 1981, Taylor, Inc.,
d/b/a Taylor Group Companies, provides integrated medical
education to the health care industry, a niche that is a solid fit with
PHC’s current strategy o f moving into the medical education
market. Taylor services include strategic consultation, content
development, accredited and non-accredited program development
and execution, logistics management and a variety o f internet-based
solutions related to these services. Taylor’s core services maintain
high profit margins and include symposia, advisory boards, sales
force education meeting support, large-scale video conferences,
patient services, publication planning, and web conferencing.
PHC reports strong performance in its Consulting Services
Segment which represents twenty percent o f PHC’s operating
earnings. Accelerated demand for patient safety applications and
the supply automation product line is an important factor driving
the performance. Revenues have increased 24 percent over the first
two quarters on increased
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
149
demand for existing proprietary automation products. Furthermore,
demand has been steadily increasing for bedside technology
products such as PatientPlus, an integrated medical information
system. Compared to fiscal year 2002, gross margins have been
increasing substantially as a result o f favorable product mix and
manufacturing efficiencies. Segment productivity and the strong
margin performance are driving operating earnings up compared to
last year.
Another PHC business Segment performing well is the
Pharmaceutical Distribution and Provider Services Segment
(PDPS), which comprises half the Company’s operating earnings.
The positive financial performance in the PDPS segment is due to
first and second quarter record highs in revenues, operating
earnings and return on sales. A good customer mix and disciplined
expense control also contributed to the Segment’s operating
performance. To date, total revenues have increased 27 percent,
caused by increased growth in the pharmaceutical distribution and
specialty distribution businesses. Strengthened revenue growth,
continued vendor margin programs, and improved productivity,
including the realized benefits from the takeover o f Emma West
Industries, were the primary reasons for the Segment’s earnings
performance.
With its high-quality products and positive recent trends in
performance, PHC is well positioned as a player in the medical
products and services market and should provide their shareholders
will strong value for years to come.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
150
Part 2 (continued) Directions:
You are free to look back at any information w hen answ ering the following
q uestions about potential results in 2003; however, p lease do not ch an g e
any an sw ers that you have previously given.
1. Considering the company background, financial statements, and press article
information presented concerning Principal Health, Inc. (PHC), what do you estimate the
2003 Net Income will be?
2. Please use the scale below to indicate how confident you are o f your estimate o f the
2003 Net Income in (1) above. (Place an X on the point o f the scale that corresponds to
your answer.)
$
W rite in a num ber
0
0%
C onfident
2 3 4 5 6 7
50%
C onfident
8 9 10
100%
C onfident
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
151
Part 3
Read the information on the following pages
only after you have read and completed Parts
18s 2.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
152
Part 3 Directions:
Please read and complete the following questions. (Place an X on the point of the
scale that corresponds to your answer.)______________________________________
As part of the case materials, you read PRESS ARTICLES5 6 about Principal
Health. Please answer the following questions concerning the press articles.
1. The information concerning Principal Health, Inc. contained in the press articles seemed reliable.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
2. The press articles were relevant for making an earning forecast for Principal Health.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
3. The press articles you read contained financial news pertaining to Principal Health, Inc. Do you
agree that this information was dependable?
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
4. The information contained in the press articles made a difference in my estimation of the annual
earnings for Principal Health.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
5 6 Depending on your form, you may have read one or two press articles. If you read only one press
article, simply answer the questions as if they read “press article” rather than “press articles”.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
153
5. In the current case materials, the press articles represent information available pertaining to
Principal Health, Inc. Do you agree that this information was credible?
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
6. The information in the press articles on Principal Health helped me to form a prediction for
annual earnings.
1 111. 1___I___I___I___I___I___I
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
7. The information presented in the press articles seemed consistent.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
8. The information contained in the press articles was pertinent when forming an earnings prediction
for Principal Health.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
9. I believe that the editors of the journals in the case materials know more information about
Principal Health than what appears in the articles and convey this fact to readers through the act
of selecting the articles.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
154
10. I understood the information presented in the press articles.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
11. 1 believe that the authors of the journal articles contained in the case materials know more
information about Principal Health than what appears in the articles and attempt to convey this
fact to readers through the act of choosing certain companies to feature.
-5 -4
Strongly
Disagree
-3 -2 - 1 0 1 2 3
Neither
Agree nor
Disagree
4 5
Strongly
Agree
12. The information in the press articles was difficult to comprehend.
I I I I I I I I I .. 1 I
-5 -4
Strongly
Disagree
-3 -2 - 1 0 1 2 3
Neither
Agree nor
Disagree
4 5
Strongly
Agree
13. The author of a financial press article highly recommends investing in a Principal Health. The
author supports the recommendation by providing facts in the article. However, there may be
“additional” information that doesn’t make it into the article. A reader of the article assumes that
this “additional” positive information exists based on the author’s choice to feature the company.
To what extent do you agree with the reader’s assumption?
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor
Disagree
Agree
14. The press articles contained pessimistic news about Principal Health, Inc.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor
Disagree
Agree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
155
15. There is a press release issued by Principal Health that contains positive information concerning
recent company events. An industry journal editor decides to assign a writer to summarize the
press release into an article on this positive information of Principal Health. It is likely that the
editor was prompted to take these actions because the editor: (1) possessed “extra” information
(information in addition to the press release about Principal Health), and (2) wished to
communicate this “extra” information to readers of the journal.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
16. The information contained in the second press article was redundant with the information
contained in the first press article.2
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
17. Overall, the Principal Health, Inc. press articles contained good news.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
18. The information contained in the first press article was very similar to the information contained
in the second press article.3
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
2 If you only had one press article in your materials, skip this question.
3 If you only had one press article in your materials, skip this question.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
156
P l e a s e a n s w e r t h e f o l l o w i n g g e n e r a l q u e s t i o n s :
19. Company A issues a press release that informs the public of a positive earnings announcement.
This press release is written up in Newspaper 1 and Newspaper 2. In this scenario, the fact the
information appeared in two newspapers rather than only one newspaper makes the information
more r e l i a b l e for decision-making purposes.
1____1 1 1 1 1 1___I___I___I____1
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
20. Company B issues a press release that informs the public of a positive earnings announcement.
This press release is written up in Newspaper 1 and Newspaper 2. In this scenario, the fact the
information appeared in two newspapers rather than only one newspaper makes the information
more r e l e v a n t for decision-making purposes.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
21. If a piece of information A is repeated in the media 4 times, and a piece of information B, is
repeated in the media only 1 time, then this is a good indication that Information A is
approximately 4 times more r e l i a b l e than Information B.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
22. If a piece of information A is repeated in the media 4 times, and a piece of information B, is
repeated in the media only 1 time, then this is a good indication that Information A is
approximately 4 times more r e l e v a n t than Information B.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
157
23. When the same story appears in the media multiple times, this repetition indicates that the
information is more reliable than if the story appeared only one time.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
24. When the same story appears in the media multiple times, this repetition indicates that the
information is more relevant than if the story appeared only one time.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
25. Imagine that two equally accurate home pregnancy tests are available, Brand A and Brand B.
Although the brands are equally accurate and valid tests, the brands use slightly different
processes to determine pregnancy. Each pregnancy kit contains only one test and both tests are
extremely easy to use. Your friend informs you that she has purchased Brand A and the result
was positive, indicating pregnancy. Your friend wants to make very sure of the accuracy of the
first test result, so she consults you regarding the purchase of a second test.
Please evaluate the following statement: If your friend wants to be very sure that she is truly
pregnant the best choice for her is to purchase Brand B for the second test as opposed to
purchasing Brand A again.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
26. I believe that journal editors in general know more information about companies than what
appears in the articles and convey this information to readers through the act of selecting the
articles.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
158
27. Suppose that you have an investment advisor, Mary, who has recommended that you purchase
ABC Company stock. However, before you make this purchase, you want to consult with one
other investment advisor from a different firm about the choice of ABC stock. At the new firm,
you meet two investment advisors, Mike and John. After talking to both advisors you have
determined that Mike is very similar to Mary on the dimensions of education, training, and
research methods. John, on the other hand, is very different from Mary on these same dimensions.
Assume that despite their differences in education, training and methods, Mike and John are
equally competent advisors.
Please evaluate the following statement: The best choice for a second opinion on the ABC stock
purchase would be John as opposed to Mike.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
28. Company A issues a press release that contains positive information. An industry journal editor
decides to assign a writer to summarize this press release for an article on Company A. The
editor probably took this action because the editor possessed “additional” information
(information in addition to the press release about Principal Health), and wished to communicate
this “additional” information to readers of the journal.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
29. Assume that you wish to research information on the purchase of a digital camera. As part of
your search, you read an article in the magazine Best Buy Electronics that evaluates various
cameras and their features. Based on this article, you decide to purchase the XI0 Camera. Just to
verify that the XI0 is truly a wise purchase, you decide to read an additional article that evaluates
digital cameras. Assume that your choices are limited to either reading another article by a
different author in a new issue of Best Buy Electronics or reading another article by a different
author in a competing magazine, Great Purchase Electronics. Furthermore, assume that both
magazines are equally credible and well written.
Please evaluate the following statement: To assist you in your camera purchase you should
choose to read the article in the new issue of Best Buy Electronics, as opposed to the article in
Great Purchase Electronics.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
159
30. I worked hard when reading the company information, financial statements, press articles and
then completing the related questions.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
31. The press articles concerning Principal Health, Inc. seemed unusual compared to typical articles
that one might find in the financial press.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
32. The author of a financial press article highly recommends investing in shares of XYZ stock. The
author supports the recommendation by providing facts about XYZ in the article. However, there
may be “additional” information that doesn’t make it into the article. A reader of the article
assumes that this “additional” positive information exists based on the author’s choice to
recommend the company. To what extent do you agree with the reader’s assumption?
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
33. The press articles concernine Principal Health. Inc. seemed realistic.
1 1 1 1 1 1 1 1 . 1
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
34. 1 concentrated as much as 1 possibly could while reading the company information, financial
statements, press articles and then completing the related questions.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission
160
35. The press articles presented in the case materials had a formatting and writing style similar to real
articles in the financial press.
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
36. What constitutes an item of information can be defined quite narrowly or more broadly. An
example of a narrow definition of information could be a single fact such as a division’s
operating income figure, the CEO of company, or the name of a product. Alternatively,
information could be defined more broadly as the launch of a new product line, the takeover of a
company, or the overall performance of a division.
Assuming the broad definition of an item of information, how many independent items
of information were contained in the set of press articles?
W rite in a num ber
A s p a r t o f t h e c a s e m a t e r i a l s , y o u r e a d P r i n c i p a l H e a l t h ’ s C o m p a n y B a c k g r o u n d
i n f o r m a t i o n a n d F i n a n c i a l S t a t e m e n t s . P l e a s e a n s w e r t h e f o l l o w i n g q u e s t i o n s :
37. Generally, how positive or negative was the Principal Health, Inc. company background and
financial statement information?
-5 -4 -3 -2 - 1 0 1 2 3 4 5
Extremely Neither Extremely
Negative Positive nor Positive
Negative
38. Overall, the Principal Health, Inc. company background information and financial statements
contained good news.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Strongly Neither Strongly
Disagree Agree nor Agree
Disagree
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
161
Parts 4 & 5
Read the information on the following page
only after you have read and completed Parts
1- 3 .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
162
Part 4 Directions:
Please read through the information and complete the questions regarding the case
materials. Please do not refer back to any of the materials when answering the
questions.______________________________ . _______ __________
1. In the information provided, PHC was attempting a takeover o f , a provider
of integrated medical education to the health care industry.
a. Taylor Inc.
b. Kerstetter Inc.
c. Frybarger Inc.
d. Stanford Inc.
2. According to the information provided, a substantial impact on the financial statements
should follow from the recent acquisition o f____________ .
a. Health Inc.
b. Harris International.
c. Genis International.
d. Montgomery Inc.
3. According to the information provided, the Pharmaceutical Distribution and Provider
Services Segment (PDPS) engaged in a previous takeover o f _____________ Industries?
a. Emma West
b. Northwest
c. Embler
d. Jackson
4. The__________ Segment is currently reporting large changes in revenue and earnings, which
is significant given that this segment comprises half of the PHC’s operating earnings.
a. Medical Education.
b. Consulting Services.
c. Hospital Management.
d. Pharmaceutical Distribution and Provider Services.
5. According to the information provided, the Consulting Services Segment is producing a
medical information systems product called___________.
a. Medical Plus
b. Pharma Med
c. Patient Plus
d. Pharma Plus
6. The Chief Financial Officer of PHC is:
a. Phil Genis
b. Tracy Davis
c. Charles Jackson
d. Chuck Fetzer
7. How did the authors of the press articles compile their information on Principal Health Care:
a. Independent research
b. Company press releases
c. An investment advisory service
d. Financial analysts
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
163
Part 5 Directions:
This is the final section. In order to help us better understand why your responses
might differ from those of other participants in this study, please answer the
following questions. (As with the rest of the form, all of your answers are completely
anonym ous.)_____________________ _ _ _ _ _ _ _ _ _ __
1 Have you ever made investments in the common stock of a
company?
If so, approximately how many stock trades (buy or sell) have
you made in the last twelve months?
7
8
Have you ever read an article in the business press? (Examples
of business press would include The Wall Street Journal,
Business Week, and also the Business Section of any major
newspaper such as the Los Angeles Times.)
If so, approximately how many articles have you read in the last
twelve months?
Prior to this study, have you ever analyzed the financial
statements of a company for any purpose?
How many Accounting classes have you taken (including those
you are taking this semester)?
How many Finance classes have you taken (including those you
are taking this semester)?
How many years of full time work experience do you have?
What is your gender ?
Please record the time that you completed the materials, e.g.,
10:35AM (use the clock on the wall or your watch).
YES NO
W rite in a n um ber
Do you plan to invest in the common stock of a company at any YES NO
time in the future?
YES NO
W rite in a num ber
YES NO
W rite in a num ber
W rite in a num ber
W rite in a num ber
Male
Female
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Information diffusion and the boundary of market efficiency

PDF

Does adding nonfinancial value drivers to a summary financial measure improve the learning and performance of managers?

PDF

Customer satisfaction, customer bargaining power, and financial performance

PDF

Beyond skill development: The effects of training and development on the attitudes and retention of employees

PDF

Essays on technological evolution and financial returns to innovation

PDF

Consumer choice under budget constraint: Why consumers overspend

PDF

Essays in supply chain management

PDF

Essays on new product preannouncements

PDF

Essays on technology innovation and new product development strategy

PDF

Demographic diversity, team process, and team performance: Assessing moderator effects of cognitive conflict management practices and task interdependence

PDF

Influencing individual innovation through technology features that support cross -departmental understanding

PDF

Board involvement in monitoring and strategy making: Antecedents and consequences

PDF

Androgyny and transformational leadership: Effects of gender and sex -role identity in the collectivistic context of Taiwan, R.O.C.

PDF

Defining the advice -seeking request in electronic solicitations in boundary spanning environments

PDF

Commercial loan default risk assessment: Predictive accuracy of alternative accounting structures in an inflationary environment

PDF

An empirical investigation of everyday low price (EDLP) strategy in electronic markets

PDF

Corporate ethical orientation and crisis management before and after September 11, 2001

PDF

Optimism and accuracy in forecasts of financial analysts: The effects of earnings predictability and experience

PDF

Feeling good about control: Design considerations for accountability systems in schools

PDF

In church with my ancestors: The changing shape of religious memory in the Republic of Armenia and the North American diaspora

Asset Metadata

Creator Hugon, J. Artur (author)

Core Title Effects of redundancy in media coverage on nonprofessional investors' earnings forecasts

School Graduate School

Degree Doctor of Philosophy

Degree Program Business Administration

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

Tag business administration, accounting,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Bonner, Sarah (committee chair), Beatty, Randy (committee member), Manegold, James (committee member), Miller, Norman (committee member), Walther, Beverly (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-390816

Unique identifier UC11340226

Identifier 3180777.pdf (filename),usctheses-c16-390816 (legacy record id)

Legacy Identifier 3180777.pdf

Dmrecord 390816

Document Type Dissertation

Rights Hugon, J. Artur

Type texts

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

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA