USF Libraries
USF Digital Collections

The impact of management's tone on the perception of management's credibility in forecasting

MISSING IMAGE

Material Information

Title:
The impact of management's tone on the perception of management's credibility in forecasting
Physical Description:
Book
Language:
English
Creator:
Slater, Robert D
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Source credibility theory
Intention
Decision-making
Public Accounting Oversight Board
Credibility scale
Auditing Standard No. 2
Dissertations, Academic -- Business Administration -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: The purpose of this study is to examine the impact of management altering its tone in communications on participants' perceptions of management credibility. Management's tone in communicating with participants was manipulated using communications from management under two treatment conditions. In period one of the study management's tone was manipulated within the management statement on internal controls as required by the Public Company Accounting Oversight Board's (PCAOB) Auditing Standards No. 2. In period one, participants had no knowledge of management's prior forecasting accuracy. Consistent with predicted hypotheses, the findings reveal that management can increase its credibility with participants by communicating its empathy, responsiveness, and understanding. Management's increased credibility was measured using both a validated credibility scale and by examining participants' reliance on management's forecasts. In period two of the study all participants had knowledge of management's forecast failure in period one. The results from period two found that tone could impact the rating of management's credibility when management had previously failed to meet a forecast but that tone had no impact on participant's changes in their earnings per share estimates after management had previously failed to meet a forecast.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2007.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Robert D. Slater.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 189 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001967732
oclc - 264637849
usfldc doi - E14-SFE0002139
usfldc handle - e14.2139
System ID:
SFS0026457:00001


This item is only available as the following downloads:


Full Text

PAGE 1

The Impact of Management’s Tone on the Perception of Management’s Credibility in Forecasting by Robert D. Slater, Jr. A dissertation submitted in partial fulfillment of the requirement s for the degree of Doctor of Philosophy School of Accountancy College of Business Administration University of South Florida Co-Major Professor: Jacqueline Reck, Ph.D. Co-Major Professor: Uday Murthy, Ph.D. Stephanie Bryant, Ph.D. Ellis Blanton, Ph.D. Date of Approval: April 30, 2007 Keywords: Source Credibility Theory, Intention, Decision-Making, Public Accounting Oversight Board, Credibilit y Scale, Auditing Standard No. 2, Copyright 2007, Robert D. Slater, Jr.

PAGE 2

Dedication I would like to dedicate this dissertat ion to my wife Beth who persevered with me through this entire process. Be th’s never ending understanding, support, and belief that I could finish were inspirat ional. I would also like to dedicate this work to Dr. Jim Hunton encouraging me to join the PhD program. This work is also dedicated to Dr. Jackie Reck and Dr. Uday Murthy both of whom took the reins from Dr. Jim Hunton and inspired me to keep going. This work is also dedicated to the many PhD students from the University of South Florida who inspired me while making this journey. To Dr. Tonya Benford and Dr. David Hayes for setting t he seeds for what may be the most collegial group of graduate students ever. Thank you to Dr. Cindy LeRouge for helping me to cope with the first semester. I would also like to dedicate this work to Anita Reed and thank her for becoming a best friend along the way. This work is dedicated to the late Rosalyn Mansour Rosalyn was only with us for a short time on this journey but she was an insp iration to everyone. I also hope this dissertation inspires the current student s at USF: Ann Dzuranin, John Chan, Norma Montague, and Chris Jones to race to the end of their path as fast as possible but to enjoy ever minute of the journey.

PAGE 3

Acknowledgments I would like to thank the co-chairs of my dissertation committee, Dr. Jackie Reck and Dr. Uday Murthy for helping me i mmensely to complete this document. Their time and dedication in mentoring me helped me to succeed. I would also like to thank the other two members of my committee, Dr. Stephanie Bryant and Dr. Ellis Blanton for their support and comments. Stephanie’s door was always open and she provided lots of motivation and encouragement. I would like to thank Dr. L. Rene "B ud" Gaiennie for establishing the Gaiennie Fund which financially supported my dissertation. I would also like to acknowledge the financial support I re ceived from the Dr. Henry Efebera scholarship fund. I would like thank my family for neve r giving up on me, or at least never publicly giving up on me! Thank you to Robert and Karen Slater Sr. for being the best parents anyone could ask for. To my brother Eric Slater, thank you for always calling to check up on me and reminding me to “keep in touch…” I would also like to thank my in-laws Teddi and Paul Wengert for their gracious support and understanding. I would also like to thank my cat Nikki for patiently sitting by my side late into many nights as I worked on this dissertation. Most of all I would especially like to thank my wife Beth fo r helping me to endure the trials and tribulations we have faced to complete this degree. Thank you all!

PAGE 4

i Table of Contents List of Tables ........................................................................................................iii List of Figures .......................................................................................................vi ABSTRACT .........................................................................................................vii 1.0 Introdu ction.....................................................................................................1 2.0. Theoretical Background and Hypotheses De velopment................................5 2.1 Credibility and Managem ent Foreca sting.............................................5 2.2 Credibility and Investor Belief Re vision...............................................8 3.0 Research Method.........................................................................................29 3.1 Introducti on........................................................................................29 3.2 Research Desi gn...............................................................................29 3.3 Ta sk...................................................................................................31 3.4 Partic ipants........................................................................................36 3.5 Measured Va riables...........................................................................36 3.6 Pilot Studies .......................................................................................44 4.0 Main Study Results.......................................................................................63

PAGE 5

ii 4.1 Introdu ction........................................................................................63 4.2 Partic ipants........................................................................................63 4.3 Manipulati on Checks..........................................................................64 4.4 Data A nalysis .....................................................................................68 4.5 Hypothesis Testing..........................................................................103 4.6 Post Hoc Analysis ............................................................................122 5.0 Summary and C onclusion ...........................................................................130 5.1 Summary of Study...........................................................................130 5.2 Limitati ons.......................................................................................136 5.3 Contributio ns...................................................................................139 5.4 Future Re search..............................................................................142 5.5 Conclu sion.......................................................................................143 Referenc es.......................................................................................................144 Appendice s.......................................................................................................150 Appendix A: Moderating Fa ctors That Influence Credi bility or Its Impact on Belief Re vision.............................................................................150 Appendix B: View of Experimental Materials Used for Pilot St udy....................159 Appendix C: Pilot Study O ne Surprise Te sting.................................................188 About the Author............ .................. .................. ................ ............... .......End Page

PAGE 6

iii List of Tables Table 1: Overview of Steps in Study..................................................................32 Table 2: Reliance on Management’s Fo recast (H2) and Difference in Reliance on Management’s Fore casts Calculat ions (H4)...................42 Table 3: Pilot Study O ne Surprise Analysi s........................................................46 Table 4: Means and Standard Deviati on for the Seven-point Likert Scale Ratings of Surpri se and Signif icance........................................47 Table 5: Comparison of Statistics for Management’s Internal Control Letter and Earni ngs Letter..................................................................51 Table 6: Main Pilot Study Results fo r Sarbanes-Oxley Internal Control Letter..................................................................................................56 Table 7: The Results of the Main P ilot Study for the Earnings Results Letter..................................................................................................58 Table 8: Reduction in Credibility Between Periods O ne and Tw o.......................61 Table 9: Demographic Co variate Q uestions .......................................................71 Table 10: Level of Involv ement with Task Questi ons..........................................74 Table 11: Reporting Bias and Knowledge Bias Questions.................................79

PAGE 7

iv Table 12: Recent Event Ques tions .....................................................................83 Table 13: Covariates Correlated with Credibility Ra ting (H1)..............................86 Table 14: Preliminary ANCOVA Te sting of Credibility Rating (H1) Using Cova riates................................................................................87 Table 15: Covariates Correlat ed with Reliance on Management’s Forecast (H2).....................................................................................88 Table 16: Preliminary ANCOVA Te sting of Reliance on Management’s Forecast (H2) Us ing Covari ates.........................................................88 Table 17: Covariates Correlated with the Difference in Credibility Ratings (H3).......................................................................................89 Table 18: Preliminary ANCOVA Test ing of Difference in Credibility Ratings (H3) Usi ng Covari ates...........................................................89 Table 19: Calculation of the Useful Covariate by Period One Treatment Conditio n............................................................................................92 Table 20: Descriptive Statistics for the Useful Covariate by Year Two Treatment C onditio n...........................................................................93 Table 21: Covariates Correlated with the Difference in Reliance on Management Foreca sts (H 4)..............................................................93 Table 22: Preliminary ANCOVA Te sting Difference in Reliance on Management Forecasts (H4) Using Cova riates..................................94 Table 23: Correlation Between D ependent Variables by Peri od.........................95

PAGE 8

v Table 24: Results for Internal C ontrol Letter’s Impact on Credibility Ratings and Reliance on M anagement’s Fo recast.............................96 Table 25: Descriptive Statistics for Credibility Rating (H 1)...............................105 Table 26: ANCOVA Results for Inte rnal Control Letter’s Impact on Credibility Rati ngs (H1) .....................................................................107 Table 27: Descriptive Statisti cs for the Reliance on Management’s Forecasts (H2)..................................................................................110 Table 28: ANCOVA Test of Reliance on Management’s Fore cast (H2)...........112 Table 29: Descriptive Statistics for the Difference in Credibility Ratings from Period One to Period Two (H3)................................................116 Table 30: ANCOVA Test of Difference in Credibility Ra tings (H3)....................117 Table 31: Descriptive Statistics for the Difference in Reliance on Management’s Foreca sts (H4) .........................................................120 Table 32: ANCOVA Results for Difference in Reliance on Management’s Foreca st (H4)...........................................................121 Table 33: Descriptive Statistics for Sub-Factors of Cr edibilit y..........................124 Table 34: Summary of Hy potheses and Fi ndings.............................................135

PAGE 9

vi List of Figures Figure 1. Investor Beli ef Revision Model............................................................13 Figure 2. Factors Affecting Managem ent Forecasts (From Davidson and Nue, 1993)..........................................................................................22 Figure 3. Eagly, Wood and Chaiken (1978)...................................................... 151 Figure 4. Moderating Variabl es to Credi bility....................................................155 Figure 5. Okeefe's Level of Involv ement Affect on Credibili ty...........................156 Figure 6. Timing of Identify ing the Communi cator............................................157 Figure 7. Position of Mess age..........................................................................158

PAGE 10

vii The Impact of Management’s Tone on the Perception of Management’s Credibility in Forecasting Robert D. Slater Jr. ABSTRACT The purpose of this study is to examine the impact of management altering its tone in communications on participants’ perceptions of management credibility. Management’s tone in co mmunicating with participants was manipulated using communications fr om management under two treatment conditions. In period one of the study management’s tone was manipulated within the management statement on internal controls as required by the Public Company Accounting Oversight Board’s (P CAOB) Auditing Standards No. 2. In period one, participants had no knowledge of management’s prior forecasting accuracy. Consistent with predicted hypotheses, the findings reveal that management can increase its credibility with participants by communicating its empathy, responsiveness, and underst anding. Management’s increased credibility was measured using both a validated credibility scale and by examining participants’ reliance on management ’s forecasts. In period two of the study all participants had knowledge of management’s forecast failure in period one. The results from period two found that tone could impact the rating of

PAGE 11

viii management’s credibility when management had previously failed to meet a forecast but that tone had no impact on participant’s changes in their earnings per share estimates after management had prev iously failed to meet a forecast.

PAGE 12

1 1.0 Introduction The stock market has placed a great deal of importance on companies meeting earnings estimates, with stock pr ices often dropping for companies that fail to meet estimates. In addition to a drop in share price, management can lose credibility for failing to meet project ed forecasts (Williams 1996; Mercer 2005). Credibility plays an important role in m anagement’s ability to signal the market about its expected earnings and the mark et’s beliefs about management’s earnings signal (Williams 1996; Hirst et al. 1999; Mercer 2001). The market will align with management’s beliefs when it re ceives signals from management that it perceives as more credible (Verrecchia 1983; Verrecchia 1990). Therefore, it is important to study ways in which management credibility can be enhanced, allowing management to communicate with the market in a more efficient manner. Developing an understanding of the factors that affect management credibility will allow us to gain a be tter understanding of how management is able to communicate its beliefs about earnings to the market. However, there is some debate in the literature about t he factors that make up credi bility. The most recent studies have assumed that credibility is a two factor construct consisting of trustworthiness and expertise (Hirst 1994; Mercer 2004; Mercer 2005), while there is evidence from the persuasion liter ature that credibility is a three factor

PAGE 13

2 construct that includes as the third factor perceived intentions of the communicator (McCroskey and Teven 1999). The objective of this research is to test whether management’s credibility is impacted by investors’ perceptions of management’s intentions toward them. Also, in an attempt to reconcile the theoret ical model of source credibility from McCroskey and Teven (1999) with the findings from accounting studies (Jennings 1987; Williams 1996; Hirst et al. 1999; Mercer 2005) the impact of changes in management’s credibility on par ticipants’ judgments will be examined. Specifically, this study examines part icipants’ ratings of management credibility and their reliance on information provided by management. To investigate participants’ percepti ons of management’s credibility, this research manipulates management’s tone in written communications and measures the ensuing impact on partici pants’ perceptions of management’s credibility. The research investigates whether changi ng the tone in written communications allows management to alter participants’ perceptions of management’s intentions; thereby, increas ing perceived credibility and also the degree to which participants rely on m anagement’s forecast guidance (Hovland et al. 1953; McCroskey and Teven 1999). The research also investigates whether management’s loss of credibility for failing to meet a projected forecast of earnings per share can be mitigated by the tone management uses to convey the news to shareholders in a written communication. The study manipulates management’s tone in two communications with participants under different circumstanc es. The first manipulation of

PAGE 14

3 management’s tone will be delivered with the financial statements, in management’s statement about the company’ s internal controls, as mandated by the Public Company Accounting Oversi ght Board’s (PCAOB)(2004) Auditing Standard No. 2, “An Audit of Internal C ontrol over Financial Reporting Performed in Conjunction with an Audit of Financial Statements.” This is a new standard that calls for management to report to invest ors on the internal controls of its company. While the report on internal controls is a mandatory communication management must make to investors, t he wording of the report has not been specified by the PCAOB. This report wa s selected for study because it presents an opportunity to examine the possible e ffect different wording of a newly mandated report may have on in vestors’ judgments. Management’s tone is also manipulat ed after participants receive actual financial results for which management has inaccurately forecasted. As stated previously, management’s credibility shoul d be reduced when it fails to meet a forecast (Jennings 1987; Hirst 1994; W illiams 1996; Mercer 2005). The tone in management’s letter informing participants about the actual results of the quarter will be manipulated and the effects will be measured on both the perception of management’s credibility (measured by a cr edibility rating) and on the reliance on information supplied by management (meas ured by participants’ reliance on management’s forecasts). The results of this research should inform policymakers, such as the PCAOB, as to the effect different wording can have in communicating with investors. As indicated, Auditing Standard No. 2 (PCAOB, 2004, §163), leaves

PAGE 15

4 the exact wording of the management statement on internal control up to management, stating that the report c an “take many forms” as long as management states a direct conclusion about the effectiveness of internal controls. If it is found that subtle wo rding differences in required reports can impact investors’ perceptions, then the PC AOB may want to consider restricting the wording that could be used in such reports. The remainder of this study proceeds as follows. Chapter 2 reviews the theoretical background of the study and develops the proposed hypotheses. Prior studies investigating management credibilit y in forecasting will be examined in reference to a model of credibility fr om persuasion studies (O'Keefe 1990; McCroskey and Teven 1999; Perloff 2003). Formal hypotheses will be developed based on the theoretical model proposed. Chapter 3 explains the research method used to test the hypotheses proposed in Chapter 2. Also in Chapter 3, each of the variables in the study is defined. A formal research model is presented as well as the proposed statisti cal analyses of each of the hypotheses. The results of three pilot studies are also discussed. In Chapter 4, H1 through H4 are statistically tested and results are pr esented to support the findings from the study. After analyzing the predicted resu lts a post hoc analysis is conducted to examine the credibility construct further Chapter 5 summarizes the findings of the study and presents the conclusions drawn from the analysi s of the study. Limitations of the study are also present ed in Chapter 5, as well as the overall contributions of the study.

PAGE 16

5 2.0. Theoretical Background and Hypotheses Development This chapter begins with a discussion of the impact of credibility on management forecasts. A formal model of the tw o factors (surprise and credibility) that have been found to affect investors’ belief revisions, with regard to management forecasts, is presented. Each of the factors (surprise and credibility) is discussed in general, and pr ior studies that have examined these factors are presented. One of the factors of this model – credibility – will be further analyzed within the context of source credibility theory (Hovland and Weiss 1951; O'Keefe 1990). A three fact or model of credibility is discussed (McCroskey and Teven 1999), as are two meas ures of credibilit y – the impact of management’s past forecast accuracy and its tone in communications with investors. Based on source credibility theory, four hypotheses are then proposed regarding the effects of altering the t one of communications with investors and how the tone can influence management’s perceived credibility and the amount by which investors rely on management forecasts. 2.1 Credibility and Management Forecasting Firms are required by the Securiti es and Exchange Commission to report certain financial information to investors. Additionally, many firms go beyond the mandated disclosures and make voluntary di sclosures to investors. One example of a voluntary disclosure is a management earnings forecast.

PAGE 17

6 Voluntary disclosures, such as ear nings forecasts, allow managers to share knowledge that is not already ava ilable to investors, and thereby reduce the amount of information asymmetry (Ajinkya and Gift 1984). For management to determine whether to release volunt ary information to in vestors, management needs to have an understanding of t he true value of the information,1 the costs of disclosing the information, and an underst anding of investors’ expectations. Management’s optimal threshold level of disclosure is simultaneously dependent on these three variables (Verrecchia 1983; Kim and Verrecchia 1991). Some analytical models of disclosure hold that once management reaches the disclosure threshold it always discloses trut hfully, since it may face lawsuits if disclosures do not match the actual resu lts (Hughes 1986). Research findings looking at management forecast discl osures are mixed as to whether management is truthful (i.e., unbiased) in its disclosures. Some studies have looked at management’s earnings forecast s and compared the forecasts to the actual results for the period and f ound that management’s forecasts were positively biased (Penman 1980; Waym ire 1984; Clarkson et al. 1992; Mcconomy 1998). While other studies, such as the one done by McNichols (1989), found no evidence that earnings fo recasts were systematically biased, positively or negatively, but did find t hat investors did not take management forecasts at face value. McNichol’s fi nding indicates that investors perceived management’s forecasts as biased, or lacking some credibility. 1 The “true value of the information” is the understanding by management of the economic and competitive advantage of the in formation (Verrecchia 1983).

PAGE 18

7 For the purposes of this study, whet her or not management forecasts are actually biased is not as important as the finding that investors perceive management’s forecasts as biased, and disc ount them. Finding th at investors do not react as if management discloses honestly does not invalidate Verrecchia’s findings that management’s optimal thres hold level is depende nt on the type of information, disclosure costs, and curr ent market expectations. Instead the threshold level of disclosure shifts to a value that can include additional disclosure costs to establish management ’s credibility with investors. The increase in disclosure costs shifts managem ent’s optimal level of disclosure to a point where greater information asymme try must be present before management benefits from disclosure. This study investigates the effe ctiveness of a low cost method of influencing management’s perceived cr edibility. Management’s goodwill or perceived intentions, a factor affect ing credibility, could potentially be manipulated at relatively low cost. Incr easing management’s di sclosure credibility with a low cost option such as modifying the tone of written communications with investors increases the quality of managemen t’s disclosure, thereby, lowering the threshold at which management chooses to reduce information asymmetry (Verrecchia 1990).

PAGE 19

8 2.2 Credibility and Investor Belief Revision 2.2.1 Investors’ Reliance on Earnings Forecasts Several empirical models offer a basis for determining how users incorporate management forecasts into their own beliefs. These models all contain management credibility or a similar construct (Patell 1976; King et al. 1990; Williams 1996; Hirst et al. 1999; Mercer 2001). The goal of this study is not to redefine these models into one single unified model; instead the goal of this study is to examine management credibili ty, one of the main factors of these models. It is generally held t hat the amount by which a user’s belief is revised is a combination of the credibility of m anagement in making the disclosure and the surprise element of the information contai ned in the disclosure. Jennings (1987) proposes that the amount of belief revisi on is modeled as an interaction between management’s credibility and the newness (sur prise) of the information contained in the forecast. The Jennings model is consistent with findings in accounting research on forecasts (Williams 1996; Me rcer 2001), and models of credibility from the psychology literature (Hovl and and Weiss 1951; Hovland et al. 1953; Hovland and Pritzker 1957; O'Keefe 1990; McCroskey and Teven 1999; Perloff 2003). The current study adapts the Jennings ( 1987) model, in that it models credibility and surprise as affecting invest or belief revision. However, since the current study holds the degree of surpri se constant for a ll participants, the interaction of surprise with credibility is not modeled. In the following section I

PAGE 20

9 briefly discuss the notion of “surprise” before going on to discuss belief revision and credibility, which are the constructs of interest in this study. 2.2.2 Information Surprise The surprise of the information contai ned in the forecast is the difference between the investor’s current level of belief based on the information set currently held and the new information in t he forecast. Surprise is a measure of the degree of information asymmetry between management and investors. Surprise also represents the maximu m belief revision management expects to generate with its disclosure, since the pur pose of its disclosure is to bring investors’ expectations in line with its own. Management’s expectation that it can change beliefs is supported by studies in psychology that have found that a portion of a subject’s opinion change is a function of the di fference between the current beliefs a subject has and the beliefs advocated by a communicator (Ewing 1942; Hovland and Pritzker 1957). Greater opinion changes occur when the difference between the subject’s bel iefs and the communicator’s advocated message is larger. Surprise has al so been referred to as the degree of conformity.2 The current study does not manipulate surprise between participants but instead seeks to hold it rela tively constant across participants in order to examine the impact of credibility. 2 The degree of conformity can be an opinion in the same direction (pro-attitudinal) or in an opposite direction (counter-attitudinal). For exam ple, both the receiver and the communicator can have the same pro-attitude toward a message but have a different level of belief.

PAGE 21

10 2.2.3 Belief Revision Belief revision in this study is defined as the difference between the current level of beliefs held by partici pants before receiving new information from management and the participants’ revised le vel of beliefs after receiving new information from management. Be lief revision will equal the amount of surprise if the participants fully revise their belie fs to the new information provided by management. According to so me analytical models of disclosure, the current level of investors’ beliefs is one of the factors management must identify when deciding to make a disclosure (Verrecchia 1983; Kim and Verrecchia 1991). Disclosure costs increase as the di fference between management’s disclosure and the market’s current level of belief increases. These costs can be actual costs paid for assurance on disclosures, the cost of compiling the information, costs associated with the loss of proprietary information, and indirect costs such as those related to credibility. Measuring the market’s current level of belief is difficult to do. Prior accounting studies looking at management disclosures have used proxies for the market’s current level of belief. One such proxy was the stock price of a security before management released its informa tion (McNichols 1989). Other studies have used earnings per share estimates of analysts, and proxy the market’s current level of belief as the composit e analyst forecast (Kasnik and Lev 1995). Behavioral studies looking at disclosure have given participants base information, such as the analysts’ composite forecast, to proxy for the market’s level of belief (Libby and Tan 1999; Mercer 2005). Gi ving participants base information is a

PAGE 22

11 weakness in prior studies because it forc es participants to integrate the given base information with their true current level of belief. Since their true or initial level of belief is not known, this leads to an anchoring and adjustment (Tversky and Kahneman 1974) effect that is not captured in the study. Hirst et al. (1999) tried to contro l for differences between participants’ initial levels of beliefs by using the partici pants’ initial predictions as a covariate in the data analysis. However, they also gave all of the participants (in each treatment) the same information regarding actual earnings. Therefore, to the extent that the difference between the initial predicti on and actual earnings would vary across participants, the degree of surprise would have varied across participants potentially conf ounding the results. This study takes a unique approach to controlling for the interaction of the participants’ current level of beliefs and management’s disclosure. Instead of controlling for the current level of beliefs by giving participants a starti ng point, participants will be able to select their initial earnings per share estimate, and the actual earnings number released by management will be revealed as a set percentage increase over each participant’s initial prediction, thereby hol ding the degree of surprise relatively constant across participants. The approach taken in this study removes the possibility of an unmeasured anchoring effect that may have taken place in prior studies. 2.2.4 Credibility The factor affecting belief revision that is of interest in this study is the communicator’s credibility. Cr edibility is a measure of the perceived believability

PAGE 23

12 of a communicator, where the percepti on of credibility is made by a message recipient (perceiver) (O'Keefe 1990). Credibili ty is not a trait t hat can be directly observed. The amount of perceived cr edibility held by any one communicator may vary from one message recipient to another (O'Keefe 1990). Some message recipients will find one communicator highly credible while other recipients do not (O'Keefe 1990). This is easily seen in t he realm of politics where one candidate can be seen as quite credible by his/her fo llowers but not very credible by those who support his/her opponent. Only one prior accounting study has looked at management’s credibility using a psychological model similar to the one used in this study. Management credibility is modeled by Mercer (2005) as the trustworthiness and expertise of management, which is consistent with pr ior models of credibility from the persuasion literature (O'Keefe 1990; Pe rloff 2003). Mercer (2005) looked at management’s failure to warn investor s of a negative news surprise, and measured the failure’s impact on inve stor’s perception of management’s credibility. The study found management’s actions did im pact credibility ratings; however, no assessment was made on how t he change in credibility impacted investors’ reliance on management’s forecast s. Participants were asked if they would rely on future disclosures, and those who were in the higher credibility manipulation (i.e., the ones who were warn ed) did state they would rely on future disclosures. Therefore, while accounti ng research suggests that credibility is composed of more than one factor (see Figure 1), the link between credibility ratings and investor judgments remains unt ested. To establish the link between

PAGE 24

13 perceived credibility and invest or judgments (specifically in a belief revision task) I use source credibility theory as applied in psychology and marketing. Source credibility theory posits that in most circumstances, the more credible a communicator, the more likel y the communication will elicit change of beliefs in message recipients (Hovland and Weiss 1951; Hovland et al. 1953; O'Keefe 1990; DeZoort et al. 2003; Perloff 2003).3 In Figure 1, as the perceived credibility increases, so does the amount of investors’ belief revision. Management’s Internal Control Letter: -High Intention Tone -No Intention Tone Management’s Earnings Results Letter: -High Intention Tone -No Intention Tone Perceived Credibility: Expertise Trustworthiness Intention Perceived Credibility: Expertise Trustworthiness Intention Investor’s Belief Revision (Change in EPS Predictions Period One, Before and After Management’s Prediction) Change In Investor’s Belief Revision (Change in EPS Predictions Period Two, Before and After Management’s Prediction) H 1 H 2 H4 Difference in Change in Investor’s Belief Revision From Period One to Period Two H3 Change in Perceived Credibility From Period One to Period Two Period OnePeriod Two No Prior Information About Management’s Forecast Accuracy Management’s Forecast Inaccuracy for Period One is Known Figure 1. Investor Belief Revision Model 3 The circumstances in which more credibility does not equal more opinion change are discussed in detail in Appendix A as moderators to the model of credibility.

PAGE 25

14 As indicated, support for the impact of perceived credibility on belief revision is found in marketing and psyc hology research (G otlieb et al. 1992; Goldsmith et al. 2000; Lafferty et al. 2002). Goldsmith et al (2000) examined both corporate credibility and endorser cr edibility and found that both impact users’ intentions to purchase products The endorser credibility impacted the users’ attitude toward the advertisement, wh ich in turn affected their intention to purchase, and attitude toward the brand. Ho wever, corporate cr edibility directly impacted users’ intention to purchase, as well as attitude toward the brand and attitude toward the ad (Goldsmith et al. 2000). Lafferty et al. (2002) tested the results of Goldsmith et al. (2000) and found support for corporate credibility directly impacting intention to purchase as well as the other paths found in Goldsmith et al. (2000). In se veral disciplines, the “intent ion to act” construct is used as a proxy for actual behavior based on the theory of planned behavior (Ajzen 2001). This link between perceived cr edibility and user behavior provides support for the belief that investors’ per ceptions of management’s credibility can impact investors’ decision-making. 2.2.5 Perceived Management Credibility Factors In order to understand w hat makes a communicator credible it is important to understand the factors that affect per ceived credibility. Most studies of credibility use a two-factor m odel of credibility similar to the one used in Mercer’s (2005) study, with the two factors being expertise and trustworthiness (McGinnies and Ward 1980). This study employs the McCroskey and Teven (1999) model of credibility, which in cludes three factors: expertise,

PAGE 26

15 trustworthiness and intentions. Since th is model has not been used in judgment and decision making research, a factor analysis was conducted to examine the three factors. Using an oblique rotati on, the factor analysis suggested the 18 questions loaded on three unique factors4. Sixteen of the questions loaded on the correct factor, while two questions lo aded higher on a factor other than their expected variable. The three original variables from McCroskey and Teven’s (1999) model were used in this study. No adjustment was made to the model for the two variables that loaded on a different factor in the factor analysis. Rather than examining the three fact ors individually, the main analyses in this study use all 18 questions of the credibility scale to measure total credibility. Thus, the discrepancy in the factor loadings does not impact the analyses in this study but should be addressed in future research. To determine if the three sub-factors were measuring the same higher level cr edibility factor a Cronbach’s Alpha was calculated for the responses to the scale in the study from period one and period two. The three sub-factors of cred ibility had a Cronbach’s Alpha of .912 for period one and .926 for period two. 2.2.5.1 Expertise Hovland et al. (1953) define expertness as “the extent to which a communicator is perceived to be a source of valid assertions” (pg. 21). The more a perceiver believes someone is an expert in his/her field, the more credible the perceiver will find the communicator’s messages as they pertain to that particular 4 Two factors had Eigenvalues above 1 (11.959 and 1.706) with a third factor having an Eigenvalue of .97.

PAGE 27

16 field. Prior research has found many differ ent attributes can signal that someone is an expert (Hovland et al. 1953). Hovland et al. (1953) identify age, position of leadership, and similarity between perceiv er and communicator as factors that may invoke a perception of expertise. The research to date (Hovland and Weiss 1951; Hovland et al. 1953; O'Keefe 1990; Perloff 2003), indicates that communicators who are perceived as havin g greater expertise will be found to be more credible. The percepti on of expertise should be held constant in period one of the study. However, the manipulation in period two should cause expertise to drop since management fails to meet its forecast. 2.2.5.2 Trustworthiness Hovland et al. (1953) define trustworth iness as “the degree of confidence in the communicator’s intent to communica te the assertions he considers most valid” (pg. 21). It is possible for a communicator to be an expert in his/her domain but decide to communicate statemen ts known to be invalid (Hovland et al. 1953). Message recipients must form an opinion as to whether they believe an expert communicator is communicati ng truthfully. Those communicators perceived to be communicating truthfully are thought to have gr eater credibility. The perception of trustworthiness should be constant in period one of the study. However, the manipulation in period two should cause trustworthiness to drop since management fails to meet its forecast.

PAGE 28

17 2.2.5.3 Goodwill (Intention Toward the Receiver) Hovland et al. (1953) also discussed a thir d factor of credibility, which they called “intention toward the receiver.” As stated previously, trustworthiness measures the degree of confidence a receiv er has in a communicator’s intention. The intention toward the receiver constr uct is the belief a receiver has about the communicator’s intention for communicati ng an assertion. Intention toward the receiver measures the belief about the communicator’s intentions, and trustworthiness measures the degree of c onfidence in the belief. McCroskey and Teven (1999) propose that the construct’s goodwill and intentions towards the receiver are the same; thus, the third fa ctor in their model of credibility is goodwill. Goodwill has been excluded in some credibility models over the years because researchers believed that goodw ill could not be m easured properly (McCroskey and Teven 1999). McCroskey and Teven (1999) blame the poor measurements on factor analytic m odels that included extraneous “person perception” variables that hindered results. In this study, I will use McCroskey and Teven’s (1999) validated instrument to measure goodwill. However, I use the term “intention” or “intention toward the receiver,” since the term “goodwill” has another connotation in the field of accounting. (In accounting, goodwill is a measure of the purchase price of a company over its fair value.) There are three distinct element s to the intention factor: understanding, empathy, and re sponsiveness (McCroskey and Teven 1999). Understanding is defined as knowi ng another person’s ideas, feelings,

PAGE 29

18 and needs. Empathy is an acceptance of another’s view as valid. Responsiveness is viewed as the a ttentiveness of one to another. 2.2.5.4 Other Factors That Can Impact the Effect of Credibility There are other factors that hav e been identified in the psychology literature that are believed to moderate credibility’s impact on belief revision. For example, the expected and actual posit ion advocated by the message have been found to affect the perception of credibi lity factors (Eagly et al. 1978). How well liked the communicator is can affect cr edibility (Heider 1958; McCroskey 1966). Contextual factors such as the message re cipient’s level of involvement with the topic (a combination of expertise and motivation), the degree of difference between current receiver beliefs and t he beliefs presented in the message, and timing of identifying the communicator have been found to im pact the effect credibility has on a message recipient’s belief revision (O'Keefe 1990). Because these factors are theoret ically important, they will be included as possible covariates in the experiment; however, only the intention factor of credibility is measured in the current study. See Appendix A for a discussion of the moderating factors. 2.2.6 Measuring Credibility Factors While only one other accounting study (Mercer 2005) has tested credibility as a perception variable, other studies in accounting have explored the concept of credibility. Most studies, even in p sychology, have difficulty separating and manipulating the first two factors of credibility, ex pertise and trustworthiness

PAGE 30

19 (O'Keefe 1990). This is especially true in accounting studies, where one measure for management credibility is management’s past forecast accuracy (Williams 1996; Hirst et al. 1999). Past accuracy in forecasting is a noisy measure of expertise or trustworthi ness since participants do not know if management failed to make its forecast bec ause it did not have the expertise to forecast correctly, or because it was intentionally misleading investors. 2.2.6.1 Forecast Accuracy Information As indicated in the prior paragraph, ex tant accounting studies looking at the relationship between m anagement’s credibi lity and investors’ reactions to forecasts have used management’s prior forecast accuracy as a proxy for management’s forecast credibility (Baginski and Hassell 1990; Williams 1996; Hirst et al. 1999; Mercer 2001; Merc er 2004; Mercer 2005). Some studies assumed that investors’ reactions to an earnings forecast are a function of management’s credibility and the surprise or newness of the information being presented (Jennings 1987; Baginski and Ha ssell 1990; Williams 1996). Mercer’s studies (2001, 2004, 2005) measured managem ent’s credibility as a perception variable with two factors, expertise and tr ustworthiness. Consistent with research findings that it is difficult to separate expertise and trustworth iness, Mercer found expertise and trustworthiness highly corre lated variables that move together. Hirst et al. (1999) found evidence that credibility, as measured by prior forecast accuracy, was a significant factor in the earnings predictions of investors who used a management forecast in their decision-making. Archival studies have also explored the relationship between management’s prior accuracy in

PAGE 31

20 forecasting and the reaction of analysts to subsequent forecasts (Hassell et al. 1988; Williams 1996). These studies differ only in the methods used and measurement of the variables. For exam ple, Hassell et al.(1988) measured the effect of management earnings forecasts on the revision in earnings estimates of analysts, where the amount of change in analyst forecasts was dependent on management’s credibility as measured by t he difference in earnings forecasts to actual earnings. They used the ex-post a ccuracy (actual accuracy) of previous forecasts to define management’s credibilit y. They found that analyst forecasts did change after the release of a managemen t projection. The study implies that analysts’ beliefs were revised based on the credibility (prior accuracy) of management forecasts. Williams (1996) also studied prior forecast accuracy of management and operationalized prior accuracy as prior fo recast usefulness, where usefulness is measured by whether a user would have been better off adjusting expectations of earnings to management’s forecast in a prior period. The relationship between the usefulness of a prior earnings foreca st and analysts’ responses to a current forecast was studied. Exogenous variables, such as timing of the forecast and market price reactions, were controlled. As predicted, management’s credibility (as determined by its prior forecasting abi lity) affected the way in which analysts reacted to current forecasts (Williams 1996). Therefore, when management’s prior forecasts were accura te, investors’ perceptions of management’s credibility increased. The increases in perceived credibi lity lead to analysts relying on future management forecasts when making their own forecasts.

PAGE 32

21 Benjamin and Strawser (1976) conduc ted a behavioral study that looked at prior forecast accuracy. They gave investors annual reports that had earnings forecasts contained within the notes to the financial statem ents. Participants were told of the prior year’s forecast and actual results. Prior forecast accuracy was found to increase investors’ perce ived credibility of management’s current forecast, as measured by participants’ earni ngs per share predictions relative to management’s forecast. In the Benjamin and Strawser study, participants were given explanations as to why prior management foreca sts may have been inaccurate. These explanations included environmental va riables that would not be directly attributable to management. Thus, Benjam in and Strawser’s results may have been affected by beliefs that management was not to blame for failing to forecast correctly. Accordingly, t he effect on investor’s perc eptions of management being at fault for an inaccurate forecast is not known. The accuracy of an earnings foreca st is a function of management’s forecasting ability and the business ri sks facing the firm (Davidson and Neu 1993). Figure 2 models this relationshi p as presented in Davidson and Neu (1993). A manager can only make a forecast based on contemporaneous information available at the time he/s he forecasts. If management uses sound procedures and valid assumptions in fo recasting and still fails to forecast accurately, unforeseeable business risk fa ctors could be contributing to actual earnings deviating from the forecast. Ba sed on Davidson and Neu (1993), it is possible management’s prior forecast accuracy is a poor proxy for

PAGE 33

22 management’s credibility. Further complicatin g the validity of forecast accuracy as a proxy for credibility is fundamental attribution error theory. This theory predicts that people are more likely to attribute outcomes of events to dispositional factors rather than to sit uational factors (Heider 1958). Dispositional factors are internal factor s associated with a person and situational factors are outside the control of an individual. Accuracy of Earnings Forecasts Ability to Forecast Business Risk Figure 2. Factors Aff ecting Management Forecasts (F rom Davidson and Nue, 1993) Two important dispositional factor s discussed by Heider (1958) are intention and ability. Intention is som eone’s desire to do and act, and ability is the individual’s power to take action. These co nstructs are similar in nature to the expertise, trustworthiness, and intention fa ctors of credibility. Expertise is a level of ability. Intention is an external perception of someone’s desires, and trustworthiness is the belief that someone will or will not attemp t to act on his/her desires. The components of credibility (ex pertise, trustworthiness, and intention) are all dispositional factor s. Therefore, attribution theory (Heider 1958; Kelley 1973; Reagan et al. 1974; Wood and Eagly 1981) suggests that investors would blame management’s failure to forecast accu rately on internal factors, absent an external reason given for management’s fa ilure. Investors may believe that a

PAGE 34

23 missed forecast was due to such inte rnal factors as management’s lack of expertise in forecasting or management’s willingness to forecast accurately (trustworthiness), thus reducing management’ s credibility. Conversely, investors’ perceptions of management’s credibilit y increase when management’s prior forecasting accuracy increases (Jennings 1987; Williams 1996; Hirst et al. 1999). Williams (1996) found that analysts re lied more upon current management forecasts when management’s forecasts were more useful in the past. Management’s forecasts were considered mo re useful if, in the past, the new information management gave in its fore cast was more accurate than the analysts’ current estimates. For example, two firms could predict and achieve their earnings per share at exactly the same amount of $4.00 per share. However, if analysts for the two firms expected that earnings would be $3.75 and $3.50, respectively, then management’s foreca st was more useful in the case of the firm with the $3.50 analysts’ expecta tion. The Williams (1996) study was conducted using actual stock market forecast revisions made by analysts following an actual disclosure from a publicly traded company. Analysts’ forecast revisions were observed and management ’s prior forecast accuracy was measured based on management’s real prio r forecasts. Since the study was conducted using only archival data, there is no way to know if analysts actually used management’s prior forecast accu racy when making their revisions. As shown by the review of prior studies, management’s accuracy in prior forecasting has a large impact on manage ment’s credibility. The impact on management’s credibility in turn impacts the amount of investor belief revision

PAGE 35

24 when management issues a forecast. In this study, two condi tions of accuracy will be used. In the first condition participants will have no knowledge of management’s prior accuracy. In the second condition participants will not be told of management’s accuracy but they wil l be given a prior forecast from management that they will subsequently lear n is inaccurate; thus, experiencing the inaccuracy first hand. 2.2.6.2 Tone of Communications There is anecdotal evidence that management’s tone in communicating with investors can affect management’s perceived credibility with investors. Articles in recent public relations jour nals tout strategies that say management can increase or regain its credibility with investors by altering its tone in communications with investors, such as the company’s annual report (Budd 2000; Calvey 2001; O'Brien 2001; Thom pson 2002; Leckey 2003). At the same time, investor publications tell investors they can learn about a company from the tone a company takes in communicati ng with its sharehol ders (Rodgers 2002). Most shareholders will never physi cally meet the management of the companies in which they invest. Managem ent’s communications to shareholders are often the only direct contact shar eholders have with a company. It would be logical to assume that these writt en communications from management may be the best opportunity management has to co nvey its goodwill (intentions) toward investors. Thompson (2002) suggests that management can regain lost credibility by issuing communications to inve stors that “shoot stra ight.” Could it be this simple for management to regai n some of its lost credibility?

PAGE 36

25 The intention of this study is to te st management’s ability to increase its rating of credibility by communicating its in tention toward investors via a written communication that conveys managem ent’s understanding, empathy, and responsiveness. In period one, half t he participants will receive an altered statement of internal control letter from management intended to convey its understanding, empathy, and re sponsiveness (intentions) at a high level. The treatment group receiving the high leve l letter will be referred to as the “high intention” treatment while the other treatment will be referred to as the “no intention” treatment. Participants who perceive that management has better intentions toward them should rate management’s credibility higher than participants without such a perception.5 Therefore, H1 states: H1: Participants who receive an inte rnal control letter from management with a high intention tone will rate management’s credibility higher than participants who are given an internal control letter with no intention tone. Hypothesis 1 predicts that participant s will rate management’s credibility higher after receiving the internal cont rol letter from management seeded with a high intention tone. Although McCro skey and Teven (1999) have already established this link using the credibility scale, credibility is also known to impact participants’ reliance on management’s forecasts. In ac counting research, management’s perceived credibility is m easured by the amount of revision participants make when receiving earnings guidance from management, ceteris 5 In period one of the study, only the intention variable of credibility has been manipulated in the communication; therefore, no predictions are made regarding the expertise and trustworthy factors of credibility.

PAGE 37

26 paribus (Patell 1976; King et al. 1990; Willi ams 1996; Hirst et al. 1999; Mercer 2001). Based on source credibility theory, higher perceived management credibility yields a greater amount of reliance by participants on management’s forecast. The greater the amount of reliance on management’s information the more credible management is thought to be. Therefore, consistent with source credibility theory, H2 pr edicts that participants who are given written communications from m anagement (management’s internal control letter in period one of this study) with a high intention tone will per ceive management as having higher credibility, and thus will rely on management’s forecast by revising their earnings per share predictions clos er (greater revision) to management’s predictions.6 H2: Participants who receive an inte rnal control letter from management with a high intention t one will rely more on management’s forecasts by revising their earnings per share estimates closer to management’s predictions than partici pants who receive an internal control letter with no intention tone. 2.2.6.2 Tone of Communications versus Inaccurate Forecast The intention factor of credibility has not heretofore been studied in a judgment and decision-making setting. This study is designed to test the impact of management’s tone in communications on management’s credib ility, including the intention factor, and then to test t he ensuing effects on participants’ belief revisions. It is expected that the result s of period one will show that the tone 6 The treatment in period one is the tone of the communication by management to shareholders. This tone is expected to impact the investors’ perception of management credibility. The letter used in period one of this study (the statement on the effectiveness of internal controls) is not expected to impact investor’s in itial earnings per share estimates, only their perception of management’s credibility, and therefore their fu ture reliance on management’s forecasts.

PAGE 38

27 management has with participants can increase its credibility. What is not known is whether that increase in credibility can substitute or outweigh the other two factors of credibility: experti se and trustworthiness. In order to test whether the different factors of credibility are substitutes for one another, the study will manipulate management’s accuracy between a level where participants have no knowledge of management’s past accuracy (period one) and a level where they know management has forecasted inaccurate ly (period two). In period two of this study all of the participants have had a pr evious experience with management in which management failed to meet an earni ngs per share forecast. It is expected that the participants will reduce their ra ting of management’s ov erall credibility, including their rating of management’s expe rtise and trustworthiness. Participants will receive a letter from management notif ying them of management’s failure to meet its earnings per share estimate. T he tone of the letter will be manipulated at two levels; one with a high intention tone and one with the no intention tone. When management forecasts inaccurately its credibility should drop and participants should rely less on management ’s forecast. But if management’s tone in letters to shareholders increases it s intention factor of credibility when using a high intention tone, it should increase overall perceived credibility and allow management to effect ively persuade participants of management’s beliefs. Therefore, H3 predict s that management’s tone in communications can mitigate lost credibility ratings from fa iling to forecast accurately. Based on source credibility theory, H4 predicts that management’s tone in communications can also lead to greater reliance on future management forecasts.

PAGE 39

28 H3: Participants who receive an inaccurate forecast from management and receive an earnings notificati on letter from management with a high intention tone will not lower management’s credibility rating as much as participants who receive an inaccurate forecast from management and who are given an earni ngs notification letter from management with no intention tone. H4: Participants who are notified of an inaccurate forecast from management by an earnings notification letter that conveys a high intention tone will redu ce the amount they rely on management’s future forecasts less than partici pants who receive an earnings notification letter from managemen t conveying no intention tone.

PAGE 40

29 3.0 Research Method 3.1 Introduction In Chapter 2 a theoretical model was developed and hypotheses were proposed. This chapter details the testing of those hypotheses. A description of the research design, including a descrip tion of the task and participants is presented. The measured and manipulated variables in this study will then be discussed. The measured variables in clude the dependent variables and the covariates. The manipulated variables are the independent variables. Following the discussion of the variables, a discussion of the three pilot studies that tested the research materials is presented. The chapter ends with a discussion regarding the overall findings fr om the pilot studies. 3.2 Research Design All participants completed the ta sk under both a condition of no prior information about management’s forecast accuracy (period one) and under a condition where participants had know ledge that management had previously forecasted inaccurately (period two). The independent variable for all four hypotheses is the tone of the communications. Tone is a between participant factor (high intention and no intention) in both period one and period two. In period one, tone is manipulated using the statement on internal control effectiveness by management. In period two, tone is manipulated using a letter

PAGE 41

30 from management regarding ac tual earnings from period one. Four dependent variables are measured in this study In period one there are two dependent variables; the rating of management’s credibility and participants’ reliance on management’s forecasts. In period two of the study the same two pieces of information are collected as in per iod one but the dependent variables are constructed as the difference between t he measured variables in period one and period two. Therefore, the two dependent variables in period two are; 1) the difference in management’s credibility ra ting, and 2) the difference in reliance on management’s forecasts. Management’s credibility rating is m easured using a previously validated 18-question credibility scale (McCroskey and Teven 1999). In H1, the credibility scores between the two treatment groups (h igh intention statement on internal controls vs. no intention statement on in ternal controls) are compared. For H3, the difference in credibility ratings (credi bility in period two less credibility in period one) between the two treatment grou ps (high intention earnings letter and no intention earnings le tter) is compared. Participant reliance on management’s forecasts is a measure of the percentage change participants make in their forecasts after receiving management’s forecast. In H2 the percentage change in forecasts made by participants from before to after managem ent’s forecast is compared between treatment groups (high intent ion statement on internal c ontrols vs. no intention statement on internal contro ls). For H4 the variable measured is the difference in the percent change in forecasts made from period one to period two, which is

PAGE 42

31 compared between treatment groups (hi gh intention earnings letter and no intention earnings letter). 3.3 Task An Internet-based laboratory experiment was conducted. The experimental materials resided on a lo cal Web site. Participants signed up for scheduled experimental runs. A facilita tor read the study’s instructions and provided each of the participants with a computer disk. The disk contained a hyperlink to the Web page, which randomized each parti cipant into one of the four treatments in the study. The study’s task required particip ants to make earnings per share estimates after reviewing the background an d financial statements of a selected company. Participants also rated managem ent’s credibility using the McCroskey and Teven (1999) scale. The experimental task was adapted from pr ior research studies by Hirst et al. (1999) and Mercer (2005). Task mate rials regarding the company, including company background, products, and financial statements were directly adapted from Mercer (2005). The ta sk of predicting earnings pe r share was adapted from Hirst, et al. (1999). Hirst, et al. (1999) used this task to study the joint effects of management’s prior forecast accuracy and th e form of a financial forecast on participants’ judgments. .

PAGE 43

32 Table 1: Overview of Steps in Study 1. General instructions about the study 2. Participant informed consent 3. Introduction to task 4. Study background information 5. Company background information 6. Overview of company products 7. Internal control letter a. High intention b. No intention 8. Financial statements (i ncome statement and balance sheet) 9. Participants provide initial earnings per share estimate for period one 10. Participants are given management’s earnings per share estimate for period one 11. Participants revise their earnings per share estimate for period one 12. Participants complete the credibility rating for period one a. Expertise factors b. Intention factors c. Trust factors 13. Participants are given the financial results letter for period one (with actual earnings per share) a. High intention version of letter b. No intention version of letter 14. Company financial statements with period 1 actual results 15. Participants provide initial earnings per share estimate for period 2. 16. Participants are given manag ement’s EPS prediction for period 2 17. Participants revise their earnings per share estimate for period 2 18. Participants complete the credibility ratings for period two a. Expertise factors b. Intention factors c. Trust factors 19. Participants answer ma nipulation check questions 20. Participants answer covariate questions – fraud 21. Participants answer covariate questions – Sarbanes-Oxley 22. Participants answer demographic covariate questions 23. Participants answer possible covariate questions – prior investing experience 24. Participants answer theoretically derived covariate questions 25. Participants are allowed to prov ide feedback regarding their experience 26. Participants view the finished scre en which thanks them for participating

PAGE 44

33 Table 1 provides an overview of the steps of the task. A step by step progression of the study is presented in Appendix B with screen shots from the study, including descriptions of each step re quired of the participants. The steps referenced in this section refer to the corresponding steps in Table 1 Before they began the task, the par ticipants read general instructions (step 1) for the experiment and were given an informed consent form (step 2). After participants elected to participate in the study (the informed consent), they were given an introduction to the task and asked to assume the role of an investor evaluating the subject company (step 3). They were then given a short description of the company, its products, and the industry in which the company operates (steps 4-6). Participants were also given a management letter regarding the company’s internal controls over financ ial reporting pursuant to PCAOB Auditing Standard No. 2 (step 7). Half of the participants receiv ed this letter in a manner that conveyed a message of high int ention tone by management (step 7a). The other half of the participants received the required letter without the intention tone manipulation (step 7b). The participants then received three years of financial statement data (step 8). The balance sheets and income statements were identical for all treatment groups in stage one. Participants were told they were looking at this information as of January 1 of the current year (period one). Once the participants had re viewed all of the background material about the company (steps 4-6), the internal control letter at either a high intention (step 7a) or no intention (step 7b), and the financial

PAGE 45

34 statements (step 8), they were asked to predict the company’s earnings per share for the first year (step 9). The participants were then given management’s earnings per share forecast for period one (step 10). At this point in the study, none of the participants had any knowledge of management’s past forecast accuracy and were, therefore, in the no prior accura cy treatment. The predi ction of earnings per share made by management was aut omatically generated based on the participant’s earnings per share predicti on in step 9. The management forecast was 132 percent higher than the partici pant’s response from step 9. After receiving management’s forecast, partici pants were given a chance to adjust their prediction of earnings per share for period one (step 11). The adjustment made by participants is the dependent variab le for H2, reliance on management forecasts. The participants then comp leted the credibility measurement instrument (step 12 a,b,c). The credibility measurement in strument is an 11-point Likert scale with 18 questions designed to measure the three factors of credibility: expertise (step 12a), intent ion (step 12b), and trustworthiness (step 12c). Credibility rating is the dependent variable for H1 in the study. The participants were given a lette r from management (step 13) that notified participants of the actual result s for the year (period one). Participants were told the difference between what management predicted and actual results. The letter stated that management failed to meet its prediction of earnings per share by 17 percent (109 percent of the participant’s earnings per share). The tone of the letter from m anagement was determined by whether participants were

PAGE 46

35 in the high intention (13a) or no int ention (13b) tone-of -the-financial-letter treatment conditions. O ne half of the participants received in the letter a statement that conveyed a message of high intentions by management (step 13a). The other half of the participants received a letter that only stated the results for the year (period one) and the difference between what management predicted and actual results (step 13b). In the next stage of the study partici pants were given the financial income statement that include d the actual results for peri od one (step 14). The income statement reflected the fi nancial results the participants were told about in the management letter (see step 13a or 13b). Th e participants were then asked to make a prediction of earnings for the sec ond year (period two) (step 15) for the same company. After making their earnings per share predictions for period two, participants were given management’s earni ngs per share prediction for period two (step 16). Management’s forecast wa s 132 percent higher than the prediction made by the participants. After receiv ing management’s estimate, participants were given an opportunity to revise their forecast for the second year (step 17). Participant’s revised forecast for y ear two was used to calculate the dependent variable for H4, difference in reliance on management’s forecast. After revising their forecasted earnings per share, par ticipants were asked to complete the credibility measurement instrument fo r the second time (step 18a-c). The credibility rating in period two was used to calculate the dependent variable for H3, difference in credibility ratings.

PAGE 47

36 After finishing the main part of the study, the parti cipants answered several different types of questions includ ing: manipulation check questions (step 19), covariate questions regarding fraud (s tep 20), covariate questions regarding Sarbanes-Oxley (step 21), demographic ques tions (step 22), and other questions that were included in the study to identify possible covariates based on the participant’s past investing experiences (step 23) and theoretical models (step 24). Participants were also given a c hance to provide feedback about the study (step 25). Finally, participants were t hanked for participating in the study and asked not to discuss the study with ot hers until they had been debriefed (step 26). 3.4 Participants The participants for the main study are discussed in the Chapter 4. All participants for the pilot studies were students, and are discussed, respectively, with each pilot study. 3.5 Measured Variables This section describes each of the independent variables, dependent variables, and covariates that are c aptured and/or measured in this study. 3.5.1 Independent Variables The independent variables for this study are forecast accuracy and management’s tone in its letter to investor s (participants). These variables are discussed in the following sections. A third variable (surprise) will also be

PAGE 48

37 discussed, since it is theoretically import ant to the study; howev er, this variable is held constant in the study. 3.5.1.1 Forecast Accuracy Management’s forecast accuracy has been used as a proxy for management’s credibility (Williams 1996; Hir st et al. 1999; Mercer 2001; Mercer 2005). In this study management’s foreca st accuracy was manipulated at two levels, no information and inaccurate inform ation. In the no information condition none of the participants had knowledge of management’s prior accuracy. The no information condition occurs in period one of the task. The inaccurate condition in this study occurs in period two. In period two, all of the participants knew the earnings forecast from management in period one was inaccurate. In prior studies, parti cipants were only told that management had been accurate or inaccurate (Hirst et al. 1999). In this st udy, the participants are aware of management’s pr edictions in period one and the actual results of period one. A decision was made not to include a th ird level where participants know that management has been accurate in foreca sting. Prior studies have found that when management’s prior accuracy is high investors perceive management’s credibility as high and few other variables can increase management’s credibility to a higher level (Hirst et al. 1999).7 The high accuracy variable level was purposefully omitted from the study to reduce the size and complexity of the study. 7 This finding may be the result of a ceiling effect.

PAGE 49

38 3.5.1.2 Management’s Tone Management’s tone was manipulated at two different times during the study. There are two levels of management ’s tone manipulation; one is a high intention tone and the second is a no in tention tone. In the high intention treatments, the wording in the letters wa s consistent with m anagement displaying the three sub-factors of the intention factor of credib ility: understanding, empathy, and responsiveness toward participants. The letters can be seen in Appendix B (step 8 and 14). The first tone manipulat ion between groups occurred in the letter from management discussing the company’s internal controls over financial reporting, as required by the PCAOB Auditing Standard No. 2. This letter was included with the company’s background and historical fin ancial statements. While the internal control letter is required by the PCAOB, the exact wording of the letter has been left to management. This study manipulates a single paragraph of this letter to convey a high intention or no in tention tone to participants. The second tone manipulation is in the form of a letter from management disclosing the actual earnings of the com pany for the prior y ear (period one) of operations. Management’s tone is manipulat ed at two levels, high intention and no intention. 3.5.1.3 Surprise Surprise measures the difference bet ween participants initial expectation of earnings per share and m anagement’s prediction of earnings per share. The amount of surprise was set to 32 percent above participant s’ initial prediction for

PAGE 50

39 all treatment conditions. To obtain a su fficient treatment ef fect, 32 percent was selected based upon the resu lts of pilot test one. Although the surprise variable is held constant across treatment conditions, it is an important variable in comparing this research model to prior studies. Past studies have used proxies to measure the market’s initial level of belief. In behavioral studies such as Hir st et al. (1999), par ticipants were given analysts’ consensus estimates as an initial level of belief and then asked to make their estimates of earnings per share. While this method seems to give all participants a similar anchor, the partici pants still have some prior belief. The difference between participant’s prior leve l of belief and the anchor given to them cannot be measured in those studies. In this study, the participants gave their earnings per share estimate without bei ng given any confounding anchors. The treatments for the remainder of the study were then based on the participant’s initial earnings per share estimates. Basing the treatments on the participant’s initial le vel of belief eliminates another type of potential bias. Studies t hat use a constant dollar amount for surprise may introduce an unintended bias If one participant’s initial level of belief is $1.00 and management’s estimate is $1.50 the surprise is $.50, which is a 50 percent increase over the initial le vel of belief. Another participant may have an initial belief of $1.25 and when gi ven management’s estimate of $1.50 he/she only has a 20 percent increase over t he initial level of belief. In an attempt to eliminate this bias, this study uses relative percentages, instead of incremental adjustments measured in dollars.

PAGE 51

40 3.5.2 Dependent Variables The dependent variables for this study are management’s credibility rating (H1), participants’ reliance on management’s forecasts (H2), the difference in credibility ratings from period one to period two (H3), and the difference in reliance on management’s forecast from pe riod one to period two (H4). Each of these variables is discussed below. Va riables H1 and H3 are discussed first followed by H2 and H4. 3.5.2.1 Management Credibility Rating (H1) The first dependent variable captured in this study is management’s credibility rating. The credibility rating was measured using the McCroskey and Teven (1999) scale. The scale is lo cated in Appendix B (steps 12a,b,c & 18a,b,c). Eighteen questions are used in the scale to determine participants’ perception of management’s cr edibility. The scale measures credibility along three factors: (a) competence, (b) intention, and (c) trustworthiness. Management’s credibility rating is repor ted as the average of all 18 scale questions. 3.5.2.2 The Difference in Credibility Ratings (H3) Participants were asked to complete t he credibility scale twice. The first time participants complete the scale is after revising their earnings per share forecasts in period one. Participants were also asked to complete the scale after revising their forecast in period two. The difference in credibility ratings measures

PAGE 52

41 the difference between participants’ rati ngs of management’s credibility from period one to period two in the study. 3.5.2.3 Reliance on Management’s Forecast (H2) Reliance on management’s forecast measures the difference between the participants’ first earnings per share estimate and their revised earnings per share estimate after receiv ing management’s forecast. The participants are asked to forecast the earnings per share of the company in period one based on the background financial data about the company. They are then given management’s forecast of earnings per share for the same time period. Management’s fore cast will always be 13 2 percent of the earnings prediction made by the participant. Participants are provided a chance to modify their earnings per share es timate after receiving management’s forecast. Table 2, Panel A demonstrates th e measurement of reliance on management’s forecast. To measure the participant’s reliance on management’s forecast, first the amount of surprise in management’s forecast is calculated. As stated above, all participants make a fore cast for period one (EPS1) and receive a forecast from management (MEPS) t hat is 32 percent higher than the participant’s initial earnings per share es timate. Thus, all participants have a surprise of 32 percent of their initial fo recast. Surprise is the denominator in the calculation of reliance. As can be seen in Table 2, Panel A, a participant with an initial earnings per share estimate of $2.00 is given information from management with $.64 of surprise ([M EPSEPS1] or [$2.00 .32]).

PAGE 53

42 Table 2: Reliance on Management’s Forecast (H2) and Difference in Reliance on Management’s Forecasts Calculations (H4) Panel A: Reliance on Management’s Forecast (H2) as Measured by the Percentage Change in Earnings Per Share Estimate Revision EPS2 – EPS1 2.40 2.00 0.40 Surprise = MEPSEPS1 = 2.64 – 2.00 = 0.64 = 63% Panel B: Difference in Reliance on Management’s Forecasts (H4) Percentage Change in EPS Pe riod One – Percentage C hange in EPS Period Two* 63% 40% Reliance dropped by 23% EPS1: Initial earnings per share es timate made by participant. EPS2: Revised earnings per share es timate made by participant MEPS: Management's estimate of earnings per share for period The reliance on management’s forecast for period two is calculated exactly as the reliance on management’s forecast for period one. H4 only uses the difference between the two reliance measures; no hypothesis was made for the reliance on management’s forecast in period two. Once the amount of management’s surpri se is calculated, the amount of revision in the participant’s forecasts is calculated. The revision in a participant’s forecast is calculated by subtracting the initial earnings per share estimate (EPS1) from the revised earnings per shar e estimate (EPS2). For example, in Table 2, Panel A, the initial earnings per share estimate is $2.00 and the revised earnings per share estimate is $2.40. Thus the revision in earnings per share is $.40. After the revision and t he surprise have been calculated, the revision is divided by the surprise to determi ne what percentage of management’s new information was relied upon in revising the pa rticipant’s earnings forecast. In the example in Table 2, Panel A the revision of $.40 was divided by the surprise of $.64. In the example, the participant made a 63 percent adjustment to

PAGE 54

43 management’s new information. If the participant had revised his/her earnings per share estimates to $2.64, relianc e on management’s forecast would have been 100 percent. 3.5.2.4 Difference in Reliance on Management’s Forecast (H4) To calculate the difference in reliance on management’s forecasts (H4), the reliance on management’s forecast is measured twice in this study, once in period one and once in period two. So t hat reliance can be measured in period two, participants will be asked to give an earnings-per-share prediction for period two. As in period one, participants will be given management’s forecast of earnings per share for period two, wh ich will again be 32 percent higher than participants’ estimates. After receiving management’s earnings per share forecast for period two, participants will be given a chance to modify their forecast. Panel A of Table 2 provided a demonstration of how the reliance on management’s forecast is calculat ed in period one. The reliance on management’s forecast in period two is ca lculated the same way. No hypotheses are given regarding the reli ance on management’s forecast in period two, instead this study examines the difference in reliance on management’s forecasts (H4) by subtracting the period one reliance measure from the period two reliance measure. Panel B of Table 2 demonstrates this calculation using the numbers for period one as calculated in Panel A. Assume in period two the participant had a reliance on management’s fo recast of 40 percent. T herefore, the difference in reliance on management’s forecasts is 23 percent (63 perc ent – 40 percent). This signifies a 23 percent drop in reliance.

PAGE 55

44 3.5.2.5 Covariates Data about possible covariates were captured and tested. A thorough discussion of the covariates measured and tested in this study appears in Chapter 4. The task in this study requi red participants to review financial statements and forecast earnings per share estimates. The covariates discussed in Chapter 4 include the par ticipants’ prior experience with financial statements or forecasting, their level of investing activity, their background (major), and gender. Other possible covariates were theoretic ally derived from persuasion studies and are discussed in both Chapter 4 and Appendix A (O'Keefe 1990). 3.6 Pilot Studies Before the main study was run, three separate pilot studies were conducted. Two pilot studies were us ed to design and test some of the experimental treatments used in the study The third pilot te st was conducted to test the overall research instrument and the effect of the treatments on the dependent variables. 3.6.1 Pilot Study One The first pilot study was designed to test the level of surprise necessary to generate a sufficient size effect. T he pilot study was conducted using 19 students enrolled in an accounting info rmation systems course at a large southeastern university. Each participant was given four treat ments which were comprised of different statements regarding differ ences between management and analysts’ forecasts. They were then a sked to use a seven-point Likert scale

PAGE 56

45 to rate the difference between analysts’ forecasts and management’s as to the significance of the difference and how surprising they found the difference. Neither significance nor surprise was defined for participants. Both terms (significant and surprising) were sele cted to determine how large management’s earnings per share estimate would have to be to create a perceived significant difference. Participants were asked if they felt the difference between management’s forecast and the analysts’ fo recasts was significant and then were allowed to draw their own conclusion. Appendix C contains the complete questionnaire used in pilot study one. A decision was made to use analysts ’ forecasts in the experimental materials because participants were not given background data regarding the company or financial statement s. Additionally, it would not be feasible to tell the participants what their forecast was and use that as a proxy for their true beliefs. Instead, analysts’ forecasts were used to examine the amount of difference needed between a believable external fore cast and management’s forecast. Management’s forecast varied between the four treatments as did the percentage of surprise (Tabl e 3). Forecasts were set at four different dollar amounts with the surprise varying fo r each amount. The dollar amounts and the surprise are indicated in Table 3:

PAGE 57

46 Table 3: Pilot Study One Surprise Analysis Management’s Forecast Percentage of Surprise Analyst Forecast Difference $1.23 10% $1.35 $0.12 $1.55 20% $1.86 $0.31 $1.75 30% $2.27 $0.53 $1.10 40% $1.54 $0.44 The dollar amount of management’s fo recasts and the percentage of surprise were arbitrarily selected to determine an approximate percentage at which participants would f eel there was a significant difference between management’s forecast and the analyst s’ forecasts. The amount of management’s forecast increases over the first three treatments, up to $1.75, and in the fourth treatment the amount of the forecast decreases to $1.10. This was done to examine whether the surpri se was being generated by the dollar amount of the difference or by the percentage difference. An ANOVA was performed to determi ne if the changes in the difference between management’s forecast and the anal yst’s forecast had an effect on participants’ ratings of surprise and signi ficance on the seven-point Likert scale. Overall, the change in percentage differ ences was significantly associated with surprise (F=11.52, two-tailed p =.001) and significance (F= 10.23, two-tailed p= .001). Table 4 presents the means and st andard deviations of the surprise and significance questions for each of the treatments.

PAGE 58

47 Table 4: Means and Standard Deviation for the Seven-point Likert Scale Ratings of Surprise and Significance Treatment N Surprise Mean (Standard Deviation) Significance Mean (Standard Deviation) 10% 19 3.263 (1.147) 4.368 (1.383) 20% 19 4.000 (1.105) 5.526 (1.172) 30% 19 5.526 (1.218) 6.368 (0.831) 40% 19 4.947 (1.615) 5.894 (1.197) Scale end points were 1 = Insi gnificant to 7 = Significant A Scheffe’s test was used to determine if there was a significant difference between the treatment groups. For the significance variable, the 10 percent group was significantly different from all other treatments. No other significant differences were found. For the surpri se variable, there were no significant differences among groups. The lack of si gnificant differences among treatment groups may have been due to the design of the instrument. The 40 percent treatment resulted in a lower dollar di fference than the 30 percent treatment. Therefore, the results were re-computed eliminating t he 40 percent treatment. In this analysis, the 30 percent treatment was significantly different from the remaining two groups for both measures. After consideration of the result s of this pilot study, 32 percent 8 was selected as the difference between participants’ initial forecast and management’s forecast for the main pilot study (pilot study 3). 8 Although 30% was the amount tested, 32% was used to limit participant’s ability to guess that management’s forecast was a percentage of t heirs. If participants predicted an even number like $1.00, they may figure out by period two that management’s prediction is exactly 30% larger.

PAGE 59

48 For the purposes of this study it wa s important to have surprise between management’s forecast and participants’ prio r expectations. There is no correct amount of surprise. All of the participa nts in the main study got the same percentage of surprise. The differenc e between treatments was examined using a percentage calculation of change. 3.6.2 Pilot Study Two The second pilot study was conducted to examine the power of the tone manipulations on participants’ ratings of management’s intention factor of credibility. As previously discussed, ther e are two separate tone manipulations in the study, each at two levels. One of the tone manipulations occurs in management’s statement about the company’ s internal controls, as mandated by the PCAOB Standard No. 2. The second t one manipulation occurs in a letter from management discussing the disappointing results from the prior year (period one). Twenty-one students were given management’s internal control letter and a letter from management statin g the prior year’s financi al results. They were also given the six questi ons from the credibility sca le measuring perceived intention (Appendix B, step 13b). There were two versions of each letter, one with a high intention tone and one with no intention tone. Each student received only a single instance of each letter. For example, a single student could have received an internal control letter wit h the high intention tone and a financial

PAGE 60

49 results letter with the no intention tone. There were four possible groups and the order in which students were assigned to each group was random.9 For the internal control letter credi bility ratings a MANOVA was conducted using the six credibility rating questions as the dependent variables and the tone manipulation as the independent variable. An overall MANOVA F-test indicated a significant difference in the rati ng of management’s cr edibility between participants in the high intention tone and no intention tone groups (F=3.73; twotailed p= .020). Not all six of the questi ons measuring perceived intention were significant. A breakdown of the individ ual ANOVA results for each question is presented in Panel A of Table 5. The tw o items measuring perceived intentions that were not statistically signific ant were self-centered and understanding. For the earnings results letter credibility ratings a MANOVA was conducted using the six credibility ra ting questions as the dependent variables and the tone manipulation as the indepen dent variable. An overall MANOVA Ftest indicated no significant difference in the ratings of management’s credibility between participants in the high intenti on and no intention tone groups (F=.84; two-tailed p=.556). All six items we re insignificant when analyzed using ANOVAs. The results are presented in Panel B of Table 5. The overall results of the second pilot study indicated that the right wording in a letter from management could influenc e participants’ perceptions and ratings of management’s credibility. With respect to the earnings lette r, the manipulation 9 The order the letters were presented was not r andom because the order of these letters cannot be randomized in the main study. All participants receive the internal control letter before receiving the financial results letter.

PAGE 61

50 was not sufficiently strong, or was not im portant to participants, as there was no statistically significant difference betw een the two groups. Before the main pilot study was conducted, the wording of the earnings letter was revised. Both the high intention tone and no intention t one earnings letters were lengthened10. 10 Ph.D. students were asked to analyze the le tters used in pilot study two and to make recommendations for increasing the impact of the letters.

PAGE 62

51 Table 5: Comparison of Statistics for Management’s Internal Control Letter and Earnings Letter Panel A: MANOVA Results of the Manageme nt Internal Control Letter on Six Intentio n Questions (Wilks’ Lambda 0.384; MANOVA F=3.73; p=.020) High Intention Treatment No Intention Treatment Upper Bound (1) Lower Bound (11) Mean (standard dev iation) Mean (standard deviation) F Stat P-value Cares about me Doesn’t care about me 4.818 (1.078) 3.600 (1.429) 4.92 0.039 Has my interest at heart Does not have my interest at heart. 4.727 (0.904) 3.400 (1.712) 5.07 0.036 Not self-centered Self-centered 4.545 (1.128) 3.800 (1.229) 2.10 0.163 Concerned with me Unconcerned with me 5.272 (0.646) 3.700 (1.337) 12.14 0.003 Sensitive Insensitive 5.181 (0.603) 4.400 (1.699) 7.57 0.013 Understanding Not Understanding 5.181 (0.751) 4.800 (0.788) 1.29 0.270 Panel B: MANOVA Results of the Earnings Results Letter on Six Intention Questions (Wilks’ Lambda 0.734; MANOVA F=.84; p=.556) High Intention Treatment No Intention Treatment Upper Bound (1) Lower Bound (11) Mean (standard dev iation) Mean (standard deviation) F Stat P-value Cares about me Doesn’t care about me 4.181 (1.887) 3.900 (1.524) 0.14 0.713 Has my interest at heart Does not have my interest at heart. 4.272 (1.272) 4.000 (1.633) 0.18 0.673 Not self-centered Self-centered 4.363 (1.362) 4.400 (1.712) 0.00 0.957 Concerned with me Unconcerned with me 4.727 (1.737) 4.000 (1.764) 0.09 0.354 Sensitive Insensitive 4.454 (1.293) 4.400 (1.265) 0.01 0.923 Understanding Not Understanding 4.818 (1.28) 4.300 (1.159) 0.90 0.355 All questions begin with the phrase “I believe that management of MBMC, Inc. …”

PAGE 63

52 3.6.3 Main Pilot Study A third pilot study was conducted to test the overall effectiveness of the research instrument and to test the main effects of t he treatments on the dependent variables. Student participants enrolled in the study online and completed the task at their own pace via the Internet. Orig inally, 33 usable responses were tested.11 Participants were solic ited from two accounting classes. One of the classes had participated in the second pilot study. Participants from this class previously r ead the letters. There is a possibility that the students in this class did not re-read the letters, as they may have believed they already knew what the letters said. Their responses were a potential threat to the internal validity of the study Data for participants who had previously participated in the second pilot study were removed and 19 overall responses were used to analyze the pilot data. Students who were retained in the study were senior level accounting student s enrolled in an Auditing II course. In the main pilot study, participants were asked to pretend they were members of an investment club. They were given background information about a fictitious company. Included in this in formation was the internal control letter (manipulated at two levels: high int ention tone and no intention tone) and financial data for the last three ye ars, including the earnings per share information. Participants were asked to estimate earnings per share for the coming year (period one). Participants were then given management’s earnings per share estimate for the same period (set at 132 percent of the participant’s 11 Two responses were eliminated because of missing or incomplete information. Several students lost connection of the host Web site but re -started the instrument and completed it in full.

PAGE 64

53 initial prediction) and then were given an opportunity to alte r their original estimate. After completing the earnings per share estimate, participants completed the credibility scale. In the second part of the main pilot study, participants were notified of the actual earnings per share for period one. No tification came in the form of a letter from management that was al so manipulated at two levels (high intention and no intention). Participants were notified that management had failed to meet its forecast. The amount of difference betw een actual earnings per share in period one and management’s prediction was hel d as a constant percentage of management’s prediction (equivalently stat ed as 109 percent of the participants’ original estimate or approximately 17. 4 percent less than management’s initial estimate). Thus, although the absolut e dollar amount of difference between management’s forecast and the actual re sults varied between participants, the percentage difference remained constant. Participants were again presented with the financial statements, including the most recent year’s per formance. They were asked to estimate earnings per share. Participants were given management ’s prediction of earnings per share for the year (period two) (132 percent of their estimate) and given the opportunity to revise their forecast. Pa rticipants then completed the credibility instrument as well as manipulation check questions and demographic questions. Appendix B contains screen shots of the experiment as it was pr esented to the participants.

PAGE 65

54 3.6.3.1 Main Pilot Study Data Analysis The data collected in the pilot study were analyzed to determine if the treatments (the letters) had an effect on participants’ judgments. The main study is based on four hypotheses. Recall that H1 and H2 predict that users who are given communications from management wi th the high intention tone will rate management’s credibility higher and have more reliance on management’s forecasts. In period two, H3 and H4 predict that communications from management with the high intent ion tone can mitigate other losses of credibility. In the main study this was measured as t he difference in credibility ratings from period one to period two and as the difference in reliance on management forecasts from period one to period two. For the pilot study only participant’s ratings of credibility and t heir adjustment of earnings per share for period two were compared between groups. The pilot study was designed to test t he effect of the treatments; however, due to the small sample size, statistically significant results were not expected in all treatments. Instead the strength and direction of the difference in means between treatments was analyzed. The main effects results from the pilot study are presented in Table 6 and Table 7. Only the six questions that measure the intention fa ctor of credibility were analyzed for both letters. While overall m anagement credibility is important, the model already lacks statistical power, adding twelve more questi ons/variables to test the impact of the le tters on the expertise and trustworthy variables would

PAGE 66

55 lower the statistical power of the analysis.12 Also, the intention factor of credibility is the primary factor that is manipulat ed in the study; whereas, the expertise and trustworthy factors are not manipulated. 3.6.3.2 The Effects of the Internal Contro l Letter Manipulation For the first year’s ratings and predi ctions, the overall model was tested using the tone of the internal contro l letter participants received as the independent variable, the six intention factor of credibility questions and the percentage change in earnings per share estimates as the dependent variables. The model was significant (F=5.31, twotailed p= .008) as seen on Table 6 Panel A. As Table 6 Panel B shows, three of t he intention questions are statistically significant (cares about me [F=4.25, one-tailed p= .027] concerned with me [F=4.42, one-tailed p= .025]), and has my interest at heart [F=2 .51, one-tailed p= .066]. The other three questions were found to be insignifi cant (not self centered [F=.300, one-tailed p= .294], sensit ive [F=.12, one-tailed p= .368], and understanding [F=.01, one-tailed p= .454]). Interestingly, the three insignificant variables were the only three questions on the instrument that were reverse coded. It is possible that participants fa iled to read the questions carefully. 12 Overall credibility is composed of the three major factors of cred ibility (intention, expertise, and trustworthiness) as measured by the 18 questions on the cred ibility scale. Each major factor is measured by 6 of the 18 questions.

PAGE 67

56 Table 6: Main Pilot Study Results for Sarbanes-Oxley Internal Control Letter Panel A: Overall Significance of Sarbanes Oxley Internal Control Letter: Wilks’ Lambda =.234, F= 5.31, p=.008 Panel B: Impact of Sarbanes Oxle y Internal Control Letter on Intent ion Factor of Credibility Ratings Question Description* High Intention Tone (N=10) No Intention Tone (N=9) More Credibility (1) Less Credibility (11) Mean (standard deviation) Mean (standard deviation) F Statistic P-value*** Cares about me Doesn’t care about me 5.200 (1.398) 6.777 (1.92) 4.25 .027 Has my interest at heart Does not have my interest at heart. 5.500 (2.461) 7.111 (1.922) 2.51 .066 Not self-centered Self-centered 6.900 (2.182) 6.333 (2.291) .30 .294 Concerned with me Unconcerned with me 5.500 (1.841) 7.222 (1.716) 4.42 .025 Sensitive Insensitive 5.400 (1.646) 5.666 (1.732) .12 .368 Understanding Not Understanding 5.100 (1.449) 5.000 (2.179) .01 .454 Panel C: Impact of Sarbanes Oxle y Internal Control Letter on Percentage Change in Calculation High Intention Tone (N=10) No Intention Tone (N=9) Variable Mean (Standard Deviation) Mean (Standard Deviation) F Statistic P-value *** Y1PERDIF** .550 (.443 ) .287 (.320) 2.22 .077 *All questions begin with the phrase “I be lieve that management of MBMC, Inc” ** Y1PERDIF: Measures the percentage change in adjustment toward management’s forecast. *** Two tailed p-values

PAGE 68

57 The reliance on management’s forecast was calculated as demonstrated in Table 2 and was compared between the two treatment groups. Table 6 Panel C, shows that participants in the hi gh intention condition relied more on management’s estimates. The mean perc entage adjustment for participants in the high intention condition was .550 co mpared to .287 for participants in the no intention condition. The per centage of difference between means was statistically significant (F=2.22, two-tailed p= .077) and in the direction expected. 3.6.3.3 The Effect of the Ea rnings Letter Manipulation The second part of the pilot study sought to lower management’s credibility by having management fail to meet the forecast from period one. The study sought to determine whether a high intention tone letter from management to shareholders could mitigate the loss of credibility as measured by the difference in credibility ratings and by t he difference in participants’ reliance on management’s forecast. All participants we re given a forecast from management in period one that was inaccurate. In th is part of the study, the number of possible treatments has now doubled and the power of the statistical tests is further reduced. Approximat ely 3-6 participants were in each treatment cell. Table 7 presents the results of the earnings letter manipulation.

PAGE 69

58 Table 7: The Results of the Main Pilot Study for the Earnings Results Letter Panel A: Overall Significance of Earnings Result s Letter: Wilks’ Lambda =.791, F= .053, p=.772 Panel B: Impact of Earnings Results Letter on Intention Factor of Credibility Ratings Question Description* High Intention Tone (N=10) No Intention Tone (N=9) More Credibility (1) Less Credibility (11) Mean (standard deviation) Mean (standard deviation) F Statistic P-value Cares about me Doesn’t care about me 6.375 (1.767) 6.182 (1.834) .05 .821 Has my interest at heart Does not have my interest at heart. 6.000 (2.000) 6.818 (2.088) .74 .403 Not self-centered Self-centered 6.750 (2.251) 6.364 (2.292) .13 .719 Concerned with me Unconcerned with me 6.000 (2.449) 6.272 (1.794) .08 .782 Sensitive Insensitive 5.875 (1.642) 5.364 (2.292) .29 .598 Understanding Not Understanding 5.500 (1.069) 5.273 (2.240) .07 .795 Panel C: Impact of Earnings Results Letter on Percenta ge Change in Calculation High Intention Tone (N=10) No Intention Tone (N=9) Variable Mean (Standard Deviation) Mean (Standard Deviation) F Statistic P-value Y2PERAJDF** .260 (.457) .058 (.188) 1.77 .201 *All questions begin with the phrase “I be lieve that management of MBMC, Inc” ** Y2PERADJ: Measures the percentage change in adjustment toward management’s forecast.

PAGE 70

59 Table 7 Panel A presents the result s of a MANOVA test on the earnings letter manipulation. There was no overa ll significance found (F=.053, two-tailed p=.772) when the main effect of the earni ngs letter treatment was tested against the six intention factors of credibility and the adjustment of earnings per share estimates. All six-intention factors of credibility were statistically insignificant (cares about me [F=.05, two-tailed p= .821] has my interest at heart [F=.74, twotailed p= .403], not self centered [F=.13, two-tailed p= .719], concerned with me F=.08, two-tailed p= .782], sensit ive [F=.29, two-tailed p= .598], and understanding [F=.07, twotailed p= .795]). Table 7 Panel C presents the results of testing the impact of the earnings letter on the participants’ adjustment of earnings per share. This was also insignificant (F=1.77, two-tailed p= .201); however, the means were in the predicted direction, suggesting that parti cipants in the high in tention treatment revised their earnings per share closer to management’s predictions. As shown in Table 7 Panel C, the mean adjus tment, stated as a percentage between participants’ initial predic tion and management’s prediction, was 26 percent for the high intention treatment and 5.8 percent for the no intention treatment. 3.6.3.4 Comparison of Earnings Pe r Share Adjustments Between Groups To compare the degree to which per ceived credibility decreased between period one and period two the difference in intention ratings and adjustments of earnings per share were compared betwe en treatment groups. In making the comparison, two new dependent variables were calculated and analyzed from the collected data. These variables m easure the difference of adjustment in

PAGE 71

60 earnings per share estimates in period one and period two and the difference in the intention factor of credibility ratings in pe riod one and period two. The first new variable is the diffe rence between the av erage intention ratings from period one and the average int ention ratings from period two. The six questions measuring the intention factor of credibility were summed and averaged for each period. Period two average ratings were then subtracted from the period one average ratings. For this variable, lower values indicate the intention factor of credibility dropped more in period two (less credibility). Panel A of Table 8 shows the mean ratings for period one and period two and the difference between them. As indicated in Panel A, participants who received the high intention earnings letter lowered m anagement’s credibility more (-1.88) than participants who received the no intenti on financial results letter (-.045). The second new variable measures reliance on management’s forecast as the difference between the percentage adj ustment in earnings per share for period one and period two. Table 2 demonstrat ed how this variable is calculated. Panel B of Table 8 compares the m ean loss of credibility for the percentage change loss of credibility from period one to period two. Consistent with the predicted findings, participants who receiv ed the high intention tone earnings letter perceived less loss of credibility (.213) than participants who received the no intention financial re sults letter (.332).

PAGE 72

61 Table 8: Reduction in Credibility Between Periods One and Two Panel A: Decrease in the Intention Factor of Credibility from Period One to Period Two Change in Intention Fact or of Credibility by Treatment Treatment Condition N Mean Period One Rating Mean Period Two Rating Mean Loss of Credibility Standard Deviation High Intention Letter 8 5.900 6.083 -.188 1.18 No Intention Letter 11 6.000 6.045 -.045 0.98 Panel B: Decrease in Percentage Change of Earnings Per Share Estimates From Period One to Period Two Percentage Change Loss by Treatment Treatment Condition N Mean Period One % Adjustment Mean Period Two % Adjustment Mean Loss of Credibility Standard Deviation High Intention Letter 8 .473 .260 .213 .431 No Intention Letter 11 .390 .058 .332 .421 3.6.3.5 Main Pilot Study Conclusions While the results of the main pilot st udy do not offer significant statistical relationships, the power of the statisti cal tests performed was low. Low power increases the possibility of a Type II error. Overall, the results of the pilot study were encouraging in that many of the manipulations seemed to be influencing the dependent variables in the hypothesized direction. Several other factors could also lead to the non-significant findings of this pilot study. First the position of the credi bility instrument in the study may have been too far removed from the letter manipulations in both treatments. In the pilot study, participants read the letter (the treatment) then read the financial statements, made their earnings per share estimates, read management’s earnings per share predictions, and then revised their forecasts before

PAGE 73

62 completing the credibility questions. For t he main study this instrument was moved up so that participants completed the credibility scale directly after reading the letters. Another significant factor that may have impacted t he results of the pilot study was the way in which the study was administered. Participants were able to participate in the study at their leisure and in the location they selected. Some participants may have had extraneous activities interfering with their attention to the study and its subtle manipulations. The only variable available to give an estimate of effort while taking the study is the time it took participants to complete the study. This is a noisy variable in that someone who takes a long time to complete the study may not have given any more effort to reading and answering the questions than someone who took a sh ort time. People or things around participants could distract them from the study for a few minutes. In an effort to help control these extraneous variables, the main study was conducted in a supervised setting. The participants comple ted the study in a computer lab with a proctor administering the study.

PAGE 74

63 4.0 Main Study Results 4.1 Introduction In this chapter, I present the result s of the main study. The task used in the study was presented in Chapter 3, Se ction 3.3. This c hapter begins with a discussion of the main study’s partici pants. Next the manipulation checks employed to ascertain the salience of the treatments to t he participants are discussed. After discussing the manipula tion checks, an analysis of potential covariates is presented. Following analysi s of covariates, the hypotheses that were presented in Chapter 2 are stat istically tested and analyzed. Using the results of the data analyses, each of t he hypotheses is examined and discussed in detail. 4.2 Participants One hundred and twenty-four graduate level students participated in the study. The participants were recruited from a large southeastern university. The recruitment pool consisted of graduate st udents enrolled either in a Masters of Business Administration (MBA) progr am, Masters of Accountancy (M.Acc.) program or post-undergraduate accounting students who were completing a fifth year of course work for professional ce rtification. Fortyfive MBA students and

PAGE 75

64 79 accounting students13 participated in the main study. There were 71 female participants and 53 male participants. The participants had an average of 1.65 years of work experience in the field of accounting. During the semester, all of the parti cipants were enrolled in an accounting course that required them to either participate in one of several available research projects or write a paper assigned by their instructor. In addition to class credit, participants who enrolled in this study were paid $10. The participants in this study repres ented an appropriate pool to test the research hypotheses since graduate bus iness students have been found to be a reasonable proxy for investors (Copeland et al. 1973; Ashton and Kramer 1980; Walters-York and Curatola 1998; ,2000; Libby et al. 2002). Also, this study focused on the judgment of investor s and relied only on general cognitive abilities. As discussed in Libby et al. (2002), students are a su itable participant pool when general cognitive abilities are re quired. This study only looks at the relative differences between participant s’ responses by treatment group. 4.3 Manipulation Checks Several manipulation check questions were included in the study to determine if the participants perceived t he treatments given to them. The manipulation check questions will be discussed below. 13 A breakdown between M.Acc. students and t he other accounting students was not made as students in the M.Acc. program were also en rolled in the fifth year accounting course.

PAGE 76

65 4.3.1 Intenti on Manipulation No manipulation check was analyzed with respect to the intention manipulation. As discussed below, there were questions included in the study that attempted to test if partici pant’s perceived the intention of the communications that they were given. Written communication was used to deliver the credibility treatment between the groups in peri od one and period two. In period one, participants received the statement of internal contro ls from management with either the high intention tone or the no intention t one treatment. In period two, participants received the letter from management inform ing them of the actual results for period one, again with either a high intention tone or no intention tone treatment. The intention factor of credibility is composed of three sub-factors: understanding, empathy, and responsive ness. An attempt was made to determine if the participants perceived t he letters as understanding, empathetic, and responsive. For each period, parti cipants answered three questions rating management on the three sub-factors of the intention factor of credibility. The manipulation check questions can be seen in Appendix B, item 19. The six manipulation check questions (three per period) dealing with the three factors of credibility were not anal yzed for this study. It was decided post hoc that the credibility scale would be a better determinant than the manipulation questions in determining if the participants perceived the letters as credible. The purpose of the manipulation check questions was to test if the participants remembered the treatment to which they were exposed. These manipulation questions asked participants to rate the credibi lity of the letters they received, but

PAGE 77

66 did not ask about the actual wording of any of the letters. All three of the manipulation questions were highly correla ted with the credib ility factors and seemed to measure the same construct as credibility, which is already measured by the credibility scale used in this study (Cronbach’s = .768). McCroskey and Tevens’ (1999) credibility scale is a va lidated instrument, while the questions used as manipulation checks have not been validated. Be tter manipulation check questions could have asked the par ticipants questions about the specific wording in the letters they received su ch as, “Did managem ent meet with focus groups of shareholders because there wa s a directive from the board of directors?” This type of question woul d have determined if the participants actively recalled the specific treatments. 4.3.2 Surprise Two manipulation questions were given to participants to determine if they were surprised by the amount of managem ent’s forecast. One question was used for each of management’s forecasts. The questions were answered using a seven-point Likert scale with answers r anging from “Lower t han I expected” to “Much more than I expected.” All but tw o of the participant s seemed somewhat surprised by management’s forecast. Thes e participants rated the surprise of management’s period one forecast as less than a four on the seven-point scale. The period one analysis was conducted with and without these two participants

PAGE 78

67 and no material differences were found in the results. Based on these findings the reported results retain the two participants. Surprise was tested for the second peri od forecast in the same manner as for period one. In period two, only one par ticipant rated management’s forecast as less than four on the seven-point sca le. The analysis was conducted both with and without the three participants who faile d the manipulation check (two from period one and one from period two) and there was no impac t on the results of the study. 4.3.3 Accuracy One question was used as a manipul ation check to determine whether participants had perceived management’s per iod one forecast as accurate. The question used a seven-point Likert scale and asked participants if management’s forecast for period one was accurate (1) to inaccurate (7). Management’s forecasts in the study were inaccura te and it was expected that participants would rate management low on the accu racy scale. Six participants rated management’s accuracy in its period one fore cast below four on the seven-point scale. Three participants answered the accuracy question with a rating of two while three others had a rating of three. The reported results for period one’s analyses and period two’s analyses were tested without these six responses and no material differences were found.

PAGE 79

68 4.3.4 Manipulation Check Summary As indicated, it was determined that removing the participants who failed manipulation check questions did not significantly add to the explanatory power of the model. By keeping all of the par ticipant responses a balanced design was achieved, which increased the power of th e statistical methods employed in the data analysis. 4.4 Data Analysis 4.4.1 Introduction This section begins with an explanati on of the selection of covariates included in the study. Follo wing the discussion of the covariates, each of the hypotheses will be discussed and statistica lly tested. Included in the write-up about hypotheses testing is a description of the type of analysis used to test each hypothesis. 4.4.2 Covariates This study was designed to m anipulate and measure management’s perceived credibility by altering the t one of a written communication from management. Management’s credib ility was also manipulated in the study by having management fail to meet a fore cast. These manipulations were purposeful and controlled in the experimen t. There are, howev er, other factors that can impact perceptions of managem ent’s credibility or the impact that perceptions of management’s credibilit y will have on decision making. While

PAGE 80

69 these factors are important not all of them can be manipulated and randomized between participants. Some of these factor s are unique to each individual in the study. It is important to identify and m easure factors that are associated with individual participants that may impac t their assessment of management’s credibility or the impact that management’s credibilit y has on their decision making. In statistical terms, thes e factors are called covariates. Good covariates will help explain so me of the variation in the dependent variables without removing any of t he power of the model. Generally, good covariates are highly correlated with the dependent variable but not with independent variables. If t he covariates are correlat ed with independent variables they reduce the ability to measure the true impact of the independent variables on the dependent variable. All covariates in this study were selected by first testing a correlated relationship between the covariate and the dependent variable. As the models are developed and tested, these covariates are further scrutinized with regard to t heir relationship to the dependent variables. In the case of the credibility rating variable the potential covariates were tested individually against the three sub-factor s of the credibility variable. The subfactors were used to ensure that no potential cova riates were omitted at a higher level of analysis since the credibility construct is analyzed by the three subfactors in the post hoc analysis section. Three types of possible covariates were measured and tested in this study: demographic, theoretical, and recent events covariates. Covariates were first tested by using a correlation anal ysis testing each covariate with the

PAGE 81

70 respective dependent variables. An alpha of .10 was used as the significance level in the correlation analysis. Once the covariates with a significant correlation to the dependent variable were identifi ed, a preliminary ANCOVA model was tested using the covariates and the appropriate independent variable. The following sections detail the testing of t he different classifi cations of possible covariates used in this study. 4.4.2.1 Demographic Covariates Demographic information was captur ed about each participant. Capturing demographic information allows testing to determine if ther e were systematic differences between similar groups of people participating in the study. The demographic questions tested as covariates in this study are presented in Table 9 and include gender, accounting experienc e, and years of work experience. Gender was selected as a possible covariate because prior research has found significant differences between males and females in stock market investing tasks (Barber and Odean 2001). This study makes no directional predictions as to the impact of gender on the task result s. Accounting experience and years of work experience were selected as covariat es due to the nature of the task. There was a potential for participants with more accounting and work experience to recognize differences between their experience with disclosures from management and those presented withi n the task in the study.

PAGE 82

71 Table 9: Demographic Covariate Questions Panel A: Variable Names, Questions and Response Format Variable Question Response Gender What is your gender? Male=1/Female=0 Accounting Experience Have you worked in the field of accounting? Yes=1/No=0 Years of Experience In total, how many years have you worked in the field of accounting? (Numeric Response) Panel B: Demographic Covariates Descriptive Data Variable N Mean Gender 124 0.427 Accounting Experience 124 0.508 Years of Experience 63 3.253 Panel C: Covariate Correlations with Dependent Variables Using Pearson Correlation Coefficients (p-values**) Variable* Credibility Rating Reliance on Management’s Forecast Difference in Credibility Ratings Difference in Reliance on Management’s Forecast Gender -0.177 (0.050) -0.093 (0.307) 0.063 (0 .486) -0.102 (0.258) Accounting Experience 0.109 (0.227) -0.014 (0 .876) 0.136 (0.132) 0.081 (0.369) Years of Experience -0.085 (0.509) -0.015 (0.904) -0.028 (0.823) -0.030 (0.816) See Panel A for a description of the variables ** P-Values are two-tailed tests. Panel A identifies the demographic ques tions asked and Panel B provides descriptive data for the demographic cova riate questions. Approximately 43 percent of the participants were male s (53) and about 57 percent were females (71). Slightly over 50 percent of t he participants had previous accounting experience (63), resulting in a mean of about 3.25 years of experience per participant with experience. As shown in Table 9 Panel C, only the gender variable was found to be significantly correlated (p = .050) wit h any of the dependent variables. The gender variable was correlated with credibility rating.

PAGE 83

72 4.4.2.2 Theoretical Covariates There are several factors that are t heoretically tied to credibility ratings and the impact credibility plays in dec ision making. These variables are discussed in detail in Appendix A, but will be briefly reviewed here as they relate to the covariates in the study. The t heoretical factors to consider are the participants’ level of involvement with t he task, the timing in identifying the communicator, the advocat ed position of the message and the perception of possible knowledge and reporting biases. 4.4.2.3 Level of Involvement Based on the theoretical model discuss ed in Appendix A, a participant’s level of involvement with the task s hould not impact his/her ratings of management’s credibility but should have an impact on the reliance on management’s forecasts. Participant’s leve l of involvement c ould be an important factor in the findings of this study so nine questions were asked to determine participants’ level of involvement with t he study (Table 10, Panel A). Participants who were more involved in the stock market might be more involved in the task. Panel B and Panel C of Table 10 present the descriptive statistics for the level of involvement questions. As can be seen in Table 10, 54 percent (67) of participants had made investments (PR EVIOUSLY INVESTED) in the stock market, while 88 percent (100) plan (PLAN TO INVEST) to invest in the stock market (some of the parti cipants who indicated they had invested in the stock market also plan to invest in the stock market). When asked on a five-point scale if they would pick their own stocks ( SELF SELECT STOCKS) rather than have a

PAGE 84

73 broker select them, the group responses indicated that parti cipants would pick their own stocks with a mean of around 3. 68, more than they would rely on a broker to pick their stocks (BROKER), with a mean around 2.53. Participants were also asked if they had previously per formed a task similar to the one in the study (TASK EXPERIENCE). Of the 124 participants, most had not performed a task similar to the one in the study (m ean 2.60). Most participants seemed to enjoy the task (TASK ENJOYMENT) as t hey rated their mean enjoyment at about 3.8, which was close to the “agree” end of the response scale. They also seemed to have some confidence in their earnings per share estimates (CONFIDENCE) as they rated the conf idence question with a mean of about 3.5, falling just between the “ neutral” and “agree” points on the response scale. Panel D of Table 10 presents the correlations between the level of involvement questions and the dependent vari ables. Several of the level of involvement questions were correlated wit h the credibility rating variables and were further tested as cova riates. Questions relating to participant’s investing behavior were highly correlated with the cred ibility rating. The following variables were significantly correlated with the credibility rating: PREVIOUSLY INVESTED, SELF SELECT STOCKS, INVESTED IN MUTUAL FUND, TASK ENJOYMENT, and CONFIDENCE. PREVIOUSLY INVESTED and CONFIDENCE were also significantly correlated with reliance on management’s forecast in period one. Only INVESTED IN MUTUAL FUND was significantly correlated with the difference in credibility ratings from per iod one to period two. The difference in reliance on management’s forecasts from period one to period two was

PAGE 85

74 significantly correlated with only one level of involvement question— PREVIOUSLY INVESTED. Table 10: Level of Involvement with Task Questions Panel A: Variable Names, Questions and Response Format Variable Question Response PREVIOUSLY INVESTED Have you ever made an invest ment in the stoc k market? Y=1/N=0 PLAN TO INVEST Do you plan to invest in the stock market? Y=1/N=0 INVESTED IN MUTUAL FUND Have you ever invested in a mutual fund? Y=1/N=0 SELF SELECT STOCKS I pick which stocks I want to purchase. [SA,A,N,D,SD] BROKER I rely on my broker to tell me which stocks to purchase. [SA,A,N,D,SD] MUTUAL FUNDS ONLY I only invest in mutual funds [SA,A,N,D,SD] TASK EXPERIENCE I have previously performed tasks similar to the one in this study. [SA,A,N,D,SD] TASK ENJOYMENT I enjoyed working on the ta sk in this study. [SA,A,N,D,SD] CONFIDENCE I am confident in my earni ngs per share predict ions. [SA,A,N,D,SD] Key: [SA,A,N,D,SD] = Strongly Agree (5), Agr ee (4), Neutral (3), Disagree (2), Strongly Disagree (1) Panel B: Covariate Descriptive Data for Dichotomous Variables Variable N Mean Previously Invested 124 0.540 Plan to Invest 124 0.880 INVESTED IN MUTUAL FUND 124 0.557 Panel C: Covariate Descriptive Data for Ordinal Variables Variable N Mean* Standard Deviation Minimum Maximum SELF SELECT STOCKS 124 3.685 1.054 1 5 BROKER 124 2.532 1.070 1 5 TASK EXPERIENCE 124 2.604 1.254 1 5 TASK ENJOYMENT 124 3.823 0.744 2 5 CONFIDENCE 124 3.508 0.791 1 5 See Panel A for scale values for each variable.

PAGE 86

75 Table 10: Level of Involvement with Task Questions (Continued) Panel D: Pearson Correlation Coefficients for Covariates and Dependent Variables (pvalues**) Variable* Credibility Rating Reliance on Management’s Forecast Difference in Credibility Ratings Difference in Reliance on Management’s Forecast Previously Invested -0.187 ( 0.037 )0.165 (0.068) 0.038(0.676) 0.190 (0.035) Plan to Invest 0.025 (0.786)-0.009 (0.923)0.070(0.435) 0.112 (0.214) INVESTED IN MUTUAL FUND -0.212 (0.018 )0.116(0.199)-0.268 (0.003) 0.119 (0.190) SELF SELECT STOCKS -0.173 ( 0.055 )-0.003(0.974)0.015(0.866) 0.013 (0.888) BROKER 0.056 (0.538)-0.007(0. 937)0.003(0.971) 0.093 (0.306) MUTUAL FUNDS ONLY 0.127 (0.159)-0.004(0.966)0.048(0.595) 0.118 (0.191) TASK EXPERIENCE -0.143 (0.114)0. 003(0.978)-0.007(0.936) 0.039 (0.671) TASK ENJOYMENT 0.221 ( 0.014) -0.117(0.195)0.032(0.718) 0.031 (0.731) CONFIDENCE 0.165 ( 0.067 )-0.154 (0.089) 0.029(0.745) 0.075 (0.408) See Panel A for a description of the variables ** P-Values are two-tailed tests. 4.4.2.4 Timing in Identifyi ng the Communicator Covariate The model in Appendix A reveals t hat for people receiving a message, timing in identifying who a communicator is can impact the role of credibility in decision making. Timing in this sense refe rs to when a person is notified of the sender of a message, which can either be before or after the message has been given. In this study, all recipients we re given communications from management and all recipients were told the co mmunication was from management. Two questions were given to parti cipants to test if they we re aware that management was responsible for the communications in this study. Consistent with the theory, since all participants were told the identity of their communicators before

PAGE 87

76 receiving the message, the responses to this question were uncorrelated with any of the dependent variables. 4.4.2.5 The Advocated Position of the Message Another factor discussed in Appendix A is the position of the message. Communicators can either give a message t hat is in the same direction as their audiences’ beliefs or in an opposite directi on. The participant’s initial forecast of earnings per share was used to compare the position of the message given by management to the participants’ beliefs about future earnings. Participants who forecasted positive earnings per share (n= 114) over the prior period were tested against those who forecasted negative earnings per share (n= 10) using a bivariate dummy variable. Since almost all (92 percent) of the participants forecasted positive earnings per share over the initial prior period the results of the comparison were insignificant. 4.4.2.6 Knowledge Bias and Reporting Bias The message delivered by any communicator will interact with the message recipient’s expectations of t he message. Knowledge bias and reporting bias refer to two such interactions that have been found in prior literature (Eagly and Wood, 1978). While these two topics ar e covered in detail in Appendix A, a brief review will be given here to clarify how these interactions were tested. It is also important to note that testing for these effects occurred near the end of the study after the treatments were given; therefore, the questions had no impact on the main treatments, thus eliminating t he possibility of confoundi ng the results of

PAGE 88

77 the main experiment with the manipula tion check questions. Unfortunately, asking the questions at the end of the st udy limited the scope of the questions. A knowledge bias would occur if the message recipient believed the communicator’s background or education would pre-dispose the communicator to advocating only one side or perspective of an issue. It would be confirmed if the message delivered by the communicator was consistent with the recipient’s pre-message expectations. A confirmed k nowledge bias can reduce credibility via the expertise factor of credibility while a disconfirmed knowledge bias can increase perceived credibility via the experti se factor of credibi lity. For example, more weight is given to a politician who gives an opinion opposite that of his political party’s message. The message re cipients are expecting the politician to take a position in line wit h the politician’s party. If the politician takes the expected position, he/she can lose credibi lity with the message recipients (unless the message takes the same position held by the recipient, then credibility means less to the decision). A reporting bias is also an expecta tion from the message recipient. The expectation is formed based on the in tended audience of the communication. That is, a recipient believes the comm unicator will alter the message to conform to the beliefs of the intended audience. When a reporting bias is confirmed, the trustworthiness factor of credibility is reduced. When a reporting bias is disconfirmed, the trustworthiness factor of credibility is enhanced. Three questions were used to determine if reporting or knowledge biases were present in the study. The three ques tions used to test for reporting and

PAGE 89

78 knowledge biases do not allow differ entiation between which bias may be present. This study was not designed to test the impact of these biases; the check is being used to test a possible co variate that may help remove some of the effects of these biases, thus isolat ing the impact on the main treatments. Table 11, Panel A presents the th ree questions used to test the knowledge bias and reporting bias in the st udy. It was importa nt to determine the expectations of the participants with regard to what management reports. The first question asked if participants expe cted management to correctly report its earnings estimates while the other two questions both tried to determine if participants expected management to inflat e its earnings estimates and if they expected those estimates to be positively inflated. The descriptive data for the reporting and knowledge bias questions are presented in Panel B of Table 11. Regarding their expectations about management reporting earnings correctly, pa rticipants were somewhat neutral in their response with a mean score slightly less than 3.26. They did, however, have a higher mean (about 3.53) with re spect to the belief that management would inflate its earnings predictions At an even higher level of agreement, (approximately 4.22) part icipants indicated that management would predict positive earnings. Panel C of Table 11 pr esents the correlation data between the reporting and knowledge bias questions and the dependent variables. The question asking participants if they expected management to inflate its earnings estimates (EXPECT MANAGEMENT TO INFLATE EPS) was the only variable that was

PAGE 90

79 correlated with any of the dependent variabl es. This variable was correlated with both reliance on management’s forecast (H 2), and the difference in reliance on management’s forecasts (H4) from period one to period two. Table 11: Reporting Bias and Knowledge Bias Questions Panel A: Variable Names, Questions and Response Format TRUST MANAGEMENT Managers in public companies are most likely to report earnings estimates correctly. [VL,L,N,U,VU] EXPECT MANAGEMENT TO INFLATE EPS I expected management to inflate their earnings predictions. [SA,A,N,D,SD] EXPECT POSITIVE EARNINGS ANNOUNCEMENTS I expected management to pr edict positive earni ngs. [SA,A,N,D,SD] Key: [SA,A,N,D,SD] = Strongly Agree (5 ), Agree (4), Neutral (3), Dis agree (2), Strongly Disagree (1) [VL,L,N,U,VU] = Very Likely (5), Likely (4), Neutral (3), Unlikely (2), Very Unlikely (1) Panel B: Covariate Descriptive Data Variable N Mean* Standard Deviation Minimum Maximum TRUST MANAGEMENT 124 3.258 0.927 1 5 EXPECT MANAGEMENT TO INFLATE EPS 124 3.532 1.199 1 5 EXPECT POSITIVE EARNINGS ANNOUNCEMENTS 124 4.218 0.704 2 5 See Panel A for scale values for each variable. Panel C: Covariate Correlations with Dependent Variables Using Pearson Correlation Coefficients (p-values) Variable* Credibility Rating Reliance on Management’s Forecast Difference in Credibility Ratings Difference in Reliance on Management’s Forecast TRUST MANAGEMENT 0.130 (0.149) -0.020 (0.822) 0.067 (0.461) -0.033 (0.714) EXPECT MANAGEMENT TO INFLATE EPS 0.127 (0.159) -0.282 (0.002) -.034 (0.706) -0.220 (0.014) EXPECT POSITIVE EARNINGS ANNOUNCEMENTS 0.109 (0.230) -0.146 (0 .105) 0.035 (0.702) -0.002 (0.984) See panel A for a description of the variables. ** P-Values are two-tailed tests.

PAGE 91

80 4.4.2.7 Recent Event Covariates The final group of variables tested as co variates related to recent events that had occurred in the financial markets and the reaction of lawmakers to those events (Sarbanes-Oxley). The exposure of participants to high profile financial frauds such as Enron and WorldCom c ould impact participant responses. Twelve questions were used to develop an understanding of each participant’s exposure to and understandi ng of the recent events regarding management fraud and the government reacti on to those recent frauds (i.e., the Sarbanes-Oxley Act of 2002). The recent event questions are presented in Table 12, Panel A. Eight of the 12 questions deal with fraud and four questions deal with participant’s knowled ge of and beliefs about the Sarbanes-Oxley Act. Table 12, Panel B and Panel C present s descriptive data regarding the recent event questions. Participants we re asked if they believed management fraud was prevalent (FRAUD IS PREVAL ENT), and their responses seem to indicate a bit of indecision with respect to the issue. Just over half of the participants about .51 responded that they believed management fraud is prevalent. Seemingly in agreement, when par ticipants were asked if they would rely on a forecast from management (TRUST MANAGEMENT FORECAST) the mean response was a neutral 3 on a scale of 1 to 5. Participants were then asked if they would rely on a fore cast from management if an independent auditor provided assurance on those forecasts (AUDIT ASSURANCE OF FORECASTS). The results were a bit more positive for the effect of auditor assurance, as the mean response to the auditor question was around 3.55.

PAGE 92

81 Of the 124 participants, 5.6 percent (7 ) indicated they had been defrauded as a shareholder of a public company (PERSONAL FRAUD HI STORY), but 22.6 percent (28) of participants k new someone who had been defrauded (ACQUAINTANCE FRAUD HISTORY) and 79.8 percent ( 99) had heard of someone being defrauded by a public company (HEARD ABOUT FRAUD). Surprisingly, 1.6 percent (2) of t he participants responded that they have committed fraud in a public company (FRAUDSTER)14. With respect to the Sarbanes-Oxley Act recent event questions the participants were asked if they had st udied the Sarbanes-Oxley Act (STUDIED SOX). They responded with a mean score around 4.45 indicating somewhere between “agree” and “strongly agree” that they had studied the Sarbanes-Oxley Act. The participants also felt that the Sarbanes-Oxley Act (FRAUD SOX RELATIONSHIP) was between “relevant” an d “very relevant” with respect to fraud (mean about 4.37). The last two ques tions asked participants if they were familiar with the Public Company A ccounting Oversight Board (PCAOB FAMILIARITY) and if they were familiar with Auditing Standard No. 2 as issued by the PCAOB (ASNO2 FAMILIARITY). T he results for these two questions were somewhat consistent as participants were close to “agree” that they knew about the PCAOB (mean about 3.82), and they were between “ neutral” and “agree” on the question regarding their familiari ty with the Auditing Standard No. 2 promulgation, with a m ean rating of 3.54. 14 Both of these participant’s responses to other questions were checked to determine if they were not taking the study seriously. Based on the time they spent on questions before and after this question it did appear the participants at least took a reasonable amount of time to answer these questions.

PAGE 93

82 Panel D of Table 12 pres ents the correlations betw een the recent events covariates and the dependent variables. The recent event covariates could be split into two sections, recent ev ents regarding managem ent fraud and the events that transpired in reaction to t hose events (e.g., the passage of the Sarbanes Oxley Act). Only one of t he questions associated with management and fraud was significantly correlated with any of the dependent variables, while all four of the Sarbanes-Oxl ey questions were correlated with at least one of the dependent variables. The fraud question that asked if the participant had been defrauded as a shareholder of a public com pany was significantly correlated with the credibility rating. For the Sarbanes -Oxley questions, the question asking participants if they had studied the Sa rbanes-Oxley Act in their accounting courses was correlated with the differ ence in management’s credibility ratings from period one to period two. The question asking the importance of the Sarbanes-Oxley Act was significantly corre lated with the credibility rating and the reliance on management’s forecast. The questions asking students about their familiarity with Auditing Standard No. 2 and the PCAOB were also significantly correlated with the credibility rating variable.

PAGE 94

83 Table 12: Recent Event Questions Panel A: Variable Names, Questions and Response Format Variable Question Response FRAUD IS PREVALENT Do you believe managem ent fraud is prevalent? Y=1/N=0 PERSONAL FRAUD HISTORY Have you ever been defrauded as a shareholder of a public company? Y=1/N=0 ACQUAINTANCE FRAUD HISTORY Do you know of someone who has been defrauded by a public company? Y=1/N=0 HEARD ABOUT FRAUD Have you ever heard of someone who was defrauded by a public company? Y=1/N=0 FRAUDSTER Have you ever committed fraud as a member of management? Y=1/N=0 TRUST MANAGEMENT FORECASTS I trust management in providing me with foreword looking forecasts. [SA,A,N,D,SD] AUDIT ASSURANCE OF FORECAST I would rely on management’s forward looking forecasts if an independent auditor provided assurance on management’s assertions. [SA,A,N,D,SD] STUDIED SOX I have studied Sarbanes-Oxley in my accounting courses. [SA,A,N,D,SD] FRAUD SOX RELATIONSHIP With regard to fraud, I believe Sa rbanes-Oxley is : [VR,R,N,I,VI] ASNO2 FAMILIARITY I am familiar with the requirements of PCAOB Auditing Standard No. 2, "An Audit of Internal Control over Financial Reporting Performed in Conjunction with an Audit of Financial Statements." [SA,A,N,D,SD] PCAOB FAMILIARITY I know what the PCAOB is. [SA,A,N,D,SD] Key : [SA,A,N,D,SD] = Strongly Agree (5), Agr ee (4),Neutral (3), Disagree (2), Strongly Disagree (1) [VR,R,N,I,VI] =Very Relevant (5), Relevant (4),Neutral (3), Irrele vant (2), Very Irrelevant (1)

PAGE 95

84 Table 12: Recent Event Questions (Continued) Panel B: Recent Event Covariates Descriptive Data for Dichotomous Covariate Variables Variable N Mean Standard Deviation FRAUD IS PREVALENT 124 0.508 0.502 PERSONAL FRAUD HISTORY 124 0.056 0.232 ACQUAINTANCE FRAUD HISTORY 124 0.226 0.420 HEARD ABOUT FRAUD 124 0.798 0.403 FRAUDSTER 124 0.016 0.126 Panel C: Recent Event Covariates Descriptive Data for Ordinal Covariate Variables Variable N Mean Standard Deviation Min Max TRUST MANAGEMENT FORECASTS 124 3.000 0.855 1 4 AUDIT ASSURANCE OF FORECAST 124 3.548 0.868 2 5 STUDIED SOX 124 4.452 0.780 1 5 FRAUD SOX RELATIONSHIP 124 4.371 0.643 2 5 ASNO2 FAMILIARITY 124 3.540 1.340 1 5 PCAOB FAMILIARITY 124 3.823 1.437 1 5 See Panel A for scale values for each variable.

PAGE 96

85 Table 12: Recent Event Questions (Continued) Panel D: Covariate Correlations with Dependent Variables Using Pearson Correlation Coefficients (p-values**) Variable* Credibility Rating Reliance on Management’s Forecast Difference in Credibility Ratings Difference in Reliance on Management’s Forecast FRAUD IS PREVALENT 0.048 (0.598)0.014 (0.879)0. 099 (0.274) 0.018 (0.846) PERSONAL FRAUD HISTORY -0.187 (0.038) 0.051 (0.575)-0.054 (0 .549) 0.055 (0.546) ACQUAINTANCE FRAUD HISTORY 0.016 (0.860)0.069 (0.449)0. 040 (0.657) 0.112 (0.215) HEARD ABOUT FRAUD -0.072 (0.424)-0.010 (0.910) 0.045 (0.616) 0.043 (0.636) FRAUDSTER 0.026 (0.774)-0.122 (0.179)0. 093 (0.300) -0.117 (0.198) TRUST MANAGEMENT FORECASTS 0.007 (0.939)0.115 (0.204)-0 .144 (0.109) 0.106 (0.243) AUDIT ASSURANCE OF FORECAST 0.138 (0.128)-0.041 (0.651)-0 .052 (0.566) -0.004 (0.968) STUDIED SOX 0.129 (0.152)-0.044 (0.627)-0.186 (0.038) -0.051 (0.575) FRAUD SOX RELATIONSHIP 0.260 (0.004) -0.159 (0.078) 0.024 (0.791) -0.048 (0.597) ASNO2 FAMILIARITY 0.283 (0.001) 0.075 (0.406)0.042 (0 .643) 0.101 (0.267) PCAOB FAMILIARITY 0.200 (0.026) 0.033 (0.715)-0.003 (0 .970) 0.101 (0.264) All participants had read about a fraud being committed. ** P-Values are two-tailed tests. 4.4.2.8 Summary of Covariate Findings The results of the covariate testing found several variables correlated with the dependent variables in this study. This section contains a brief discussion of the covariates included in the model us ed to test the hypotheses. The discussion relates each covariate to the dependent variables with which they were correlated.

PAGE 97

86 4.4.2.8.1 Covariates Correlat ed with Credibility Rating (H1) The covariates that were correlated with the credibility rating (H1) were determined by testing each potential co variate’s correlation with the credibility rating from period one of the study. Table 13 summarizes the correlated covariates with credibility rating by type of covariate. Table 13: Covariates Co rrelated with Credibility Rating (H1) Variable* CorrelationP-Value** Type of Covariate Reference Table Gender -0.177 0.050 Demographic Table 9 PREVIOUSLY INVESTED -0.187 0.037 Level of Involvement Table 10 SELF SELECT STOCKS -0.173 0.055 Level of Involvement Table 10 INVESTED IN MUTUAL FUND -0.212 0.018 Level of Involvement Table 10 TASK ENJOYMENT 0.221 0.014 Level of Involvement Table 10 CONFIDENCE 0.165 0.067 Level of Involvement Table 10 PERSONAL FRAUD HISTORY -0.187 0.038 Recent Events Table 12 FRAUD SOX RELATIONSHIP 0.260 0.004 Recent Events Table 12 PCAOB FAMILIARITY 0.200 0.026 Recent Events Table 12 ASNO2 FAMILIARITY 0.283 0.001 Recent Events Table 12 ** P-Values are two-tailed tests. Once the significantly correlated va riables were identified they were further tested for inclusion in the final model by running a preliminary ANCOVA analysis. This analysis included the i ndependent variable tone of management letter (Internal Control Letter) used to te st H1 as well as all of the identified covariates. Using an alpha of .10, only f our of the correlated variables were significant in the model as presented in Table 14. Participants’ prior history with investing in mutual funds (INVESTED IN MUTUAL FUND) (F=3.53, two-tailed p= .063), their confidence in their earnings per share estimates (CONFIDENCE) (F =

PAGE 98

87 5.82, two-tailed p= .018), their fam iliarity with Sarbanes-Oxley (FRAUD SOX RELATIONSHIP) (F = 9.47, two-tailed p= .002), and their familiarity with Auditing Standard No. 2 (ASNO2 FAMILIARITY), (F =2.97, two-tailed p= .087) were included in the final model used to test H1. Table 14: Preliminary ANCOVA Testing of Credibility Rating (H1) Using Covariates Variable* DF Sum of Squares Mean Squares F Statistic P-Value** Internal Control Letter 1 10.755 10.755 7.15 0.009 PERSONAL FRAUD HISTORY 1 1.845 1.845 1.23 0.270 FRAUD SOX RELATIONSHIP 1 14.246 14.246 9.47 0.002 PCAOB FAMILIARITY 1 2.177 2.177 1.45 0.231 ASNO2 FAMILIARITY 1 4.472 4.472 2.97 0.087 Gender 1 1.327 1.327 0.88 0.350 PREVIOUSLY INVESTED 1 2.981 2.981 1.98 0.162 INVESTED IN MUTUAL FUND 1 5.305 5.305 3.53 0.063 SELF SELECT STOCKS 1 3.587 3.587 2.38 0.125 TASK ENJOYMENT 1 1.707 1.707 1.13 0.289 CONFIDENCE 1 8.753 8.753 5.82 0.018 Model 11 84.858 7.714 5.13 0.0001 Error 112 168.430 1.503 Corrected Total 123 253.289 n = 124 *See tables 9-12 for variable descriptions ** P-Values are two-tailed tests. 4.4.2.8.2 Covariates Correlated wit h Reliance on Management’s Forecasts The four variables correlated with the participants’ reliance on management’s forecasts are presented in Ta ble 15. The covariates were further tested against the reliance on management’s forecasts by running a preliminary ANCOVA analysis. As Table 16 demonstrates, all four of the potential covariates

PAGE 99

88 were significant using an alpha of .10. Consequently, all four covariates are used in the hypothesis testing. Table 15: Covariates Correlated with Reliance on Management’s Forecast (H2) Variable* CorrelationP-Value** Type of Covariate Reference Table PREVIOUSLY INVESTED 0.1650.068Level of Involvement Table 10 CONFIDENCE -0.1540.089Level of Involvement Table 10 EXPECT MANAGEMENT INFLATE EPS -0.2820.002 Reporting and Knowledge Bias Table 11 FRAUD SOX RELATIONSHIP -0.1590.078 Recent Events Table 12 ** P-Values are two-tailed tests. Table 16: Preliminary ANCOVA Testing of Reliance on Management’s Forecast (H2) Using Covariates Variable DF Sum of Squares Mean Squares F Statistic P-Value** Internal Control Letter 1 0.031 0.031 4.27 0.041 FRAUD SOX RELATIONSHIP 1 0.046 0.046 6.31 0.013 CONFIDENCE 1 0.050 0.050 6.88 0.010 PREVIOUSLY INVESTED 1 0.025 0.025 3.39 0.067 EXPECT MANAGEMENT INFLATE EPS 1 0.071 0.071 9.71 0.002 Model 5 0.214 0.042 5.86 <0.000 Error 118 0.862 0.007 Corrected Total 123 1.077 *See tables 9-12 for variable descriptions. ** P-Values are two-tailed tests. 4.4.2.8.3 Covariate for the Diffe rence in Credibility Ratings H3 measures the difference in credib ility ratings from period one to period two by examining the difference in participants’ ratings of management’s credibility on the McCroskey and Teven ( 1999) credibility scale. Table 17 shows that only two of the potential covariates were correlated with the difference in credibility ratings.

PAGE 100

89 Table 17: Covariates Correlated with the Difference in Credibility Ratings (H3) Variable* Correlation P-Value** Type of Covariate Reference Table INVESTED IN MUTUAL FUND 0.271 0.002Level of Involvement Table 10 STUDIED SOX -0.186 0.039Recent Events Table 12 ** P-Values are two-tailed tests. A preliminary ANCOVA was run using both the independent variable and both of the significantly correlated covari ates. The results presented in Table 18 show that both the participants’ mutual fund investing experience (F = 9.06, twotailed p= .032) and their studying of Sa rbanes-Oxley (F = 4.17, two-tailed p= .043) were significant (alpha = .10) vari ables in the model. Thus, both covariates are included in the tests of H3. Table 18: Preliminary ANCOVA Testing of Difference in Credibility Ratings (H3) Using Covariates Variable* DF Sum of Squares Mean Squares F Statistic P-Value** FINANCIAL RESULTS LETTER 1 3.866 3.866 2.66 0.105 INVESTED IN MUTUAL FUND 1 13.149 13.149 9.06 0.032 STUDIED SOX 1 6.059 6.059 4.17 0.043 Model 3 24.147 8.049 5.55 0.0013 Error 120 174.160 1.451 Corrected Total 123 205.486 *See tables 9-12 for descriptive statistics of the variables ** P-Values are two-tailed tests. 4.4.2.8.4 Covariates for the Dif ference in Reliance on Management’s Forecasts (H4) In this study, the difference in re liance on management forecasts (H4) is measured as the difference between t he reliance on management’s earnings per

PAGE 101

90 share estimates made in period one and the reliance on management’s earnings per share estimates in period two. To c ontrol for the residual experimental effects of period one in this study, a variable wa s created to measure the usefulness of management’s prediction in period one15. The useful variable was measured as the difference between the participants’ final earnings per share prediction in period one and the actual results of ear nings per share in period one. High (positive) values of the useful variable in dicate that participants’ final estimates of earnings per share were above the actual earnings per share for the period. Low (negative) values of the useful variabl e indicate that t he participants’ final estimates of earnings per share were bel ow the actual earnings per share for period one. For example if a participant selected $2.00 as his/her initial earnings per share estimate, management would pred ict earnings per share for the period of $2.64. The surprise in management ’s prediction is $.64. On average participants in the high intention tone treat ment revised their earnings per share estimates about 33 percent of the surpri se in management’s forecast. So this participant would have adjusted their earni ngs per share estimate to $2.21 ($2.00 initial prediction plus 33 per cent of the $.64 surprise). Actual earnings per share for this participant’s example would be 1. 09 percent of the in itial earnings per share or $2.18 ($2.00 1.09) The useful measure for this example would be $.03 as the participant’s estimate was $. 03 higher than the actual earnings per share. In period two this participant was less likely to rely on management’s forecast than someone with a negative useful score. This was seen in correlation testing where the useful variable was highly correlated in a negative 15 See the discussion in the post hoc analysis se ction as well as in the studies limitations.

PAGE 102

91 direction with the difference in relianc e on management forecasts variable. The higher the useful variable the less reli ance on management’s forecast in period two. Table 19 below presents the actual useful calculations based on the period one treatment conditions. As can be seen from the table, participants in the high intention condition had a mean re vised earnings per share estimate of about $2.21, while participants in the no intention treatment condition had a mean revised estimate of earnings per share of around $2.11. The mean actual earnings per share for the high intenti on group was just over $2.17, while the actual earnings per share for the no intention group was about $2.15. After rounding, the usefulness of management’s forecast was $.04 for participants in the high intention treatment while it was $.04 for participants in the no intention treatment condition.

PAGE 103

92 Table 19: Calculation of the Useful Covariate by Period One Treatment Condition Participants’ Period One Revised EPS Estimate Mean (Std. dev.) Min Max Period One Actual Earnings Per Share Mean (Std. dev.) Min Max Usefulness of Period One EPS Estimate ** Mean (Std. dev.) Min Max Treatment Condition (Period One) $ 2.206 (.264) $ 2.173 (.125) $.037 (2.11) High Intention $1.550 – $2.870 $1.635 – $2.452 $-.616 – $.185 $ 2.113 (.279) $ 2.148 (.230) $-.038 (.153) No Intention $.500 – $2.600 $.545 – $2.398 $-.429 – $.189 Calculated as Participant’s Initial Earnings Per Share 1. 09 ** Calculated as Actual Period One EPS – Part icipant’s Revised Period One EPS Estimate Table 20 presents the descriptive statis tics for the useful variable based on the period two treatment conditions. Participants’ who received the high intention tone financial st atement letter had a positiv e difference (mean =.026) between their earnings per share estimate s in period one and actual earnings per share. The positive difference is indica tive of estimating earnings per share above the actual earnings per share for the period. The participants who received the no intention tone financia l statement letter had a negative difference (mean = -.024) between their earnings per share es timates in period one and the actual period one results. The negative difference indicates that the participants had underestimated actual earnings per shar e. This relationship between the usefulness of period one’s forecast and t he amount of difference in reliance on management’s forecasts in period two is further examined in the post hoc analysis.

PAGE 104

93 Table 20: Descriptive Statistics for the Useful Covariate by Year Two Treatment High Intention Letter No Intention Letter Covariate Mean (Standard Dev) Min Max Mean (Standard Dev) Min Max Useful* 0.026 (0.177) -0.616 0.189-0.024 (0.195) -.494 0.190 The useful variable measured the difference between participants’ second earnings per share estimate in period one and the actual period one earnings per share for the company As can be seen in Table 21, the new variable, useful, was found to be a significantly correlated with the differenc e in reliance on management’s forecasts (H4). In addition to the useful covariate, two other potential covariates were found to be significantly correlated with the depend ent variable at an alpha of .05. The variable that measured participants’ belief that management would inflate its earnings, along with the question asking if participants planned to invest in the stock market. Table 21: Covariates Correlated with the Difference in Reliance on Management Forecasts (H4) Variable* CorrelationP-Value** Type of Covariate Reference Table Useful -0.7880.000 Control Variable for Period One Table 19 EXPECT MANAGEMENT INFLATE EPS -0.2200.014 Reporting and Knowledge Bias Table 11 PREVIOUSLY INVESTED 0.1900.035Level of Involvement Table 10 ** P-Values are two-tailed tests. The results of a preliminary ANCOVA test are presented in Table 22. An ANCOVA was used to examine the covari ates in relationship to the dependent variable while controlling for the rela tionship to the independent variable. Only the useful variable was significant (F = 177.65, two-tailed p< .000), using an

PAGE 105

94 alpha of .10. The two other potential cova riates were insignificant and will not be included in the final test ing of the hypothesis. Table 22: Preliminary ANCOVA Testing Difference in Reliance on Management Forecasts (H4) Using Covariates Variable* DF Sum of Squares Mean Squares F Statistic P-Value** Financial Results Letter 10.0360.0401.3 0.257 Previously Invested 10.0330.0321.08 0.301 Expect Management Inflate EPS 10.0040.0030.12 0.733 Useful 15.4165.412177.65 0.000 Model 46.1331.53350.29 0.000 Error 1193.6280.030 Corrected Total 1239.761 See tables 9-12 for variable descriptions ** P-Values are two-tailed tests. 4.4.3 MANCOVA Testing The dependent variables in this study are examined to determine if they are correlated. When using multiple depe ndent variables that are correlated the MANCOVA model is used to determine the main and interaction effects of the independent variables to the combined dependent variables. Table 23 presents the results of the correl ation analysis on the dependent variables. There are two dependent variables for both periods of the study. In period one the dependent variables are management’s credibility ra ting and the participants’ reliance on management’s forecast. As can be seen in Panel A of Table 23, these two dependent variables are correlated (Pearson -.195, p = .029) indicating that using a MANCOVA model is appropriate. In period two there are also two dependent variables. The first dependent variable is the change in credibility rati ngs and the second dependent variable is the change in reliance on manage ment’s forecasts. As ca n be seen in Panel B of

PAGE 106

95 Table 23, these two dependent variables are not correlated, and therefore are not examined using a MANCOVA model. Table 23: Correlation Between Dependent Variables by Period Panel A: Correlation Between Period One Dependent Variables Variable Correlation P-Value Credibility Rating (H1) & Reliance on Management’s Forecast (H2) -.195 .029 Panel B: Correlation Between Period Two Dependent Variables Variable Correlation P-Value Difference in Credibility Rati ngs (H3) & Difference in Reliance on Management’s Forecast (H4) -.014 .873 Table 24 Panel A presents the result s of the MANCOVA model examining the significance of the internal cont rol letter on both dependent variables from period one. As the table shows, the F statisti c is significant for the internal control letter (Wilks’ Lambda .899, F = 6.46, P= .0 02). Since the intention factor is significant in the MANCOVA it is appr opriate to conduct a separate ANCOVA analysis for each dependent variable. P anels B and C of Table 24 present the separate ANCOVA models for each of the dependent variables. In Panel B of Table 24 the impact of t he internal control letter on credibility rating is significant (F = .7.19, two-ta iled p = .008). Those covariates that are significant (p=.10) will be retai ned for the analysis of H1. As can bee seen in Panel C of Table 24, the impact of the internal control letter on the reliance on managem ent’s forecast is also si gnificant (F = 3.19, twotailed p = .077). Again, those covariat es that are significant (P=.10) will be retained for the analysis of H2.

PAGE 107

96 Table 24: Results for Internal Control Letter’s Impact on Credibility Ratings and Reliance on Management’s Forecast Panel A: MANCOVA Results for Internal Control Letter on Credibility Ratings & Reliance on Management’s Forecast Wilks’ Lambda .899, F Statistic = 6.46, P = .002 Panel B: ANCOVA Results for Internal Control Letter on Credibility Rating Variable DF Sum of Squares Mean Squares F Statistic P-Value* Internal Control Letter 111.03311.0337.19 .008 FRAUD SOX RELATIONSHIP 117.14517.14511.17 .001 CONFIDENCE 18.1268.1265.29 .023 ASNO2 FAMILIARITY 15.9165.9163.85 .052 INVESTED IN MUTU AL FUND 14.6624.6623.04 .084 Previously Invested 19.0379.0375.89 .016 Expect Management Inflate EPS 14.1844.1842.73 .102 Model 775.19110.7417.0 <.000 Error 116178.0971.535 Corrected Total 123253.289 Panel C: ANCOVA Results for Internal C ontrol Letter on Reliance on Management’s Forecast Variable DF Sum of Squares Mean Squares F Statistic P-Value* Internal Control Letter 1.227.2273.19 .077 FRAUD SOX RELATIONSHIP 1.538.5387.56 .007 CONFIDENCE 1.503.5037.06 .009 ASNO2 FAMILIARITY 1.097.0971.35 .247 INVESTED IN MU TUAL FUND 1.071.071.99 .321 Previously Invested 1.125.1251.76 .188 Expect Management Inflate EPS 1.722.72210.14 .002 Model 72.255.3224.52 <.000 Error 1168.262.071 Corrected Total 12310.518 Internal Control Letter: Treatment given with ei ther high intention tone or no intention tone. FRAUD SOX RELATIONSHIP: Question asked t he relevance of Sarbanes-Oxley to fraud. CONFIDENCE: Asked participants their confiden ce in their earnings per share predictions. ASNO2 FAMILIARITY: Asked participants about t heir familiarity with Auditing Standard No. 2. INVESTED IN MUTUAL FUND: Asked participants if they had invested in mutual funds. Previously Invested: Asked participants if they had previously invested in the stock market. Expect Management Inflate EPS: Asked if they expected management to inflate earnings. *P-Values are all two-tailed tests

PAGE 108

97 4.4.4 Statistical Assumptions Testing Since the overall MANCOVA model was significant, the analysis proceeded with ANCOVA models for testi ng the hypothesis for each dependent variable. As a first step, the statisti cal assumptions associated with ANCOVA were evaluated. Therefore, in this sect ion the statistical assumptions regarding each of the dependent variables are tested and discussed. Although different statistical methods may be used to analyze each of the hypotheses, most of the multivariate procedures have simila r assumptions regarding the dependent variables. After examining whether there ar e violations of any of the assumptions underlying the statistical procedure for each hypothesis, the hypotheses will be discussed and tested. ANCOVA tests are most appropriate when looking for main and interaction effects of a categorical independent variable and co variates on a dependent variable. There are three assumptions that should be met for the ANCOVA procedure: 1) each observation s hould be independent, 2) the dependent variables should follow a normal distribut ion, and 3) the variances between the groups should be equal (Hair et al 1998). The accuracy of the ANCOVA procedure is also sensitive to data that is not representative of the sample population (outliers) (Hair et al. 1998). This section proceeds as follows. Each of the assumptions are discussed and tested for each of the depend ent variables in the model. A discussion of the assumption of independent observations is next followed by a discussion of the tests of normality, heteroscedasticity, and outliers.

PAGE 109

98 4.4.4.1 Independent Observations With respect to the assumpti on of independent observations, all participants in this study were randomly assigned to one of t he treatments. No participants were allowed to participate in this study more than once and all participants worked individually. Therefore, each observation is independent of all others. 4.4.4.2 Multivariate Normality Several methods were used to test if the dependent variables followed a normal distribution. Box and whisker plot s and normal probability plots were used to graphically analyze the data. Normal probabi lity plots allow a visual inspection of the data against a theoretic ally normal distribution pa ttern. Statistical methods measuring the skewness and kurtosis of the data were also examined for each dependent variable. Reported kurtosis numbers indicate the peak of the distribution, while skewness numbers i ndicate if the observations fall disproportionately to the left or the right of the distribution. Two statistical tests were also used to determine if a va riable is normally distributed. The Kolmogorov-Smirnov and the Anderson-Darli ng statistics both test if data come from a normal distribution. With respect to the credibility rating (H1), normal probability plots indicated the data were slightly skewed to the right, which was consistent with the skewness statistic of .43. The kurtosis statistic was -.369, which was represented in a graph as a higher peak around the midpoint of the data. The graphical results were consistent with the Anders on-Darling (p=.012) st atistic but not the

PAGE 110

99 Kolmogorov-Smirnov (p= .133) statistics. Wh ile the plots seemed to indicate the data were not normally distributed, the Kolmogorov-Smirnov test failed to find non-normality in the data set. In testing for normality in the reliance on management forecast variable (H2), normal probability plots, a histogram and box and whisker plots indicated the reliance on management forecast dat a were positively skewed (skewness .96), with an overall low peak or very m ild kurtosis (.13). Consistent with the graphical observations, the Kolmogorov-S mirnov (p=.010) and Anderson-Darling (p=.005) tests indicated the data were not normally distributed. An examination of the difference in credibility ratings (H3) variable for normality indicated that the distribution of responses for this variable was not normally distributed. This is indicated graphically by normal probability plots, and a histogram that shows the data follow a normal distribution with a peak showing larger observations above the mean (kur tosis .43) and negative skewness (-.59). Statistical testing also indicated the dat a did not follow a normal distribution with the Kolmogorov-Smirnov (p = .010) and the Anderson-Dar ling (p=.005) tests rejecting the hypothesis that the data were normally distributed. The plots related to the distribution of participant’s difference in reliance on management’s forecasts (H4) indicated the data were positively skewed (.35) with slightly more observations in the upper end of the tail (kurtosis .21). The conclusions reached by the graphical anal ysis were supported by a statistical analysis of the data using the Kolmogor ov-Smirnov (p = .010) and AndersonDarling (p = .005), which rejected the hy pothesis of normally distributed data.

PAGE 111

100 While only one of the four dependent variables met the assumption of normality, the ANCOVA method is robust to even critical violations of the normality assumption (Keppel 1982). The robustness of the ANCOVA methodology with respect to violations of normality is even greater when an equal number of observations per treatm ent group is compar ed, which is the case in this study. For these reasons, no adjustments were made to the data to address the non-normality found in the data. 4.4.4.3. Variance Between Groups Testing the variances between groups invo lves looking at all of the levels of the independent variables to determine if the variance is similar at all levels. When the variance in the dependent variable is similar for all levels of the independent variable, the data are said to be homoscedastic. When there is a different amount of variance in the dependent variable at each level of the independent variable, the data are said to be heteroscedastic. Two tests were conducted to check for heteroscedastici ty. A Levene’s test for equality of variances was used, and a linear relationship was examined between the squared residuals and the predicted va lues for the dependent variables. The second test was conducted since it has already been determined the data are non-normal and a Levene’s test is sensitive to non-normality. Using a Levene’s test of equal vari ances, it was determined that the credibility rating (H1) data did not displa y equal variance at all levels of the independent variable (p =.089) However, less than 1 perc ent of the variation in

PAGE 112

101 the squared residuals was found to be asso ciated with the predicted period one credibility rating variable, indicating no significant variance problems. The Levene’s test was also run for the reliance on management’s forecast (H2) variable. Test results indicated that the distribution of responses displayed unequal variances across treatment conditions (p = .024). This was also found by testing the linear relationship between the squared residuals and the predicted values of the dependent variables. The resu lts indicated that 13.7 percent of the variation in the squared residuals was associated with the variation in the predicted values. In testing the variance between groups fo r the difference in credibility ratings (H3) variable, the Levene’s test indica ted the variance in the data was not consistent at all levels of the i ndependent variable (p = .014). Evidence contradicting the Levene’s test indicated less than 1 per cent of the variance in the squared residuals was associated with t he predicted differ ence in credibility ratings variable. Finally, when testing the variance betw een groups for difference in reliance on management’s forecast (H4), the Lev ene’s test indicated the variance displayed was equal for all four treatment groups (p =.169). Only one of the four variables consis tently displayed heteroscedasticity (reliance on management’s forecast). While it is prudent to exercise caution when interpreting results involving heter oscedasticity, the heteroscedasticity involving reliance on management’s foreca sts has been mitigated by the use of

PAGE 113

102 equal cell sizes (Glass and Hopkins 1996; Garson 2006). Therefore, the heteroscedasticity found in the variables in this study is not a concern. 4.4.4.4. Testing for Outliers Outliers are data points that seem to indicate they may not be representative of the sa mple population. An a ssumption of the ANCOVA procedure is that the data ar e representative of the sample population. To test the data regarding representativeness in relation to the sample population, statistical tests were conducted to fi nd potential outliers in the dataset. A Studentized Residual statistic was used to determine if there were outliers. The Studentized Residual procedure looks at the influence of each data point by removing it from the analysi s and then examining the influence the individual observation had on the over all significance of the model. The calculation for the Studentized Residual statistic divides the deleted residual value by its standard error. A cutoff residual value of +/3.641 is used as a rule of thumb to identify outliers (Mendenhall and Sincich 1996). No outliers were found with respect credibility ratings (H1), t he reliance on management’s forecast (H2), and for the difference in credibility rating (H3). Two outliers were identified for the difference in reliance on management’s forecasts (H4). The model was tested with both values eliminated. There was little effect on the overall model’s significance after removing these observa tions but the observations did have a significant effect on the results of the ANCOVA for the ma in treatment (the financial results letter). The significance leve l of the t-test for the financial results letter went from .125 to 235. Since the results for H4 are insignificant, both with

PAGE 114

103 and without these two observations the obs ervations were kept in the model leaving the model with an equal number of observations per cell. 4.4.4.5 Assumptions Testing Summary While some of the assumptions test ed were found to be violated, the statistical method (ANCOVA) employed in this study is fairly robust with respect to violations of normality and heterosc edasticity when there are equal treatment groups as was the case in this st udy (Glass and Hopkins 1996; Garson 2006). The one instance where outliers were observed (H4) did not impact the interpretation of the results. 4.5 Hypothesis Testing In this section the statistical results for the four hypothes es are presented. Descriptive statistics for the relevant va riables are presented first. Conclusions about the degree of support for the hypothes es are presented in this chapter. The overall conclusions regarding the results of the hy potheses tests are discussed in Chapter 5. 4.5.1 Testing of H1 and H2 The effect of the report on internal c ontrol (either high intention tone or no intention tone) was tested on the dependent variables, cr edibility rating (H1) and the reliance on management’s forecast (H2). The significant covariates identified in section 4.4 were included in each model to account for their potential impact on the dependent variables.

PAGE 115

104 4.5.1.1 Management’s Cr edibility Rating (H1) H1 predicted that users given management’s letter on internal controls with a high intention tone would rate management’s credibility higher than participants given management’s letter on in ternal controls with a no intention tone. To determine the credibility ra ting (H1), participants completed the McCroskey and Teven (1999) credibility sca le. The scale consists of 18 items with six questions for each of three fact ors: intention, tr ustworthiness, and expertise. To simplify the data analysis, the average credibility score was used for testing differences between the groups. Table 25 presents the descriptive statistics for the dependent variable, credibility rating. As can be seen in Table 25, participants who were given management’s report on internal controls with the high intention tone rated management’s credibility higher (mean = 7. 886) than participants who received the report with the no int ention manipulation (mean = 7.088). These means are in line with H1, which predict ed that participants in the high intention treatment would rate management’s credibility hi gher than would participants in the no intention treatment.

PAGE 116

105 Table 25: Descriptive Statisti cs for Credibility Rating (H1) Descriptive Statistics for the Credibility Rati ng by Management Internal Control Letter Treatment High Intention Letter No Intention Letter Covariate Mean (Standard Dev) Min Max Mean (Standard Dev) Min Max Credibility Rating 7.886 (1.517) 5.055 11.0007.008 (1.235) 4.222 11.000 Credibility Rating: The average of the 18 questions from the McCroskey and Teven (1999) Credibility Scale To see if the participants in the hi gh intention internal control letter treatment rated management’s credibility higher than par ticipants who were given the no intention internal c ontrol letter (H1) an ANCOVA was run using the internal control report as the treat ment and credibility rating as the dependent variable. Four covariates identified in Table 14 as possibly impacting credibility rating were also included in the model: participant ’s confidence in completing the task (CONFIDENCE), their views on the Sarbanes-Oxley Act (FRAUD SOX RELATIONSHIP), their prio r mutual fund investment s (INVESTED IN MUTUAL FUND), and their familiarity with Auditing Standard No. 2 (ASNO2 FAMILIARITY). Table 26 shows the resu lts of the ANCOVA model. The overall model is statistically significant (F = 7.94, two-tailed p< .001). Table 26 also indicates t hat the impact of the diffe rent internal control letters on credibility ratings was signifi cant (F = 6.94, one tail p=.006). The significant effect was in the direction hy pothesized (see Table 22), supporting H1. All four of the covariates were signifi cant factors in the model. Three of the covariates were positively correlated with management’s credibility ratings and one was negatively correlated with management ’s credibility rating. Participants’

PAGE 117

106 opinions regarding the Sarbanes-Oxley Ac t (FRAUD SOX RELA TIONSHIP) were significantly (F = 9.20, two-tailed p = .003) associated with credibility ratings, indicating that more knowledge of Sar banes-Oxley led to higher ratings of management’s credibility. Participant s’ confidence (CONFIDENCE) in completing the task was also significant ly (F =4.66, two-tailed p = .033) associated with the ratings of management’s credibility, indicating that more confidence led to higher ratings of management’s credibility. Understanding of Auditing Standard No. 2 (ASNO2 FAMILIARIT Y) was also significantly (F =4.31, two-tailed, p = .040) associated with manag ement credibility ratings, as greater understanding of Auditing Standard No. 2 led to higher credibility ratings. Investment history with mutual funds was significantly (F =8.91, two-tailed p = .004) associated with management’s credibility ratings as well, but higher values on this question led to lower management credibility ratings.

PAGE 118

107 Table 26: ANCOVA Results for Internal Cont rol Letter’s Impact on Credibility Ratings (H1) Variable DF Sum of Squares Mean Squares F Statistic P-Value Internal Control Letter 110.42210.4226.49 .006* FRAUD SOX RELATIONSHIP 114.77014.7709.20 .003 CONFIDENCE 17.4857.4854.66 .033 ASNO2 FAMILIARITY 16.9146.9144.31 .040 INVESTED IN MUTUAL FUND 114.30214.3028.91 .004 Model 563.79012.7587.94 <.000 Error 118189.4991.606 Corrected Total 123253.289 Internal Control Letter: Treatment given with ei ther high intention tone or no intention tone. FRAUD SOX RELATIONSHIP: Question asked t he relevance of Sarbanes-Oxley to fraud. CONFIDENCE: Question asked participants thei r confidence in their earnings per share predictions. ASNO2 FAMILIARITY: Question asked participants about their familiarity with Auditing Standard No. 2. INVESTED IN MUTUAL FU ND: Question asked participants if they had invested in mutual funds *P-Value adjusted for one-tailed test for Internal Control Letter only. 4.5.1.2 Reliance on Management’s Forecast (H2) H2 predicts that users gi ven management’s letter on internal controls with a high intention tone will rely more on management’s forecasts by revising their EPS forecast closer to management’s forecast than participants given the internal control letter with no intention tone. Before te sting H2, the participants’ initial earnings per share estimates were compared between groups. The tone of the internal control letter should not have an effect on the initial earnings per share estimates made by participants. It was important to te st the differences between the groups’ initial earnings per s hare estimates because the tone of the internal control letters is expected to impact the participants’ perception of

PAGE 119

108 management’s credibility, and therefore, t he amount by which participants rely on management’s forecast. The hypothesized relationship is between the two versions of the internal control lett er and participants’ reliance on management’s forecast and not the partici pants’ initial predictions. The results of the study could be impacted if the different internal contro l letters systematically resulted in user’s estimating different initial earnings per share estimates for the company. As seen in Table 27 both groups had similar predict ions for period one. The mean for the high intention treatment was $1.99 and t he mean for the no intention treatment was $1.97, resulting in a statistically insi gnificant (t = .55, two-tailed p = .461) mean difference of $.02. In addition to the period one predicti on by participants, Table 27 also shows management’s predicti on of earnings per share estimates for period one and the participants’ revisions of earnings per share after receiving management’s prediction. Since m anagement’s period one prediction is mathematically derived based on the partici pants’ initial period one prediction, no significant (t = .55, p = .461) difference exists between the two groups concerning management’s predictions of earnings per share for the groups in period one. H2 examines the reliance of par ticipants’ on management’s forecast. Table 2 demonstrated how the participants’ reliance on management’s forecast is calculated. First the difference between the participants’ revised earnings per share estimate and their initial earnings per share estimate is calculated, this is called the difference in EPS estimates. The difference in EPS estimates is then divided by the amount of surprise in management’s forecast s. The amount of

PAGE 120

109 surprise in management’s forecast is measured as the difference in management’s prediction and the participant’s initial earnings per share estimate. The result is the reliance on management’s forecast, which is used as a measure of management’s credibility. For example, if management’s prediction was $.50 higher than the participant ’s initial earnings per share estimate and the participant raised his/her earnings per shar e estimate by $.25, then the reliance on management’s forecast would be 50 percent.16 As Table 27 demonstrates participant s in the high intention treatment relied on management’s forecast and adjust ed their earnings per share estimates by about 34.3 percent ($.22/$.64) of management’s advocated change, while participants in the no intent ion treatment only adjusted their earnings per share estimates by 22.2 percent ($.14/$.63) of management’s advocated change. The greater reliance on management’s forecast found in the high intention treatment was consistent with th e prediction of H2. 16 The percentage change from participant’s first and second estimate was considered as an alternate measure of reliance. However, I belie ve measuring reliance as a percentage of the change advocated by management makes the prac tical explanation of the results clearer. If management’s advocated adjustment (the surprise) is 100%, this variable is a measure of how much of that adjustment the participants believed was necessary.

PAGE 121

110 Table 27: Descriptive Statistics for the Reliance on Management’s Forecasts (H2) Participants’ Initial EPS Estimate Mean (Std. dev.) Min – Max Management's EPS Forecast Mean (Std. dev.) Min – Max Surprise in Management’s Forecast ** Participants’ Revised EPS Estimate Mean (Std. dev.) Min Max Participants’ Reliance on Management’s Forecast Mean (Std. dev.) Min – Max *** Treatment Condition High Intention $1.99 (.115) $2.63 (.151) $.64 (.037) $2.21 (.264) .343 (.328) $1.50 -$2.250 $1.980 – $2.970 $.480 -$.720 $1.550 – $2.870 0 1.225 No Intention $1.97 (.211) $2.60 (.279) $.63 (.068) $2.11 (.279) .222 (.242) $.500 $2.20 $.660 $2.90 $.160 $. 704 $.500 – $2.600 0 .987 Calculated as Participant’s Initial Earnings Per Share 1. 32 ** Calculated as Management’s EPS Forecast – Participant’s Initial EPS Estimate *** Calculated as (Participant’s Revi sed EPS – Participant’s Init ial EPS) / Surprise in Management’s Forecast

PAGE 122

111 The impact of the different internal control letters on the reliance on management’s forecast was tested using AN OVA. Four covariates were included in the model. Recall from Table 16 that participants’ familiarity with SarbanesOxley (FRAUD SOX RELATIONSHIP), conf idence in their earnings per share estimates (CONFIDENCE), their prior history with investing (PREVIOUSLY INVESTED), and their expectation of management inflating earnings (EXPECT MANAGEMENT INFLATE EPS) were signifi cantly correlated with the reliance on management’s earnings per share estimate s; thus, they were included as controls for the test of H2. Table 28 indicates the overall AN COVA model used to test H2 was statistically significant (F = 5.86, two-ta iled p <.000). As reflected in Table 28, the impact of the internal control letters was a statistically significant factor (F = 4.27, one-tailed p = .021) in t he difference in reliance on management’s forecast between participants in the high int ention treatment (34.3 percent) and participants in the no intent ion treatment (22.2 perc ent), supporting H2. In agreement with the information provided in Table 16, all four covariates were significantly (alpha of .10) associat ed with reliance on management’s forecast.

PAGE 123

112 Table 28: ANCOVA Test of Reliance on Management’s Forecast (H2) Variable DF Sum of Squares Mean Squares F Statistic P-Value Internal Control Letter 10.3050.3054.27 0.021* FRAUD SOX RELATIONSHIP 10.4510.4516.31 0.014 CONFIDENCE 10.4910.4916.88 0.010 EXPECT MANAGEMENT INFLATE EPS 10.6930.6939.71 0.002 PREVIOUSLY INVESTED 10.2420.2423.39 0.066 Model 52.0900.4185.86 0.000 Error 1188.4270.071 Corrected Total 12310.518 Internal Control Letter: Treatment given with ei ther high intention tone or no intention tone. FRAUD SOX RELATIONSHIP: Question asked t he relevance of Sarbanes-Oxley to fraud. CONFIDENCE: Question asked participants thei r confidence in their earnings per share predictions. EXPECT MANAGEMENT INFLATE EPS: Question asked participants if they expected management to inflate their earnings per share predictions. PREVIOUSLY INVESTED: Question ask ed participants if they had ev er made an investment in the stock market. P-Value adjusted for one-tailed test. Three of the covariates were negativ ely correlated with the participants’ reliance on management’s forecast and one covariate was positively correlated with reliance on management’s forecast. The three negatively correlated covariates were all significant with res pect to the model and in cluded participants’ feelings regarding the Sarbanes Oxle y Act (FRAUD SOX RELATIONSHIP) (F=6.31, two-tailed p= .014), their c onfidence (CONFIDENCE) in their earnings per share predictions (F= 6. 88, two-tailed p= .010), and their expectations that management would inflate earnings (EXPECT MANAGEMENT INFLATE EPS) (F= 9.71, two-tailed p= .002). As the response values to these questions increased, participants’ reliance on management’s forecasts decreased.

PAGE 124

113 Intuitively this made sense for the c onfidence (CONFIDENCE) variable and the expectation of management to inflate it s earnings per share estimates (EXPECT MANAGEMENT INFLATE EPS). The more confident participants were in their selection of earnings per share the less likel y they were to revi se their forecast. Also, the more participants expected m anagement to inflate its earnings per share estimates, the less likely they were to revise their own earnings per share estimates after receiving management’s There is no intuitive reasoning for participant’s beliefs about the relev ance on the Sarbanes Oxley Act (FRAUD SOX RELATIONSHIP) to reduce their reliance on management’s forecasts. Participants’ prior investing expe rience (PREVIOUSLY INVESTED) was significant (F= 3.39, two-tailed p=.066) and positively correlated with the reliance on management’s forecast, indicating parti cipants with more investing experience tended to rely more on management’s forecasts. 4.5.2 Testing of H3 and H4 For period one, it was predicted that a letter from management with the high intention tone would lead to great er reliance on management’s earnings per share forecast. In part, this was due to t he fact that the parti cipants in period one had no prior information regarding managem ent’s past forecast accuracy. Prior research has shown that management’s past forecast accuracy can impact management’s credibility ratings (W illiams 1996; Hirst et al. 1999). When management fails to forecast accurately it s credibility drops and so too does the market’s reliance on management’s forecasts. All of the participants in period two of the study knew of management’s failure to accurately forecast its earnings per

PAGE 125

114 share in period one. Therefore, it was expected that all of the participants’ ratings of management’s credibilit y would drop because managem ent failed to meet its predicted earnings per share estimate fr om period one. Period two of the study examines the impact of manipulating management’s tone in a communication with investors when investors have prior knowledge of management’s inability to forecast accurately. For H3 and H4 the intention factor of management’s credibility is manipulated using the tone of a communication from management to recipients who have received only inaccu rate past forecasts from management. In period one, half of the participants in the study either received management’s statement of internal contro ls with a high or no intentio n manipulation. In period two, the participants received a le tter from management communicating the actual earnings results from period one of the study. The intention factor of credibility was manipulated in the actual earnings letter at either a high intention or no intention tone. 4.5.2.1 The Difference in Credibility Ratings from Period One to Period Two (H3) Hypothesis three examines the impact of the communication tone (high intention vs. no intention) in the financ ial results letter on the difference in participants’ rating of management’s credibi lity from period one to period two. When management failed to meet its fo recast, the prediction was that management would experience a smaller di fference in credibility ratings from participants who received an earnings letter with the high intenti on manipulation. Each participant completed the credib ility scale twice; once after reading the internal control letter from managem ent and once after reading the letter from

PAGE 126

115 management explaining the comp any’s failure to forecast earnings per share. To test H3, participants’ ratings of managem ent’s overall credibility in period two were subtracted from participants’ overall ratings of credibility in period one. Table 29 presents the credibility ra ting scores for period one and period two, as well as the difference in overa ll credibility ratings from period one to period two. As can be seen in Table 29, participants who received the high intention financial results letter in period two reduced their rating of management’s credibility less than partici pants who received the no intention financial results letter. T he participants who received t he high intention financial results letter in period two rated managem ent’s mean credibility 7.513 in period one and 7.009 in period two; thus, the decli ne in management’s credibility ratings is .504. The participants who received t he no intention financial results letter in period two rated management’s credibi lity 7.461 in period one and 6.594 in period two; thus, the decline in credibility ratings is .867. The participants in the no intention treatment reduc ed their ratings of management credibility by more than participants in the hi gh intention treatment.

PAGE 127

116 Table 29: Descriptive Statisti cs for the Difference in Credib ility Ratings from Period One to Period Two (H3) Credibility Ratings for Period One and Peri od Two and the Difference in Credibility Ratings from Period One to Period Two Treatment Condition Period One Credibility Rating Mean (Std. dev.) Min Max Period Two Credibility Rating Mean (Std. dev.) Min Max Difference in Credibility Ratings Mean (Std. dev.) Min Max 7.513 (1.450) 7.009 (1.876) 0.504* (1.466) High Intention 4.222 11.000 3.281 10.940 -4.333 2.166 7.461 (1.431) 6.594 (1.764) 0.867* (1.017) No Intention 5.000 10.611 3.333 10.500 -3.500 .500 62 participants in each treatment condition Positive means indicate a gain of credibility Prior testing (Table 17 and 18) rev ealed that two covariates should be included in the analysis of H3. The two co variates are study of Sarbanes-Oxley (STUDIED SOX) and investment in mutual fund (INVESTED IN MUTUAL FUND). These covariates are included in the ANCOVA model reported in Table 30, which displays the results of testi ng the difference in credibility ratings between the high intention and no intention tone financial results letters for period two. The model is significant (F= 5.55, two-tail p=.002). In support of H3, the financial results letter indicates a signi ficant (F= 2.66, one-tailed p = .053) difference in credibility ratings. Both covariates were significant with respect to the model. Student’s history with studying Sarbanes-Oxley was significantly (F= 4.17, two-tailed p= .044) associated with the difference in cr edibility ratings, indicating that more knowledge of Sarbanes-Oxley led to la rger drops in rating management’s credibility. There was also a significant (F = 9.06, two-tailed p= .032) difference in responses for students who had previously invested in mutual funds. Those with

PAGE 128

117 previous investment experi ence were also more likel y to drop their rating of management’s credibility from period one to period two. Table 30: ANCOVA Test of Differ ence in Credibility Ratings (H3) Variable DF Sum of Squares Mean Squares F Statistic P-Value Financial Results Letter 13.8663.8662.66 0.053* STUDIED SOX 16.0596.0594.17 0.044 INVESTED IN MUTU AL FUND 113.14913.1499.06 0.032 Model 324.1478.0495.55 0.002 Error 120174.1601.451 Corrected Total 123205.486 One tailed P-Value Financial Results Letter: Treatment given with either high intention tone or no intention tone. STUDIED SOX: Question asked partici pants if they studied Sarbanes-Oxley. INVESTED IN MUTUAL FU ND: Question asked participants if they had invested in mutual funds 4.5.2.2 The Difference in Relianc e on Management’s Forecast (H4) Hypothesis four examines the difference in reliance on management’s forecast from period one to period two of the study. It was predicted that the communication tone (high intent ion vs. no intention) of t he financial results letter would impact the differenc e in reliance on management’s forecast from period one to period two. While H3 examines t he loss in the creditability rating scores from period one to period two, H4 ex amines the difference in reliance on management’s forecast from period one to period two. It was predicted that participants who received the high intenti on financial results letter would have a smaller decline in reliance on managem ent’s forecasts after receiving management’s estimate than the participant s who received the financial results letter with no intenti on manipulation.

PAGE 129

118 In each period of the study parti cipants’ reliance on management’s forecasts was indicated by the revision in their initial earnings per share estimates after receiving management’s fore cast. As defined in the testing of H2 and demonstrated in Table 2, the relianc e on management’s forecasts is first measured by subtracting the participants’ revised earnings per share estimate from their initial earnings per share esti mate, and then dividing that difference by the amount of surprise in management’s forecasts. The amount of surprise in management’s forecast is measured as the difference in management’s prediction and the partici pant’s initial earnings per share prediction.17 The result of the calculation is the percentage change in forecasts, which is used as a measure of reliance on management’s fo recasts. The reli ance on management’s forecast was made for both periods one and two and the difference in reliance on management’s forecast (H4) is th en measured as the percentage change (revision) from period two subtracted from the percentage change (revision) in period one. Figure 1 demonstrates the research model used in the study. Only two factors influence the amount of investors’ belief revision: surprise and credibility. By holding the amount of surprise const ant and then comparing the difference in the amount of belief revision between peri od one and period two of the study it is possible to attribute the belief revision to the decrease or loss in management’s credibility. For example, if a partici pant had relied on management’s forecast and 17 The percentage change in participant’s earnings pe r share estimates is calculated as (Revised earnings per share Initial earning per share) / (Management’s predicted earnings per share – Initial earnings per share.)

PAGE 130

119 revised his/her earnings per share esti mate by 30 percent of what management advocated in period one and only 20 percent in period two, the participant’s reliance on management’s information dr opped. Since the amount of surprise was held constant the change in beliefs must be due to a drop in credibility as less reliance was placed on management’s surprise information in period two. Table 31 provides the descriptive statis tics for the difference in reliance on management’s forecast between groups. T he participants who received the financial statement letter with the high intention m anipulation revised their earnings per share estimates in per iod one by 24.0 percent of the change advocated by management. In period two t hey revised their earnings per share estimates by only 10.8 percent of managem ent’s advocated change. Overall, the participants who received the high int ention financial statement letter had a decline in reliance on management’s forecast s of 13.2 percent. In contrast, the participants who received the financial statement letter with no intention manipulation revised their earnings per share estimates in period one by 31.3 percent of the change advocated by manage ment, and in period two they revised their earnings per share estimates by 15.7 percent of management’s advocated change. Overall the participants who receiv ed the no intention financial statement letter had a decline in reliance on managemen t’s forecasts of 15.5 percent from period one to period two.

PAGE 131

120 Table 31: Descriptive Statistics for the Difference in Reliance on Management’s Forecasts (H4) Reliance on Management’s Forecast for Period One & Period Two and the Reduction in Reliance Between Periods Treatment Condition Period One EPS Adjustment Mean (Std. dev.) Min – Max Period Two EPS Adjustment Mean (Std. dev.) Min Max Mean Difference in EPS Adjustment Mean (Std. dev.) Min Max 0.240 (.278) 0.108 (.149) 0.132* (.267) High Intention 0 1.225 0 0.614 -0.510 0.864 .313 (.304) .157 (.220) .156* (.297) No Intention 0 – 1.000 0 1.007 -0.625 0.792 62 participants in each treatment condition Positive means indicate a gain of credibility To test H4, an ANCOVA model was us ed to determine if the difference between the groups receiving the high int ention letter and the no intention letter was statistically significant. The dependent variable used in the model was the difference in reliance on management’s fo recasts from period one to period two and the independent variable was the financ ial statement letter. One variable was included as a covariate (useful) to control for differences found between participants’ estimates of earnings per s hare and the actual earnings per share for period one.18 Table 32 displays the results of the ANCOVA model. The model is significant (F= 100.77, two-tailed p-val ue <.000). While the model is significant, the financial statement letter is insigni ficant (F = 1.34, one-tailed p=.125). 18 For more information on the useful variabl e refer to section 4.4.2.8.4 and Table 19.

PAGE 132

121 Table 32: ANCOVA Results for Difference in Reliance on Management’s Forecast (H4) Variables DF Sum of Squares Mean Squares F Statistic P-Value Financial Results Letter 10.0400.0401.34 .125* Useful 16.0836.083200.99 < .000 Model 26.0993.049100.77 < .000 Error 1213.6620.030 Corrected Total 1239.761 Financial Results Letter: Given at two le vels high intention and no intention tone. Useful: The usefulness of the period one forecast. One tailed P-Value The covariate (useful) was found to be significant (F=200.99, two-tailed p<.000) and had the greatest influence on the participant’s loss of credibility in period two. The useful variable meas ured the difference between participants’ final earnings per share estimate in peri od one and the actual earnings per share for the company. Positive val ues of the useful variable indicate that participants’ final earnings per share estimate was abov e the actual earnings per share for the company. Negative values of the useful variable indicated participants’ final earnings per share estimates were below the actual earnings per share for the company. Participants with la rger positive values of the useful variable relied less on management’s forecast in period two.

PAGE 133

122 4.6 Post Hoc Analysis 4.6.1 Overview of Post Hoc Analysis In this section of the study I will ex amine other impacts of the treatments on participants’ judgment and decision making. The post hoc analysis will proceed as follows. In period one, H1 predi cted that the internal control letter would impact the rating of management’s credibilit y. Since the “management credibility” construct actually comp rises three sub-factors—intention, trustworthiness, and expertise—the impact of the internal control letter (high intention vs. no intention) on each of thes e three sub-factors will be examined in the post-hoc analysis. Also discussed will be the unintended tr eatment created by the design of the study. The impact of the unintended treatment on t he results of period two was included in the model for H4 as a cova riate called useful. The useful variable will be examined further in this section. 4.6.2 The Impact of Altering T one on the Perception of Management’s Credibility When the participants filled out the credibility scale in period one, they completed 18 questions that load on three sub-factors of credi bility. In period one, an average credibility score compri sed of all 18 questions was used to examine the differences between groups. In this section a MANOVA test is used to determine the impact of the management internal control letter on each of the average scores for the three sub-factors of credibility (intention, trustworthiness and expertise). The results of the ANOVA tests are pr esented in Table 33. Recall

PAGE 134

123 from Table 25, participants in the high intention tr eatment gave management a mean credibility rating of 7.89 compared to those in the no intention treatment who rated management’s average credibility at 7.01. These results were similar for the three sub-factors of credibility. With respect to the intention factor of credibility participants in the high int ention treatment ra ted management’s intention an average of 7.66 while participa nts in the no intention treatment rated management’s intention an average of 6. 66. With respect to management’s trustworthiness, participants in the hi gh intention treatment rated management’s trustworthiness 7.68 on average while par ticipants in the no intention treatment rated management’s trustworthiness 7. 02 on average. There was also a significant difference with respect to management’s expertise ratings. The participants in the high intention group rated management’s expertise an average of 8.31 while participants in the no intention group rated management’s expertise 7.58 on average. The letters from managem ent with a manipulated intention tone significantly (p = .05) impacted each of the sub-factor s of management’s credibility.

PAGE 135

124 Table 33: Descriptive Statistics for Sub-Factors of Credibility Treatment High Intention No Intention Comparison Credibility Question (1=Less 11=More) Mean (Std. Dev.) Mean (Std. Dev.) F Statistic p-value* Average Intention 7.669 (1.768) 6.664 (1.311) 12.93 < 0.000 Average Trust 7.682 (1.679) 7.024 (1.303) 5.95 0.008 Average Expertise 8.306 (1.522) 7.578 (1.400) 7.7 0.003 One tailed p-value 4.6.3 Examination of Period Two Result s as a Function of the Usefulness of Period One’s Prediction In period two, it was expected that the participants who received the financial statement results le tter with the high intention tone of credibility would have a smaller change in reliance on management’s forecast than the participants who received the financial resu lts letter with no intention tone. Upon further examination it was determined that the most significant factor impacting the results for H4 was t he period one forecast. The study was originally designed as a one period study examining the impact of altering the tone of the comm unication from management to investors on investors’ ratings of management’s credibility and reliance on management’s forecast. The design of the study was then expanded to a second period. For period two, it was decided that for both treatment groups, management would fail to meet its forecasted re sults for period one, thus lowering the credibility of management. A written communication fr om management would be used to attempt to reduce the loss of credibility resulting from failing to meet forecasted

PAGE 136

125 earnings. It was expected that altering the tone of the written communication would reduce the loss of credibility. In period one of the study, participants were allowed to freely pick their initial earnings per share estimates, and management’s prediction of earnings per share, as well as the actual period one results, were based on a constant percentage of each participant’s init ial earnings per share estimate. Management’s forecasts were always 132 per cent of the parti cipants’ initial earnings per share estimates and actual re sults were always 109 percent of the initial earnings per share estimate. Fo r example, if one par ticipant predicted earnings per share in period one of $1. 00, management’s prediction of earnings per share would be $1.32 and t he actual results for the period would be reported as $1.09. A second participant could select an initial earnings per share estimate of $2.00 and be told management had predict ed earnings per sh are of $2.64 with the actual results for period one at $2.18. The comparisons across all treatment groups were made based on percent age changes and not absolute dollar amounts. So if the participant in example one had revised his/her initial estimate to $1.16 and the participant in example two had revised his/her initial estimate to $2.32 then both participants had revised t heir predictions by 50 percent of management’s recommendation. A comparis on of initial earnings per share estimates by participants in period one revealed no difference between the groups in the high and low condition. Therefore, both groups effectively began period two at the same point since the actual period one earnings were based on the participant’s initial earnings per s hare estimate from period one.

PAGE 137

126 While participants in period one began t he study at about the same place with respect to earnings per share estima tes (no difference in their earnings per share between groups), they were not at the same pl ace at the conclusion of period one of the study. The participants in the high intention treatment had revised their initial earnings per share estimates (34 percent) more than the participants in the no intent ion treatment (22 percent). An example of the difference caused by the revision in per iod one can be explained as follows. Using two participants “X1” and “X2,” assume X1 is assigned to the high intention treatment in period one and X2 is assigned to the no intention treatment in period one. Both X1 and X2 predict their initia l earnings per share estimate as $1.00. They will both receive the same earni ngs per share estimate from management of $1.32 and actual earnings for period one will be reported as $1.09. Suppose, participant X1 revised his earnings per share by 34 percent of the surprise information given by management or $. 11 (34 percent x $.32 = $.11) while participant X2 revised her initial predi ction by 22 percent of the surprise information given by management or $.08 (22 percent x $.32= $.08). When the actual results for period one are report ed at $1.09, participant X1, in the high intention group, had an estimate that was hi gher ($1.11) than the actual results of $1.09 while X2 had a prediction that was lo wer ($1.08) than the actual results. It was assumed that all participants woul d begin the period using the actual earnings from period one as a basis for the period two tasks. The difference between each participant’s period one revi sed estimate and the actual earnings per share for period one might inadvertent ly have impacted the effectiveness of

PAGE 138

127 the treatments in period two. The di fference found between the revised period one estimates and actual period one result s impacted both participants’ initial earnings predictions for period two and the amount of reliance on management’s forecasts for period two. The impact of this unintended variable can actually be explained with prior findings from accounting literature. W illiams (1996) examined the relationship between the usefulness of prior earni ngs forecasts by management and analyst revisions to current forecasts. In her study, Williams give s an example of the usefulness of a prior forecast where two companies (firm A and firm B) made earnings predictions of $2.75 and $2.50, respectively. The actual earnings per share of each company was $3.00, so Fi rm A’s forecast was deemed more accurate. To differentiate accuracy fr om usefulness, W illiams furthers the example by supposing that analysts had es timates of earnings for Firm A of $2.90 ($.10 lower than actual earnings) and Firm B of $2.00 ($1.00 lower than the actual earnings). For Firm A, managem ent’s forecast was $.15 lower than analysts’ forecast and for Fi rm B management’s forecast was $.50 higher than analysts’ forecasts. Thus, in this example, Firm B’s forecast wa s more useful to analysts since it provided more informati on based on analysts’ current level of belief, even if it was not more accurate. Data from the current study can be us ed to test Williams (1998) notion of a difference between forecast usefulness and forecast accuracy. In period one of this study management’s forecasts and the actual results for period one were a fixed percentage of each parti cipant’s initial earnings per share estimate. The

PAGE 139

128 accuracy of management’s forecast was c onstant between all of the participants in the study. What varied between partici pants in period one of the study was the participants’ revised expectations of ear nings per share. T herefore, for each participant we can determine the usef ulness of management’s prediction by examining the difference between each participant’s revised (or expected) earnings per share and the actual ear nings per share for period one. Participants who were in the high in tention treatment for period one had a mean difference of .033 above the actual earnings per share, while participants who were in the no intention treatment in period one had a mean difference of .035 below the actual earnings per shar e. This difference between the two treatments was statistically signif icant (F= 4.15, one-tailed p=.022). In examining the percentage of a ccuracy for management’s prediction, both of the treatment gr oups had management predictions that were identical. Management’s initial prediction of earnings was 132 percent of participants’ initial earnings per share estimate and the act ual earnings were 109 percent of the participants’ initial earnings per share estimate for both groups. The expectation of earnings per shar e for the participants in the high intention group was above the actual earni ngs per share for period one, while the expectations of earnings per share for the no intention tone group was lower than the actual earnings per share for period one The data indicate that participants whose expectations were above the ac tual earnings per share in period one predicted lower earnings per share estimate s for period two but they also did not revise their earnings per share estima tes in period two as much as the

PAGE 140

129 participants whose expectations of earni ngs per share were below the actual results for period one. Stated differ ently, participant’s who had experienced management’s earnings per share estimates that were below t he actual earnings per share estimates in period one found t he period two management forecast of earnings per share to be more useful Since accuracy was held constant between treatment conditions, these results indicate that the usefulness of the forecast and not the forecast accuracy was driving the results in period two. These findings add some support for t he Williams (1998) proposition that forecast usefulness and forecast a ccuracy are two separate constructs.

PAGE 141

130 5.0 Summary and Conclusion 5.1 Summary of Study This study was designed to examine t he impact of the int ention factor of management credibility on investors’ decis ion making. Credibility is a latent variable with three sub-factors: experti se, trustworthiness, and intentions. The intention factor of credibility is a per ception variable that measures perceived understanding, empathy, and responsiveness of a communicator. The study manipulated management’s intention via written communications with participants. The impact of manipulat ing the percepti on of management’s intention factor of credibility wa s then examined using both ratings of management’s credibility and by exam ining participant’s reliance on management’s forecasts of earnings per share estimates. There were four hypotheses tested in this study. In H1 and H2 participants had no information regarding management’s prio r forecast history. In H3 and H4 the changes from period one to period tw o were examined after participant’s experienced management failing to meet its forecast from period one. As stated above, in period one of the study, participants had no prior knowledge of management’s forecast accura cy. The participants, representing average investors, were given backgr ound financial information regarding the company. Included in the financial in formation was a letter from management

PAGE 142

131 regarding the effectiveness of internal c ontrols for the company as required by the PCAOB. One half of the participants received a letter that had a high intention tone, while the other half rece ived a letter with no intention tone. H1 predicted that participants who received the high intention tone letter would rate management’s credibility higher than parti cipants who received the letter from management with no intention manipulati on. The findings from the study supported the prediction of H1, in t hat participants who received the high intention letter from management rated management’s credibility significantly higher than the participant’s who received the letter wit h no intention tone. These findings were statistically significant at an alpha of .10. This suggests that management can increase its credibility by communicating with a tone that implies understanding, empathy, and res ponsiveness to investors’ concerns. In addition to testing the rating of management’s credibility, a second hypothesis was also tested in period one. Participants in both groups (high intention vs. no intention tone) completed a task in which they predicted earnings per share for the company for period one. After making their earnings per share predictions they receiv ed management’s prediction of earnings per share. Participants were then given an opportunity to revise their earnings per share prediction. H2 predicted that participant ’s who had received the high intention letter would rely more on management’s pr edictions by revising their earnings per share estimates closer to ma nagement’s than those participants who received the letter from management with the no intentions manipulation. Support was found for H2 as participant s who had received the high intention

PAGE 143

132 treatment revised their earnings per shar e to a greater degree than those who received the letter with no intention tone. In period two of the study, H3 and H4 we re tested. Period two of the study begins with the participants from period one receiving a letter from management stating the actual earnings per shar e for period one. For all participants management failed to meet its forecasted estimate. Failing to meet its forecasted estimate of earnings per share should r educe management’s cred ibility (Williams 1996; Hirst et al. 1999). The manipulatio n for H3 and H4 had one half of the participants receive the earnings letter with a high intention tone while the other half of the participants received a letter wit h the no intention tone. Differences in credibility ratings and reliance on management ’s forecasts from period one to period two of the study were tested. Hypothesis three predicted that t he participants who received the financial results letter from management with the high intention tone would not lower their rating of management’s credibility as much than those who received the financial results letter from management with the no intention tone. The hypothesis was tested by having each participant rate management using the same credibility scale in both periods one and two. To compare the difference in credibility ratings, a difference score was calculated using period one and two credibility ratings. Higher difference scores indicated great er loss of credibility. In looking at the difference in credibility ratings bet ween the treatment groups (Table 29) a statistically significant difference was found, providing suppor t for H3 in that participants who received the high intent ion financial results letter reduced their

PAGE 144

133 perceived credibility of management less than participant’s receiving the no intention tone letter. In addition to testing the difference in credibility ratings between periods one and two, the difference in the relianc e on management’s forecast was also tested between period one and period two (H4). It was expected that when management failed to meet its earnings per share estimates that participants would reduce the amount by which they relied on management’s forecasts. H4 predicted that participants who received th e earnings letter with the high intention tone would reduce their reli ance on management’s forecast less than participants who received the earnings letter with no in tention tone. However, the results indicated that the tone of t he financial results letter was an insignificant factor in determining the loss of credibility from period one to period two of the study. Thus, there was no support for H4. Table 34 summarizes the overall resu lts of the study. The study found support for H1 and H2 suggesting that management can influenc e the perception of its credibility by the tone it uses when communicating with investors. The increase in credibility was seen in both participants’ ratings of management’s credibility and the reliance on management’s forecast when predicting earnings per share. The study also found s upport for H3, which posited that when management fails to meet its earnings per share estimates, it can mitigate its loss of credibility by altering the tone of its written communications with investors. The impact of the financial results letter di d significantly impact the difference in credibility ratings from periods one and tw o of the study. However, the financial

PAGE 145

134 results letter was not a significant factor in the amount of reliance on management’s forecasts (H4) as there wa s no significant difference between the treatment groups in period two. An unintentional finding of the st udy involves the usefulness of management’s forecast. After rece iving information from management, participants in the high intention treatment in year one had revised their earnings per share estimates above the actual ear nings per share for the period while participants in the no intent ion treatment had revised their earnings per share estimates below the actual earnings per share estimates for the period. The impact of the difference between the year one treatment groups could be seen in the revisions of earnings per share predict ions in year two. Participants who over-relied on management by selecting an estimate of earnings per share higher than the actual earnings per shar e for period one seemed to rely less on management’s forecast in period tw o. Participants who under-relied on management’s forecast, and who subsequ ently had earnings per share estimates lower than the actual earnings per share for period one, tended to rely more on management’s forecast in period tw o. This finding gives support to the findings from Williams (1996) who us ed market data to determine if the usefulness of management’s forecast was more important than the accuracy of management’s forecast.

PAGE 146

135 Table 34: Summary of Hypotheses and Findings Hypothesis Dependent Variable Independent Variable Covariates Overall Model* Hypothesized Effect** Table Reference H1 Credibility Rating Internal Control Letter FRAUD SOX RELATIONSHIP CONFIDENCE ASNO2 FAMILIARITY INVESTED IN MU TUAL FUND F= 7.94 p<.000 F = 6.49 p =.006 Table 26 H2 Reliance on Management’s Forecast Internal Control Letter FRAUD SOX RELATIONSHIP CONFIDENCE EXPECT MANAGEMENTINFLATE EPS F = 5.86 P <.000 F = 4.27 p = .021 Table 28 H3 Difference in Credibility Rating Earnings Results Letter STUDIED SOX INVESTED IN MU TUAL FUND F = 8.049 p = .002 F = 2.66 p = .053 Table 30 H4 Difference in Reliance on Management’s Forecast Earnings Results Letter Useful F= 100.77 p = <.000 F = 1.34 p = .125 Table 32 Two tailed test ** One tailed test (adjusted for directional hypotheses)

PAGE 147

136 5.2 Limitations This study was designed as a labor atory experiment. With appropriate controls for the effects of extraneous vari ables, the laboratory experiment sought to maintain high internal validity. Ho wever, laboratory experiments may have lower external validity than field studies in which a real business task is being performed by experienced decision makers. The lower external validity might limit the generalizability of the findings. The experiment was designed as an abstraction of a task an investor might face. Participants were given limited background regarding a fict itious company. Although the limited information in the task reduces the external validity of the study’s findings, it is necessary to reduce the amount of variati on for each participant to maintain internal validity at as high a level as possible. In this ta sk all participants were given the same background information and financial stat ements. The only differences between experimental materials were the treatm ent effects. Theref ore, given that participants were randomly assigned to tr eatments, any observed variation in participants’ responses should be due to ei ther random (uncontro lled) individual differences between participants or the treatment conditions. Individual differences can be controlled thr ough randomizing the participants into treatments and analyzi ng the tested differences post hoc using demographic questions as possible covariates. The indi vidual differences measured and tested as covariates in this study did not alter the study’s findings.

PAGE 148

137 The use of students as participants in experiments can sometimes pose a threat to validity. However, in this study the use of student subjects does not represent a limitation. The study was des igned to examine a theoretical link between management’s perceived intenti ons based on the tone of its written communications and investors’ willingness to rely on management’s guidance. Prior research (Ashton and Kramer 1980; Li bby et al. 2002) indicates that when a theoretical link is being examined st udents can be appropriate participants. Even if this study employed professional st ock analysts, the amount or percentage of their adjustments to their earnings per share predicti ons could not be used to measure or predict future adjustment percentages but could only be used to show support for the theory that future judgments will be impact ed by perceptions of management’s credibility. Another limitation in this study is a pr oblem in the design of the study that was found only after all of the data were collected. The study was originally designed as a one period study to test the impact of altering the intention tone on the perceptions and amount of revision in estimates when participants had no prior knowledge of management’s prior fore cast accuracy. It was decided that since the participants would complete the study in a short amount of time, it would be reasonable to ext end the study to examine the impact of altering the tone of communications when the parti cipants have prior knowledge that management has failed to forecast accura tely. In period one of the study the participants were allowed to select their initial earnings per share estimates. To control for the amount of surprise in management’s information between

PAGE 149

138 participants, the forecast given by management was calculated as a percentage of the participant’s initial earnings per share estimate (132 percent). Participants were then allowed to alter their forecast after receiving management’s estimate. When the study was expanded to include the second period, a decision was made to hold the accuracy of management’s period one prediction to 109 percent of the participant’s initial earnings per share estimate. Thus, all participants would receive forecasts from management that were inaccurate by a constant percentage of management’s earnings per share estimate. What was not considered in the design of the study was how the difference between participants’ revised earnings per shar e estimates and the actual results of earnings per share for the period would im pact participants’ future reliance on management’s forecast. The participants who were in the high intention condition in period one revised their earnings per share estimate closer to management’s earnings per share estimate than the par ticipants in the no intention treatment. Management’s estimates of earnings per shar e as well as the actual results of the period were calculated based on the par ticipants’ initial earnings per share estimates. There was no difference between the treatment groups in the initial earnings per share estimates, but there we re differences in the revised earnings per share estimates. When the actual re sults for period one were reported, the mean earnings per share estimate for the participants in the high intention treatment were higher than t he actual earnings per share for the period, while the mean earnings per share estimates for the participants in the no intention treatment were lower than the actual earnings per s hare for the period. As a

PAGE 150

139 result, the participants did not begin peri od two on an even basis. The difference between the actual earnings per shar e for period one and the participants’ revised earnings per share estimates in period one drove the behavior of the participants in period two more than any of the treatments. The flaw in the design of the study may explain the lack of findings in period two. The unintended effect was much more powerful than the treat ment effect designed in the study. 5.3 Contributions The findings from this study offer c ontributions to accounting research, accounting policy makers and to the psychology literature on persuasion and credibility. In accounting research, this study expands upon two prior research streams in accounting. With respect to the introduction of credibility scales in accounting, this study builds upon the findings of Mercer (2004, 2005) by expanding the two factor model of cr edibility, consisting of expertise and trustworthiness, to the three factor m odel of credibility, which includes the intention factor. The findings from per iod one of this study suggest that the intention factor of credibility shoul d also be considered when measuring management’s credibility. Also with respect to accounting re search, an unintended consequence of the design of this study (as discussed in the limitations and post hoc analysis sections) offers experimental support for the findings of Willi ams (1996) in which the usefulness of a prior earnings fore cast issued by management impacts the reliance on future forecasts. While prio r research had used the accuracy of prior forecasts, Williams (1996) s uggested that the usefulness of a forecast and not its

PAGE 151

140 accuracy would determine future belie f revision. Williams (1996) defined usefulness as a forecast which improved upon initial earnings expectations. In this study the accuracy of management’s earnings per share estimates were identical between treatment groups, howev er, the participants who had predicted the earnings per share higher than the actual results (the high intention treatment group) relied less on management predictions in period two of the study than the participants with predictions of earnings per share less than the actual results (the no intention period one group). These findings indicate that forecast usefulness as opposed to accuracy is a better indicator of future forecast revisions when management issues guidance. This study contributes to accounti ng policy making. The PCAOB’s (2004) Auditing Standard No. 2, “An Audit of In ternal Control over Financial Reporting Performed in Conjunction with an Audit of Financial Statements” leaves the wording of the report up to management. The choice to leave the wording of a mandatory report to management could lead users to different decisions based on the wording used in the management reports In this study altering the tone of communications with management was enough to impact management’s credibility ratings and the amount of reliance participants placed on management’s forecast estimates. These findings suggest that more research should be conducted to determine the impact of wording on the manner in which decision-makers use accounting reports. It is important to note that the different reports did not impact the par ticipant’s initial earnings per share estimates, only the amount by which participants relied on management forecasts. While this

PAGE 152

141 seems insignificant, it suggests that res earch looking at the impact of different types of reports, such as differences in audit reports versus reviews and compilations, may not have been properly de signed to measure the impact of the reports on investors’ decision-making.19 Many of these studies failed to find differences between decision makers when making immediate decisions such as the decision to lend, the interest rate at which to lend, and the maximum loan amounts (Strawser 1994). However, thes e studies did not examine how the wording of the reports impacted other constructs such as management’s credibility. In this study management’s credibility was affected by the letters included in the financial st atements, yet there were no differences between groups in their initial estimates of ear nings per share for the company. The participant’s did however, show differences in their reactions to future decisions when given information by management. The lending decisions in prior studies were found to be based on the solvency of the company more than any of the reporting formats (Blackwell et al. 1998). Had the prior studies examined the impact of the different accounting reports on the bankers’ reactions to future events they may have found differences in how the bankers’ reacted to explanations of these future events bas ed on the type of repor t the accountants issued to management. This study also contributes to the persuasion and credibility literature in psychology by further validating the three factor model of credi bility presented by McCroskey and Teven (1999). Additiona lly, a task was examined where the perception of credibility (a s measured by the McCroskey and Teven Credibility 19 See Strawser 1991,1994 for a review of prior research.

PAGE 153

142 Scale) and actual decisions were co mpared when participants viewed reports that differed in the tone used to comm unicate information. The task used in presented an example of a si ngle factor of credibility (intention) was manipulated and measured successfully. Prior psychological research has had difficulty in finding tasks to manipulate that test the individual sub-factors of the credibility model (O'Keefe 1990). 5.4 Future Research Other scales of credibility could be us ed to examine how different factors may impact auditor credibility. Testing c ould be done to determine if wording in auditor reports could impact the perceived intentions of the auditors and thus impact auditors’ overall credibility. The standard wording in audit reports may reduce the impact those reports have on decision-makers. This study found support for the a ssertion that different forms of management’s statement on internal cont rols impact investors’ judgments. Currently, under Auditing Standard No. 2 management is allowed considerable latitude in choosing its wording in the required statement on internal controls. Future research could be conducted in this area to determine if standard reports with prescribed (or constrained) wording woul d allow for more consistent investor interpretations of the reports. Also, diffe rent standard reports could be explored to determine which format investors prefer. Future research could also focus on testing the impact of tone on actual investors and/or institutional investors. This study was conducted using students

PAGE 154

143 who are a proxy for investors. Actual investor responses may alter the conclusions reached in this study. The model of credibility including the factors known to m ediate the impact of credibility as discussed in the appendix should be examined further. Path analyses can be used to examine if the covariates in the model follow the direction and strengt h of the theory. 5.5 Conclusion This study presents strong evidence t hat the tone of communications used by management may impact the participants’ ratings of management’s credibility and the amount of reliance participants pl ace on management’s forecasts. The robustness of this finding was also test ed after management had failed to meet its earnings per share forecast. After fa iling to make a forecast, the tone of communications was able to mitigate the reduction in management’s credibility ratings by participants. The tone of co mmunications was not able to mitigate the loss of credibility as measured by participants’ relianc e on management’s forecasts in period two of the study.

PAGE 155

144 References Ajinkya, B. and M. Gift (1984). "Corpor ate manager's earnings forecasts and symmetrical adjustments of market expectations." Journal of Accounting Research 22(2): 425-444. Ajzen, I. (2001). "Nature and operation of attitudes." Annual Review of Psychology 52: 27-58. Ashton, R. H. and S. S. Kramer (1980). "Students as surrogates in behavioral accounting research: Some evidence." Journal of Accounting Research 18(1): 1-15. Baginski, S. and J. Hassell (1990). "T he market interpretation of management earnings forecasts as a predictor of subsequent fi nancial analyst forecast revision." The Accounting Review 65(1): 175-190. Barber, B. and T. Odean ( 2001). "Boys will be boys: Gender, overconfidence, and confidence, and common stock investment." The Quarterly Journal of Economics Feb: 261-292. Blackwell, D. W., T. Noland and D. B. Winters (1998). "The value of auditor assurance: Evidence from loan pricing." Journal of Accounting Research 36(Spring): 57. Budd, J. (2000). "The incredible credibility dilemma." Public Relations Quarterly 45(3): 22-26. Calvey, M. (2001). Annual reports attempt to spin straw into gold. San Francisco Business Times San Francisco : 22. Clarkson, P., A. Dontoh, G. Richardson and S. Sefcik (1992). "The voluntary inclusion of earnings forecasts in ipo prospectuses." Contemporary Accounting Research 8(2): 601-626. Copeland, R., A. Francia and H. Straws er (1973). "Students as subjects in behavioral business research." The Accounting Review 48(2): 365-372. Davidson, R. A. and D. Neu (1993). "A note on the association between audit firm size and audit quality." Contemporary Accounting Research 9(2): 479489.

PAGE 156

145 DeZoort, F., D. Hermanson and R. Hous ton (2003). "Audit committee member support for proposed audit adjustments:A source credibility perspective." Auditing: A Journal of Theory and Practice 22(2): 189-205. Eagly, A., W. Wood and S. Chaik en (1978). "Causal inferences about communicators and their e ffects on opinion change." Journal of Personality and Social Psychology 36(4): 424-435. Ewing, T. (1942). "A study of certai n factors involved in change of opinion." The Journal of Social Psychology 16: 63-88. Garson, D. (2006). Statnotes: Topics in multivariate analysis. 2006. Glass, G. and K. D. Hopkins (1996) Statistical methods in education and psychology Needham Heights, Allyn & Bacon. Goldsmith, R. E., B. A. Lafferty and S. J. Newell (2000). "The impact of corporate credibility and celebrity credibility on consumer reaction to advertisements and brands." Journal of Advertising 29(3): 43. Gotlieb, K., J. Schlacter and R. St. Loui s (1992). "Consumer decision making: A model of the effects of involvement, source credibility, and location on the size of the price difference requi red to induce consumers to change suppliers." Psychology and Marketing 9(3): 191-208. Hair, J. F., R. E. Anders on, R. L. Tatham and W. C. Black (1998). Multivariate data analysis Englewood Cliffs, NJ, Prentice Hall. Hassell, J., M. M. Jennings and Lasser (1988). "Management earnings forecasts: Their usefulness as a source of fi rm specific information to security analysts." Journal of Financial Research 11(2): 303-319. Heider, F. (1958). The psycholog y of interpersonal relations John Wiley & Sons, Inc. Hirst, D. E. (1994). "Auditors' sens itivity to source reliability." Journal of Accounting Research 32(1): 113-126. Hirst, D. E., L. Koonce and J. Miller ( 1999). "The joint effect of management's prior forecast accuracy and the form of its financial forecasts on investor judgment." Journal of Accounting Research Vol 37(Supplement): 101-124. Hovland, C., J. Irving and H. Kelley (1953). Communication and persuasionpsychological studies of opinion change New Haven, CT, Yale University Press.

PAGE 157

146 Hovland, C. and H. Pritzker (1957). "Ext ent of opinion change as a function of amount of change advocated." Journal of Abnormal Psychology 54: 257261. Hovland, C. and W. Weiss (1951). "T he influence of source credibility on communication effectiveness." Public Opinion Quarterly 15: 635-650. Hughes, P. J. (1986). "Si gnaling by direct disclos ure under asymmetric information." Journal of Accounting & Economics 8(2): 119-142. Jennings, R. (1987). "Unsystematic se curity price movements, management earnings forecasts, and revisions in consensus anaylsts earnings forecasts." Journal of Accounting Research 25(1): 90-110. Kasnik, R. and B. Lev (1995). "To warn or not to warn: Ma nagement disclosures in the face of an earnings estimate." The Accounting Review 70(1): 113134. Kelley, H. (1973). "The process of causal attribution." American Psychologist 28: 107-128. Kim, K. J. and R. E. Verrecchia (1991). "Trading volume and price reactions to public announcements." Journal of Accounting Research Autumn: 302321. King, R. R., G. Pownall and G. Waymir e (1990). "Expectations adjustment via timely management forecasts: Revi ew, synthesis, and suggestions for future research." Journal of Accounting Literature 9: 113-144. Lafferty, B. A., R. E. Goldsmith and S. J. Newell (2002). "The dual credibility model: The influence of corporate and endorser credibility on attitudes and purchase intentions." Journal of Marketing Theory and Practice 10(3): 1. Leckey, A. (2003). As clum p continues, annual reports turn candid. JSOnline Milwaukee Journal Sentinel Milwaukee, WI. Lev, B. and S. H. Penman (1990). "Voluntar y forecast disclosure, nondisclosure, and stock-prices." Journal of Accounting Research 28(1): 49-76. Libby, R., R. Bloomfield and M. W. Nels on (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 of Accounting Research 37(2): 415-435.

PAGE 158

147 Mcconomy, B. (1998). "Bia s and accuracy of management earnings forecasts: An evaluation of the impact of auditing." Contemporary Accounting Research 15(2): 167-95. McCroskey, J. (1966). "Scales fo r the measurement of ethos." Speech Monographs 33: 65-72. McCroskey, J. and J. Teven (1999). "Good will: A reexamination of the construct and its measurement." Communications Monographs 66: 90-103. McGinnies, E. and C. D. Ward (1980). "B etter liked than right: Trustworthiness and expertise as factors in credibility." Personality and Social Psychology Bulletin 6: 467-472. McNichols, M. (1989). "Evidence of in formation asymmetries from management earnings forecasts and stock returns." The Accounting Review 64(January): 1-27. Mendenhall, W. and T. Sincich (1996). A se cond course in statistics regression analysis Upper Saddle River, Simon & Schuster. Mercer, M. (2001). The credibility c onsequences of managers' disclosure decisions. Dissertation, School of Accounting Austin, The University of Texas at Austin. Mercer, M. (2004). "How do investor s assess the credibility of management disclosures?" Accounting Horizons 18(3): 185. Mercer, M. (2005). "The fleeting effect s of disclosure forthcomingness on management's reporting credibility." The Accounting Review 80(2): 723744. O'Brien, B. (2001). Monthly mutual funds review --' dear shareholder: The year 2000 was... how best to say it? 'not r eally very good' --if fund managers ever faced a time when bluntness is needed, this could be it. Wall Street Journal New York : 1. O'Keefe, D. J. (1990). Pe rsuasion theory and research Newbury Park, California, Sage Publications. Pastore, N. and M. Horowitz (1955). "The influence of attributed motive on the acceptance of a statement." Journal of Abnormal Psychology 51(331-332). Patell, J. (1976). "Corporate forecasts of earnings per share and stock price behavior: Empirical tests." Journal of Accounting Research 14(Autumn): 246-276.

PAGE 159

148 PCAOB, P. A. O. B. (2004). "an audit of internal control over financial reporting performed in conjunction with an audit of financial statements", washington, d.C. Penman, S. H. (1980). "An em pirical investigation of t he voluntary disclosure of corporate earnings forecasts." Journal of Accounting Research 18(Spring): 132-160. Perloff, R. M. (2003). The dynamics of persuasion : Communication and attitudes in the 21st century Mahwah, NJ, Lawrence Erlbaum Associates, Inc. Reagan, D., E. Straus and R. Fazio (1974). "Liking and the attribution process." Journal of Experimental Social Psychology 10: 385-397. Rodgers, J. and P. Stocken (2002). "C redibility of management forecasts." Working Paper Rodgers, J. W. (2002). "Talking point #2: The tale of the telltale letters." Journal of Financial Planning 15(10): 22. Skinner, D. J. (1994). "Why firms voluntarily disclose bad news." Journal of Accounting Research 32(1): 38-60. Stocken, P. (2000). "Credibilit y of voluntary disclosure." RAND Journal of Economics 31(2): 359-374. Strawser, J. R. (1994) "An investigation of the effe ct of accountant involvement with forecasts on the decis ion and perceptions of commercial lenders." Journal of Accounting, Auditing and Finance 9: 553-560. Thompson, C. (2002). Rega in credibility with an annua l report that shoots straight. PR News 58 : 1. Tversky, A. and D. Kahneman (1974). "J udgment under uncertainty: Heuristics and biases." Science 185: 1124-1131. Verrecchia, R. E. (1983). Discretionary disclosure." Journal of Accounting & Economics 5(3): 179-194. Verrecchia, R. E. (1990). "Information quality and discr etionary disclosure." Journal of Accounting & Economics 12(4): 365-380. Walters-York, M. and A. Curatola ( 1998). "Recent evidence on the use of students as surrogate subjects." Advances in Accounting Behavioral Research 1: 123-143.

PAGE 160

149 Walters-York, M. and A. Curatola (2000). "Theoretical reflections on the use of students as surrogate subjects in behavioral experimentation." Advances in Accounting Behavioral Research 3: 243-263. Waymire, G. (1984). "Additional evi dence on the information content of management earnings forecasts." Journal of Accounting Research 22(Autumn): 703-718. Williams, P. A. ( 1996). "The relation between a prior earnings forecast by management and analyst response to a current management forecast." Accounting Review 71(1): 103-115. Wood, W. and A. Eagly (1981). "Stages in the analysis of persuasive messages: The role of causal attributions and message comprehension." Journal of Personality and Social Psychology 40(2): 246-259.

PAGE 161

150 Appendices Appendix A: Moderating Factors That Infl uence Credibility or Its Impact on Belief Revision The model of credibility’s impact on belie f revision used in the study accounts only for factors that are introduced in the study. There are other factors known to affect credibility or to impact the effect credibility has on belief revision. This Appendix describes these factors as m ediators and moderator s. Mediators are variables that will affect the level of credibility. Moderators will not affect credibility but rather impact the role cr edibility plays in belief revision. Mediating Variables of Credibility Two variables that can impact a pers on’s credibility are knowledge biases and reporting biases. Eagly and Wood (1978) propose that the position taken by a communicator will interact with the me ssage recipient’s expectations of the communicator position. This interaction affects the message recipient’s perceived credibility of the communicator. These expectancy biases are referred to as knowledge bias and reporting bias (Eagly et al. 1978). Both of these biases and their expected effect on credibility are diagramed in Figure 3 and will be discussed in the following sections.

PAGE 162

151 Reporting Bias Knowledge Bias Disconfirmed Disconfirmed Confirmed Disconfirmed Reduced Credibility Via Expertise Perceived Biases Enhanced Credibility Via Trusworthiness Increase Credibility Via Expertise ExpectanciesResult Reduced Credibility Via Trustworhtiness Figure 3. Eagly, Wood and Chaiken (1978 ) Reporting Bias and Knowledge Bias Knowledge Bias Message recipients may expect a comm unicator to advocate a certain biased position based on the communica tor’s background. Eagly and Wood (1978) note the influences on a communica tor can be internal or external. Internal influences are influences such as a person’s biological makeup (skin color, body composition, and gender). Fo r example, people with a particular ethnic background are expected to favo r programs and policies that benefit people similar to them. If, for instance, a law maker was a member of a protected class of individuals, via affirmative acti on, some message recipients of this law maker might believe he/she will be in favo r of a particular affirmative action legislation that benefits his/her group. If he/s he were to vote for this action people would assume it was because of their herit age. If they violated that expectation

PAGE 163

152 and voted against the legislation, some message recipients would believe the action was taken based on the merits of the l egislation (in spite of the legislator’s background) and therefore, more credible. External influences are environmental factors that are expected to bias a communicator. Observers may feel that a communi cator’s education about a specific issue is non-representative of t he whole issue. For example, a person with a degree from a Chri stian college might be expected to have a more conservative stance on some issues. If he/ she were to take a more progressive stance we would assume he/she was taki ng this stance in spite of his/her background; that, therefore, his/her knowledge in general must be greater than previously expected. Knowledge biases can be thought of as a stereotype. People expect certain people to act a certain way. W hen this expectancy is confirmed, the persuasiveness of the communicator’s message may be reduced. It is believed that this reduction in persuasiveness is due to a reduction in the perceived expertise of the communicator (O'Keefe 1990). Eagly and Wood (1978) reject the argument that confirma tion of knowledge bias corre sponds to the notion of expertise; the very notion that a person with a biased set of information would be thought of as an expert wit h respect to a field where his/her knowledge about the topic is biased seems counterintuitiv e. O’Keefe (1990) seems to be in agreement, stating that a communicator with a perceived knowledge bias is perceived as less competent.

PAGE 164

153 Reporting Bias A reporting bias occurs when a message recipient believes communicators may alter t heir message to conform to the audience to which they are speaking (Eagly et al. 1978). Mess age receivers may discount an otherwise credible source when they perceive a reporting bias. Reporting biases are believed to occur when a communicator delivers a message that differs from his/her true beliefs because of perce ived external pressures (Pastore and Horowitz 1955; Eagly et al 1978). One example of a reporting bias is a message from a young Republican pr esidential candidate to members of the AARP (formerly the American Association of Retired Persons) about the need for expanding Medicare benef its. A person judging th e credibility of the communicator may believe that this y oung Republican may not really feel that Medicare needs to change in as much as th ey believe the communicator is telling the audience what it wants to hear. When a reporting bias is violated message recipients should deem the communicator more credible. In the example of the young Republican, if the message were consistent with the expectations of a young Republican and inconsistent with the desires of the AARP audience the communicator would be perceived to be more trustworthy with respect to communi cating his/her true beliefs. Eagly and Wood (1978) believe that reporting bias corresponds fairly well with the trustworthiness factor of credibility in that viol ations of the reporting bias indicate a propensity to communicate a ssertions the communicator believes are valid without regard to the audience.

PAGE 165

154 With respect to financial forecast s, it is possible that negative news forecasts are a violation of a reporti ng bias. The expectation of positive information by investors is a perceived reporting bias. Investors expect management to release positive informa tion, since the release of negative information causes the stock price of the firm to drop (Lev and Penman 1990; Skinner 1994; Kasnik and Lev 1995; Libby and Tan 1999). Therefore, when management releases negative information, it is opposite investor’s expectation’s (a disconfirmation of the reporting bias) i ndicating the credibilit y of the disclosure should rise. This appears to be s upported by research, negative news announcements cause greater revision in market expectations and are therefore more credible (Skinner 1994; St ocken 2000; Rodgers and Stocken 2002). 2021 Variables That Moderate the Effect of Credibility Other factors can influence the magnitude of the effect that credibility can have on a recipient. Factors such as the recipient’s level of involvement with the task and the timing of identifying the communicator along with the position of the message can influence the magnitude of the effect of sour ce credibility (O'Keefe 1990) It is important to note that these factors s hould not affect credibility, only the amount of influence credibility has. I will discuss each of thes e factors. A flowchart is also presented ( 20 There is a possibility the market reacts to negative forecasts because of their form. Skinner (1994) found that negative news forecasts were usually point estimates. Pownall and Waymire (1989) found that point forecasts revisions in gen eral were more informative than other types. 21 The studies mentioned were not specifically lo oking at reporting bias although the theory would help support their findings.

PAGE 166

155 Figure 4) to aid in understanding ho w these variables interact with credibility. Credibility Level of InvolvementHigh Level of Involvement Timing in Identifying the CommunicatorAfter Communication Before Communication Position of the MessagePro-attitudinal Counter-attitudinal Knowledge BiasDisconfirmed Confirmed Communicator Credibility Means Less Increase Credibility Via Expertise Reporting Bias Disconfirmed Increase Credibility Via TrustworthinessLow Level of Involvement Decrease Credibility Via Expertise Confirmed Decrease Credibility Via Trustworthiness Message Characteristics Hypothesis One Hypothesis Two Information Suprise Expertise Goodwill Trustworthiness Forecast Revision Figure 4. Moderating Vari ables to Credibility

PAGE 167

156 Level of Involvement Recipients with a high level of invo lvement with respect to the message are less likely to be influenced by differences in credibility while perceivers with a low level of involvement are more likely to be affected by source credibility (Figure 5) (O'Keefe 1990). The relations hip between involvement and sensitivity to source credibility is believed to ex ist because perceivers that are highly involved with respect to a particular subj ect are more likely to form their own opinions about the validity of the communi cator’s statements. Recipients who are involved in the task at a high-level are more likely to pay attention to the details of the message and to pick apart the message and base the importance of the message on its content. In the current study, all parti cipants were believed to have a low level of involvem ent with the particular task. None of the participants in the study were expected to be experts in the field. Receiver’s Level of Involvement Low Level of Involvement High Level of Involvement Credibility Means More Credibility Means Less Message Characteristics Mean More Message Recipient’s Level of Involvement Figure 5. O’Keefe’s Level of In volvement Affect on Credibility

PAGE 168

157 Timing of Identifying the Communicator The timing of identifying a communicato r is also believed to impact the magnitude of effect from source credibility. When a communicator is identified before a message is received source credi bility will have a gr eater impact on the perceiver. When the source is identified after the message is received the perceiver is believed to be more affected by the message characteristics than by the source credibility (O'K eefe 1990). Figure 6 shows the relationship between identifying the communicator before and after a message. When a communicator is identified after a message has been giv en receivers are likely to pay more attention to message details as the message is delivered. Timing in Identifying the Communicator After Message Before Message Credibility Means Less Credibility Means More Message Characteristics Mean More Message Characteristics Mean Less Figure 6. Timing of Iden tifying the Communicator Influence of Message Direction The position a communicator takes wit h his/her message can also affect the influence of credibility on the percei ver. A communicator can send a message that is pro-attitudinal or counter-attitudinal with respect the perceiver’s position on a topic (O'Keefe 1990). Pro-a ttitudinal messages are in line with prior beliefs of the perceiver. Counter-attitudinal mess ages are opposite what the perceiver

PAGE 169

158 believes. Pro-attitudinal messages diminish the role of credibility in judgments; whereas, counter-attitudinal messages usually require a more credible communicator (Figure 7). It is import ant to understand that pro-attitudinal messages do not actually reduce the amount of perceived credibility of a communicator; they just di minish the role credibility plays in the judgment process for the perceiver. The diminished role of credibility could be a type of ceiling effect, in that a message recipient who already holds strong beliefs about a topic has less room for opinion change. Whereas when the message is in the opposite direction, the participant has more room for opinion change, and therefore, the role of cr edibility is enhanced, or at least appears to be as there will be more room for an effect to be found. Position of the Message Pro-attitudinal message Counterattitudinal Message High Credibility Effects Diminish High Credible has advantage Figure 7. Position of Message

PAGE 170

159 Appendix B: View of Experimental Materials Used for Pilot Study The following pages show a screen shot of each stage of the experiment with a brief description of the picture. 1. General Instructions: Participants are given the general instructions for completing the task. They are notified of their right to leave the exam at any time as well as their compensation for participating.

PAGE 171

160 Appendix B Continued 2. Informed Consent: Participants have an opportunity to read the Informed Consent form. They have a yes/no button to indicate voluntarily consent to parti cipate in the experiment.

PAGE 172

161 Appendix B Continued Informed Consent Continued

PAGE 173

162 Appendix B Continued Informed Consent Continued

PAGE 174

163 Appendix B Continued 3. Instructions: An overview of the task is then given to each participant.

PAGE 175

164 Appendix B Continued Instructions: An overview of the ta sk is then given to each participant. Continued

PAGE 176

165 Appendix B Continued 4. Background Information: Participants are instructed as to their role in the task. 5. Company Background: Participant s are given a brief background about the company.

PAGE 177

166 Appendix B Continued 6. Products: Participants are told about the products the company produces.

PAGE 178

167 Appendix B Continued 7. Internal Control Letter: Part icipants are given the letter from management, “Assessment of Inter nal Controls Over Financial Reporting.” At this point partici pants are broken into two groups: a. Internal Control High Intention letter.

PAGE 179

168 Appendix B Continued b. Internal Control No Intention letter.

PAGE 180

169 Appendix B Continued 8. Financial Statements: All partici pants are given the prior income statement and balance sheet information. 9. Earnings Per Share Prediction On e Period One: Participants give an estimate of earnings per share.

PAGE 181

170 Appendix B Continued 10. Management’s Earnings Per Share Prediction Period One: Management provides participants with its earnings per share prediction for the same period. Management’s prediction is 132 percent of the particip ant’s earnings per share estimate for the same period. 11. Earnings Per Share Prediction Two Period One: Participants are given an opportunity to revise their initia l earnings per share predictions.

PAGE 182

171 Appendix B Continued 12. Credibility Rating One: Participants fill out the credibility instrument comprised of 18 Likert scale questions on an eleven-point scale. a. Six expertise factors

PAGE 183

172 Appendix B Continued b. Six intention factors

PAGE 184

173 Appendix B Continued c. Six trustworthiness factors

PAGE 185

174 Appendix B Continued 13. Earnings Letter: Management informs the participants about the actual earnings per share for period one. This amount is 109 percent of the participant’s initial earnings per share estimate. This is 23 percent less than management’s prediction. There ar e two different earnings letters: a. High intention earnings letter.

PAGE 186

175 Appendix B Continued b. No intention earnings letter.

PAGE 187

176 Appendix B Continued 14. Financial Statements Two: Participants are given a copy of the income statement for the first period in which they predicted earnings per share. This also included the three prior years of data. 15. Earnings Per Share Prediction One Period Two: After reading the financial statements the participants are asked to make an earnings per share prediction for the next year.

PAGE 188

177 Appendix B Continued 16. Management’s Earnings Per Share Prediction Period Two: Management provides participants with their earnings per share prediction for the same period. Management’s prediction is 132 percent of the particip ant’s earnings per share estimate for the same period. 17. Earnings Per Share Prediction fo r Period Two: Participants are given an opportunity to revise their initia l earnings per share predictions.

PAGE 189

178 Appendix B Continued 18. Credibility Rating Two: Participants fill out the credibility instrument comprised of 18 Likert scale questions on an 11 point scale. a. Six intention factors

PAGE 190

179 Appendix B Continued b. Six expertise factors

PAGE 191

180 Appendix B Continued c. Six trustworthiness factors

PAGE 192

181 Appendix B Continued 19. Manipulation Questions: Partici pants answer manipulation questions about both letters they have received.

PAGE 193

182 Appendix B Continued Manipulation Check Questions Continued 20. Covariate Questions – Fraud

PAGE 194

183 Appendix B Continued Covariate Questions: Fraud Continued 21. Covariate QuestionsSarbanes Oxley

PAGE 195

184 Appendix B Continued 22. Demographic Questions

PAGE 196

185 Appendix B Continued 23. Possible Covariate Questions: Prior Investment Experience

PAGE 197

186 Appendix B Continued 24. Theoretical Covariat es Based on Appendix A

PAGE 198

187 Appendix B Continued 25. Feedback Screen 26. Finished Screen

PAGE 199

188 Appendix C: Pilot Stud y One Surprise Testing Please rate the follo wing items independently Industry analysts have predicted earnings per share of XYZ Company at $1.23 per share. Management has issued a foreca st predicting earni ngs per share of $1.35. Please rate the difference between t he analysts’ predictions and management predictions: Insignificant Very Significant |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7 Not Surprising Very Surprising |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7 Industry analysts have predicted earnings per share of XYZ Company at $1.55 per share. Management has issued a foreca st predicting earni ngs per share of $1.86. Please rate the difference between the analysts’ predictions and management predictions: Insignificant Very Significant |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7 Not Surprising Very Surprising |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7 Industry analysts have predicted earnings per share of XYZ Company at $1.75 per share. Management has issued a foreca st predicting earni ngs per share of $2.28. Please rate the difference between t he analysts’ predictions and management predictions: Insignificant Very Significant |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7 Not Surprising Very Surprising |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7

PAGE 200

189 Appendix C Continued Industry analysts have predicted earnings per share of XYZ Company at $1.10 per share. Management has issued a foreca st predicting earni ngs per share of $1.54. Please rate the difference between t he analysts’ predictions and management predictions: Insignificant Very Significant |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7 Not Surprising Very Surprising |------------|------------|------------|-----------|-----------|------------| 1 2 3 4 5 6 7

PAGE 201

About the Author Robert David Slater Jr. was born in New Haven, Connecticut, the son of Karen and Robert Slater. Robert graduated from Gateway High School in Kissimmee, Florida in 1989 and entered the University of South Florida in Ta mpa, Florida. In December 1995, he graduated with a Bachel or of Science degree in Business Administration majoring in accounting. He worked as a financial manager for Michael Steinberg and Associates fr om 1994 until 1999. In August 1999, he entered the College of Business at the Univ ersity of South Florida, earning his Master’s degree in Accountancy (M.Acc. ) in 2003 and a PhD in Accounting in 2007. In June of 2000 he married Elizabeth Hemphill. Robert is a licensed CPA in the State of Florida and is currently a faculty member at Texas A&M Corpus Christi.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001967732
003 fts
005 20081029083508.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 081029s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002139
035
(OCoLC)264637849
040
FHM
c FHM
049
FHMM
090
HF71 (ONLINE)
1 100
Slater, Robert D.
4 245
The impact of management's tone on the perception of management's credibility in forecasting
h [electronic resource] /
by Robert D. Slater.
260
[Tampa, Fla.] :
b University of South Florida,
2007
3 520
ABSTRACT: The purpose of this study is to examine the impact of management altering its tone in communications on participants' perceptions of management credibility. Management's tone in communicating with participants was manipulated using communications from management under two treatment conditions. In period one of the study management's tone was manipulated within the management statement on internal controls as required by the Public Company Accounting Oversight Board's (PCAOB) Auditing Standards No. 2. In period one, participants had no knowledge of management's prior forecasting accuracy. Consistent with predicted hypotheses, the findings reveal that management can increase its credibility with participants by communicating its empathy, responsiveness, and understanding. Management's increased credibility was measured using both a validated credibility scale and by examining participants' reliance on management's forecasts. In period two of the study all participants had knowledge of management's forecast failure in period one. The results from period two found that tone could impact the rating of management's credibility when management had previously failed to meet a forecast but that tone had no impact on participant's changes in their earnings per share estimates after management had previously failed to meet a forecast.
502
Dissertation (Ph.D.)--University of South Florida, 2007.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 189 pages.
Includes vita.
590
Co-adviser: Jacqueline Reck, Ph.D.
Co-adviser: Uday Murthy, Ph.D.
653
Source credibility theory,
Intention.
Decision-making.
Public Accounting Oversight Board.
Credibility scale.
Auditing Standard No. 2.
0 690
Dissertations, Academic
z USF
x Business Administration
Doctoral.
773
t USF Electronic Theses and Dissertations.
856
u http://digital.lib.usf.edu/?e14.2139