<%BANNER%>

Understanding Audit Committee Effectiveness

Permanent Link: http://ufdc.ufl.edu/UFE0021182/00001

Material Information

Title: Understanding Audit Committee Effectiveness
Physical Description: 1 online resource (168 p.)
Language: english
Creator: Macgregor, Jason Eugene
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: audit, committee, incentives, risk
Accounting -- Dissertations, Academic -- UF
Genre: Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Many prior studies have examined the determinants of audit committee effectiveness with mixed results. I examined why an audit committee would allow a manager to opportunistically report earnings by using both an analytical model and a series of empirical tests. Using signal detection theory, a widely accepted psychological theory, I modeled the audit committee?s decision to allow or reject earnings management as a two-step process: detecting evidence and determining a response. From this model, I identified ten factors that influence a committee?s judgment, many of which have not previously documented. In particular, I identified that a committee's decision depends upon the members' incentives. While prior work has considered incentives as a determinant of audit committee effectiveness, the results have been inconsistent. Some researchers found that incentives motivate the members to be effective monitors, while other researchers found that incentives motivate the members to focus on short-term stock appreciation. My model suggests that the effect of incentives depends upon the expected payoff from allowing aggressive reporting, given the likelihood of a reporting problem. I empirically tested how incentives influenced the committee's judgment by examining how equity and reputation incentives influence the committee's willingness to allow earnings management when earnings are near key earnings? benchmarks. When earnings are near key earnings' benchmarks, the committee's willingness to allow earnings management depends upon the committee's equity and reputation incentives and the CEO's incentives to opportunistically report, a proxy for the likelihood of a reporting problem. The inconsistencies in prior research may be attributable to the omission of a key moderator variable (CEO incentives) and an ineffective proxy for equity incentives. My results were robust to the three common earnings' benchmarks (zero earnings level, prior year?s earnings level, and analyst forecasts) documented in the literature, a variety of specification of the incentives, and a variety of control for audit committee ability and firm-specific attributes.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jason Eugene Macgregor.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Asare, Stephen K.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2017-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021182:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021182/00001

Material Information

Title: Understanding Audit Committee Effectiveness
Physical Description: 1 online resource (168 p.)
Language: english
Creator: Macgregor, Jason Eugene
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: audit, committee, incentives, risk
Accounting -- Dissertations, Academic -- UF
Genre: Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Many prior studies have examined the determinants of audit committee effectiveness with mixed results. I examined why an audit committee would allow a manager to opportunistically report earnings by using both an analytical model and a series of empirical tests. Using signal detection theory, a widely accepted psychological theory, I modeled the audit committee?s decision to allow or reject earnings management as a two-step process: detecting evidence and determining a response. From this model, I identified ten factors that influence a committee?s judgment, many of which have not previously documented. In particular, I identified that a committee's decision depends upon the members' incentives. While prior work has considered incentives as a determinant of audit committee effectiveness, the results have been inconsistent. Some researchers found that incentives motivate the members to be effective monitors, while other researchers found that incentives motivate the members to focus on short-term stock appreciation. My model suggests that the effect of incentives depends upon the expected payoff from allowing aggressive reporting, given the likelihood of a reporting problem. I empirically tested how incentives influenced the committee's judgment by examining how equity and reputation incentives influence the committee's willingness to allow earnings management when earnings are near key earnings? benchmarks. When earnings are near key earnings' benchmarks, the committee's willingness to allow earnings management depends upon the committee's equity and reputation incentives and the CEO's incentives to opportunistically report, a proxy for the likelihood of a reporting problem. The inconsistencies in prior research may be attributable to the omission of a key moderator variable (CEO incentives) and an ineffective proxy for equity incentives. My results were robust to the three common earnings' benchmarks (zero earnings level, prior year?s earnings level, and analyst forecasts) documented in the literature, a variety of specification of the incentives, and a variety of control for audit committee ability and firm-specific attributes.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jason Eugene Macgregor.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Asare, Stephen K.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2017-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021182:00001


This item has the following downloads:


Full Text

PAGE 1

1 UNDERSTANDING AUDIT COMMITTEE EFFECTIVENESS By JASON EUGENE MACGREGOR A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 Copyright 2007 Jason Eugene MacGregor

PAGE 3

3 I dedicate this work to my Lord Jesus Christ, wh o gave himself for my sins to deliver me from the present evil age according to the will of my God and Father, to whom be the glory forever and ever, to Kelly my love, and to Paul Alis tair, who was not here when this was written.

PAGE 4

4 ACKNOWLEDGMENTS I wish to express my deep appreciation to Joe Alba, Gary McGill, Robert Knechel, Victoria Dickinson, Jenny Tucker, Jesse Itzkowitz, and most notably my chair Stephen Asare for their honest feedback and encouragement. I wo uld like to thank the wo rkshop participants at University of Florida, Baylor University and McGill University for their helpful comments. I would like to acknowledge my peers, Carlos Jime nez, Monika Causholli, and Liang Fu for their patience listening to the many ideas that were co nsidered before I settled upon audit committees. Lastly, I would like my wife for her tireless efforts editing this dissertation and learning more about accounting than she ever dreamed possible.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES................................................................................................................ .........9 ABSTRACT....................................................................................................................... ............10 CHAPTER 1 INTRODUCTION..................................................................................................................12 Background Information on Audit Committees.....................................................................13 The DAHR Model of Audit Committee Effectiveness...........................................................14 Inputs: Composition, Resources, and Authority..............................................................14 Process: Diligence...........................................................................................................16 Limitations of the DAHR Model.....................................................................................17 2 A MODEL OF AUDIT CO MMITTEE DECISION-MAKING............................................22 Model Development.............................................................................................................. .23 Simplified Decision-Making Model.......................................................................................25 Detection Ability.............................................................................................................26 Evaluating the Evidence..................................................................................................26 Determining the Response...............................................................................................27 Reputation incentives...............................................................................................27 Equity incentives......................................................................................................28 Calculating zc............................................................................................................29 Multi-Person Audit Committee..............................................................................................30 Committee Decision Statistic..........................................................................................31 Non-Optimal Individual Decision Statistics....................................................................33 Observations from the Model.................................................................................................33 Reconciling the Model and the Literature..............................................................................42 Reconsidering the DAHR Model............................................................................................47 Consistency with DAHR Model......................................................................................47 External Influences..........................................................................................................48 Expanded Processing Section..........................................................................................49 Individual contributions...........................................................................................49 Group processing......................................................................................................50 Limitations.................................................................................................................... ...52 3 SETTING AND DATA..........................................................................................................59 Variables of Interest.......................................................................................................... ......61

PAGE 6

6 The Dependent Variables................................................................................................61 The Independent Variable: Audit Committee Incentives................................................61 The Control Variables.....................................................................................................62 Audit committee ability............................................................................................62 CEO incentives.........................................................................................................65 Firm control variables..............................................................................................65 The Growth Sample.............................................................................................................. ..67 Sample Selection.............................................................................................................67 Descriptive Statistics.......................................................................................................68 The Profit Sample.............................................................................................................. .....69 Sample Selection.............................................................................................................69 Descriptive Statistics.......................................................................................................69 The Analyst Sample............................................................................................................. ...70 Sample Selection.............................................................................................................70 Descriptive Statistics.......................................................................................................70 4 EQUITY HOLDINGS AND TH E GROWTH THRESHOLD..............................................84 Hypothesis Development........................................................................................................84 Case 1: Total Equity Holdings........................................................................................86 Case 2: Partitioned Equity Holdings..............................................................................87 When Earnings are Near the Previous Years Earnings Level........................................90 Method......................................................................................................................... ...........93 Preliminary Analysis.......................................................................................................93 Test of H1..................................................................................................................... ...93 Test of H2..................................................................................................................... ...94 Secondary Analysis.........................................................................................................94 Other Findings.................................................................................................................96 Discussion..................................................................................................................... ..........97 5 DIRECTORSHIPS AND THE GROWTH THRESHOLD..................................................106 Hypotheses Development.....................................................................................................106 Empirical Tests................................................................................................................ .....109 Discussion..................................................................................................................... ........110 Governance Expertise....................................................................................................110 Diligence Effect.............................................................................................................111 6 INCENTIVES AND THE PROFIT THRESHOLD.............................................................115 Hypotheses Development.....................................................................................................115 Empirical Tests................................................................................................................ .....116 Primary An alysis...........................................................................................................116 Secondary Analysis.......................................................................................................117 Discussion..................................................................................................................... ........118

PAGE 7

7 7 INCENTIVES AND THE ANALYST THRESHOLD........................................................123 Hypotheses Development.....................................................................................................123 Empirical Tests................................................................................................................ .....124 Primary An alysis...........................................................................................................124 Secondary Analysis.......................................................................................................126 Discussion..................................................................................................................... ........126 8 INCENTIVES: BIAS OR MOTIVATE...............................................................................130 Effort or Bias................................................................................................................. .......130 Signal Detection Theory.......................................................................................................132 Results........................................................................................................................ ...........134 Growth Sample..............................................................................................................135 Profit Sample.................................................................................................................137 Discussion..................................................................................................................... ........138 9 CONCLUSIONS AND FUTURE WORK...........................................................................150 Audit Committee Incentives and Seasoned Equity Offerings..............................................150 Compensation................................................................................................................... ....152 Group Dynamics................................................................................................................. ..152 APPENDIX....................................................................................................................... ...........155 LIST OF REFERENCES.............................................................................................................156 BIOGRAPHICAL SKETCH.......................................................................................................168

PAGE 8

8 LIST OF TABLES Table page 3-1 Sample selection for growth sample..................................................................................73 3-2 Pearson correlation of key independe nt variables for the growth sample.........................74 3-3 Descriptive statistics of growth sample.............................................................................75 3-4 Sample selection for profit sample....................................................................................76 3-5 Pearson correlation of key independe nt variables for the profit sample............................77 3-6 Descriptive statistics of profit sample................................................................................78 3-7 Sample selection for the analyst sample............................................................................79 3-8 Pearson correlation of key independe nt variables for the analyst sample.........................80 3-9 Descriptive statistics of the analyst sample.......................................................................81 4-1 Logistic regression models of growth sample.................................................................100 4-2 Logistic regression models with growth sample partitioned by CEO equity holdings....101 4-3 Logistic regression models of growth samp le of firms with who lly independent audit committees..................................................................................................................... ..102 4-4 Logistic regression models with growth sample partitioned by auditor size...................103 4-5 Logistic regression models with growth sample and narrow EPS range.........................104 4-6 Logistic regression models of growth sample considering auditor materiality...............105 5-1 Logistic regression models includi ng directorships for growth sample..........................113 5-2 Logistic regression model for alternative explanation.....................................................114 6-1 Logistic regression models for profit sample...................................................................120 6-2 Logistic regression model for profit sa mple with clients of Big 4/5 auditors.................122 7-1 Logistic regression model for analyst sample..................................................................127 7-2 Logistic regression model for analyst sample with narrow range EPS...........................129 8-1 Applying signal detection theory.....................................................................................140

PAGE 9

9 LIST OF FIGURES Figure page 1-1 DAHR model of audit committee effectiveness................................................................21 2-1 Individual decision-making pro cess with signal de tection theory.....................................53 2-2 Group decision-making process with signal detection theory...........................................54 2-3 Modified model of audit committee effectiveness.............................................................55 2-4 Committee size and d`..................................................................................................... ..56 2-5 Correlation and group d`.................................................................................................. ..57 2-6 Group d` and weighting method........................................................................................22 8-1 Growth Sample: d` Audit committee equ ity holdings, and CEO short-term equity incentives..................................................................................................................... ....142 8-2 Growth Sample: Bias, audit committee equity holdings, and CEO short-term equity incentives..................................................................................................................... ....143 8-3 Growth Sample: d`, audit committee di rectorships, and CEO short-term equity incentives..................................................................................................................... ....144 8-4 Growth Sample: Bias, audit committee directorships, and CEO short-term equity incentives..................................................................................................................... ....145 8-5 Profit Sample: d`, audit committee equ ity holdings, and CEO short-term equity incentives..................................................................................................................... ....146 8-6 Growth Sample: Bias, audit committee equity holdings, and CEO short-term equity incentives..................................................................................................................... ....147 8-7 Profit Sample: d`, audit committee dir ectorships, and CEO short-term equity incentives..................................................................................................................... ....148 8-8 Profit Sample: Bias, audit committee di rectorships, and CEO short-term equity incentives..................................................................................................................... ....149

PAGE 10

10 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy UNDERSTANDING AUDIT COMMITTEE EFFECTIVENESS By Jason Eugene MacGregor August 2007 Chair: Stephen Asare Major: Business Administration Many prior studies have examin ed the determinants of audit committee effectiveness with mixed results. I examined why an audit comm ittee would allow a manager to opportunistically report earnings by using both an analytical m odel and a series of empirical tests. Using signal detection theory, a widely accepte d psychological theory, I modeled the audit committees decision to allow or reject earnings management as a two-step process: detecting evidence and determining a response. From this model, I identified ten factors that influence a committees judgment, many of which have not previously documented. In particular, I identified that a committees decision depends upon the members incentives. While prior work has considered incentives as a determinant of audit committee effectiveness, the results have been inconsistent. Some researchers found that incentives motivate the me mbers to be effective monitors, while other researcher s found that incentives motivate the members to focus on shortterm stock appreciation. My model suggests th at the effect of incentives depends upon the expected payoff from allowing aggressive repor ting, given the likelihood of a reporting problem. I empirically tested how incentives influe nced the committees judgment by examining how equity and reputation incentives influence the committees willingness to allow earnings management when earnings are near key earnings benchmarks. When earnings are near key

PAGE 11

11 earnings benchmarks, the committee's willingness to allow earnings management depends upon the committees equity and reput ation incentives and the CEOs incentives to opportunistically report, a proxy for the likelihood of a reporting problem. The inc onsistencies in prior research may be attributable to the omission of a key moderator variable (CEO incentives) and an ineffective proxy for equity incentives. My re sults were robust to th e three common earnings benchmarks (zero earnings level, prior years ea rnings level, and analyst forecasts) documented in the literature, a variety of specification of the incentives, and a variety of control for audit committee ability and firm-specific attributes.

PAGE 12

12 CHAPTER 1 INTRODUCTION Are audit committees the remedy for the curr ent epidemic of corporate scandals? Regulators claim that an effective audit comm ittee can eliminate managerial opportunism (Pitt [2001], Ruder [2002]). Yet, the ev idence supporting this claim is mixed (DeZoort et al. [2002]). In my study, I considered why audit committees may be ineffective at preventing managerial opportunism. I modeled the audit committee decisi on-making process, empirically tested how incentives influence the comm ittees judgments, and examined whether incentives induce increased effort or bias judgme nt. From the model, I identified ten factors, many previously undocumented, that may influence audit committee e ffectiveness. From the empirical evidence, I found that equity and reputation incentives infl uence the audit committees judgment. Lastly, I found that incentives both bias and motivate audit committee members. My study makes three contributions to the lit erature on audit committees. I presented a model of how audit committees decide if the mana ger is opportunistically reporting. From this model, I identified many factors th at may influence the audit committees decisions that have not previously been documented, and I presented a comprehensive model of audit committee effectiveness. Further, I presented evidence th at audit committees are sensitive to the influence their decisions have on the value of their equity holdings and their reputa tions. I also presented evidence that the influence of audit committee incentives depends upon the risk environment. Lastly, I presented evidence that audit committ ee incentives induce both increased effort and biased judgments. This was the first work th at documented why incentives influence an audit committees judgment.

PAGE 13

13 Background Information on Audit Committees While audit committees have been central to many of the recent reforms to prevent managerial opportunism, corporations have used them for over a century.1 The audit committee is a sub-committee of the board of directors charged with overseeing the financial reporting process (Sarbanes-Oxley Act [ 2002]). Typically, the committee consists of three to five independent members who meet two to four times per year with managers, internal auditors, and external auditors in order to determine the ac ceptability of the accountin g policies, select the auditor, approve non-audit servic es, and resolve auditor-manageme nt disputes (Anderson et al. [2004]). How audit committees actually operate is largely unknown. For this reason, Gendron and Bedard [2004] referred to audit committees as a black box. The purpose of audit committees is to alleviat e the agency conflict between managers and shareholders.2 An audit committee works with the auditors and the managers to ensure that the financial statements are free of material misstat ement. For example, the audit committee works with auditors and managers to resolve disputes over capitalization policie s. Despite the recent accounting scandals, on average, audit committees app ear to achieve this purpose. For example, effective audit committees are associated with im proved reporting quality, reduced incidence of reporting problems and irregularities, reduced earnings management, improved disclosure quality, and increased earnings informativen ess (DeFond and Jiambalvo [1991], Carcello and Neal [2000], Peasnell et al. [ 2000], Klein [2002], Carcello and Neal [2003], Xie et al. [2003], Abbott et al. [2004]). However, what makes an audit committee effective is still uncertain. 1 The first report of a British company (Great Railway) having a small group of directors acting as an audit committee was in 1872 (Tricker, [1978]). 2 An agency conflict exists because the separation of corporate management and ow nership potentially creates incentives for managers to work in their own interest, not in the shareholders (Fama and Jensen, [1983]).

PAGE 14

14 The DAHR Model of Audit Committee Effectiveness From 1970 to 2002, there were dozens of articles in academic journals that explored the benefits of an effective audit committee, the determinants of audit committee effectiveness, and the limitation of audit committees. Yet, absent from this literature was a model that explained audit committee effectiveness. To address this deficiency, DeZoor t et al. [2002] summarized the research into an input-process-output model (henceforth called the DAHR model). The DAHR model identified three inputs (c omposition, authority, and resources ) and one process (diligence) (Figure 1-1). This model provided an intuitiv e framework for discussing the determinants of audit committee effectiveness. Using this framework, I discuss the determinants of audit committee effectiveness in relation to reporti ng quality (i.e. prevention of managerial opportunism).3 Given the comprehensive nature of th e DeZoort et al. [2002] review and the significant regulatory changes in the last five years, I focus upon research from 2002 to present. Inputs: Composition, Resources, and Authority Composition refers to member level at tributes that influence the committees effectiveness. The two most common member-level attributes are independe nce and expertise. While independence has long been a requirement for auditors, only recently have regulators required that audit committee members be inde pendent as well (Sarbanes-Oxley Act [2002]).4 Companies with audit committees having a high er proportion of independent members are less likely to experience fraud, SEC enforcement action, material restat ements, and earnings 3 DeZoort et al. [2002] also considered two additional outp uts, internal control and risk management, that are beyond the scope of this paper. 4 To ensure that members are independent, Sarbanes-Oxl ey (SOX) prohibits current employees, former employees within the last 3 years, or a close re lative of an executive offi ce, directors who accept nondirector compensation in excess of $60,000 or whose em ployer receives at least $200,000 in any of the past 3 years from serving on audit committees SOX, however, failed to provide a definition of compensation. This omission makes it uncertain how boards are to account for stock grants, stock options, and future retirement benefits.

PAGE 15

15 management than companies with audit comm ittees having a lower proportion of independent members (Defond and Jiambalvo [1991], McMu llen and Raghunandan [1996], Beasley et al. [2000], Klein [2002], Bdard, et al [2004], Bradbury et al. [2004]). While regulations ensure that the members are independent in appearance, there is still s ubstantial doubt as to whether or not the members are independent in fact. For ex ample, Carcello et al. [2006] found that when the CEO is involved in the nominating process, the audit committee effectiveness decreases. Sarbanes-Oxley requires each committee to have at least one financial expert.5 The documented benefits of having a financial expert on the committee include reduced incidence of restatements, fraud, and earnings management (Abbott et al. [2004], Bedard et al. [2004], Agrawal and Chadha [2005], Farber [2005]). Two additional dimensions of expertise that may influence the committees effectiveness are governance and firm expertise. Governance expertise relates to the members experience serving on other boards (Bedard et al. [2004]). Bedard et al. [2004] found that the more directorships a member holds the lower the incidence of ear nings management. There is, however, some evidence that additional directorsh ips may limit the directors availability (Morck et al. [1988], Beasley [1996]). Firm expertise relates to the members ability to identify opportunistic behavior through knowledge of the companys history, financial policies, and executives (Hermanlin and Weisbach [1991]). There appears to be a negative association between the average member tenure and earnings ma nagement (Xie et al. [2003], Bedard et al. [2004], Yang and Krishnan [2005]). It is, howev er, possible that as tenure increases, the members lose their independence as they become more familiar with management. 5 SOX defines a financial expert as a person who has, through education and experience, an understanding of the generally accepted accounting principles and financial statements and experien ce in preparing or auditing financial statements, applying accounting principles related to accoun ting estimates, accruals, and reserves, internal controls, and audit committee functions. The Blue Ribbon Committee identified a financial literate as one who has the ability to read and understand fundamental financial statements.

PAGE 16

16 An audit committee can only accomplish what it has the authority to carry out. The research on audit committee authority primar ily consists of gauging how well members understand their responsibilities (e.g. Kalbers and Fogarty [1993], DeZoort [1997], Lee and Stone [1997]). The most transpar ent measure of an audit committ ees authority is the presence of a charter. A charter is a written document that explicitly defines the audit committees authority. Bedard et al. [2004] found a negative association between a company having a charter and earnings management. The most researched measure of resources is committee size. A larger audit committee should have a greater knowledge base to draw from, a greater ability to delegate tasks, and be less vulnerable to a single member dominating di scussions (Felo et al. [2003]). Larger audit committees are negatively associated with the in cidence of suspicious auditor switches and restatements (Carcello, et al. [2006]). Howe ver, recent work has found no evidence of an association between committee si ze and internal control problems, restatements, and earnings management (Archambeault and DeZoort [2001], Abbott et al. [2004], Bedard et al. [2004], Krishnan [2005], Vafeas [2005]). These null re sults may be caused by the homogenization of audit committee size by regulations, rather than a reduction in the importance of committee size. Process: Diligence Diligence refers to the willingness of members to work together to prepare, ask questions, and pursue answers when dealing with management external auditors, in ternal auditors and other relevant constituents (DeZoort et al. [2002]). The most common proxy for diligence is meeting frequency. Meeting frequency is negativ ely associated with the level of earnings management and incidences of fraud and SEC sa nctions (Abbott et al. [2000], Beasley et al. [2000], Xie et al. [ 2003], Farber [2005]).

PAGE 17

17 Audit committee diligence is also linked to th e members incentives. Two incentives that may motivate the members are equity incentive s and reputation incenti ves. Arguably, when members have equity holdings in the firm th ey have greater incentiv es to focus on improved monitoring. However, the evidence is mixe d. Some researchers have found a negative relationship between equity hol dings and earnings management (Beasley [1996], Klein [2002], Vafeas [2005]), while others have found a posi tive relationship (Bedard et al. [2004], Yang and Krishnan [2005]). Audit committee members have incentives to develop their reputation to maximize their directorship opportunities. Re putation incentives motivate the member to be an effective monitor (Fama [1980], Fama and Jensen [ 1983], Milgrom and Roberts [1992], Srinivasan [2005]). However, reputation incentives also mo tivate the members to de velop reputation as a pro-management (Fama and Jensen [1983], Va ncil [1987], Gilson [1990], Kaplan and Reishus [1990], Shivdansani [1993], Shivdasani and Yerm ack [1999], Ferris et al [2003], Petra [2005]). Limitations of the DAHR Model I summarized the major findings in the lite rature over the last decade in Table 1-1. Specifically, I identified eight audit committee attributes that influence reporting quality: independence, accounting expertise, directorships, tenure, committee size, charter, meetings, and equity holdings. As highlighted by the frequent ly conflicting evidence, despite its significant face validity, the DAHR m odel still has significan t potential for improvement. Specifically, the model has two limitations: no explicit role for managerial influence and an overly simplified process section. The ability and incentives of management to act opportunistically directly influence the likelihood that the audit committee is effective. If management is forthright, then audit committee effectiveness is high, regardless of the i nputs and processes. If management is highly

PAGE 18

18 motivated to undermine audit committee effectivene ss, then the likelihood that the committee is effective diminishes. The failure to include a role for managers limits th e validity of the model. The DAHR models process secti on is inadequate since it fails to appreciate that audit committees are groups and effectiveness is a choi ce. The research on audit committee operations has focused on how individual members reach deci sions, yet how committees make decisions is fundamentally different from ho w individuals make decisions (Davis [1992]). Audit committee operations are influenced by the procedural mechanisms and the group dynamics, yet the models only process factor was di ligence. The failure to consid er how the group nature of audit committees influences undermines the validity of the model.

PAGE 19

19 Table 1-1. Research on audit committee effectiveness and reporting quality Paper Independent variable Measure of reporting quality Findings Beasley [1996] Independence Fraud Negative (not significant) Bedard et al. [2004] Independence Earnings management Negative (not significant) Vafeas [2005] Independence Earnings management Negative (not significant) Klein [2002] Independence Earnings management Negative (*) Krishnan [2005] Independence Internal control problems Negative (***) Bradbury, Mak, and Tan [2004] Independence Earnings management Negative (***) Carcello, Neal, Palmrose, and Scholz [2006] Independence Restatement Negative (**) Abbot et al. [2004] Independence Restatement Negative (***) Farber [2005] Independence Fraud Positive (not significant) Bedard et al. [2004] Accounting Expertise Earnings management Negative (***) Song and Windram [2004] Accounting Expertise Reporting problems Negative (not significant) Yang and Krishnan [2005] Accounting Expertise Internal control problems Negative (not significant) Krishnan [2005] Accounting Expertise Internal control problems Negative (**) Abbot et al. [2004] Accounting Expertise Restatement Negative (***) Qin [2006] Accounting Expertise Earnings Quality Positive Carcello, Neal, Palmrose, and Scholz [2006] Accounting Expertise Restatement Negative (***) Carcello, Hollingsworth, Klein, and Neal [2006] Accounting Expertise Earnings management Negative (**) Bedard et al. [2004] Firm Expertise (Tenure) Earnings management Negative (not significant) Xie, Davidson, and DaDalt [2003] Firm Expertise (Tenure) Earnings management Negative (**) Vafeas [2005] Firm Expertise (Tenure) Earnings management Positive (not significant) Yang and Krishnan [2005] Firm Expertise (Tenure) Earnings management Negative (**) Bedard et al. [2004] Directorships Earnings management Negative (**) Xie, Davidson, and DaDalt [2003] Directorships Earnings management Negative (**) Song and Windram [2004] Directorships Reporting problems Negative (not significant) Vafeas [2005] Directorships Earnings management for small earnings increases Negative (not significant) Vafeas [2005] Directorships Earnings management to meet analyst forecasts Positive (not significant) Yang and Krishnan [2005] Directorships Earnings management Negative (***) p-value <0.1, ** p-value <0.05, *** p-value <0.01

PAGE 20

20 Table 1-1. Continued Paper Independent variable Measure of reporting quality Findings Bedard et al. [2004] Committee Size Earnings management Positive (not significant) Vafeas [2005] Committee Size Earnings management for small earnings increases Negative (not significant) Vafeas [2005] Committee Size Earnings management to meet analyst forecasts Positive (not significant) Yang and Krishnan [2005] Committee Size Earnings management Negative (***) Krishnan [2005] Committee Size Internal control problems Negative (not significant) Archambeault and DeZoort [2001] Committee Size Suspicious auditor switches. Abbott, Parker, and Peters [2004] Committee Size Restatemen t Negative (not significant) Qin [2006] Committee Size Earnings quality Positive Carcello, Neal, Palmrose, and Scholz [2006] Committee Size Restatement Negative (***) Farber [2005] Committ ee Size Fraud Negative (Not Signiticant) Bedard et al. [2004] Charter Earnings management None Bedard et al. [2004] Meetings Earnings management Negative (not significant) Xie, Davidson, and DaDalt [2003] Meetings Earnings management Negative (**) Vafeas [2005] Meetings Earnings management for small earnings increases Negative (**) Vafeas [2005] Meetings Earnings management to meet analyst forecasts Positive (not significant) Song and Windram [2004] Meetings Reporting problems Negative (not significant) Yang and Krishnan [2005] Meetings Internal control problems Positive (not significant) Xie et al., 2003; Meetings Earnings management Negative (**) Farber [2005] Meetings Fraud Negative (*) Carcello, Neal, Palmrose, and Scholz [2006] Meetings Restatement Negative (not significant) Abbott, Parker, and Peters [2004] Meetings Restatement Negative (***) Farber [2005] Meetings Fraud Negative (**) Song and Windram [2004] Equity Holdings Reporting problems Positive (not significant) Vafeas [2005] Equity Holdings Earnings management for small earnings increases Positive (not significant) Vafeas [2005] Equity Holdings Earnings management to meet analyst forecasts Negative (**) Klein [2002] Equity Holdings Earnings management Negative (*) Yang and Krishnan [2005] Equity Holdings Earnings management Positive (***) Bedard et al. [2004] Equity Holdings Earnings management Positive (not significant) Bedard et al. [2004] Equity Holdings Earnings management Positive (not significant) p-value <0.1, ** p-value <0.05, *** p-value <0.01

PAGE 21

21 Out p ut CompostionAuthorityResourcesACEDiligenceProcess Input Figure 1-1. DAHR model of a udit committee effectiveness

PAGE 22

22 CHAPTER 2 A MODEL OF AUDIT CO MMITTEE DECISION-MAKING How do audit committees decide whether the manager is opportunistically reporting or not? The audit committee must not only find evidence of opportunistic reporting, but also correctly interpret the ev idence and decide the appr opriate response. I explored how an audit committee decides whether or not to allow aggr essive reporting using a decision-making model based on signal detection theory. A model of the audit committee decision-maki ng has many benefits. It can identify the determinants of audit committ ee effectiveness and understand their influence on the decisionmaking process. Prior research has largely overlooked how the committee operates, making it difficult to understand how contextual factors may influence decisions. For example, it is unclear how different risk environments infl uence the effect of incentives on committee decisions. Further, a model may reconcile the fre quent inconsistencies in the prior research on the influence of audit committee at tributes. For example, some evidence that suggests equity holdings improve monitoring (Beasley [1996], Klei n [2002], Vafeas [2005] ) and other evidence suggests equity holdings promote a focus on shor t-term performance (Bedard et al. [2004], Yang and Krishnan [2005]). The difficulty in modeling the audit comm ittee decision-making process is two-fold. First, the model needs to inco rporate two processes: detecti ng evidence and determining the appropriate response. If the model only considers one stage, then it woul d trivializes the nuances of the decision-making process. For example, if the model ignores the response decision, then this wrongly assumes that the committees response is automated and professional judgment is irrelevant. If the model ignores the detection process, then this wr ongly assumes that audit committees always have the information to make the correct decision, regardless of its resources

PAGE 23

23 or effort. Second, the decision-making process ha ppens at individual level and group level. If the model only considers the individual level, then it ignores the poten tial for group synergy or group dysfunction. Further, if the model only cons iders the group level, then it ignores that the group is the product of indi vidual contributions. To address th ese difficulties, I based my model on signal detection theory. Signal detection theory provides a theore tical framework of a diagnostic task that separates the decision-making pro cess into a discrimination task and a decision (Swets [1986]). Further signal detection theory can analyze how groups integrate individual contributions into a group decision (Sorkin and Dai [1994] ). While accounting researcher s have advocated the use of signal detection theory to analy ze empirical data (Ramsay and Tubbs [2005]), there have been no models presented using signal de tection theory to analyze an accounting decision. Further, no accounting researcher to date has employed group-level signal de tection theory. Model Development Signal detection theory examines how a decisi on-maker determines the source of a signal from a finite set of sources. It is widely used in the electrical engineering and psychology literature because of its ability to calculate accuracy and bias contemporaneously, ability to calculate optimal accuracy and its ability to analyze individual contribution to group decisionmaking (Tanner and Swets [1954], Swets, Tanner, and Birdsall [1961], Sorkin and Dai [1994], Ramsay and Tubbs [2005]). Signa l detection theory can be used to analyze any binary decision. Despite these benefits, accounti ng researchers have rarely empl oyed this theory (Brown [1981], Blocher et al. [1986], Sprinkle and Tu bbs [1998], Ramsay and Tubbs [2005]).6 6 Ramsey and Tubbs [2005] argues the benefits of signal detection theory for accounting researchers but does not conduct any orig inal research.

PAGE 24

24 While many audit committee decisions are comp lex, most decisions boil down to a binary decision, such as whether or not the manager re porting opportunistically. Consider the setting where the audit committee must determine whether the reported earnings are the result of productive effort alone, or are th e result of productive a nd opportunistic effort. Further, assume a typical agency setting involving shareholders, managers, auditors, and an audit committee. At T0, shareholders hire a manager to provide productive services and agree to compensate the manager according to reported earnings (R). Th e manager is required to report truthfully. At T1, the manager exerts effort to pr oduce genuine earni ngs (E). At T2, the manager reports R to the shareholders where R is a f unction of E and S, where S is a distortion the manager may elect to include. For simplicity, assume that S always inflates earnings and is either present or not.7 At T3, the audit committee hires an audito r to confirm that R=E. At T4, the auditor reports to the audit committee evidence that R>E and the mana ger presents evidence that R=E. At T5, the audit committee determines its response to the auditor-provided evidence. To determine its response, the members review the evidence presented by both the auditor and the manager. Each members response on whether R=E or R > E depends upon R, his detection ability, and his bias. If the members communicate a di chotomous response (i.e. agree or disagree with the auditor), then the chair tallies the votes to determine if there is sufficient support to conclude opportunistic reporting. If the members co mmunicate a continuous decision statistic, then the committee combines the i ndividual statistics into an aggregate decision 7 It is possible that the manager could desire to manage earnings downward. The audit committee would sequentially analyze the potential earnings management strategies such as big bath, cookie jarring, and managing earnings to meet thresholds (expectations).

PAGE 25

25 statistic. A decision rule will translate the gr oup decision statistic into an outcome. If the committee concludes opport unistic reporting, then R is reported to the market. For simplicity, assume that if R =R, then the market interp rets this as good news and responds positively, and if R 0, E can represent a number of economic constructs including earnings or

PAGE 26

26 normal accruals. S can represent the effect of an inappropriate polic y, accelerated revenue recognition, or accrual management. Detection Ability A member's ability to discriminate between truthful and opportunist ic reporting depends upon the difference between the means of the distri butions and the standard deviation. Define detection ability as: 10' d (Tanner and Birdsall [1958]) d varies between 0 for a chance level performance, and approximately 4 for errorless performance. d is a function of both skill and effort. If a member has greater training, experience, or natural ability, or exerts greater effort, then he is better able to discriminate between truthful and opportunistic reporting. Evaluating the Evidence The decision statistic is the members estimat e of the likelihood of opportunistic reporting given R. The optimal decision statistic is the likelihood ratio (Hoba llah and Varshney [1989]): 8 (/) (/) f RES z f RE This statistic represents the members ev aluation of the strength of evidence suggesting that the manger is opportunistically reporting. Th e higher the value of z, the more compelling is the evidence that the manager is opportunistically reporting. The value of z depends upon the members ability to discriminate between E and ES and the magnitude of S under consideration. 8 This will be optimal under a number of different criteri a (Hoballah and Varshney [1989]): Maximum hit rate at fixed false alarm rate, maximum expect ed value, maximum information, and maximum difference between posterior distributions. Alternatively, the optimal decision statistic can be a monotonic transformation of the likelihood ratio.

PAGE 27

27 Determining the Response After evaluating the evidence, the member must decide if there is sufficient evidence to conclude opportunistic reporting. A member reaches his conclusion by comparing his decision statistic (i.e. evaluation of evid ence) with his evidence threshold. The evidence threshold is called the decision criterion (zc). The member conclude s opportunistic reporting if z>zc, otherwise he concludes tr uthful reporting. A high zc indicates a management (liberal) bias, and a low zc indicates an auditor (conservative) bias. If the member is rational, then the decision criterion is a function of his incentives. Since audit committee members rarely receive performan ce related remuneration, their incentives are indirect (Black et al. [2005]). Specifically, audit committee incentiv es fall into two categories: reputation and equity. I electe d not to model other incentives (opportunities to enhance their knowledge, networking opport unities, litigation risk9, and prestige of serving) because they have minimal significance relative to the other incen tives (Fama and Jensen [1983], Lorsch and MacIver [1989], Olson [1999], Ki rk [2000], Black et al. [200 5], Vera-Munoz [2005]). Reputation incentives Audit committee members have incentives to develop two reputations : a reputation as an effective monitor and a reputation as pro-manage ment. A members reputation as an effective monitor is valuable because it influences the members opportunities to serve on boards (Milgrom and Roberts [1992]). Yet, a members reputation as pro-mana gement is important because it too influences the members opportuni ties to serve on other boards (Vancil [1987], 9 Between 1968 and 2003, there were only four cases of di rectors contributing their own money to settle a lawsuit, since market forces and political dynamics may have kept member liability risk low (Black et al., [2005]). Srinivasan [2005] found only 78 of the 2,016 directors of firms that restate earnings between 1997 and 2001 are named as defendants in lawsuits and there is no evidence that any director paid out any money.

PAGE 28

28 Petra [2005], Shivdasani and Yermack [1999]). These different, likely conflicting, reputation concerns motivate different behaviors. To develop a reputation as an effective monitor, the member makes decisions that minimize the risk of an irregularity going undete cted (i.e. conservative manner). However, minimizing the risk of an irregularity also incr eases the probability of a false alarm. These incentives are also not necessarily consistent w ith maximizing reporting qual ity, as earnings are likely to be understated. Ones reputation as an effective monitor improves when there are no reporting failures and diminishes with failure s (Fama [1980], Fama and Jensen [1983]). For example, members will lose directorships if their companies experiences bankruptcy, restatements, or SEC investigations (Gilson [1990], Srinivasan [2005]). To develop a reputation as pro-management the members allow the manager greater discretion over reported earnings A members pro-management reputation increases when the firm performs well, but decreases when th e firm underperforms (Fama and Jensen [1983], Shivdansani [1993], Ferris et al. [2003]). When there is a false alarm, the members risk developing a reputation as one who always clai ms the manager acts opportunistically. Define the reputation incentives for a decision as the function:(RV)=(RV) f where RV is the present value of current and future directorships prior to the decision, and is the magnitude of the effect. If is greater than zero, the member s reputation improves. Otherwise, the members reputation is damaged. captures the net effect on both reputations as an effective monitor and as pro-management. Equity incentives Corporations effectively tie an audit co mmittee members personal wealth to firm performance by compensating members with st ock and stock options (Yermack [2004]).

PAGE 29

29 Intuitively, it seems problematic to have an i ndependent member financia lly vested in a firms reported performance. Critics allege that equity holdings motivate directors to artificially boost stock-related performance (Mills tein [2002], Pitt [2002]). Proponent s argue that equity holdings provide greater incentives for the members to be effective monitors (Klein [2002]). Define the equity value-maximizing incen tives for a decision as the function: (H)=(EH) fE where EH is the value of all equity holdings prior to the decision and is the magnitude of the effect. Calculating zc zc is calculated as a function of the reputation and equity ince ntives for each of the four outcomes. Let V be the utility the member derives from each outcome: 10 ()() VRVEH such that Manager Audit Committee Reputation Equity Reports Concludes Reference Effect Effect Opportunistically Opportunistic Hit 11 01 11 0 Reporting V11 Truthfully Opportunistic False Alarm 01 0 01 11 Reporting V01 Truthfully Truthful Correct Rejection 00 0 00 0 Reporting V00 Opportunistically Truthful Miss See Below Reporting V10 Reputation Equity Miss Payoffs Probability Effect Effect Not detected by market q4 104 0 104 0 Detected Restated earnings q1 101 0 101 0 SEC Action q2 102 101 102 101 Bankruptcy q3 103 102 103 102 where q1 + q2 + q3 +q4=1 and q4>q1 q2 q3. Assume that the loss probabilities are exogenous. 10 This is a similar payoff function as Barkan, Zohar, and Erev [1998].

PAGE 30

30 zc can be calculated using maxi mum expected value criteria: 0001 1110()() ()()cVV pEpE zV VVpESpES Barkan, Zohar, and Erev [1998]. where p(E) and p(ES) are the prior probability of the truthful and opportunistic reporting. If the member is risk neutral, then 00000101 4 11111010 1() () ()c iii iRVSHRVSH p E z p ES RVSHqRVSH If the member is risk averse with utility function U(x)=-ex+K ( is a measure of risk aversion and K is an arbitrary constant), then 00000101 1010 1111()() 4 () () 1() () ()iiRVSHRVSH c RVSH RVSH i ieepE z p ES eqe The committee concludes oppor tunistic reporting if z> zc and truthful reporting otherwise (Figure 2-1). Multi-Person Audit Committee Now assume that the committee has n memb ers, with each member communicating a continuous variable that indicates how co mpelling they found the evidence suggesting opportunistic reporting. While each member observes the same R, they interpret it differently depending upon his d. Define vector X as the list of decision statistics generated by each of the members: . If each member communicates the optimal decision statistics, then X consists of likelihood ratios. Define vector as the difference between the members mean estimate of X when there is truthf ul and opportunistic reporting: < 1, 2,, n>. Each members decision statistic is subject to several noise factors. For simplicity, assume that all noise sources are additive, indepe ndent of whether or not the signal is present,

PAGE 31

31 and normally distributed random variables with mean 0. Unique noise (2 i ) is noise that is exclusive to one member. Common noise is noise that is shared by some members (2 ij for members i and j) or all members (2 common) (Figure 2-2). The variance of a members decision statistic is the sum of all the noise. In Figure 2-2, member 1s variance is: 222 111,3()commonVarX. The covariance of the estimates of any pair of detectors is equal to the sum of the variance of the noise sources common to those detectors. Therefore, the covariance for members 1 and 3 in Figure 2-2 is 22 131,3(,)commonCovXX. The entries of the covariance matrix, summaries the values of the variances and covariances. 1121 1222 12()(,)(,) (,)()(,) (,)(,)()n n nnnVarXCovXXCovXX CovXXVarXCovXX CovXXCovXXVarX Committee Decision Statistic To reach a decision, the committee must aggregate all the individual decision statistics into a single decision statistic. Ashby and Madd ox [1992] showed that as long as the covariance matrix has the same form for the truthful and opportunistic reporting, then the optimal decision statistic is a linearly weighted sum of the detectors estimates. 1' Z Xk (Sorkin and Dai [1994]) where X` is a row matrix, -1 is the inverse of the covariance matrix, is a column vector and k is a constant. For simplicity define a= -1 a represents the relative we ight a members opinion is given in the decision-making process. Therefore, the decision statistic is 1 n ii i Z aXk (Sorkin and Dai [1994])

PAGE 32

32 The difficulty in this definition of Z is obtaining the general form for the inverse of the covariance matrix. To address this issue, rese archers typically make simplifying assumptions. Durlach et al. [1986] assumed that the noise is either common to all members or unique to one member. With this assu mption, all of the off-diagonal elem ents of the covariance matrix are equal. Further, Sorkin and Dai [1994] assumed that the magnitude of individual variances are equal (22 ij for all i and j) with as the uniform pair-wise correlation 2 22com comind The particular value of a depends upon the weighting sche me. Weighting represents how efficiently the group assimilates th e individual opinions into a decision. Two common schemes are optimal weighting and uniform weighting. Optimal Weighting The optimally weighting scheme is defined as: 1 [1(1)]'(')n iij jandd 22 ''' 11(1)dd idealdn n (Sorkin and Dai [1994]) where' d is the average individual d and 2 d is the variance of individual d. Under this method, each members opinion is weighted in pr oportion to the members ability. This means that each members ability is incorporated into the decision regardless of the members ability. Therefore, the optimal weighting is not to de fer entirely to the member with the highest d. Uniform Weighting If each members contribution is given equal weight, then 1ia n 2 '' 1(1)d uniformdn n (Sorkin and Dai [1994])

PAGE 33

33 This approach assumes that each members opinion is equally valid. This weighting approach under-weights the opi nions of members with high d and over-weights the opinion of members with low d. Threshold The committee evidence threshold is the combination of individual member bias, group norms, and/or regulatory standards. Non-Optimal Individual Decision Statistics If the members communication is limited to a dichotomous vari able, then the audit committee is essentially a condorce t voting group. With this re striction, the members can only indicate whether or not they believe there is oppo rtunistic reporting, and not how strongly they hold their belief. Each member calculates his z and compares it with his zc to determine his vote. If the members communicate a dichotomous signal, then it is reasonable to assume that each opinion is equally weighted. The chair tallies the number of member s that believe there has been opportunistic reporting ( ka) The committee will have a decisi on rule that says to conclude opportunistic reporting if ka k. Otherwise, conclude truthful reporting If k equals n, then the committee requires unanimous agreement to conclude opportunistic reporting. If k equals n/2, then the committee requires a majo rity of members to agree to c onclude opportunistic reporting. If k equals 1, then the committ ee only requires a single vote to c onclude opportunist ic reporting. This approach to decision-making fails to fully realize the skills of members with higher d. Observations from the Model 1. Committee Size (n)

PAGE 34

34 If the committee uses optimal weighting or even uniform weighting, then larger committees outperform smaller committees.11 Assuming the number of members does not affect 'd or 'd each additional member improves d, but the amount of improvement declines with each additional member. Specifically, the change from the committee going to n+1 members from n members under optimal weighting is: 22 ''(')(1)0 (1)1(1)dd idealdnn n Specifically, the change from the committee goi ng to n+1 members from n members uniform weighting is: 2 '(')(1)0 1(1)d unifromdnn n If the new members differ from the current members in d, then the effect of the new member is more difficult to predict. Under optimal weighting, even if new members have equivalent d, the magnitude of the improvement depends upon the interaction between and 2 'd (Figure 2-4). Predicting the effect of a new member is mo re problematic if non-optimal weighting is used. For example, if the committee uses unifo rm weighting, then the influence depends upon the new members d relative to the other members. Specifically, if the committee adds a member with above the median d, then the committees effectiveness improves. Otherwise, the committees effectiveness diminishes. The committee could use several other non-optim al weighting schemes. For example, if members accept the recommendation of the member with the greatest d, then an additional 11 A simple proof by contradiction shows that under optimal weighting, a committee of n+1 members has a higher d then a committee of n members.

PAGE 35

35 member will only influence the de cision if he has the greatest d on the committee. Alternatively, if the committee weights a membe rs opinion in proportion to the members length of service, then new members will only have sign ificant influence after years of service. If the members only communicate a dichotom ous signal, then the benefits of new members are highly dependent upon the decision rule and current composition. For example, if a committee uses a simple majority decision rule and the committee is e qually divided between liberal and conservative members with the cons ervative chair casting the tiebreaking vote, then an additional liberal member changes the committee from conservative to lib eral. Alternatively, if the committee consists of only conservative members, then the addition of a liberal member has minimal influence. Therefore, the influence of increasin g committee size depends upon the weighting method, communication method, bias of the curr ent committee members, and the new members d relative to the other members. 2. Mean of Members Detection Ability (' d ) An audit committee can increase'd by increasing the amount of time discussing issues, engaging in member training progr ams, employing consultants to pr ovide additional evidence, or gathering additional evidence from the external or internal auditor. Trivially, observe that improving 'd improves the committees d when a continuous signal is communicated independent of the weighting scheme: 1 2 2 22 ''(')(1) 0 1(1)(1(1))(1)ideald d ddddn d nn

PAGE 36

36 1 2 '(') 0 1(1)uniform ddd n dn If the members communication is limited to a dichotomous signal, then the effect on committee effectiveness from increasing d is more complex. Increasing d means that each member is more likely to correctly differentiate between opportunistic and truthful reporting, hence the probability that the members communicat e the correct signal increases. However, it is possible that the increase in d will not be sufficient to cause the members to alter their conclusion. Therefore, regardless of the weighting sc hemes or communication method, increasing d either has no effect or improves committee effectiveness. 3. Diversity (Inter-mem ber correlations of ability) Diversity refers to differences in knowledge bases and perspectives that members bring to the committee based on differences in education, experience, and expertise (Jackson [1992]). Diversity promotes divergent th inking (Nemeth [1986]), reduces conformity pressure (DeDreu and deVries [1996]), and promotes thoughtful co nsideration (Nemeth [1986]). While a high promotes an efficient discussion, consen sus of opinion, and limited conflict, high is not necessarily optimal since extended discussi on and conflicts increase decision accuracy. 2 d exists among members with high d because some members ma y have greater accounting expertise while others may have greater industry-speci fic knowledge. The effect of increasing under optimal weighting is:

PAGE 37

37 22 '' 22 1 22 2 ''(1) 11(1) (') 11(1)dd ideal ddn n n dd d n There is a range where increasing will enhance d and a range where it will not (Figure 2-4 and 2-5). If non-optimal weighting is used, then may be undesirable. Observe that with uniform weighting, performance only increases from reducing the among members with uniform weighting: 3 2(') (1) 0 21(1)uniform ddd nn d n The benefits of diversity are conditional. Non-experts will not enhance audit committee effectiveness if the experts do not consider the non-expert s opinions, or if non-experts consistently defer to experts. Experts will not consider non-experts opinions if they believe non-experts opinions are not relevant, or if the non-experts are unable to articulate their opinions (Matz and Wood [2005]). Non-expert s opinions may be discounted because they deviate significantly from the rest of the committee. If interpersonal influence decays exponentially as a non-experts opinion deviates fr om the experts opinion, then a non-experts weight in the decision is a function of the total discrepancies betw een the non-expert and everyone else. Mathematically, this is: 1'n ii i Z bX where '1 1'1(|') for jj' (|')r jj j i rr jj ijexx b exx where is a positive constant. Under this approa ch, a non-experts opinion enhances decision

PAGE 38

38 quality, provided that the opinion is not too different from the rest of the committee. This is consistent with a trimming heuristic where expe rts ignore outlying opinions (Yaniv and Hogarth [1993]). It is likely that the more technical the issue, the more li kely that expert and non-expert opinions will conflict. Non-experts may defer to experts because of a feeling of social pressure to comply, a desire to reduce their responsibilit y if a problem occurs, or a bias of assuming expert equals correct analys is (Harvey and Fischer [1997]).12 Therefore, the influence of nonexpert opinions may be proportionate to the technical difficulty of the issue. 4. Weighting The weighting scheme determines how effec tively the committee combines the individual contributions into a decision. Under optimal weighting, the committee considers not only a members opinion, but also his ability. Under uniform weighting, ones opinion matters, but not ones competence. A more effective weighting scheme increases the committees d (Figure 26). While there are many additional non-optimal weighting schemes, the one constant of any non-optimal weight is that the committ ee fails to realize its potential. 5. Procedural Mechanisms Procedural mechanisms govern how the committee operates. The procedural mechanisms include discussion protocols, agenda setting protocols, and decision rules. The model specifically highlights th e importance of decision rules. A committee can employ a number of different decision rules: unanimous ag reement before action, simple majority, supermajority, etc. Changing the decision rules can cha nge the decision. For example, if the decision rule requires unanimous approval, then the committ ee is liberal if a single member has a liberal 12 This is similar to the stereotype biases identified in Cialdini [1984] where consumers would assume jewelry is of higher quality simply if the price were higher.

PAGE 39

39 bias or low d. Under this scenario, the members wi th the most liberal bias or lowest d determine the outcome and not the more skilled, unbiased members. However, if the committee employed a simple majority decision rule, then the influence of any one individual declines. Therefore, in order to predict the influen ce of members it is impor tant to understand how procedural mechanisms integrate indivi dual contributions in to a group decision. 6. Communication Method Committees have flexibility with what information the members communicate to each other. Members can communicate a binary (i.e agree or disagree comment), ordinal (i.e. strongly agree, agree, disagree, strongly disagree ), or continuous signal (i.e. agree/disagree and exact degree of confidence). If members co mmunicate a more informative signal, then the likelihood of reaching the correct decision increase s as the members have more information to jointly consider. As identified in the model, the most informative signal the members can communicate is the likelihood ratio. When tr ansmitting a dichotomous signal, the committees ability to efficiently weigh the members opin ion is limited. Therefore, while traditional decision-making is predicated upon members tall ying votes for or against a motion, decisionaccuracy can be improved if the members communi cate a degree of support for or against a motion. 7. Member Incentives (Committee Bias) The ideal incentive structure yields a high util ity for correctly identifying truthful and opportunistic reporting, and a high disutility for failing to iden tify opportunistic reporting and false alarms. If the penalty for failing to iden tify when the manager reports opportunistically is too high, then the member will always concl ude that the manager re ports opportunistically, thereby inflating false alarms. If the relati ve reward for correctly concluding opportunistic

PAGE 40

40 reporting is too low, then the member will always conclude truthful repor ting. Therefore, to understand the bias it is necessary to understand how these incentives influence the members judgment. Further, the bias depends upon the risk envir onment. In a high-risk environment, the members will be more likely to conclude opportunist ic reporting than in a low-risk environment. Many factors may influence the audit committee me mbers assessment of the risk environment, such as managerial incentives to opportunistical ly report and auditor qual ity. If the manager has strong incentives to opport unistically report, then the risk of opportunistic re porting is greater than if the manager has weak incentives to oppor tunistically report, assuming the managers are utility maximizers. When the firm employs a low quality auditor, then the probability of opportunistic reporting is greater than when the firm employs a high quality auditor, assuming high quality auditors are more effective at preventing opportunist ic reporting. 8. Independence (Non-incentive based bias) The model assumed that committee members ma ke their decisions, in part, by evaluating their incentives. However, other biases may dist ort the evidence threshold. For example, if the members trust management, then they may require a significantly greater amount of evidence before concluding the manager is acting opportunisti cally. Alternatively, other biases may lead the members to adjust their effort level (i.e. adjust d ). The presumption among regulators is that a bias reduces d and increases the evidence threshold, but it is possible th at a conservative bias will have the opposite effect. For example, if the members dislike th e manager, then they may reduce their evidence threshold. Regardle ss, any bias will have an influence on the decision. 9. Strength of Signal

PAGE 41

41 The smaller the difference between the E and ES the more difficult it is to differentiate between E and ES. For example, it is much more difficult to detect evidence of a quantitatively immaterial distortion than to detect evidence of a quantitatively material distortion. Therefore, the effectiveness of the committee depe nds upon the magnitude of the signal under consideration. A secondary factor that influences the commi ttees ability to differentiate between the E and ES is the ability of the ma nager to act opportunistically whil e appearing to act truthfully. The audit committee must evaluate the evidence presented by management. The more effective managers are at disguising inappropriate behavior, the more the behavior will go undetected. While it is impossible to capture this managerial attribute, se veral factors may influence the managers ability to distort earnings, such as accounting expertise, tenure, and executive influence over the board. 10. The Status Quo If the committee fails to fi nd sufficient evidence that R E, then it will conclude R=E, but it would be equally valid to conclude R E unless there is sufficient evidence to conclude that R=E. Some may argue that Sarbanes-Oxley changed th e status quo from the manager having free reign to implement policies unless the audit committee interceded, to the mana ger requiring approval before he can implement any policy. The signifi cance of the status quo is especially important when considered jointly with th e decision rule. For example, if the status quo is to reject managerial actions unless the audit committee unani mously allows it, then a single member can compel conservative behavior. However, if th e status quo is to allow managerial action unless the audit committee unanimously allows it, then a single member can compel libel behavior.

PAGE 42

42 Reconciling the Model and the Literature To examine the validity of my model, I compared the recent literature on audit committees with models observation to determine if the observations are consistent with the empirical evidence. Observation 1: Committee Size Audit committee size is one of the most re searched determinants of audit committee effectiveness. Larger audit committees are less li kely to have suspicious auditor switches, and restatements (Archambeault and DeZoort [2001], Car cello et al. [2006]). However, the recent findings were more ambiguous. For example, recent work found no association between committee size and internal control problems, re statements, and earnings management (Abbott et al. [2004], Bedard et al. [2004], Krishnan [2005], Vaf eas [2005]). In additi on, several studies did not even consider committee size (Klein [ 2002], Song and Windram [ 2004], Yang and Krishnan [2005]). While these null results suggest that th e importance of audit committee is declining, it is more likely that the regulatory changes that established a minimum committee size have homogenized the data, making it difficult to detect a relationship. Therefore, despite the lack of evidence in some recent work, there is strong evidence that audit committee effectiveness depends upon the size of the committee. Observation 2: Member Ability The importance of an audit committees techni cal skills has been the focal point of many studies. For example, audit committees with experts are less likely to experience suspicious auditor switches, fraud, restatements, earnings management, or internal control problems (McMullen and Ragjhunandam [199 6], Beasley et al. [2000], Archambeault and DeZoort [2001], Xie et al. [2003], Abbott et al. [2004], Krishnan [2005], Yang and Krishna n [2005], Carcello, et al. [2006]). However, Song and Windram [ 2004] found no evidence in the U.K. of an

PAGE 43

43 association between audit committee expertise and financial reporting proble ms. Carcello et al. [2006] found that accounting expe rts reduce the incidence of ea rnings management, but strong governance negates the benefits of a financial expert. Therefore, there is strong evidence that audit committee effectiveness depends upon the members d. Observation 3: Diversity Only recently have researchers considered the implications of diversity on audit committee effectiveness. Initial work by McDaniel et al. [2002] experimentally examined the differences in how experts and nonexperts (literates) consider i ssues. They found that experts focus on less-prominent, recurring issues, while literates focus on more prominent, nonrecurring issues. Subsequent archival research examined how different experts infl uence reporting quality. Xie et al. [2003] found that accounting experts, corporate experts, and investment bankers enhanced a committees ability to prevent earnings management. Bedard et al. [2004] found that financial experts and governance experts influenc ed an audit committee s willingness to allow aggressive earnings management. Carcello, Ho llingsworth, Klein and Neal [2006] found that both accounting experts and non-experts reduce the incidence of earnings management. Therefore, there is preliminary evidence that suggests audit committee effectiveness depends upon at least one type of diversity of the committee. Observation 4: Weighting The research to-date has yet to consider how the committee weighs individual member opinions. The interviews by Gendron and Beda rd [2006] suggested that the members do consider a variety of factors when weighing individual member opinions. In particular, one interviewee claimed that long-tenure members were more influential. Furt her, the interviewees suggested that new members and non-experts da mage their credibility by asking questions

PAGE 44

44 perceived as irrelevant. Therefore, there is only anecdotal evidence that audit committee effectiveness depends upon how the commi ttee weighs individual member opinions. Observation 5: Procedural Mechanism Spira [2002] argued that ceremonial features surrounding audit committee meetings play a role in the committees effectiveness. Gendr on and Bedard [2006]s interviews of audit committee members confirmed this intuition. Ther efore, there is anecdotal evidence that audit committee effectiveness depends upon the procedural mechanisms. Observation 6: Communication Method While it is doubtful that audit committee co mmunication would ever be restricted to a dichotomous signal, how well the co mmittee communicates information is critical to its success. Spira [2002] and Gendron and Bedard [2006] claimed that audit committees operated by consensus and discussion. If the committee spends more time in discussion, then the members will share more information. Theoretically, m eeting time should be positively associated with audit committee effectiveness. Unfortunately, au dit committees do not disclose the meeting time in the proxy statement and no survey to date has linked meeting length to audit committee effectiveness. Therefore, while there is strong theory to suggest audit committee effectiveness depends upon the communication method, there is no empirical evidence to support this assertion. Observation 7: Incentives Empirical researchers have focused upon th e influence of equity incentives and reputation incentives. The resear ch on the relationship between audit committee incentives and audit committee effectiveness is mixed. Klei n [2002] found that the presence of a blockholder13 13 Klein [2002] defines a blockholder as a member owning greater than 5 percent of the company.

PAGE 45

45 on the audit committee reduced the level of abnorm al accruals. Bedard et al. [2004] found that the greater the audit committee members ratios of options to stock holdings, the more likely the audit committee members were to tolerate extr eme earnings management. Vafeas [2005] found no evidence of a relationship be tween equity holdings and the likelihood that a firms earnings met the previous years earnings level. He found that equity holdings were negatively related to the likelihood that a firm meets analysts foreca sts. Yang and Krishnan [2005] found that equity holdings of independent and nonindependent members were positiv ely related to the level of abnormal accruals. Collectively, these results de monstrate that equity holdings influence the committees judgment, but the direction is not cons istent. To date, there is no explanation for these inconsistencies. Similarly, prior research has found mixed ev idence on the influence of audit committee directorships. Directorships are positively a ssociated with monitoring effectiveness (Fama [1980], Fama and Jensen [1983], Milgrom and Roberts [1992], Sriniv asan [2005], Yang and Krishnan [2005]). However, directors have stro ng incentives to deve lop a reputation as promanagement because CEOs nominate members with a pro-management bias (Vancil [1987], Petra [2005], Shivdasani and Yermack [1999]). These conflicting effects may explain the null results in Song and Windram [2004] and Vafeas [2005]. Observation 8: Independence Audit committee independence is one of the first determinants of audit committee effectiveness considered by researchers. C onsistent with expecta tions, independence was positively associated with effectiveness. For example, companies with independent audit committees were less likely to experience earnin gs management, internal control problems, SEC sanctions for fraudulent or misleading financial reporting, fraud, and su spicious audit switches

PAGE 46

46 than companies without independent audit co mmittees (McMullen and Ragjhunandam [1996], Abbott et al. [2000], Beasley et al. [2000], Archambeault an d DeZoort [2001], Klein [2002], Bedard et al. [2004], Krishnan [2005]). Howe ver, Yang and Krishnan [2005] found no evidence of an association between a udit committee independence and quarterly earnings management. While the current regulations require that all audit committee members be independent, not all independent members are equally unbiase d. For example, some independent members have a pro-management bias (high zc) (Shivdasani and Yermack [1999] ). Carcello et al. [2006] found that if the CEO were involved in th e nominating process, then the likelihood of restatement is increases even if the members are independent. Therefore, there is strong evidence that audit committee effec tiveness depends upon the committee members independence. Observation 9: Strength of Signal While research has considered the effectiveness of audit committees in preventing many different types of financial re porting problems (fraud, SEC sanc tions, restatements, earnings management), no study has examined the relative effectiveness of audit committees in differing settings. Therefore, there is no evidence that audit committ ee effectiveness depends upon the magnitude of reporting e rror under consideration. Observation 10: Status Quo While it is impossible to identify individua l member preconceptions, many of the recent reforms have shifted the status quo from ma nagerial freedom unless the audit committee intervenes to managerial constr aint unless the audit committee a llows discretion. For example, before 2002, there were no restri ctions on auditor provided nonaudit services, but SarbanesOxley prohibited all non-audit serv ices unless the audit committee specifically allows it.

PAGE 47

47 Therefore, there is strong evidence suggesting that audit committee effectiveness depends upon the status quo. Reconsidering the DAHR Model In this final section, I will examine how my model compares to the DAHR Model and will propose to extensions to the DAHR Model. Consistency with DAHR Model Each of the four core fact ors in the DAHR Model is consistent with my model. If the committee does not have the authority to make a deci sion, then it is not effective since the status quo always remains (observation 10). If the co mmittee has greater resources (i.e. larger committee size), then it is more likely to be eff ective (observation 1). If a committee has greater expertise (i.e. higher d ), then it is more likely to be effective (observation 2). If the committee lacks independence, then there are three possible effects. Fi rst, non-independent members may calculate their zc independent of their incentives. Second, non-independent members may be less risk averse. If the members are trusting ma nagement, then they underestimate the risk of a reporting problem. Third, d may decline because the nonindependent members are less diligent. If the members are trusting manageme nt, then they may reduce their effort or accept their bias assessment of the mana ger as evidence of the truthful ness of the financial reporting. The net result of these three eff ects is that the committee is less effective (observation 8). If a committee is more diligent (i.e. higher d ), then it is more likely to be effective (observation 2). Collectively, this indicates that the DAHR m odel is consistent with the decision-making processing model theorized in my study. While the DAHR Model is consistent with my model, it fails to capture all the factors. Therefore, I proposed two modifications to better reflect the de cision-making process (Figure 2-

PAGE 48

48 3). First, expand the model to consider th e influence of managers on the decision-making process. Second, divide the pr ocess section into two factors: individual contri butions and group processing. Individual contri butions are a function of di ligence and incentives. Group processing is a function of procedur al mechanisms and group dynamics. External Influences The DAHR Model failed to consider how any ex ternal factor influences the committees effectiveness. Yet audit committee ineffect iveness is only possibl e if the manager acts inappropriately because the audi t committee cannot eliminate opportuni stic behavior if there is none. Therefore, to understand when the audit co mmittee is ineffective, it is necessary to understand managerial behavior. Nelson et al. [2003] identified many factors that influence the managerial decisions such as i ndustry factors, supplie rs, and the regulatory environment. Two specific factors that may affect the managers influence on audit committee effectiveness are managerial ability and manageri al incentives (Figure 2-3). Managerial ability represents the managers capability to manage earnings in such a way as to benefit himself and be allowed by the aud itor and audit committee. This capability depends upon the managers accounting skills and organiza tional power. When the manager has greater accounting skills, then he is more capable of designing transactions to subvert the accounting principles. Aier et al [2005] found that the likelihood of re statement was negatively related to the CFOs accounting expertise. The managers power represents the managers ability to influence the reporting system. The managers power depends upon many factors such as the strength of the internal audit function, the managers role in the nomination pr ocess, and the managers personality. A strong internal audit function will li mit the managers ability to act without the audit committee knowing. For example, an increase in internal audit functions status may prevent or detect

PAGE 49

49 material misstatements (Raghunandan and McHugh [1994]). Further, if the internal audit function reports to the CEO, then the intern al audit function effectiveness diminishes (Raghunandan et al. [2001], Raghuna ndan et al. [1998]). When th e manager has a role in the nominating process, he has the ability to place pro-management members on the audit committee and remove unaccommodating members. Carcello et al. [2006] found CE O involvement in the nominating process comprises audi t committee effectiveness. According to agency theory, managers will oppo rtunistically report if it is in their best interest to do so. The mana gerial incentives to opportunist ically report include more job security, maximizing bonuses, and maximizing th e value of personal equity holdings. The likelihood of an irregularity is positively associated wi th the CEOs incentives to opportunistically report (Efendi et al. [2004], Burns and Ke dia [2005], Harris and Bromiley [2005]). Specifically, CEO equity incentives are positively associat ed with opportunistic behavior (Klein [2002], Ba rtov and Mohanram [2004], Cheng and Warfield [2005]). Expanded Processing Section The second modification divides the process section into individual contributions and group processing. Individual contributions repres ent the committees potentia l to be effective. The group processing represents how efficiently the committee converts individual contributions into a decision. It is a function of pr ocedural mechanisms and group processing. Individual contributions Individual contributions are a function of d iligence and incentives. While DeZoort et al. [2002] discussed diligence factor s at length, audit committee incentives have received only passing consideration. Audit committee member s have a variety of incentives, including maximizing equity value and reputation developmen t. These incentives can motivate the audit committee to be liberal or conservative.

PAGE 50

50 Audit committee incentives are only relevant when three conditions are satisfied. First, the manager must have the potential to act opport unistically. If the ma nager has no opportunity to misreport, then the manager will truthfully report. Second, the audi tor must be willing to allow a potentially opportunistic tr ansaction. If the auditors do not allow an opportunistic transaction, then the audit comm ittee has limited ability to allow it. Third, the decision must impact the audit committee member personally. If the decision does not influence the audit committee member personally, then there are no incen tives. Numerous settings that satisfy these conditions, including prior to stock market transactions and to m eet earnings targets (Shivakumar [2000], Teoh et al. [1998], Ranga n [1998], Dechow et al. [1996]). Group processing How committees operate is as important to predicting committee performance as individual member attributes (Dri skell, and Sala [1991]). Groups differ from individuals in that they have higher confidence in wrong answers a nd lower confidence in co rrect answers, are less risk averse, and suffer from groupthink (Mos covici and Zavalloni [1969], Janis [1982], Puncochar and Fox [2004]).14 Group processing is a function of the procedural mechanism and group dynamics. Procedural mechanisms are the rules and customs that govern the committee operations. Group dynamics are th e psychological factors that in fluence committee operations. Jointly, these factors determine how effectively the committee comb ines individual contributions into a single group decisions. Procedural mechanisms organi ze and regulate the interperso nal process through agendas, discussion rules, and decision rule s. The recognition of procedural mechanisms as fundamental to the group decision-making process is a relativel y recent development (Davis [1992]). While 14Groupthink is the psychological drive for consensus at any cost that suppresses disagreement and prevent the appraisal of alternatives in cohesive decision-making groups (Janis [1982]).

PAGE 51

51 procedural mechanisms should facilitate decisions, they also influence, frequently in unexpected and undesirable ways, the committees de cisions (Davis et al. [1997]). To illustrate the potential influence of proce dural mechanism, consider the influence of voting protocols. Voting prot ocols are the formal rules that govern how members translate opinions into a decision. While some decisions may not require a formal voting protocol, on contentious issues, a voti ng protocol is essential.15,16 The most crucial vo ting protocol is the decision rule. In a sense, it is the decision rule that determines the outcome, not the votes. Typical decision rules include majority, two-thirds majority, and unanimous agreement. Understanding the decision rule is critical to kno w which members to study. Previous studies focus on the average member, but if unanim ity is required, then knowing what the average member believes is not meaningful since it is the most demanding member that determines the outcome. Group dynamics refers to the effects created by combining the opinions and analysis of several individuals into a committee decision. Group dynamics include discussion strategies, minority influence, group effects, and, notably member diversity. With the exception of diversity, audit committee researchers have not considered any of these factors. Typically, diversity is captured as the mix between experts an d non-experts. While experts are viewed as preferable to nonexperts because they make judgme nts that are more consistent, have greater insights, have greater consensus, an d have greater technical skills than non-experts, the inference that a committee of all experts is strictly superior to a committee with non-experts is dubious since diversity enhances decisi on-making (DeZoort [1998], De Dreu and West 15 For example, if there is only one viable option, or all members are in agreement, then a formal vote is likely not required. 16 If there are no contentious issues then one must question why there is an audit committee. If all the issues have a clearly preferred position then only a single decision-maker is needed.

PAGE 52

52 [2001]). Non-experts enhance a committees ability to identify additiona l issues and promote discussion of different issues (M cDaniel et al. [2002]). Braiotta ([2004]: 56) believed that the committee should consist of both financial and nonfinancial people so that the board can draw on members from various professions, such as accounting, economics, education, psychology and sociology. Kosnik ([1990]: 138) contended that diversity reduces the probability of complacency and narrow-mindedness. If non-experts ar e able to identify additional issues or to provide industry-specific insight s, then a committee with a mix of experts and non-experts should identify at least as many issues as a committee of all experts. Limitations I modeled the audit committee decision-maki ng process for a single decision. This approach failed to capture the multi-period nature of the audit committees existence. Not only do audit committees make a series of decisions in a single meetings but the implications of their decisions can last several years. For example, the decision to allow aggr essive reporting in one period may limit the committees ability to allow aggressive reporting. Future models should account for the cost of loss flexibility as a cost of allowing aggressive reporting. Further, the model also assumed that the audi tor and manager accept the audit committees decision. Yet, it is likely that the auditor and manager will alter their behavior in response to any decision. Future models need to consider any potential cost or benefits to audit committee members based on changes in auditor or manager behavior.

PAGE 53

53 Calculate z Compare Z with zc DecisionManager reports either E or ES2 1Member with d` Figure 2-1. Individual deci sion-making process with signal detection theory

PAGE 54

54 Decision Member 1 with d`12 22 12 1,2,32 1,3Group Decision Making ProcessManager reports either E or ESMember 3 with d`3Member 2 with d`22 3 Figure 2-2. Group decisionmaking process with signal detection theory

PAGE 55

55 Audit Committee Incentives Managerial Ability External Influences Process Committee Membership Managerial Incentives Results Resources Authority Diligence Inputs Procedural Mechanisms Group Dynamics Reporting Quality Figure 2-3. Modified model of audit committee effectiveness

PAGE 56

56 Group d` as a function of Committee Size0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 12345678910 Number of Committee MembersGroup d' given average d`=1 Variance=.5 and Correlation = 0.1 Variance=.5 and Correlation = 0.3 Variance=.5 and Correlation = 0.5 Variance=0 and Correlation = 0.1 Variance=0 and Correlation = 0.3 Variance=0 and Correlation = 0.5 Figure 2-4. Commi ttee size and d`

PAGE 57

57 Group d` as a function of the Correlation Among Members0 1 2 3 4 5 6 7 8 00.10.20.30.40.50.60.70.80.9 CorrelationGroup d` given averge d`=1 Committee Size= 3, Variance=.5 Committee Size= 6, Variance=.5 Committee Size= 9, Variance=.5 Committee Size= 3, Variance=0 Committee Size= 6, Variance=0 Committee Size= 9, Variance=0 Figure 2-5. Correlation and group d`

PAGE 58

58 Group d` as a function of the Correlation Among Members and Weighting Method 0 0.5 1 1.5 2 2.5 3 3.5 4 12345678910 Committee SizeGroup d` given Mean d`=1 and Variance=.5 Correlation = 0.1 and Optimal Weighting Correlation = 0.3 and Optimal Weighting Correlation = 0.5 and Optimal Weighting Correlation = 0.1 and Uniform Weighting Correlation = 0.3 and Uniform Weighting Correlation = 0.5 and Uniform Weighting Figure 2-6. Group d` and weighting method

PAGE 59

59 CHAPTER 3 SETTING AND DATA Audit committee incentives only influen ce the committees decision to allow opportunistic reporting when three conditions are satisfied. Fi rst, the manager must report opportunistically. If the manager reports truthfu lly, then the audit committee is unable to allow opportunistic reporting since their ro le is to oversee the financial reporting process, not actively initiate policies. Second, the a uditor must be willing to tolerate the opportunistic reporting. If the auditor resigns over an issue, then th is nullifies any potential gains from allowing opportunistic reporting.17 Third, ex post the decision to allow opportunistic reporting must influence outsiders perception of the firm. If the audit committees decision to allow opportunistic reporting did not infl uence anyones judgment, then there would be no incentive to allow opportunistic reporting. In my study, I focused upon the audit co mmittees decision to allow opportunistic reporting when earnings are near a key threshold because it sati sfies these three conditions. When earnings are near a threshold, managers ha ve strong incentives to opportunistically report (Burgstahler and Dichev [1997], Burgstahler a nd Eames [1998], DeGeorge et al. [1999], Myers and Skinner [2000], Dechow et al. [2003]). Since many firms just m eet rather than just miss an earnings threshold, researchers cl aim that managers must opportunistically report when earnings are near thresholds (DeGeorge et al. [1999]). The auditor only has weak incentives to prohibit quantitatively immaterial levels of earnings management to meet th e earnings threshold.18 17 If the audit committee were to allow a policy rejected by the auditors, then the auditor will either issue a going concern or resign. If the auditor issues a going concern opinion or resigns, then any gains from allowing the policy are likely lost. 18 I assume any earnings management is quantitatively immate rial because auditors are le gally bound to prevent all quantitatively material misstatements. While it is possible that some firms do have quantitatively material misstatements, they are likely very few unless ther e is a systematic problem with the auditor.

PAGE 60

60 Caneghem [2004] found no evidence of an associ ation between auditor qu ality and the likelihood that a firm meets the previous years earning level.19 The perceptions of firms that just meet relative to firms that just miss an earnings thresh old vary dramatically. There is a significant stock premium (discount) for firms that meet (miss) thresholds and that firms that meet (miss) a threshold are viewed as solid performers (underperformers) (DeFond and Jiambalvo [1991], Wild [1994], Dechow et al. [1996], McMullen [1996], Wild [ 1996], Barth, Elliott, and Finn [1999], Skinner and Sloan [2002]).20 I examined the three thresholds documente d by Degeorge et al. [1999]: the previous years earnings level, the zero earnings leve l, and the consensus analyst forecast. For convenience, I refer to firms that have an earning change near zero as the growth sample, firms that have earnings near the zero as the profit sa mple, and firms that have an earnings surprise near zero (i.e. earnings less the consensus analys t forecast near zero) as the analyst sample. While I formed the profit and growth samples us ing a combination of hand-collected data from proxy statements and data from commercial database s, I formed the analyst sample solely with data from commercial databases. I used diffe rent methods because using only commercial databases would produce too small a sample to conduct my analysis for the growth and profit sample. 19 If auditors were motivated to prevent earnings management near key thresholds, then one would expect high quality auditors to be more likely to detect evidence of earnings management. 20 While there is some controversy regarding the existence of the kink in earnings, there is no dispute that there are incentives for managers to manipulate earnings to meet thresholds. The absence of a kink may raise as many questions as the existence of a kink.

PAGE 61

61 Variables of Interest In the empirical tests, I classified the variab les under consideration into three categories: the dependent variables, the independent variable s (audit committee incentives), and the control variables (audit committee ability, CEO in centives, and firm control variables). The Dependent Variables For the growth/profit/analyst sample, the depe ndent variable equaled one if the earnings change/earnings/earnings surprise is greater than 0, and zero otherwise. This implicitly assumes that firms manage earnings if they just met a th reshold and did not manage earnings if they just miss an earnings threshold. Further, this suggest s that the audit committee is effective if a firm just misses a threshold and ineffective if the firm just meets a threshold. In robustness tests, I considered different definitions of near a threshold. For example, I initially considered near the growth threshold to be an earning change per share between -$0.05 and $0.05, but in the robustness test I considered near the growth threshold to be an earnings change per share between -$0.025 and $0.025. The Independent Variable: Audit Committee Incentives Two incentives that influence a committ ees decisions are equity holdings and directorships. In Chapters 4 and 5, I discussed the expected influen ces of these factors at length. I calculated equity hold ings incentives as 100 times the average of the members equity holdings divided by total shares ou tstanding at fiscal year-end. This calculation is similar to Yang and Krishnan [2005]s measure of audit committee equity holdings.21 To examine the effects of the nature of equ ity ownership, I parti tioned equity holdings into direct holdings and indir ect holdings. Direct holdings ar e the shares that the members 21 The results do not change if I define audit committee equity holdings as the natural log of the average approximate market value of the audit committee equity holdings.

PAGE 62

62 personally own in the company. Indirect holdin gs are all other holdings that the member is deemed to own by the SEC. For the growth and profit samples, I differe ntiated between direct and indirect holdings from the de tails included in the proxy statemen ts. For the analyst sample, I classified equity holdings as indirect if the member is a designated director from another corporation. Like Bedard et al. [2004], I calculated director ships as the members average current directorships held in other corporat ions as disclosed in the proxy statement. The Control Variables Audit committee ability Audit committee effectiveness depends upon ma ny audit committee attributes and the larger governance environment. I expected audit committee ability to be negatively related to the likelihood that a firm meets an earnings th reshold. For the growth and profit samples, I focused upon the audit committee ability attribut es. For the analyst sample, I focused upon a subset of these audit committee attributes and the larger governance environment. Profit and growth samples Audit Committee Ability Proxy I considered six determinants of an audit committees ability: independence, accounting expertise, tenure, committee size, charter, and meetings.22 I expected each of these factors to be positively associated with effectiveness (i.e. are negatively associated with the likelihood that a firm just meets a key earnings threshold) I defined the variable as follows:23 Independence equaled 1 if the audit committee consists of only independent members, and 0 otherwise. 22 While equity holdings and directorships may also influence a committee diligence, I consider them separately, as their influence is complicated. 23 Reviews of the particular variables are in Chapter 2 and 3.

PAGE 63

63 Accounting Expertise equaled 1 if the audit committee ha s an independent accounting expert24, and 0 otherwise. Tenure equaled the average number of years of se rvice on the board of di rectors for the audit committee members. Committee size equaled the number of memb ers serving at fiscal year end. Charter equaled 1 if the audit committee op erates under a charter, and 0 otherwise. Meetings equaled the number of meetings th e audit committee held throughout the year. Analyst Sample Audit Committee Ability Proxy In the analyst sample, I considered a vari ation of the Dhaliwal Naiker, and Navissi [2006] 2-factor governance proxy. Specifically, I ex amined board and audit committee abilities. I expected strong governance to be negativel y associated with earnings management. Like Dhaliwal et al [2006], I applied a fou r-step process to calculate the governance scores. First, I identified seve ral attributes for eac h factor. Second, for each attribute, I determined if a firm met the necessary conditions to conclude that a firm has strong governance with respect to that attribute. Third, I summed the individual attri butes scores into a total factor score. Lastly, I determined whether the firm has strong or weak governance based upon the total factor score for each factor. The specific board attributes were: Board size: There is a nega tive association between board size and governance quality (Yermack [1996], Eisenberg, et al. [1998], L oderer and Peyer [2002]). Jensen [1993] suggested that large boards are li kely to be less effective monito rs and are easier for CEOs to control. I coded firms 1 if the board size is less than median, and 0 otherwise. Board independence: Although prio r research is mixed, the prev ailing view is that the higher the proportion of independent directors the mo re effective the board is at monitoring management (e.g.,Weisbach [1988], Rosenstein and Wyatt [1990], Rosenstein and Wyatt [1990], Brickley, Coles, and Terry [1994], Dech ow, et al. [1996], Beasle y [1996], Core et al. 24 An accounting expert has a profession al accounting designation or degree, or experience working at a senior level financial position.

PAGE 64

64 [1999]). Arguably, independent directors have more incentives to carry out their monitoring tasks and objectively question and evaluate management performance (Fama and Jensen [1983], Carcello and Neal [2003]). Following prio r studies, I coded firms 1, at least, 50% of the directors are independent, and 0 otherwis e (Weisbach [1988], Denis, Denis, and Sarin [1997]). Board experience: A members ability to monitor increases with the members experience on boards (Fama [1980], Kaplan and Reishus [1990], Ferris et al. [2003]). There is a negative association between directorsh ips and financial reporting pr oblems (Bedard et al. [2004], Yang and Krishnan [2005]). I coded firms 1 if the average number of directorships per board member is greater than 1, and 0 otherwise. Board turnover: Longer tenured directors have greater experience and knowledge of the firm so they can more effectively constrain manage ment than shorter tenured directors (Dhaliwal et al. [2006]). However, since there is like ly declining knowledge gains for each year of service, director effectiveness wi ll likely not increase significantl y after a few years. There is also a risk that long serving members devel op an inappropriate closeness to management (Vafeas [2003], Beasley [1996]). I coded firms 1 if the board ha s less than 20 pe rcent of its members with less than 3 years experience, and 0 otherwise. CEO-chair separation: CEO-Chair duality is associated with higher instances of SEC enforcement actions, higher control risk, lo wer sensitivity of CEO turnover to firm performance, and less voluntary disclosure (Dechow, Sloan and Sweeney [1996], Tsui, Jaggi and Gul [2001], Goyal and Park [2002], Gul a nd Leung [2004], Farber [2005]). I coded firms 1 if the CEO and board ch airperson positions are not held by the same individual, and 0 otherwise. Independent nominating committ ee: While the current independe nce standards prevent some inappropriate relationships between board me mbers and management, there are still many potentially problematic rela tionships. To ensure inde pendent members are unbiased, corporations can use an inde pendent nominating committee. Independent Nominating Committees are negatively associat ed with financial reporting pr oblems such as restatements (Nabar, Kim, and Heninger [2006], Carcello et al [2006]). I coded firms 1 if the firm has a nominating committee consisting of only i ndependent members, and 0 otherwise. If the total board score was above 3, then I coded it as 1, and 0 otherwise. I consider three dimensions of audit commit tee quality: size, independence, and tenure. The related literature on these factors is discu ssed in Chapter 1. The specific audit committee attributes are coded as following: Audit committee size: Consistent with th e recommendation of the Blue Ribbon Committee on Audit Committees [1999], I coded firms 1 if the audit committee size has at least three

PAGE 65

65 members, and 0 otherwise. Audit committee independence: Consistent with the Blue Ribbon Committee [1999], I coded firms 1 if the committee is composed of sole ly independent members and 0 if the committee includes at least one affiliated member. Audit committee tenure: I coded firms 1 if the average tenure of the audit committee is greater than the median, and 0 otherwise. I coded audit committee quality as 1 if the score was 3, and 0 otherwise. CEO incentives While CEOs have many incentives to inflate earnings, such as job security and to maximizing future bonuses, the most important in centive is likely to maximize the value of equity holdings (Klein [2002], Bartov and Moha nram [2004], Burns and Kedia [2004],Efendi et al. [2004], Harris and Bromiley [2005] ). I partitioned CEO equ ity incentives into long-term incentives and short-term incentives. 25 Consistent with Safdar [2003], I expected short-term equity incentives to provide gr eater motivation to manage earn ings than long-term equity incentives. I viewed shares as long-term incentiv es since managers have limited ability to sell the stock.26 I classified non-exercisa ble options as long-term incen tives because they may not vest for years (Cheng and Warfie ld [2005]). I calculated long-term equity incentives as 100 times the sum of shares owned and non-exercisabl e options divided by tota l shares outstanding at fiscal year-end. I calculated s hort-term equity incentives as 100 times the exercisable options divided by total shares outstand ing at fiscal year-end. Firm control variables Company size is a key determinant of a firms operating environment and the quality of internal controls (DeFond and Ji ambalvo [1991], Beasley et al. [2000] ). I calculated size as the 25 The results do not change substantially if I do not combine unexercisable stock options and stock holdings, or if I consider unexercisable stock as short-term incentives. 26 Stocks are more difficult to sell because of legal restri ctions and the negative market response to large sells.

PAGE 66

66 natural log of the market value of equity (F rankel et al. [2002], Ashbaugh et al. [2003], Vafeas [2005]). Since the managerial discretion is nega tively related to internal control quality, I expected firm size to be negativ ely associated with th e likelihood that a firm meets an earnings threshold. Leverage measures the closeness to the viol ation of debt contro ls (Yang and Krishnan [2005]). I calculated leverage as long-term debt divided by total assets. Since firms closer to debt covenant violation face great er pressure to report higher earni ngs, I expected leverage to be positively associated with the likelihood that a firm meets an earnings threshold. I included the market-to-book ratio as a proxy for growth. I calculated market-to-book ratio as the stock price at fiscal year, multiplied by total shares outstanding divided by the value of total assets. Since stocks that have higher market-to-book ratio have greater pressure to meet earnings thresholds to avoid th e torpedo effect documented in Skinner and Sloan [2002], I expected book-to-market ratio to be positively associated with the likelihood that a firm meets an earnings threshold. I included a control for litigation risk. While the empirical evidence indicates that audit committee members do not affect litigation risk (B lack et al. [2005]), auditors may be more conservative in high-risk industr ies. A dummy variable equa led 1 for high-risk industries (pharmaceuticals, computer, electronics, and retail industries), and 0 otherwise.27 I expected litigation risk to be negatively associated to the likelihood that the firm meets an earnings threshold. 27 Matsumoto [2002] uses the following set of industries as high-risk industries SIC codes: 283336, 3570577, 737074, 3600674, and 520061.

PAGE 67

67 I included a control variable for losses. DeGe orge et al. [1999] theo rized that there is a hierarchy of thresholds. Specifically, they claimed that firms strive to achieve profitability, then to exceed the prior years earnings, and, then to meet analyst forecast. Consistent with this hierarchy, I included a proxy for mi ssing a lower threshold. For the growth sample, I included a loss proxy that equals 1 if a firm failed to achi eve profitability in the current year. For the analyst sample, there were two loss proxies: failur e to the meet previous years earnings level and failure to achieve profitabil ity. I expected the likelihood that a firm meets a higher threshold declined when the firm failed to achieve a lower threshold. For the profit sample, I included a loss dummy variable that equaled 1 if the firm had negative earnings the previous year, and 0 otherwise.28 This captured the reduced incentives to achie ve profitability in the given year if the string of positive earnings is sna pped in the prior year (Healy and Wahlen [1999], Myers et al. [2005]). The Growth Sample Sample Selection I formed the growth sample in three stages First, I identified all firms on Compustat with an annual EPS change between -$0.05 a nd $0.05 from 1999 to 2003. I calculated EPS as earnings, excluding extraordinary it ems divided by total stock outstan ding at year-end including convertible securities.29 I calculated EPS change as the diffe rence in EPS from year t and year t1. I considered the annual earnings figure be cause it is the only earnings figure that audit committees are required to review I restricted my study to firms near the previous years earnings level rather than th e whole population because, like Dechow et al. [2003], I am 28 This is capturing the pressure a firm has to continually report positive earni ngs. When a string of positive earnings is snapped, there is less pressure to have positive earnings this year. 29 I report the results using diluted EPS but the results are unchanged if basic EPS is used.

PAGE 68

68 attempting to understand the systematic differences between firms that just miss and firms that just meet the threshold. I excluded financial in stitutions (Standard Indus trial Classification [SIC] codes 60006999) and utilities (SIC codes 4400) because of their unique regulatory environments (Table 3-1).30 Second, I identified which firms in the sample filed an electronic DEF 14A proxy statement with the SEC within 12 months of year-end (Table 3-1). Third, I hand-collected all audit committee and CEO info rmation from the proxy statements. To be included in the sample, the firm must have disclo sed all required information, have a single CEO throughout the year31, a minimum of one audit committee meeting32, an unqualified auditor opinion, a minimum of two members on the audit comm ittee, and a single class of shares (Table 3-1). The sample was divided relativel y evenly across the years (Table 3-1). The largest sample year was 2003 with 23 percent of the observation s and the smallest sample year was 1999 with 17 percent of the observations. For every year, th e number of firms that met the previous years earnings level is greater than the number of firms that did not. This is consistent with the kink documented by previous researchers (e .g. Degeorge et al. [1999]). Descriptive Statistics An analysis of the correlation among key variables showed no evidence of multicollinearity among key va riables (Table 3-2).33 Inconsistent with prior research, my results indicate that companies with charters are more li kely to achieve earnings th resholds (Table 3-3). 30 This is consistent with (Cheng and Warfield, [2005]). 31 I exclude multiple CEO firms because it is unclear which CEO will be responsible for the financial reporting process. I exclude firms with CEO changes because new CE Os have different incentives to manage earnings (Wells [2002], Godfrey, et al. [2003]) 32 If a committee does not meet, then it is incapable of having any influence. 33 In unreported results, I examine the variance inflation factors and found no evidence of multi-collinearity.

PAGE 69

69 The relative homogeneity for all ot her attributes between meet and miss firms is consistent with Dechow et al. [2003]. The Profit Sample Sample Selection I formed a sample of firms that have earnings near zero using the same three stages as for the growth sample. The only difference is that I identified all firms on Co mpustat with an annual diluted EPS between -$0.05 and $0.05 (Table 3-4) rather than an EPS change between -$0.05 and $0.05.34 The sample was divided re latively evenly across the year s (Table 3-4). The largest sample year was 2003 with 24 percent of the obser vations and the smallest sample year was 1999 with 17 percent of the observat ions. For every year except 20 03, the number of firms that achieved profitability was greater than the number of firms that did not. The anomalous finding in 2003 may be caused by changes in the governance culture that motivated audit committees to be more zealous at eliminating income-increas ing transactions and an economic recession. Descriptive Statistics An examination of the correlation between ke y variables of interest and descriptive statistics revealed no evidence of multi-collinearity among key vari ables (Table 3-5). Counter to expectations, the descriptive statistics indicate th at companies with charte rs are more likely to achieve earnings thresholds (Table 3-6).35 This suggests that more sophisticated audit committees are more willing to allow earnings ma nagement around thresholds. Alternatively, high performing companies may adopt high quality a udit committees to signal earnings quality. Overall, the relative homogeneity be tween meet and miss firms is c onsistent with Dechow et al. [2003]. 34 This is consistent with (Cheng and Warfield [2005]). 35 In unreported results, I examine the variance inflation factors and found no evidence of multi-collinearity.

PAGE 70

70 The Analyst Sample Sample Selection I formed a sample of firms that had an earnings surprise near zero in three stages. First, using Institutional Brokers Estimate System (I/ B/E/S), I identified all firms that between 1998 and 2004 with an earnings surprise between $0.01 and $0.01. An earnings surprise is the difference between reported earnings and the co nsensus analyst forecast. Of these firms, I identified which ones are included in Compustat, Investor Research a nd Responsibility Center (IRRC) Database, and Execucomp. To be included in the sample, the firm must disclose all required information, have a single CEO throughout the year36, an unqualified auditor opinion, a minimum of two members on the audit comm ittee, and a single class of shares. The sample was divided relatively evenly across the years (Table 3-7). The largest sample year was 2002 with 18 percent of the obser vations and the smallest sample year was 1998 with 9 percent of the observations. For every year, the number of firms that met analyst forecast was greater than the number of firms that did not This is consistent with the kink documented by previous researchers (e.g. Degeorge et al. [1999]). Descriptive Statistics An examination of the correlation between ke y variables of interest showed no evidence of any multi-colinearity among key variables (Table 3-8).37 The descriptive statistics showed that the sample attributes did not vary between fi rms that met analyst expectations and firms that 36 I exclude multiple CEO firms because it is unclear which CEO will be responsible for the financial reporting process. I exclude firms with CEO changes because new CE Os have different incentives to manage earnings (Wells [2002] Godfrey, et al. [2003]). 37 In unreported results, I examine the variance inflation factors and found no evidence of multi-collinearity.

PAGE 71

71 failed to meet analyst expectations (Table 3-9).38 The homogeneity between meet and miss firms is consistent with Dechow et al. [2003]. Differences in the Samples There are some significant differences betw een the samples. The growth and profit samples have more detailed audit committee variables than the analyst sample. This difference exists because the IRRC did not collect many of th e audit committee attribut es, such as expertise and literacy. The analyst sample did consider a larger set of govern ance variables such as nominating committee and board si ze, because I hand collected fe w board level attributes. I collected few board level factors because of the lack of significant results on board level attributes and the exclusion of board level variab les in the recent audit committee research (e.g. Bedard et al. [2004], Song and Windram [2004], Yang and Krishna n [2004], Vafeas [2005]), and the additional time requirements to collect board level data. I elected to present the different approaches to demonstrate that my results are ro bust to significantly di fferent proxies for audit committee ability. The samples also differed in the period under consideration. Specifi cally, the growth and profit sample covered the years 1999 to 2003, wh ile the analyst sample considered the years 1998 to 2004. The growth and profit sample was re stricted to a shorter time because of the time requirements associated with ha nd-collected data. My results did not vary significantly if I restricted the analyst samp le to the years 1999 to 2003. The samples also differed in the firms under c onsideration. The growth and profit sample needed to be listed on Compustat North Ameri ca and have an electroni c proxy filed with the 38 An examination of the particular gove rnance attributes also revealed no di fferences between miss and meet firms.

PAGE 72

72 SEC. The analyst sample firms had to be in Compustat North America39, IRRC database40, I/B/E/S, and ExecuComp.41 The most significant variati on between sample firms was the difference in firm size. Specifically, the aver age market value of the equity was $3.25 million for the profit sample, $80.13 million for the grow th sample, and $1,318.82 million for the analyst sample. While sample selection may attribute to some of these differences, the variation may also be driven by difference in the type of fi rms that are near each threshold. A related difference was the average equity held by audit committee members (direct equity holdings: 0.23 percent in analyst sample, 1.19 percent in the gr owth sample, 1.34 percent in the profit sample) and CEOs (short-term equity holding: 0.88 percen t in analyst sample, 1.82 percent in growth sample, 2.12 percent in the profit sample). This is consistent with larger companies having more diversified ownership. 39 Compustat North America covers St andard and Poors North America. 40 IRRC covers Russell 3000, Standard and Poors 500/Midcap/Small Cap. 41 Executive Data is for Standard and Poors 1500.

PAGE 73

73 Table 3-1. Sample selection for growth sample Year EPS Change 1999 2000 2001 2002 2003 Total -$0.05 to $0.00 373 363 421 393 459 2009 $0.00 to $0.05 433 452 457 508 605 2455 Firms in Compustat -$0.05 to $0.05 806 815 878 901 1064 4464 -$0.05 to $0.00 233 208 206 211 257 1115 $0.00 to $0.05 252 277 241 256 278 1304 Firms in Compustat with proxy statement on EDGAR -$0.05 to $0.05 485 485 447 467 535 2419 -$0.05 to $0.00 135 119 116 124 144 638 $0.00 to $0.05 144 151 120 144 170 729 Firms with all data requirements satisfied -$0.05 to $0.05 279 270 236 268 314 1367 EPS is the earnings per share calculated on a diluted basis.

PAGE 74

74 Table 3-2. Pearson correlation of key inde pendent variables for the growth sample (1) (2) (3) (11) (1)Audit committee direct holdings 0.17 ** (2)Audit committee indirect holdings -0.00 ** (3)Audit committee total holdings 0.10 ** (4)Committee size -0.12 ** -0.04 -0.10 ** -0.08 ** (5)Meetings -0.22 ** -0.03 -0.16 ** -0.07 (6)Charter -0.11 ** 0.01 -0.06 0.01 (7)Ind. acc. expert -0.08 ** -0.03 -0.07 -0.01 (8)Average directorships -0.14 ** 0.05 -0.06 -0.10 ** (9)Tenure 0.08 ** -0.09 -0.01 0.03 (10)Wholly independent -0.16 ** -0.09 ** -0.17 ** -0.03 (11)CEO short-term incentives 0.17 ** -0.00 0.10 ** -0.00 (12)CEO long-term incentives 0.17 ** 0.05 0.15 ** 0.03 (13)Size -0.31 ** -0.12 ** -0.28 ** -0.21 ** (14)Market to book -0.03 0.06 0.02 0.04 (15)Leverage -0.03 -0.01 -0.02 -0.00 (16)Litigation risk -0.03 0.03 0.01 -0.02 (17)Loss current year 0.09 ** 0.03 0.08 ** 0.04 *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 All variables are defined in the appendix.

PAGE 75

75 Table 3-3. Descriptive stat istics of growth sample Total sample Missed threshold Met threshold Mean Std. dev. Mean Std. dev. Mean Std. dev. Difference between miss and meet firms Audit Committee Incentives AC direct holdings 1.19 2.35 1.152.331.212.36 AC indirect holdings 0.62 2.70 0.773.210.492.15 ** AC total holdings 1.81 3.70 1.934.131.703.26 Audit Committee Controls Committee size 3.08 0.79 3.070.783.080.80 Meetings 3.67 2.32 3.662.343.682.31 Charter 0.73 0.44 0.700.460.750.43 ** Ind. acc. expert 0.41 0.49 0.400.490.420.49 Average directorships 0.97 0.93 0.960.920.980.94 Tenure 4.80 2.56 4.942.634.682.50 Wholly independent 0.87 0.33 0.860.350.890.31 CEO Attributes CEO STEI 1.82 3.031.753.001.883.06 CEO LTEI 11.58 16.8211.8016.5311.3917.08 Control Variables Size 18.20 2.21 18.222.1518.182.27 Market to book 3.66 17.15 4.4118.343.0116.02 Leverage 0.14 0.21 0.130.200.150.23 Litigation risk 0.40 0.49 0.410.490.390.49 Loss current year 0.40 0.49 0.430.500.380.49 Difference between miss and meet firms: t-tests for continuous variable and Mann -Whitney U for dichotomous variables *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 (2 Tailed) All variables are defined in the appendix.

PAGE 76

76 Table 3-4. Sample selection for profit sample Year EPS 1999 2000 2001 2002 2003 Total -$0.05 to $0.00 377 428 496 531 589 2722 $0.00 to $0.05 282 251 267 282 254 1650 Firms in Compustat -$0.05 to $0.05 659 679 763 813 843 4372 -$0.05 to $0.00 137 151 168 179 222 857 $0.00 to $0.05 170 146 155 171 136 778 Firms in Compustat with proxy statement on EDGAR -$0.05 to $0.05 307 297 323 350 358 1635 -$0.05 to $0.00 57 76 72 77 115 397 $0.00 to $0.05 97 90 109 106 104 506 Firms with all data requirements satisfied -$0.05 to $0.05 154 166 181 183 219 903

PAGE 77

77 Table 3-5. Pearson correlation of key inde pendent variables for the profit sample (1) (2) (3) (11) (1) Audit committee direct holdings 0.08 (2) Audit committee indirect holdings -0.00 ** (3) Audit committee total holdings 0.03 (4) Committee size -0.10 ** -0.01 -0.05 -0.12 (5) Meetings -0.13 ** 0.01 -0.05 -0.10 (6) Charter -0.12 ** -0.01 -0.09 ** -0.03 (7) Ind. acc. expert -0.03 -0.01 -0.01 0.03 (8) Average directorships -0.09 ** 0.05 0.01 -0.07 (9) Tenure 0.09 ** -0.08 -0.03 0.01 (10) Wholly independent -0.12 ** 0.02 -0.03 0.03 (11) CEO short-term incentives 0.08 -0.00 0.03 0.02 ** (12) CEO long-term incentives 0.05 -0.00 0.02 -0.02 (13) Size -0.26 ** -0.08 -0.18 ** -0.18 (14) Market to book -0.07 0.05 0.02 0.02 (15) Leverage 0.03 -0.00 0.01 0.04 (16) Litigation risk -0.01 0.09 ** 0.06 -0.02 (17) Loss last year 0.01 0.03 0.04 0.08 *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 All variables are defined in the appendix.

PAGE 78

78 Table 3-6. Descriptive statistics of profit sample Total sample Missed threshold Met threshold Mean Std. dev. Mean Std. dev. Mean Std. dev. Difference between miss and meet firms Audit Committee Incentives AC direct holdings 1.342.61 1.443.291.251.90 AC indirect holdings 0.984.504 0.994.700.974.35 AC total holdings 2.315.20 2.435.672.224.80 Audit Committee Controls Committee size 2.950.68 2.960.702.940.66 Meetings 3.632.25 3.632.273.632.23 Charter 0.750.43 0.770.420.740.44 Ind. acc. expert 0.400.49 0.420.490.380.49 Average directorships 0.820.94 0.790.880.840.98 Tenure 4.512.54 4.322.554.662.53 ** Wholly independent 0.860.35 0.850.350.870.34 CEO Attributes CEO STEI 2.123.502.023.192.193.73 CEO LTEI 11.9416.8310.7915.8212.8517.55* Control Variables Size 15.052.01 15.122.0814.991.96 Market to book 3.0715.89 2.4319.953.5611.77 Leverage 0.150.31 0.180.420.130.18 ** Litigation risk 0.430.50 0.440.500.420.49 Loss current year 0.540.50 0.610.490.490.50 *** Difference between miss and meet firms: t-tests for continous variable and Mann-Whitney U for dichotomous variables *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 (2 Tailed) All variables are defined in the appendix.

PAGE 79

79 Table 3-7. Sample selection for the analyst sample EPS Surprise 1998 1999 2000 2001 2002 2003 2004 Total -$0.05 to $0.00 39 55 70 70 94 91 79 498 $0.00 to $0.05 169 269 261 300 325 300 245 1869 Firms with all data requirements satisfied -$0.05 to $0.05 208 324 331 370 419 391 324 2367

PAGE 80

80 Table 3-8. Pearson correlation of key inde pendent variables for the analyst sample (1) (2) (3) (4) (1) Audit committee direct holdings (2) Audit committee indirect holdings (3) Audit committee designate (4) Audit committee total holdings (5) AC score -0.07 0.00 ** -0.04 0.03 -0.06 0.00 ** 0.41 0.00 ** (6) Board score -0.09 0.00 ** -0.01 0.61 -0.01 0.62 0.37 0.00 ** (7) CEO short-term incentives 0.04 0.03 0.01 0.73 0.02 0.39 -0.14 0.00 ** (8) CEO long-term incentives 0.07 0.00 ** -0.01 0.52 -0.01 0.50 -0.15 0.00 ** (9) Size -0.18 0.00 ** -0.03 0.21 0.00 0.85 0.38 0.00 ** (10) Market to book -0.06 0.00 ** -0.00 0.83 0.05 0.02 0.09 0.00 ** (11) Leverage -0.03 0.18 -0.00 0.85 0.02 0.24 0.01 0.75 (12) Litigation risk -0.02 0.27 -0.00 0.88 -0.00 0.89 -0.01 0.51 (13) Loss 0.01 0.61 0.00 0.90 -0.01 0.51 -0.02 0.38 (14) No growth 0.04 0.07 0.02 0.40 0.01 0.76 -0.04 0.03 *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 All variables are defined in the appendix.

PAGE 81

81 Table 3-9. Descriptive statistics of the analyst sample Total sample Missed thresholdMet threshold Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Difference between miss and meet firms Audit Committee Incentives AC direct holdings 0.23 0.65 0.190.580.240.66 AC indirect holdings 0.02 0.45 0.000.040.020.50 AC designate 0.01 0.12 0.010.120.010.12 AC total holdings 0.25 0.800.190.580.260.85 AC directorships 0.93 0.78 0.950.760.930.78 Governance Controls AC control 0.71 0.45 0.730.440.710.46 Board control 0.34 0.48 0.350.480.340.47 CEO Attributes CEO STEI 0.88 1.19 0.901.200.871.18 CEO LTEI 3.13 6.09 3.146.473.135.99 Control Variables Size 21.46 1.5821.501.6221.451.57 Market to book 1.02 1.161.001.061.111.45** Leverage 1.02 1.16 1.111.461.001.07 Litigation risk 0.31 0.46 0.300.460.320.47 Loss 0.13 0.34 0.140.350.130.33 No Growth 0.43 0.50 0.430.500.430.49 Difference between miss and meet firms: t-tests for continous variable and Mann-Whitney U for dichotomous variables *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 (2 Tailed) All variables are defined in the Chapter 3

PAGE 82

82 Table 3-9 Continued. Total sample (n=2367) Met threshol d (n=1868) Missed threshold (n=499) Mean Std. Dev. Fraction of firms with score =1 Mean Std. Dev. Fraction of firms with score =1 Mean Std. Dev. Fraction of firms with score =1 Board size (number of members) 9.04 2.46 0.73 9. 03 2.47 0.73 9.08 2.42 0.73 Board independence (fraction of independent members) 0.65 0.17 0.69 0.65 0. 17 0.69 0.66 0.16 0.71 Board experience (average number of directorships) 0.86 0.61 0.39 0.85 0. 61 0.38 0.89 0.62 0.42 Board turnover (% members with less than 3 years experience 0.22 0.17 0.53 0.22 0.17 0.53 0.23 0.17 0.52 CEO-chair separation 0.20 0.40 0.20 0.20 0.40 0.20 0.20 0.40 0.20 Independent nominating committee 0.44 0.50 0.44 0. 43 0.50 0.43 0.46 0.50 0.46 Audit committee size (number of members) 3.59 1.04 0.92 3.58 1. 04 0.92 3.61 1.05 0.94 Audit committee independence (Fraction of members) 0.90 0. 18 0.72 0.90 0.18 0.71 0.91 0.17 0.73 Audit committee tenure (average years of service) 8.86 4.35 0.48 8.94 4.36 0.49 8.53 4.27 0.47 Board score 2.97 1.12 0.31 2.95 1.12 0.31 3.05 1.10 0.34 Audit committee score 2.12 0.68 0.29 2.12 0.68 0.29 2.14 0.68 0.30

PAGE 83

83 Table 3-9 Continued. Distribution of governance variable scores Score Board score (6 point scale) Audit committee score (3 point score) Board binary score Audit committee binary score 0 15 18 16251674 1 192 365 742693 2 601 1291 3 817 693 4 538 5 186 6 18

PAGE 84

84 CHAPTER 4 EQUITY HOLDINGS AND THE GROWTH THRESHOLD Hypothesis Development How do audit committee equity holdings infl uence an audit committees willingness to allow earnings management? Much of the li terature on audit committee effectiveness has focused upon the audit committees ability to be e ffective. Yet, audit committees are composed of self-interested agents who make decisions to maximize th eir personal utility, not their effectiveness. The literature on how audit comm ittee equity holdings influence the committees willingness to allow aggressive reporting is limited and frequently inconsistent in its expectations and findings. Some researchers argued that equity holdings increase the committees incentives to be effective monitors (Shivdasani [ 1993], Short et al. [1999]), while others argued it increases the committees incentives to focus on short-term st ock appreciation (Forke r [1992], Carcello and Neal [2003]). Klein [2002] found that having a blockholder42 on the audit committee reduces the level of abnormal accruals. Vafeas [2005] found that audit committee equity holdings are negatively related to the likelihood that a firm meets analysts forecasts, as a proxy for earnings management. Bedard et al. [2004] found that the greater the audit committee ratios of options to stock holdings, the more likely the audit comm ittee members are willing to tolerate extreme earnings management. Yang and Krishnan [2005] found that equity holdings of independent and non-independent members are positively related to the level of abnormal a ccruals. Collectively, these results demonstrate that equity holdings influence the committ ees judgment, but the direction is not consistent. To date, there is no theoretical explan ation for these inconsistencies. 42 Klein [2002] defines a blockholder as a member owning greater than 5 percent of the company.

PAGE 85

85 Researchers have been unable to reconcile thes e inconsistencies, in part, because there is minimal theoretical work on how audit committees make decisions (DeZoort et al. [2002]). By not understanding how audit committees make deci sions, it is problematic to predict how particular factors influence the decision-making process. To rec oncile the previous research, I used the model presented in Chapter 2 to predic t how audit committee equi ty holdings influenced an audit committees decision to allow aggressive reporting. Cost-Benefit Model If the audit committee members have an econom ic interest in the decision, then their decision to allow or not allow aggressive reporting depends u pon how the decision affects the members personally. According to the model, the more the committee stands to gain from allowing aggressive reporting, the more willin g the committee is to tolerate evidence of aggressive reporting, and vice versa. Ther efore, I examined how a committees decision influences a members wealth. How audit committee equity holdings infl uence the committees decision may depend upon how equity holdings are defined. It is po ssible that certain type s of equity holdings motivate one behavior, while other types motivate a different behavior. For example, Guay [1999] found that the influence of stock options on CEO behavior was significantly different than the influence of stock holdings.43 I considered two definitions of audit committee equity hol dings. The first definition defined equity holdings as the members tota l deemed ownership. The second definition partitioned equity holdings into direct holdings a nd indirect holdings. Direct holdings represent the shares and options that the members personally own. With direct holdings, the member has 43 I do not consider the influence of options vs. stock holdings on audit committee behavior because the proxies provide limited details on the options of board members.

PAGE 86

86 full voting, dividend, and disposal rights. Indirect holdings repres ents holdings that the shares the member is deemed to own by virtue of his relationship with anothe r corporation. With indirect holdings, the member has conditional vot ing rights, no rights to dividends, and no right of disposal. Case 1: Total Equity Holdings How equity holdings influence the co mmittees decisions depends upon how the committees decisions influence the stock price. There are two potential outcomes if a committee allows aggressive reporting: there ar e no irregularities and th e stock value increases, or there is an irregularity and the stock value d ecreases. An irregularity is any event that results from the market detectin g evidence of opportunistic reporting including restated earnings, fraud, SEC sanction, and subsequent qualified audito r opinion. By assumption, any irregularity nullifies the gains from allowing aggressive repor ting. If the manager opportunistically reports and the audit committee allows it, then the likelihood of an irregularity increases. The influence of incentives depends upon the expected benefits if there are no irregularities, the expected costs if there is an irregularity, and the likelihood of an irregularity. The committees expected benefits of allowing aggressive reporting is: Expected net benefits of agressive reporting = (1-p)(p) R EGainRELoseEHEH where: GainEH represents the committees expectations regarding the relative increase in the value of the equity holdings if higher short-term earnings are re ported rather than lower shortterm earnings, assuming there are no irregularities. LoseEH represents the committees expectations re garding the decrease in the value of the equity holdings if there is an irregularity.

PAGE 87

87 pRE represents the committees assessment of the probability of an irregularity resulting from adopting aggressive reporting ra ther than conservative reporting. GainEH is positive because, ceteris paribus, the market prefers higher reported earnings to lower reporting earnings. LoseEH is negative because the stock price decreases when there is an irregularity (Wu [2002], Palm rose, et al. [2004], Karpoff, et al. [2004]). For example, Palmrose et al. [2004] documented a ten percent de cline in stock price following a restatement. pRE is strictly positive becaus e the committee can never know with certainty whether or not the manager opportunist ically reported. pREdepends upon the magnitude of the aggressive reporting, since the larger the manipulation the more likely it is that aud itors or other outsiders will eventually detect evidence of the manipulation.44 Also, pREdepends upon the CEOs incentives to report opportunistical ly. The likelihood of an irregul arity is positively associated with the CEOs incentives to opportunistically report (Efendi et al. [2004], Burns and Kedia [2004], Harris and Bromiley [2005]). Specifically, there is an associati on between CEO equity incentives and opportunistic repo rting (Klein [2002], Bartov and Mohanram [2004], Cheng and Warfield [2005]). As the CEO short-term equity incentives increase the perceived risk of an irregularity increases and the net benefits of allowing aggressive reporting decreases. Case 2: Partitioned Equity Holdings Direct holdings represent the equity that the members personally own. If the audit committee makes a decision that increases (decreases) the value of the holdings, then the member receives all the benefits (costs) of that decision. With direct holdings, the member has 44 This is consistent with Law of Conservation of Income, wh ich stipulates that total income over a lifetime is fixed so earnings management simply shifts when income is recognized and does not generate additional income.

PAGE 88

88 the ability to sell his holdings to realize any gains from his decision. The net expected benefit of direct holdings can be stated as: Expected net benefits of agressive reporting = (1-p)(p) R EGainRELoseDHDH where: GainDH represents the committees expectations a bout the relative increa se in the value of the direct holdings if higher short-term earni ngs are reported rather than lower short-term earnings, if there are no irregularities. LoseDH represents the committees expectations about the decrease in the value of the direct holdings if there is an irregularity. pRE represents the committees assessment of th e likelihood of an irre gularity resulting from adopting aggressive reporting ra ther than conservative reporting. The properties of GainDH and LoseDH should be similar to properties of GainEH and LoseEH from Case 1 since both di rect holdings and total holdings assume the members receive all the benefits of ownership. Indirect holdings represents holdings that the member is de emed to own by virtue of his position with another corporation or ownership in another corporat ion. Members with indirect holdings are agents of the equity owner with a fi duciary responsibility to represent the owner. While prior research has not e xplicitly considered the influe nce of indirect holdings, Klein [2002] may have inadvertently cap tured the effect. She examin ed the influence of equity holdings by considering the infl uence of blockholders on audit committee effectiveness. She noted that large non-management shareholders are given seats on audit committees to monitor their investments. She found that having block holders on the audit committee reduces earnings management. However, she did not consider whet her the amount of indire ct holdings influences the audit committees decisions, or if blockholders would allo w small levels of earnings

PAGE 89

89 management when there are strong incentives to do so. The effect of i ndirect holdings depends upon the members expectations about how the equi ty owner will reward (discipline) them for their effectiveness (ineffectiveness). As in Case 1, the members expected payoff is: Expected net benefits of agressive reporting = (1-p)(p) R EGainRELoseIHIH where: GainIH represents the committees expectations a bout the increase in the value of their personal wealth gained if there are no irregu larities and the firm aggressively reports. LoseIH represents the committees expectations a bout the decrease in the value of their personal wealth gained if there are no irregu larities and the firm aggressively reports. pRE represents the committees assessment of th e likelihood of an irre gularity resulting from adopting aggressive reporting ra ther than conservative reporting. The absence of an irregularity likely pr ovides minimal benefits to the members with indirect holdings. The equ ity owner has employed the audit committee member to provide monitoring service and their compensation should be linked to that function. It is unlikely that the equity owner would award additional compen sation to the audit committee member for the firms operational success because the members ro le was to oversee the re porting process, not to participate in the operations. Likewise, it is unl ikely that an equity ow ner would discipline the audit committee member for requiring conservative reporting because their responsibility was to oversee the process, not inflate earnings to meet earnings thresholds. Therefore, GainIH is positive but near zero. If there is an irregularity and the firms value declines, then the members may be disciplined by the equity owners for the failure. For example, the member may lose his or her board appointments following a restatement (S rinivasan [2005]). Depending upon the magnitude of the failure, the members may also damage their relationship with equity owners. If the

PAGE 90

90 member is an employee of the equity owner, th en a serious failure could lead to his or her dismissal. Therefore, LoseIH is negative.45 The magnitude of the net benefits varies with the volume of in direct holdings because the more the equity owner has invested, the greater is their losses if there is an irregularity. Further, the more the owner loses, then likely the gr eater the owners motiva tion to discipline an ineffective monitor. pRE does not vary between the cases since th e likelihood of an irregularly does not depend upon the audit committees equity holdings. When Earnings are Near the Previous Years Earnings Level Using this cost-benefit model, I examined th e association between audit committee equity holdings and the likelihood that current earnings meet or exceed the previous years earnings level. If the audit committee allowed aggressive reporting, then, ceteris paribus, a company should be more likely to meet the previous year s earnings level than if the committee rejected aggressive reporting. According to Case 1, the influence of equity holdings depends upon the values of GainEH ,LoseEH andpRE. Since the value of these factors is the members expectations, it is impossible to identify specific values. However, it is possible to estimate the relative relationship between these fact ors using prior research. 45 Alternatively, a member owning equity in a company with an ownership interest in the firm may cause indirect holdings. In this case, committee members represent not only their personal interests, but also the interests of the parent company. While personal holdin gs may create incentives to allow aggre ssive reporting, the effect is likely muted because the members have limited ab ility to benefit from short-term pri ce variations and it is unclear how the parent companys stock price will vary if a s ubsidiary aggressively reports. Therefore, GainIH is positive but likely low. If there is an irregularity, then the member may su ffer reputation losses from proving to be an ineffective monitor. Further, the members may incur relationship difficulties with the fellow owners in the parent company for failing to protect their mutual investment. Therefore, L oseIH is negative.

PAGE 91

91 GainEH should be significant and positive since ther e is a pricing premium for firms that consistently exceed the previ ous years earning level. LoseEH is significant and negative. While there is no evidence to s uggest that allowing aggressive reporting to meet the previous years earnings level would lead to Enron-type reporting problems, there is evidence to suggest that allowing aggressive reporti ng could lead to earnings being subsequently restated and stock value declining (Richard son et al. [2003]). GainEH is likely smaller than LoseEH because there is less potential upside associated with meeting an earnings target than downside associated with a reporting failure. pREis likely low because if there were a high risk of a reporting e rror, then auditors would be vigilant in preven ting aggressive reporting. pREdepends upon the CEOs short-term equity incentives. When the CEO has low incentives, pREis near zero so the net benefits of allowing aggressive reporting is positive. As CEO incentives increase, the expected benefits decline so the committee should be increasingly biased against al lowing aggressive reporting. Given the likelihood that a firm meets the previo us years level depends upon the committees willingness to allow aggressive reporting, I hypothesized: H1: Audit committee equity holdings a re positively associated with the likelihood that the firm meets the previous years earnings level, but CEO short-term equity incentives moderate this effect. For Case 2, I examined separately the effect of direct and indirect holdings. The preceding discussion on Case 1 assumed that the members accr ued all the benefits of ownership, which is the defining quality of direct holdings. Hen ce, the influence of direct holdings on the committees willingness to allow aggressive reporti ng should be similar to the influence of total equity holdings.

PAGE 92

92 For indirect holdings, it is unclear that any unique benefits would accrue to audit committee members for allowing aggressive repo rting when earnings ar e near the previous years level. Hence, GainIH is positive but near zero. LoseIH is significant and negative because the member will likely be disciplined fo r failing to fulfill the fi duciary responsibility. The influence of indirect holdi ngs is also sensitive to the additional risk presented when the CEO has high incentives. When CEO incentives are low, the expected net benefits is near zero, so the committee has no particular incentives to allow or not allow aggressive reporting. As the CEO incentives increase, the net benefits decline so the committee should be increasingly biased against allowing aggressive reporting. Gi ven the likelihood that a fi rm meets the previous years level depends upon the committees wi llingness to allow aggr essive reporting, I hypothesized: H2A: Audit committees direct holdings are positively associated with the likelihood that the firm meets the previous years earnings level, but CEOs short-term equity incenti ves moderate this effect. H2B: Audit committees indirect holdin gs are negatively associated with the likelihood that the firm meets the previous years earnings level, with the magnitude of the effect increasi ng with the CEOs short-term equity incentives. To summarize, for the extreme scenarios I expected: Effect of increasing holdings on the likelihood that a firm meets the previous year's earnings Total Holdings Direct Holdings Indirect Holdings CEO High Incentives Decreases Decreases Decreases CEO Low Incentives Increases Increases No Effect

PAGE 93

93 Method Using the growth sample, I examined the in fluence of audit commi ttee equity holdings on the likelihood that a firm meets the previous years earnings level with following logistic regression: year tyear t-1(earningsearnings)=f(AC equity holdings,controls)->0pr where the variables are defined in the appendix. Preliminary Analysis As reported in Tables 4-6, the pseudo-r2s are relatively low (approximately 0.04), yet comparable to the pseudo-r2 of similar studies (Cheng and Warf ield [2005]). As a preliminary examination, I replicated one of the tests of Vafeas [2005]. Vafeas [2005] examined the relationship between audit committe e characteristics and the likeli hood that a firm just meets the previous years earnings. Like Vafeas [ 2005], I found no evidence of a relationship between audit committee equity holdings and the probabi lity that a firm meets the previous years earnings level (Wald 0.36, p-value> 0.1) (Table 4-1 Column A). Test of H1 I found that the coefficient on audit committ ee equity holdings was positive and weakly significant (Wald 2.24, p-value <0.1) and the coefficient on th e interaction between audit committee equity holdings and CEO short-term equity incentives was negative and significant (Wald 8.33, p-value <0.01) (Table 4-1 Column B) My results indicate that when CEO equity incentives were less than 1.37 pe rcent of total shares outstanding, the audit committee equity

PAGE 94

94 holdings were positively associated with the firm meeting the prev ious years earnings level and negatively associated otherwise46. Test of H2 I found that the coefficient on direct holdings was positive and significant (Wald 5.68, pvalue <0.01) and the coefficient on the interactio n between direct holdings and CEO-short term equity incentives was negative and significant (Wald 2.92, p-value <0 .05) (Table 4-1 Column C). My results indicate that when CEO short-term e quity incentives were under 5.02 percent of total stock outstanding, then the audit committees direct holdings were positively associated with the firm meeting the previous years earnings level, and negatively associated otherwise. The coefficient on indirect holdings was negative but not significant (Wald 0. 89, p-value >0.1), and the coefficient on the interaction between direct holdings and CEO-short term equity incentives was negative and significant (Wald 6.97, p-value <0.01). This i ndicates that indirect holdings have no influence on the committees decisions un less there was evidence that the manager has incentives to opportuni stically report. Secondary Analysis I examined the robustness of my results by partitioning the sample according to CEO short-term equity incentives. My results were the same for both specifi cations (Table 4-2). Consistent with expectations, when CEO short-term equity incentives were below the median, the coefficient on total equity holdings was posit ive (Wald 3.98, p-value <0.05), and when CEO short-term equity incentives were above the median, the coefficient on total holdings was negative (Wald 4.32, p-value <0.05). When CEO short-term equity incentives were below the median, the coefficient on the direct holdings wa s positive (Wald 5.45, p-value <0.01), and when 46 I calculate the 1.37 percent by examine the B s of the main effect and interaction effect.

PAGE 95

95 CEO incentives were above the median, the coeffi cient on indirect holding s was negative (Wald 7.24, p-value <0.01). Consistent with expect ations, the coefficient on indirect holdings was negative but not significant (Wald 0.28, p-value >0 .1). Unexpectedly, the coefficient on direct holdings when CEO short-term equity incentives were above the median was positive, but not significant (Wald 0.30, p-value >0.0 1). This result is consistent with CEO short-term equity incentives moderating the infl uence of direct holdings. To verify that non-independent members di d not drive my results I replicated the analysis using a sample of firms with audit co mmittees that were composed only of independent members. My results were mostly consistent with H1 and H2. My results indicate that total and indirect holdings vary with CEO incentives (Wald 3.29, p-value <0.05 and Wald 6.20, p-value <0.01 respectively) (Table 4-3). There is no ev idence that the influence of total or direct holdings independent of CEO incentives (Wald 0.72, p-value >0.1 and Wald 0.94, p-value >0.1 respectively). Collectively, this suggests that for committees with onl y independent members, direct holdings promote a focus on short-term st ock appreciation and indi rect holdings promote focus on improved monitoring. To examine whether or not difference in audi tor quality drove my re sults, I partitioned the sample according to auditor quality. Consistent with the prior literature I used auditors size as a proxy for auditor quality. The sample of fi rms that employed a Big 4/5 auditor yields mixed results (Table 4-4). I found that increasing di rect holdings made the audit committee more likely to allow aggressive reporting, but there was no evidence that the committee was responsive to CEO short-term equity incentives. Further, as predicted, the audit committees indirect holdings were negatively associated with the likelihood that the firm met the previous years earnings level, with the magnitude of the effect increasing with the CEOs short-term equity incentives.

PAGE 96

96 My analysis using companies with non-Big 4/5 auditors was consistent with expectations. Collectively, these results suggest that members with direct holdings ar e more willing to rely upon the auditors to prevent earnings management when the auditor is high quality than when the auditor is low quality. My results also s uggest that members with indirect holdings do not rely upon auditor quality to determine their decisions. To verify that outliers did not drive my results, I considered two alternative samples. I considered a sample of companies with earn ings per share between -$0.025 and $0.025 and a sample of companies with earnings per share be tween -$0.02 and $0.02. These results with the restricted samples were consistent with the results from the full sample (Table 4-1 and Table 45). Given the potential noise in assuming that all firms that have small positive earnings growth managed earnings, I considered two alte rnative samples. Specifically, I limited the sample of firms that I claim managed earnings to firms that have earnings per share between $0 and $0.05 and a EPS change of less than five percent of sales (assets). This restriction filtered out firms that likely could not have achieved their EPS through earnings management without having a quantitative misstatement. My results a ppeared stronger under this specification than the original specification (Table 4-6). Other Findings In all models, I documented that CEO incentives were positively associated with the likelihood that a firm meet s the previous years earnings leve l. This is consistent with CEO incentives either promoting greater effort or opport unistic reporting. I al so found evidence that confirmed the Degeorge et al. [1999] hierarchy of incentives. Specifically, the coefficient on the loss dummy was negative and significant.

PAGE 97

97 Counter to expectations, the relationshi p between audit comm ittee ability and the likelihood that a firm meet s the previous years earnings leve l were, with one exception, either null or not in the expected direction. Th e coefficient on audit committee independence was positive (Table 4-1). This is consistent w ith Vafeas [2005] and Yang and Krishnan [2005], although their results were not si gnificant at conventional levels. The coefficient on charter was unexpectedly positive and significant (Wald >5, p-value <0.01 for all models). This suggests that audit committees with greater delegated re sponsibilities were not necessarily enhancing reporting quality. Collectively, these results al so suggest that firms adopt high quality audit committees to compensate for low quality earnings. Discussion I used a cost-benefit model to predict how audit committee equity holdings influence the audit committees willingness to allow aggressive reporting when earnings are near the previous years level and found results consis tent with the prediction. This model can also reconcile four recent studies on the influence of audit committee s equity holdings. Klein [2002] found that large blockholders reduced abnormal accruals. Since blockholders are more likely to have indirect equity holdings than direct holdings (K lein [2002]), this is co nsistent with indirect holdings reducing earnings management. Vafeas [2005] found that equity holdings in Fortune 500 companies were either not relate d to or negatively related to aggressive reporting. Given his sample consisted of Fortune 500 firms, audit committee members with high levels of equity holdings were likely representati ves of institutions, so the audit committee members likely had high indirect holdings. Also, since the author did not cons ider how CEO incentives may moderate the influence of equity holdings, any null results may be at tributable to this omission. Yang and Krishnan [2005] found that independen t and non-independent equity holdings were positively related to quarterly a bnormal accruals. Since quarterly earnings are not audited, the

PAGE 98

98 likelihood of an irregularity is likely nil. He nce, the audit committees have little downside to allowing aggressive reporting in quarterly earnings. Bedard et al. [2004] found that the ratio of options to stock for audit committee member s was positively related to extreme earnings management. Since the value of options is more sensitive to small changes in stock price than stock, members with more options than stock stand to gain more from allowing aggressive reporting and less to lose if there is an irregularity. This is cons istent with the models prediction that the greater the potential gain from allo wing aggressive reporting, the more likely the committee is to allow aggressive reporting. Over all, the cost-benefit model provides a plausible reconciliation of conflicting fi ndings and a prescriptive framework for future research. My evidence indicated that audit committee beha vior is sensitive to its incentives and the risk factors present. This result was akin to the Hackenbrak and Nelson [1996] finding that auditors respond to their incentives according to the risk factors pres ent. My results added to the growing evidence that equity holdings induce se lf-interested behavior, even among independent audit committee members. These results make three contributions to th e literature on audit committees. First, I showed that the CEOs incentives to misrepor t tempered a members propensity to allow aggressive. This was evidence that audit co mmittees decisions depended upon the risk factors present. Second, I showed that the influence of equity holdings de pends upon the nature of equity ownership. Specifically, direct holdings induced self-interested behavior and indirect holdings induced improved monitoring. Third, I presented and validated a model that harmonized the inconsistencies in the previous wo rk on the influence of audit committees equity holdings. Specifically, I presented evidence that suggests the incons istencies are at tributable to

PAGE 99

99 an omitted moderator, namely CEO incentives, and a failure to disaggregate equity holdings by ownership type. These results have two regulatory implicat ions. It demonstrated the potentially problematic nature in audit committee members being financially vested in their decisions. While the committee appeared sensitive to the ri sk associated with CEO short-term equity incentives, it was willing to allow aggressive re porting when the risk is low, even if the committee consists of only independent members. This suggests that audit committee members were seeking to maximize their wealth, not audit committee effectiveness. I also showed that indirect holdings were positive ly associated with effective mo nitoring. I found no evidence that indirect holdings impaired independence. Th erefore, it would be unwarranted to prohibit committee members from having in direct equity holdings.

PAGE 100

100 Table 4-1. Logistic regression models of growth sample Test of H1 Test of H2 Column A Column B Column C Exp. sign B Wald B Wald B Wald Constant ? 0.410.42 0.260.16 0.070.01 Audit Committee Incentives AC direct holdings + 0.085.68*** AC indirect holdings ? 0.030.89 AC total holdings + -0.010.36 0.032.24* Interaction Effects AC direct holdings X CEO STEI -0.022.92** AC indirect holdings X CEO STEI -0.056.97*** AC total holdings X CEO STEI -0.028.33*** Audit Committee Controls Committee size -0.010.02 0.000.00 0.000.00 Meetings -0.020.28 -0.020.30 -0.010.09 Charter 0.435.23** 0.445.35** 0.465.80*** Ind. acc. expert 0.010.00 0.020.03 0.020.03 Average directorships 0.040.58 0.040.54 0.050.78 Tenure -0.033.66** -0.023.26** -0.033.87 Wholly independent 0.241.82* 0.221.41 0.343.31** CEO Attributes CEO STEI + 0.042.55* 0.099.33*** 0.086.69*** CEO LTEI ? 0.000.33 0.000.43 0.000.86 Control Variables Size -0.030.86 -0.030.68 -0.030.67 Market to book + -0.012.28* -0.011.95* 0.001.30 Leverage + 0.321.45 0.351.69* 0.361.80* Litigation risk -0.040.12 -0.050.21 -0.050.19 Loss current year -0.305.61*** -0.295.53*** -0.326.42*** Year controls included -2 Log likelihood 1850 1861 1839 Pseudo R square 0.04 0.03 0.05 Sample size 1367 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 101

101 Table 4-2. Logistic regression models with gr owth sample partitioned by CEO equity holdings Total holdings Partitioned holdings Column A Column B Column C Column D Low incentive High incentives Low incentive High incentives Exp. sign B Wald B Wald B Wald B Wald Constant ? 0.290.11 0.56 0.32 0.140.02 0.160.03 Audit Committee Incentives AC direct holdings 0.105.45*** 0.020.30 AC indirect holdings 0.020.28 -0.107.24*** AC total holdings 0.063.98** -0.054.32** Audit Committee Controls Committee size -0.100.95 0.141.19 -0.111.03 0.151.39 Meetings -0.040.66 0.010.05 -0.030.51 0.020.18 Charter 0.422.55* 0.52 3.51** 0.452.80** 0.553.83** Ind. acc. expert 0.242.01* -0.201.40 0.242.09* -0.201.42 Average directorships 0.050.27 0.020.07 0.050.37 0.030.13 Tenure -0.031.70* -0.020.71 -0.032.25* -0.021.14 Wholly independent 0.220.66 0.402.47* 0.301.18 0.544.02** CEO Attributes CEO STEI + -0.260.87 0.087.00***-0.230.72 0.086.77*** CEO LTEI ? 0.000.16 -0.013.69** 0.000.22 -0.014.27** Control Variables Size 0.000.00 -0.092.61* 0.000.00 -0.082.20* Market to book + -0.010.50 -0.011.81* -0.010.52 -0.011.38 Leverage + 0.561.75* 0.240.45 0.571.79* 0.230.43 Litigation risk -0.140.62 0.000.00 -0.140.63 0.020.01 Loss current year -0.486.52***-0.150.77 -0.496.88*** -0.170.92 Year controls included -2 Log likelihood 914916 911 912 Pseudo R square 0.050.06 0.06 0.07 Sample size 680 686 681 686 High incentive defined as CEO STEI > .95 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Ch ange>0, 0 otherwise All variables are defined in the appendix.

PAGE 102

102 Table 4-3. Logistic regression models of growth sample of firms with w holly independent audit committees Total holdings Partitioned holdings Exp. sign B Wald B Wald Constant ? 0.030.01 0.540.57 Audit Committee Incentives AC direct holdings + 0.103.15** AC indirect holdings 0.040.94 AC total holdings 0.020.72 Interaction Effects AC direct holdings X CEO STEI 0.000.01 AC indirect holdings X CEO STEI -0.076.20*** AC total holdings X CEO STEI -0.023.29** Audit Committee Controls Committee size 0.010.02 0.020.06 Meetings -0.020.57 -0.020.29 Charter 0.322.35* 0.332.33* Ind. acc. expert -0.030.06 -0.050.19 Tenure 0.030.24 0.030.21 Wholly independent -0.034.70** -0.046.15*** CEO Attributes CEO STEI + 0.086.08***0.062.56* CEO LTEI ? -0.012.60 -0.013.02* Control Variables Size -0.030.57 -0.020.21 Market to book + 0.000.33 0.000.01 Leverage + 0.130.20 0.120.18 Litigation risk -0.050.19 -0.070.31 Loss current year -0.253.40** -0.263.80** Year controls included -2 Log likelihood 1,620 1,609 Pseudo R square 0.03 0.04 Sample size 1195 Sample includes only firms with w holly independent audit committees p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 103

103 Table 4-4. Logistic regression models with growth sample partitioned by auditor size Large Auditor Small Auditor Exp. sign B Wald B Wald Constant ? 0.460.30 0.440.08 Audit Committee Incentives AC Direct Holdings + 0.092.45* 0.104.36** AC Indirect Holdings ? 0.040.75 0.020.16 AC Total Ownership + Interaction Effects AC Direct Holdings X CEO STEI -0.010.09 -0.035.42*** AC Indirect Holdings X CEO STEI -0.064.37** -0.052.74** AC Total Holdings X CEO STEI Audit Committee Controls Committee Size 0.000.00 0.030.02 Meetings -0.010.12 0.010.03 Charter 0.494.43** 0.451.59 Ind. Acc. Expert -0.130.90 0.372.51* Average Director 0.040.29 0.110.56 Tenure -0.021.07 -0.053.72** Wholly Independent 0.331.70* 0.351.33 CEO Attributes CEO STEI + 0.020.23 0.199.56*** CEO LTEI ? 0.001.08 0.000.01 Control Variables Size -0.040.94 -0.080.87 Market to Book + 0.000.83 -0.025.31** Leverage + 0.130.10 0.581.93* Litigation Risk -0.080.29 -0.110.20 Loss -0.375.81*** -0.251.13 Year Controls Included -2 Log likelihood 1320 496 Pseudo R square 0.04 0.10 Sample size 977 377 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 104

104 Table 4-5. Logistic regression models w ith growth sample and narrow EPS range EPS between -$0.025 and $0.025 EPS between -$0.02 and $0.02 Exp. Sign B Wald B Wald Constant ? 0.110.01 -0.320.09 Audit Committee Incentives AC direct holdings + 0.083.38** 0.115.17** AC indirect holdings ? 0.061.45 0.050.73 AC total holdings + Interaction Effects AC direct holdings X CEO STEI -0.011.67* -0.034.13** AC indirect holdings X CEO STEI -0.042.55* -0.052.56* AC total holdings X CEO STEI Audit Committee Controls Committee size 0.070.38 0.070.30 Meetings -0.030.56 -0.030.40 Charter 0.443.02** 0.705.42*** Ind. acc. expert 0.181.21 0.251.76* Average directorships 0.020.04 0.070.55 Tenure -0.033.29** -0.031.62 Wholly independent 0.311.27 0.170.29 CEO Attributes CEO STEI + 0.062.40* 0.125.16** CEO LTEI ? 0.000.01 0.000.01 Control Variables Size -0.020.25 -0.020.13 Market to book + 0.000.05 0.000.07 Leverage + 0.250.46 0.250.36 Litigation risk -0.060.14 0.020.01 Loss current year -0.487.82*** -0.527.03*** Year controls included -2 Log likelihood 974 750 Pseudo R square 0.06 0.07 Sample size 977 566 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 105

105 Table 4-6. Logistic regression models of gr owth sample considering auditor materiality Sample excludes positive earnings change firms with earnings greater than 5% of sales Sample excludes positive earnings change firms with earnings greater than 5% of total assets Exp. Sign B Wald B Wald Constant ? -0.781.23 -1.092.54 Audit Committee Incentives AC direct holdings + 0.096.94***0.085.56*** AC indirect holdings ? 0.051.84* 0.051.79* AC total holdings + Interaction Effects AC direct holdings X CEO STEI -0.022.88** -0.012.15* AC indirect holdings X CEO STEI -0.077.42***-0.067.07*** AC total holdings X CEO STEI Audit Committee Controls Committee size 0.040.23 0.050.31 Meetings 0.031.00 0.010.09 Charter 0.577.02***0.618.73*** Ind. acc. expert 0.060.26 0.040.10 Average directorships 0.060.81 0.071.16 Tenure -0.033.83** -0.033.49** Wholly independent 0.140.49 0.302.23* CEO Attributes CEO STEI + 0.085.82***0.075.60*** CEO LTEI ? 0.000.22 0.000.06 Control Variables Size -0.010.02 0.010.06 Market to book + -0.011.33 0.000.10 Leverage + 0.060.03 -0.030.01 Litigation risk -0.151.24 -0.120.85 Loss current year -0.9344.12***-0.6020.43*** Year controls included -2 Log likelihood 1592 1685 Pseudo R square 0.10 0.07 Sample size 1219 1262 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 106

106 CHAPTER 5 DIRECTORSHIPS AND THE GROWTH THRESHOLD Hypotheses Development How do directorships held in other corporat ions by audit committee members (henceforth audit committee directorships) influence an a udit committees willingness to allow earnings management? While some research indicates th at there is no associati on between directorships and financial reporting problems (Song and Windr am [2004], Vafeas [2005], Krishnan [2005]), other research indicates that th ere is a negative association be tween directorships and financial reporting problems (Bedard et al. [2004], Yang and Krishnan [2005]). Collectively, these results demonstrated directorships in fluence a committees judgment but how or why is unclear. I considered how audit committee director ships influence the committees decision to allow aggressive reporting when earnings are n ear a key threshold by ex amining how the audit committees decision influence the value of the memb ers directorships. I defined the effect of allowing aggressive reporting on the value of directorships as: Expected net benefits of agressive reporting = (1-p)(p) R EGainRELoseDIRDIR where: GainDIR represents the committees expectations rega rding the relative increase in the value of directorships if higher earn ings are reported rather th an lower earnings, assuming no irregularities. LoseDIR represents the committees expectations regarding the decrease in the value of directorships if there is an irregularity. pRE represents the committees assessment of the pr obability of an irregularity resulting from adopting aggressive reporting rath er than conservative reporting.

PAGE 107

107 If the expected net benefit is positive, then directorships bias the member in favor of allowing aggressive reporting. Otherwise, di rectorships bias the member against allowing aggressive reporting. To estimate the values of GainDIR and LoseDIR I examined how members gain and lose directorships. Typically, nominating committ ees are responsible for presenting directorial candidates to shareholders for their ratification. Presumably, the nominating committee will not recommend a new candidate for appointment or a current director for reappointment if the committee is not satisfied with his or her effectiv eness as a director, either with the company or with other companies. To evaluate a nominees effectiveness as a director, the nominating committee reviews the performance of the companies on which the no minee serves on the board. Specifically, the committee examine whether or not the companies are meeting earnings exp ectations and whether or not the companies experience any reporting irre gularities. Nominating committees prefer to select members from the boards of well-perform ing companies rather than from the boards of underperforming companies (Gil son [1990], Kaplan and Reishus [1990], Fama and Jensen [1983], Shivdansani [1993], Ferris et al. [2003]) a nd from boards of companies that have not experienced an irregularity (Fama [1980], Fama and Jensen [1983], Milgrom and Roberts [1992]). The CEO undermines the nominating process by promoting directorial candidates with a pro-management bias (Vancil [1987], Petra [ 2005], Shivdasani and Yermack [1999]). Gendron and Bedard [2006: 229]s interviews suggested that the CEO encour age a scratch my back, Ill scratch yours relationship with audit committee members. To gain the CEOs approval, audit

PAGE 108

108 committee members are willing to sacrifice monitoring effectiveness for improved operating results (Nabar, Kim, and Heninger [ 2006], Carcello et al. [2006]). In summary, if the audit committee allo ws aggressive report ing and there are no irregularities, then the members expect to gain new directorsh ip opportunities and reappointments at current dire ctorships for being an eff ective monitor and being promanagement. If the audit committee rejects aggre ssive reporting, then the members risk current and future appointments for being overly conserva tive. If an irregula rity occurs, then the members lose current and future appointments fo r being an ineffective monitor. Therefore, GainDIR is positive and LoseDIR is negative. pREis strictly positive because the audit comm ittee cannot know with certainty whether or not the manager opportunisti cally reported. If pREis near zero, then the expected gains from allowing aggressive reporting outweigh the expe cted losses should an irregularity occur. Further, as pREincreases, the expected gains from al lowing aggressive re porting decrease relative to the expected loss if an irregularity occurs. Therefore, ceteris paribus, when the risk is high (low), members with many directorships are le ss (more) likely to al low aggressive reporting than members with few directorships. Specifi cally, the audit committee should consider is the risk associated with CEO short-term equity incentives.47 Therefore, I hypothesized: H1: Audit committee member directorships are positively associated with the likelihood that a firm meets the prior years earnings level, but CEO shortterm equity incentives moderate this effect. 47 See Chapter 4 for an extended discussion of the re lationship between CEO incentives and the risk of an irregularity.

PAGE 109

109 Empirical Tests Using the growth sample, I empirically tested H1. I expanded the regression model from Chapter 3 to include an interaction term betw een directorships and CEO short-term equity incentives. Specifically, I used the following logistic regression: year tyear t-1(earningsearnings)=f(AC equity holdings,AC directorships, controls)->0pr where the variables are defined in the appendix. As an initial test, I replic ated Vefeas [2005]. Consistent with Vefeas [2005], when considering the influence of dir ectorships without the associated interaction terms, there was no association between the likeli hood that a firm meets the prio r years earnings level and directorships (Table 5-1 Column A). To test H1, I considered two different mode ls. The first model excluded equity holdings while the second included equity holdings and the associated inte raction terms. The first model found a positive coefficient on directorships (0.15 Co efficient, Wald 3.84, p-value <0.05) and a negative coefficient on the interaction term (-0.06 Coefficient, Wald, 4.44, p-value <0.05). This indicates that when the CEO short-term equity in centives represented more than 2.58 percent of total shares outstanding than di rectorships induced conservative decision-making. Otherwise, directorships induced liberal decision-maki ng. This is consistent with H1. The second model found a positive coefficient on directorships (0.15 Coefficient, Wald 3.93, p-value <0.05) and a negative coefficient on the interaction term (-0.06 Coefficient, Wald 3.71, p-value <0.05). This indicates that when the CEO short-term equity incentives represented more than 2.74 percent of total shares outstan ding, then directorship s induced conservative decision-making. Otherwise, dire ctorships induced liberal decisi on-making. Again, this is

PAGE 110

110 consistent with H1. Since my results did not vary with the inclusion of the equity incentives, this suggests equity incentives and reputation incentives ar e independent incentives. Discussion I argued that directorships capture the aud it committee members reputation incentives. There is a possibility that direct orships also capture governance expe rtise and diligence level. It is difficult to discriminate among these three explan ations with an archival dataset. While my results are consistent with a ll three explanations, I discounte d the alternative explanations because they were inconsistent with other evidence and the prior literature. Governance Expertise Governance expertise represents the member s knowledge of eff ective governance and board experience (Bedard et al. [2004]). Members with greater governance experience should be more capable of detecting and correctly interpreting the evidence of opport unistic reporting by managers. According to the model in Chapter 2, as governance expertise increases, the members have a higherd'. Fama and Jensen (1983) asserted that mu ltiple directorships is evidence that the director is capable and talented. Assuming directorships capture governance expertise, my results indicate when the risk of opportunistic reporting is high, audit committe es with greater governance expertise are more likely to prevent opportunistic reporting than audi t committees with less governance expertise. This is consistent with prior work. My results also indicate that when the risk of opportunistic reporting is low, audit committees with govern ance expertise are more likely to allow opportunistic reporting than audit committees without governance expertise. This suggests that audit committees with governance experience r ecognize the low risk of reporting problems associated with allowing aggres sive reporting to meet an ear nings threshold and are more

PAGE 111

111 tolerant of aggressive reporti ng than audit committees withou t governance expertise. This implies members with greate r governance expertise assess pRE more accurately. To investigate the possibility that dir ectorships capture gover nance expertise, I considered two other variables that should also capture governance expertise: accounting expertise and firm-specific knowledge have a simila r effect as directorships. Theoretically, a committees ability to detect and interpret evidence of opportunistic reporting and assess the risk of reporting problems should also be related to a committees accounting knowledge and firmspecific knowledge. To test this explanation, I replicated the primary analysis including two additional interaction terms: CEO short-term equity holdings with the presence of an independent accounting expert and CEO short-term equity holdings and tenure of audit committee members. I found no evidence that a ccounting expertise or firm-specific knowledge are related to the likelihood of a firm meeting the earnings threshold, regardless of the risk level (Table 5-2). These results do not support the explana tion that directorship s capture governance expertise. Diligence Effect Diligence represents the amount of time the members exert to disc harge their duties. Each directorship requires a significant amount of tim e to prepare for, travel to, and participate in regular meetings (Booth and Deli [1996]; Kaplan & Reishus [1990]). Lipton and Lorsch (1992) claimed that the most common problem facing direct ors is lack of time. Consistent with this explanation, researchers find di rectors with many directorships are less effective (Beasley [1996]). My evidence indicates that members with ma ny directorships are more likely to allow aggressive reporting when the risk of opportunistic reporting is low th an when the risk is high. If

PAGE 112

112 directorships captured a committees diligence, then there are two explanations for my results: more arbitrary decision-making or an efficient time allocation. If increasing directorships results in reduced time spent on each audit committee, then the members make decisions based on le ss evidence and the decisions are more arbitrary. It is possible that with insufficient time, the comm ittee infers the presence of incentives to opportunistically report is eviden ce of opportunistic reporting. However, this theory is inconsistent with the prior work that found in creasing directorships impair audit committees effectiveness at preventi ng opportunistic reporting. Alternatively, increasing time pressure caused by having many directorships may cause the members to allocate their time among directorship s according to the risk factors. Under this explanation, my results suggest that when there is a high risk of opportunistic reporting, the members spend more time investigating for evid ence of opportunistic re porting than when the risk is low.48 Therefore, assuming that a positiv e relationship between incentives to opportunistically report and incide nce of opportunistic reporting, the audit committee is more likely to find evidence of opportunist ic reporting when the risk is hi gh than when the risk is low. However, this theory is also in consistent with the prior work that found increasi ng directorships impair audit committees effectiveness at preventing opportunistic reporting. 48 According to the model in Chapter 2, d' should be positively associated with time spent investigating for evidence of opportunistic reporting.

PAGE 113

113 Table 5-1. Logistic regression models in cluding directorships for growth sample Test of H1 Complete model Column A Column B Column C Exp. Sign B Wald B Wald B Wald Constant ? 0.32 0.275 0.470.57 0.210.10 Audit Committee Incentives AC direct holdings + 0.086.17*** AC indirect holdings ? 0.020.50 AC total holdings + AC directorships + 0.05 0.712 0.153.84** 0.153.93** Interaction Effects AC direct holdings X CEO STEI -0.023.94** AC indirect holdings X CEO STEI -0.055.70*** AC total holdings X CEO STEI AC directorships X CEO STEI -0.064.44** -0.063.71** Audit Committee Controls Committee size -0.010.03 -0.010.03 -0.010.00 Meetings -0.020.28 -0.020.24 -0.010.09 Charter 0.435.13** 0.414.79** 0.455.57*** Ind. acc. expert 0.010.01 0.000.00 0.020.02 Tenure -0.033.67** -0.022.67* -0.023.13** Wholly independent 0.272.49* 0.292.84** 0.353.45** CEO Attributes CEO STEI + 0.042.57* 0.075.90*** 0.119.83*** CEO LTEI ? 0.000.37 0.000.55 0.001.01 Control Variables Size -0.030.74 -0.041.51 -0.041.36 Market to book + -0.012.37* -0.012.37* 0.001.28 Leverage + 0.321.47 0.331.56 0.371.88* Litigation risk -0.040.14 -0.030.06 -0.040.10 Loss current year -0.305.65*** -0.305.95*** -0.336.68*** Year controls included -2 Log likelihood 1857 1861 1835 Pseudo R square 0.03 0.03 0.05 Sample size 1367 1367 1367 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 114

114 Table 5-2. Logistic regression m odel for alternative explanation Exp. sign B Wald Constant ? 0.26 0.15 Audit Committee Incentives AC direct holdings + 0.085.89*** AC indirect holdings ? 0.020.42 AC total holdings + AC directorships + 0.15 3.76 ** Interaction Effects AC direct holdings X CEO STEI -0.023.71** AC indirect holdings X CEO STEI -0.055.34** AC directorships X CEO STEI -0.053.38 ** AC ind. acc expert X CEO STEI 0.040.58 AC tenure X CEO STEI 0.000.02 Audit Committee Controls Committee size -0.010.00 Meetings -0.010.08 Charter 0.455.46*** Ind. acc. expert + -0.050.10 Tenure + -0.032.35* Wholly independent 0.343.32** CEO Attributes CEO STEI + 0.092.86** CEO LTEI ? 0.000.98 Control Variables Size -0.041.37 Market to book + 0.001.27 Leverage + 0.361.83* Litigation risk -0.040.11 Loss current year -0.326.51*** Year controls included -2 Log likelihood 1835 Pseudo R square 0.05 Sample size 1367 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 115

115 CHAPTER 6 INCENTIVES AND THE PROFIT THRESHOLD Hypotheses Development To test the generalizablity of the results in Chapter 4 and 5, I reconsidered the three hypotheses presented in these chapte rs when earnings are near th e profit threshol d rather than when earnings are near the growth thre shold. Specifically, I hypothesized: H1: Audit committee equity holdings are positi vely associated with the likelihood that the firm has small positive earnings, but CEO short-term equity incentives moderate this effect. H2A: Audit committees direct holdings are po sitively associated with the likelihood that the firm has small positive earnin gs, but CEO short-term equity incentives moderate this effect. H2B: Audit committees indirect holdings are negatively associated with the likelihood that the firm has small positive earnings, with the magnitude of the effect increasing with the CEOs sh ort-term equity incentives. H3: Audit committees directorships are positive ly associated with the likelihood that the firm has small positive earnings, but CEO short-term equity incentives moderate this effect. Based on the cost-benefit mode ls, the only reasons why the results should vary between the growth and profit settings is if there is a va riation in the incentives or in the risk of an irregularity. I assumed that the litigation risk was trivial; however, this assumption may not be valid when earnings are near zero. There may be significant litigation risk because sustained profitability may be a critical element of many debt contracts. If debt holders discovered that a firm only satisfied the profitability requirement through earnings management, then they may hold the audit committee members accountable for their ineffective monitoring. This may provide a strong incentive for audit committee members to comp el conservative reporting. The likelihood of an irregularity may al so be higher in this setting if managers are willing to engage in more egregious earnings management to ac hieve profitability than to achieve growth.

PAGE 116

116 Empirical Tests I tested the hypotheses with the pr ofit sample and this regression: (earnings>0)=f(AC equity holdings, AC directorships, controls) pr where the variables are defined in the appendix. Primary Analysis As a preliminary analysis, I replicated the Vafeas [2005] analysis. As expected, I found no evidence of an association be tween the likelihood that a firm meets the profit threshold and audit committee equity holdings (Table 6-1 Column A). I found that an audit committee s total equity holdings influenced the likelihood that a firm meets a threshold and CEO short-term equity incentives moderate this effect (Table 6-1 Column B). However, the signs of the coe fficient were opposite e xpectations and only marginally significant: specifically the coeffici ent on main term was -0.03 (Wald>2.41, p-value <0.10) and the coefficient on the interaction te rm was 0.01 (Wald>2.49, p-value <0.10). This indicates that audit committee equi ty holdings induce a conservative behavior in the absence of CEO short-term equity incentives. However, when the CEO has incentives to meet a threshold, increasing audit committee equity holdings makes the company more likely to meet a threshold. This is inconsistent with H1. When I tested H2A and H2B, I found results simi lar to the results of the test of H1 (Table 6-1 Column C). Specifically, wh en I ignored directorships, the coefficient on direct holdings was negative and significant (coeffi cient -0.09, Wald 4.72, p-value <0.05) and the coefficient on the interaction between CEO incentives and direct holdings was positive and significant (coefficient -0.02, Wald 3.72, p-value <0.05). Wh en I considered directorships, the coefficient on direct holdings was negative a nd significant (coefficient -0.10, Wald 4.79, p-value <0.05) and

PAGE 117

117 the coefficient on the interaction between CEO incentives and direct holdings was positive and significant (coefficient -0.02, Wald 3.50, p-value <0. 05). This is inconsistent with H2A. I found no evidence that indirect holdings ha ve an effect (Table 6-1). This is inconsistent with H2B. When I ignore equity incentives, the influen ce of directorships was partially consistent with expectations (Table 6-1 Column D). Speci fically, the coefficient on average directorships was 0.12 (Wald 1.66, p-value <0.10), but the inte raction term between average directorships and CEO short-term equity incentives was not sign ificant (Wald 1.16, p-value >0.10). When I simultaneously considered equity holdings and dire ctorships, I found marginal ly stronger results. Specifically, the coefficient on average directorsh ips was 0.14 (Wald 2.09, p-value <0.10) and the coefficient on the interaction term between av erage directorships and CEO short-term equity incentives was -0.04 (Wald 2.00, p-value <0.10) Therefore, there is weak support for H3. Secondary Analysis When I analyzed the significance of the other audit committee attributes and control variables, there were three notab le findings. First, counter to expectations, meeting frequency was positively associated with the likelihood that firms achieved profitability (coefficient 0.07, Wald 3.01). This suggests that when earnings are near zero, the a udit committee met more frequently to verify that the earnings were leg itimate or that the board was more active in its monitoring and this motivated th e manager to work harder. Second, CEO long-term equity incentives a nd not short-term equity incentives were positively associated with the likelihood that a fi rm achieves profitability. This suggests either the auditor and/or the audit committee is attentive to the risk of opportunistic behavior associated with short-term equity incentives and rejects all potential earnings management when the CEO has high short-term equity incentives, or shortterm equity incentives do not induce the manager to engage in opportunistic behavi or or more productive effort wh en earnings are near zero.

PAGE 118

118 Third, consistent with expect ations, I found that firms that had negative earnings in the previous year were more likely to have negative earnings in the current ye ar than firms that had positive earnings in the previous year. This is consistent with the CEOs incentives to achieve profitability declining after the firm experiences a loss. Discussion While the archival evidence did not indicat e an obvious explanation as to why the influence of equity holdings opposed my hypotheses, I suggest three possible explanations. First, the audit committee members may have othe r incentives influencing their decisions, such as litigation risk49, that were more significant when earnings were near zero than when earnings were near the previous years level. By omit ting these additional factors, I failed to capture accurately the influence of incentives.50 To address this issue, I considered a sample of firms that employed a Big 4/5 auditor. If a firm em ployed a Big 4/5 auditor, then the audit committee likely faced less litigation risk than when a fi rm employed a non-Big 4/5 auditor for two reasons. First, since a Big 4/5 auditor is less likely to miss evidence of a going concern than a non-Big 4/5 auditor, the likelihood of a reporting problem and related litigation is lower. Second, if a reporting failure does occur, then the audit co mmittee members have a more legally defendable position if they employ a Big 4/5 audito r rather than a non-Big 4/5 auditor.51 When I reconsidered H1, H2, and H3, the results were in the expected direction (Table 6-2). However, 49 While the model does include a control for industry litigation risk, there may be higher litigation risk for all firms when earnings are near zero relative to the litigation ri sk when earnings are near the previous years level. 50 I cannot consider the other proxies for litigation risk such as those suggest in Krishnan and Krishnan [1997] (financial distress, going concern variance of abnormal returns, auditor independence, and tenure) for a variety of reasons. First, all the miss firms in the sample are, by definition financially, distressed so there can be no proxy for financial distress. Second, I exclude all firms that have received a going concern opinion from the sample. Lastly, the data requirements for the other proxies are prohibitive given the sample period. 51 Audit committee members may also face lower litigation risk when employing a Big 4/5 auditor rather than a nonBig 4/5 auditor, since shareholders may be more likely to sue the auditors under these circumstances.

PAGE 119

119 some of the results were not significant at conv entional levels. The lack of significance may be attributable to the reduced power as sociated with the smaller sample.52 The second explanation is that the expected impact on e quity holdings, if a reporting problem occurs, may be greater wh en earnings are near zero compar ed to when earnings are near the previous years level. The third explanation is that if there is a higher pREfor firms near the profit threshold than firms near the growth threshold, then the increasing equity holdings may make the audit committee more conservative. However, increasing equity incentives may also give the audit committee members increasing motivat ion to verify that the firm met the threshold because of greater managerial effort and not opportunistic reporting. Since CEOs are more likely to exert greater effort when they have greater incentives to do so, the positive coefficient on the interaction term may indica te that when the CEO has incentives to work harder and the audit committee has high incentives to verify the work the firm is more likely to meet the profit threshold. To test the second and third explanati on, it is necessary to do either a field study or a behavioral experiment, both of wh ich are beyond the scope of my study. 52 The sample of firms that employ a non-Big 4/5 auditor is too small to conduct an analysis.

PAGE 120

120 Table 6-1. Logistic regression models for profit sample Replication of Vefeas (2005) Test of H1 Test of H2 Column A Column B Column C Exp. Sign B Wald B Wald B Wald Constant ? 0.99 2.00 0.98 1.97 1.20 2.79 ** Audit Committee Incentives AC direct holdings + -0.094.72 ** AC indirect holdings ? -0.010.44 AC total holdings + -0.010.54 -0.032.41* AC directorships + 0.06 0.72 0.06 0.54 0.06 0.59 Interaction Effects AC direct holdings X CEO STEI 0.023.72 ** AC indirect holdings X CEO STEI 0.000.58 AC total holdings X CEO STEI 0.012.49 AC directorships X CEO STEI Audit Committee Controls Committee size -0.050.22 -0.060.26 -0.060.27 Meetings 0.072.97** 0.073.24** 0.073.07 ** Charter 0.140.36 0.140.37 0.120.24 Ind. acc. expert -0.070.24 -0.070.22 -0.060.18 Tenure 0.010.68 0.010.68 0.020.98 Wholly independent 0.200.89 0.210.95 0.150.50 CEO Attributes CEO STEI + 0.020.56 0.000.00 -0.010.22 CEO LTEI ? 0.011.70* 0.011.90* 0.012.04 Control Variables Size -0.094.54** -0.094.38** -0.104.95 ** Market to book + 0.011.71* 0.011.90* 0.011.64 Leverage + -0.573.98** -0.584.06** -0.594.11 ** Litigation risk 0.010.01 0.010.00 0.000.00 Loss last year -0.5112.69*** -0.5112.70*** -0.5112.79 *** Year controls included -2 Log likelihood 1,196 1193 1189 Pseudo R square 0.06 0.07 0.07 Sample size 903 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 121

121 Table 6-1 Continued. Test of H3 Complete model Column D Column E Exp. Sign B Wald B Wald Constant ? 0.95 1.94 1.32 3.31 Audit Committee Incentives AC direct holdings + -0.104.79** AC indirect holdings ? -0.020.73 AC total holdings + AC directorships + 0.12 1.66 0.14 2.09 Interaction Effects AC direct holdings X CEO STEI 0.023.50** AC indirect holdings X CEO STEI 0.011.32 AC total holdings X CEO STEI AC directorships X CEO STEI -0.031.16 -0.042.00 Audit Committee Controls Committee size -0.050.20 -0.060.27 Meetings 0.062.87** 0.073.01** Charter 0.120.27 0.090.13 Ind. acc. expert -0.070.25 -0.070.20 Tenure 0.021.00 0.021.38 Wholly independent 0.241.31 0.160.56 CEO Attributes CEO STEI + 0.031.28 0.000.02 CEO LTEI ? 0.011.58 0.011.87* Control Variables Size -0.094.82** -0.115.88*** Market to book + 0.011.62 0.011.65* Leverage + -0.584.05** -0.604.25** Litigation risk 0.010.00 0.010.00 Loss last year -0.5112.54*** -0.5112.62*** Year controls included -2 Log likelihood 1195 1188 Pseudo R square 0.07 0.07 Sample size 903 903 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 122

122 Table 6-2. Logistic regression model for profit sample with clients of Big 4/5 auditors Auditor Quality High Auditor Quality High Test of H1, H3 Test of H2, H3 Exp. Sign B Wald B Wald Constant ? 2.51 6.27 ** 2.24 4.45 ** Audit Committee Incentives AC direct holdings + 0.131.77* AC indirect holdings ? 0.000.01 AC total holdings + 0.030.38 AC directorships + 0.15 1.99 0.16 2.37 Interaction Effects AC direct holdings X CEO STEI -0.031.52 AC indirect holdings X CEO STEI 0.010.86 AC total holdings X CEO STEI 0.000.04 AC directorships X CEO STEI -0.052.93 ** -0.063.81 ** Audit Committee Controls Committee size -0.181.60 -0.161.33 Meetings 0.030.49 0.030.44 Charter -0.080.06 -0.090.07 Ind. acc. expert -0.010.00 -0.020.01 Tenure 0.054.56** 0.054.65** Wholly independent 0.230.54 0.270.67 CEO Attributes CEO STEI + -0.010.13 0.010.06 CEO LTEI ? 0.000.23 0.000.20 Control Variables -0.146.28*** Size 0.000.10 -0.135.06** Market to book + -0.190.15 0.000.02 Leverage + -0.120.42 -0.210.19 Litigation risk -0.282.51* -0.120.43 Loss last year -0.292.61* -0.292.61* Year controls included -2 Log likelihood 768 765 Pseudo R square 0.06 0.06 Sample size 585 585 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 123

123 CHAPTER 7 INCENTIVES AND THE ANALYST THRESHOLD Hypotheses Development Consistent with the prior literature, I also consider whether the three hypotheses from the prior chapters generalize to when earnings are ne ar analyst forecast (i.e Degeorge et al. [1999], Vefeas [2005]). Speci fically, I hypothesized: H1: Audit committee equity holdings are positi vely associated with the likelihood that the firm has a small earnings surprise but CEO short-term equity incentives moderate this effect. H2A: Audit committees direct holdings are po sitively associated with the likelihood that the firm has a small earnings surpri se, but CEO short-term equity incentives moderate this effect. H2B: Audit committees indirect holdings are negatively associated with the likelihood that the firm has a small earnings surprise, with the magnitude of the effect increasing with the CEOs sh ort-term equity incentives. H3: Audit committees directorships are positive ly associated with the likelihood that the firm has a small earnings surprise but CEO short-term equity incentives moderate this effect. Audit committee incentives may not be relevant when earnings are near analyst forecasts if the presence of an audit committee may cause managers to engage in greater expectation management with analysts rather than earni ngs management (Francis, Lafond, Olsson, and Schipper [2003]). This is consistent with Karamanou and Vafeas [2005], who found that the presence of an audit committee influenced managerial forecast s and Liu [2004], who found that audit committee attributes influenced anal yst responses to earnings information.53 If the manager does not manage earnings, then the aud it committee does not have the option to allow it. 53 I discount the possibility that difference in litigation risk may cause different results between when earnings are near analyst forecasts and when earnings are near previous years level or zero, because there is no evidence of such risk.

PAGE 124

124 Empirical Tests I tested the hypotheses with the anal yst sample and this regression: (earnings>analyst forecast)=f(AC equity holdings, AC directorships, controls) pr where the variables are defined in the appendix. Primary Analysis As an initial test, I replic ated Vefeas [2005]. I found no evidence of an association between audit committee equity holdings and th e likelihood of meeting analyst forecast, but Vefes [2005] found a negative association (Table 7-1 Column A). This difference may be a result of sample differences. Since, Vefeas [ 2005]s sample only include d Fortune 500 firms, his audit committee equity holdings proxy may be cap turing indirect holdings. If Vefeas [2005]s equity holdings proxy captures equity hol dings, then there is no contradiction. I found strong support for the H1 and H2A. Sp ecifically, the coeffi cient on total holdings was positive (Wald 3.93, p-value <0.05) and the coefficient on the interaction between CEO incentives and total holdings was negative (Wald 3. 12, p-value <0.05) (Table 7-1 Column B and C). Further, the coefficient on direct holdings was positive (Wald 3.41, p-value <0.05) and the coefficient on the interaction between CEO incen tives and direct holdings was negative (Wald 3.32, p-value <0.05). There is partial support for H3. I found st rong evidence of an in teraction between CEO incentives and directorships (Wald 4.52, p-value <0 .05), but only weak evidence of a main effect on directorships (Wald 2.26, p-value<0.10). I found no support for H2B on the effect of i ndirect holdings (Wald <1.70, p-value>0.10 for both main and interaction term). Since I used a more noisy proxy in this sample than in the growth or profit sample, I considered an alternat ive proxy to shares of de signated directors. I

PAGE 125

125 examined the influence of the presence of a desi gnated director, which is defined as a member with any indirect holdings. Even with this ne w specification, there was only weak support for H2B (Wald 1.97, p-value<0.10 for the main term Wald 1.93, p-value<0.10 for the interaction term) (Table 7-1 Column D). I tested the robustness of my results by considering a more restrictive sample. Specifically, I excluded firms that had an earnings surprise per share between $0.005 and $0.01.54 Under this specification, there was strong evidence supporting H1, H2A and H3 (Table 7-2). The coefficient on total holdings was 0. 61 (Wald 6.26, p-value<0.01) and coefficient on the interaction between total holdings and CEO short-term equity incentives was -0.31 (Wald 5.79, p-value<0.01). The coefficient on direct holdings was 0.62 (Wald 6.10, p-value<0.01) and coefficient on the interaction be tween direct holdings and CEO shor t-term equity incentives was 0.31 (Wald 5.82, p-value<0.01). The coeffici ent on directorships was 0.19 (Wald 3.44, pvalue<0.05) and coefficient on th e interaction between directorsh ips and CEO short-term equity incentives was -0.14 (Wald 3.75, p-value<0.05). Again, I found no support for H2B. I examined a number of alternat ive approaches to verify the robustness of my results. Specifically, I redefined CEO short-term equity incentives, audit committee direct holdings, audit committee indirect holdings, an d directorships as a binary va riable (i.e. high-low). Under these definitions, in unreported results, I found sim ilar results. Lastly, I considered alternative specifications of audit committee and board ability. I used the 4 level audit committees ability score and 7 level boards ability score rather than the binary versions. In unreported results, 54 Unlike the other samples, this sample has an asymmetric miss-meet range (Miss: -$0.01 to 0, Meet $0 to $0.05). This design choice is necessary because only 87 firms in my sample have an earnings surprise per share between $0.05 and $0, which makes any analysis problematic.

PAGE 126

126 regardless of the variable specification, I f ound the primary conclusions were unchanged. Combined with the primary results, there is strong evidence to support H1, H2A, and H3. Secondary Analysis When I analyzed the significance of the a udit committee variables and control variables, there were three notable findings. First, regard less of how I measured audit committee ability, I found no evidence that audit committee ability in fluenced the likelihood that a firm meets analysts forecasts (Table 7-1). This is consistent with Chapters 4 through 6 and with prior work (Vafeas [2005], Yang and Krishnan [2005]). Second, I found no evidence that whether or not a firm met the growth threshold or the profit thre shold influenced the likelihood that a firm met analyst forecast. This is incons istent with the Degeorge et al. [1999] hierarchy of thresholds. Lastly, I found weak evidence that CEO incentives influence the likelihood that a firm meets analysts forecast. This suggests that analysts incorporate the effect of short-term equity incentives on executive behavior when making forecasts. Discussion Consistent with the prior chapter, I find strong evidence that audit committees total holdings, direct holdings, and dire ctorships influence the likeli hood that a firm meets analyst forecasts. Combined with the evidence in Chapters 4 through 6, I believe there is persuasive evidence that audit committees re spond to the incentives, reputati on and equity, in accordance with the risk environment.

PAGE 127

127 Table 7-1. Logistic regression model for analyst sample Replication of Vefeas (2005) Test of H1, H3 Test of H2, H3 Column A Column B Column C Exp. Sign B Wald B Wald B Wald Constant ? 2.065.53**2.216.13** 2.20 6.02 *** Audit Committee Incentives AC direct holdings + 0.333.41** AC indirect holdings + -16.291.67* AC total holdings + 0.141.87* 0.343.93** AC directorships + 0.020.04 0.131.93* 0.14 2.26 Interaction Effects AC direct holdings X CEO STEI -0.163.32** AC indirect holdings X CEO STEI 51.931.68* AC total holdings X CEO STEI -0.153.12** AC directorships X CEO STEI -0.144.14** -0.154.52 ** Audit Committee Controls Audit committee quality -0.030.08 -0.030.08 -0.020.04 Board quality -0.110.97 -0.090.65 -0.100.73 CEO Attributes CEO STEI + -0.020.14 0.102.11* 0.102.16* CEO LTEI ? -0.010.39 -0.010.40 -0.010.36 Control Variables Size -0.051.55 -0.051.50 Market to book + -0.040.92 -0.093.62** -0.094.07** Leverage + -0.083.31**0.020.00 0.010.00 Litigation risk 0.010.00 0.121.15 0.121.07 Loss current year 0.121.00 -0.080.25 -0.070.19 No growth -0.090.32 -0.070.41 -0.070.42 Year controls included -0.060.31 -2 Log likelihood 2415 2412 2401 Pseudo R square 0.01 0.02 0.02 Sample size 2367 2367 2367 p-value based on one-tailed test s where relation is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 128

128 Table 7-1 Continued Designated Director Audit committee control Column D Column E Exp. Sign B Wald B Wald Constant ? 2.276.44** 2.185.91** Audit Committee Incentives AC direct holdings + 0.343.56** 0.333.43** AC indirect holdings + -16.231.64 AC designate + -0.921.97* AC directorships + 0.142.27* 0.131.85* Interaction Effects AC direct holdings X CEO STEI -0.163.44** -0.163.33** AC indirect holdings X CEO STEI 51.81.65* AC designate X CEO STEI 0.941.93* AC directorships X CEO STEI -0.154.58** -0.154.84** Audit Committee Controls Audit committee quality -0.030.08 -0.030.09 Board quality -0.100.78 CEO Attributes CEO STEI + 0.102.15* 0.112.29* CEO LTEI ? -0.010.42 0.000.31 Control Variables Size -0.051.70* -0.051.51 Market to book + -0.093.93** -0.094.01** Leverage + 0.020.00 0.010.00 Litigation risk 0.131.23 0.121.10 Loss current year -0.080.23 -0.080.22 No growth -0.070.47 -0.070.39 Year controls included -2 Log likelihood 2406 2401 Pseudo R square 0.019 0.02 Sample size 2367 2367 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Change>0, 0 otherwise All variables are defined in the appendix.

PAGE 129

129 Table 7-2. Logistic regression model for analyst sample with narrow range EPS Column A Column B Test of H1, H3 Test of H2, H3 Column A Column B Exp. Sign B Wald B Wald Constant ? 2.214.92** 2.14 4.60 ** Audit Committee Incentives AC direct holdings + 0.626.10*** AC indirect holdings + -15.691.09 AC total holdings + 0.616.26*** AC directorships + 0.183.22** 0.193.44** Interaction Effects AC direct holdings X CEO STEI -0.315.82*** AC indirect holdings X CEO STEI 50.541.12 AC total holdings X CEO STEI -0.315.79*** AC directorships X CEO STEI -0.143.69** -0.143.75** Audit Committee Controls Audit committee quality 0.000.00 0.000.00 Board quality -0.090.54 -0.100.62 CEO Attributes CEO STEI + 0.122.69* 0.122.72** CEO LTEI ? 0.000.26 0.000.28 Control Variables Size -0.083.17** -0.082.92** Market to book + -0.124.50** -0.124.61** Leverage + -0.060.02 -0.070.03 Litigation risk 0.171.83* 0.171.77* Loss current year -0.181.02 -0.170.97 No growth -0.070.31 -0.060.28 Year controls included -2 Log likelihood 1902 1897 Pseudo R square 0.032 0.036 Sample size 1139 p-value based on one-tailed tests where relati on is predicated, otherwise two-tailed test *** p-value < 0.01, ** p-value < 0.05, p-value < 0.10 Dependent variable equals 1 if EPS Surprise>0, and 0 otherwise. Sample restricted to firms w ith EPS surprise between -$0.01 an d $0.05 per share. All variables are defined in the appendix.

PAGE 130

130 CHAPTER 8 INCENTIVES: BIAS OR MOTIVATE Do audit committee incentives motivate the co mmittee to increase its effort or bias the committees judgment? Understanding how incenti ves influence audit co mmittees is important for three reasons. First, unders tanding how incentives influence an audit committees judgment enhances our ability to interpret the results of the prior chapters. For example, it allows us to determine if an audit committees response to increasing CEO incentive s is to increase its effort or to develop a conservative bias. Sec ond, understanding how incentives influence the committees judgment may help us understand w hy researchers found contradictory results for the influence of incentives on an audit committees willingness to allow opportunistic reporting. For example, if an audit committees equity ho ldings bias the committees judgment, then the prior research may conflict because, in some se ttings, the evidence of opportunistic behavior is weak enough that the committees bi as allows the committee to i gnore the evidence. However, in other settings, the evidence is too compelling to be ignored, regardless of the bias. Third, there is considerable regulatory interest in improving audit committees as a monitoring tool. Understanding how incentives infl uence the committees judgment is critical to making policy recommendations regarding restricti ons on audit committee incentives. Effort or Bias In the prior chapters, I used a logistic regr ession analysis to document evidence of an association between audit commit tee incentives and the likelihood that a firm met an earnings threshold. One limitation of a regression analys is is that it cannot e xplain why there is an association. The association may exist because incentives motivated the committee to exert greater effort or because incentives biased the committees judgment.

PAGE 131

131 If incentives motivate the committee to exer t greater effort, then, according to the model in Chapter 2, the committee has a greater ab ility to differentiate between truthful and opportunistic reporting (i.e. higher d'). This is consistent with equity holdings motivating audit committee members to exert greater effort in orde r to protect their invest ment and directorships motivating audit committee members to exert greater effort in order to protect their reputation as an effective monitor. However, increasing dire ctorships may also reduce the committees effort, since there is likely a negative association be tween directorships and time availability. If incentives induce a conservative (liberal) bias, then the committee requires less (more) evidence before concluding oppor tunistic reporting (i.e. zc is affected) A liberal bias is consistent with equity holdings motivating a focus on short-term stock appreciation and with directorships motivating a focus on developing a reputation as being pro-management. A conservative bias is consistent with equity holdings motivating a higher aversion against accounting irregularities and with directorships motivating a focu s on developing a reputation as an effective monitor. Ex ante, it is unclear whether equity holdings or directorships induc e an increased effort, bias, or both. Therefore, I asked the following questions: Q1: Do audit committee equity holdings induce increased effort or bias judgment? Q2: Do audit committee directorships i nduce increased effort or bias judgment? Why is an audit committee more likely to reject aggressive re porting when the CEOs short-term equity incentives are high than when th e CEOs short-term equity incentives are low? In the presence of high CEO short-term equity incentives, do audit committee members increase their effort (higher d) or do the members develop a cons ervative bias (i.e. lower evidence threshold)? Therefore, the final question was:

PAGE 132

132 Q3: How do the CEOs short-term equity in centives influence the effect of the audit committees incentives? Signal Detection Theory To differentiate between the bi as and effort explanations, I reexamined the data with signal detection theory. Signal detection theo ry calculates detection ability and bias simultaneously. Signal detection theory dete rmines decision accuracy by comparing the committees decision with the correct decision. The difficulty in analyzing audit committee decisions with signal detection theo ry is identifying what the correct decision is. To address this issue, I considered the differe nces between the initially repor ted earnings and the restated earnings. Specifically, I examined whether or not a firm would have met an earnings threshold if the CEO had reported restated earnings rather than the initially reported earnings. Restatements occur for several reasons includ ing accounting for an acquisition, the sale of a subsidiary, a change in the reporting company's accounting po licies, errors in reporting procedures, and earnings manipulat ions (Agrawal and Chadha [2005] ). If a restatement occurs because the audit committee allowed opportunistic reporting, then the restatement decreases earnings. This is consistent with the likelihood of a restatement being positively associated with CEO incentives (Burns and Kedia [2004]) and nega tively associated with the presence of an effective monitor (Agrawal and Ch adha [2005]). However, if a re statement occurred because the audit committee pro-actively requires conservati ve reporting, then the restatement increases earnings. Hence, I assumed that if a restatem ent reduces earnings, then the audit committee made an incorrect decision by allowing the opportunist ic reporting, and if a restatement increases earnings, then the audit committee made an inco rrect decision by reje cting the opportunistic reporting.

PAGE 133

133 I applied signal detection theory with a fou r-step process. First, I identified whether or not each firm met an earnings th reshold according to the initiall y reported earnings. Second, I identified whether or not each firm would have met the earnings threshold according to restated earnings.55 If a firm did not issue restated earnings then I assumed that the manager accurately reported earnings and the audit committee made th e correct decisions. Third, I classified each firm according to this framework: Restated Earnings indicate: Miss Correct Rejection Miss Threshold Meet ThresholdFalse Alarm Meet Threshold Initial Earnings Indicate: Miss ThresholdHit Lastly, I calculated the empirical surrogates for d and : d=()() Z HZF and =22.5(()()) ZHZFe where Z represents the Z-score transformation of the probability argument, H (hit rate) is the number of hits divided by the sum of hits and misses, and F (false alarm rate) is the number of false alarms divided by the sum of false alarms and correct rejections (Swets [1986] Ramsey and Tubbs [2005]). To answer Q1, Q2, and Q3, I calculated the dand for a number of sub-samples of the profit and growth sample. Specifically, I partitio ned each sample according to the CEOs shortterm equity incentives (a proxy for the risk of opportunistic reporting), audit committee equity holdings, and audit committee directorships. I c onsidered four levels of CEO incentives, two levels of audit committee equity holdings, and two levels of audit committee directorships. I did 55 Using the restatement information from Compustat, I calcu late the restated EPS change for the growth sample and I calculate the restated EPS for the profit sample.

PAGE 134

134 not reconsider the analyst sample because re stated I/B/E/S earnings information was not available.56 Results There appeared to be a significant difference in dand between the profit and growth samples (Table 8-1). Specifically, fo r the non-partitioned profit sample d was 3.46 and was 2.04, and for the non-partitioned growth sample d was 2.48 and was 1.24. This indicates that audit committees are much more diligent a nd are more likely to err against requiring conservative reporting when earnings are near the profit threshold than when earnings are near the growth threshold. Independent of audit committee ince ntives, both samples indicate that dvaries positively with the CEO short-term equity incentives, although the effect was more pronounced for the profit sample than for the growth sample. Specifically, for the growth (profit) sample, when CEO incentives were above the median, the dwas 2.57(3.79), but when incentives were below the median the dwas 2.39(3.21). This suggests that an audit committee increased its effort only slightly when the CEOs short-term equity incentives were high, relative to when the CEOs short-term equity incentives were low. Independent of audit committee incentives, the varied positively with the CEO shortterm equity incentives for the profit sample, but did not vary with the CEO short-term equity incentives for the growth sample. Specifically, for the growth (profit) sample, when CEO shortterm equity incentives were above the median, the was 1.25 (2.57), but when the CEO shortterm equity incentives were below the median, the was 1.23 (1.78). Counter to expectations, 56 There is significant variation between earnings info rmation from I/B/E/S and Computsat because I/B/E/S information is adjusted for some items (Thompson Financial [2003]). Also, I/B/E/S does not disclose the basis for this adjustment.

PAGE 135

135 this suggests that for the profit samp le (not the growth sample), when the risk of an irregularity is high, the committee requires more evidence to c onclude opportunistic re porting than when the risk of an irregularity is low. One possible explanation for this a nomalous result is that when the CEO has high incentives, he/she is more effectiv e at persuading the a udit committee to ignore evidence of opportunistic reporting. Growth Sample Equity Holdings For each CEO incentive level, when an aud it committee had above the median equity holdings, it had a higher d than when an audit committee had below the median equity holdings (Table 8-1 and Figure 8-1). This suggests that an audit committee with high equity holdings exerts more effort than an audit committee with low equity holdings. However, I found no evidence that d varies systematically with CEO short-term equity incentives, regardless of whether audit committee equity holdings are above or below the median (Figure 8-1). This suggests that an audit committees effort level does not vary with CEO short-term equity incentives. Audit committee equity holdings also infl uenced the committees bias (Table 8-1 and Figure 8-2). For each of the CEOs short-term e quity incentives levels, an audit committee with above the median equity holdings had a higher bias than an audit committee with below the median equity holdings. This suggests that e quity holdings bias a committees judgment against concluding opportunist ic reporting. The committees bias did not vary much with the CEOs short-term incentives if the audit committee had below the median equity incentives (Figure 8-2). This is consistent with low levels of equity holdings not compromising a co mmittees judgment. However, when the audit committee had above the median equity holdings, the committees bias decreased as CEO short-

PAGE 136

136 term equity incentives increased. This suggests th at in a low-risk environment, audit committee equity holdings have a greater biasing effect on the committee s judgment than in a high-risk environment. Directorships When the CEO had low incentives, there di d not appear to be any variation in d between committees with above the median direct orships and committees with below the median directorships (Table 8-1 and Fi gures 8-3 and 8-4). This sugge sts that directorships did not influence an audit committees effort level in a low risk environment. When a CEO had high short-term equity incen tives, the influence of directorships was complicated. If the CEO had above the medi an short-term equity incentives, then donly varies from the d value in the low-risk setting if the audit committee has below the median directorships. Specifically, if the CEO had above the median short-term e quity incentives, then d was 2.98 when the audit committee has below the median directorships and 2.39 when the audit committee had above the median directorships. This suggests that when the CEO had above the median short-term equity incentives, audit committees with fewer directorships are more attentive to the risk associated with CEO short-term equity incentives than audit committees with many directorships. This is co nsistent with directorships limiting the audit committees time availability and suggests director ships reduce audit committee effectiveness. When the CEO had extremely high incentives (75th percentile or above), if the audit committee had below (above) the medi an directorships, then the d is lower (higher) than the d in the low risk setting. Specifically, if the CEO had extremely high incentives, then d was 2.88 when the audit committee had above the medi an directorships and 1.97 when the audit committee had below the median directorships. This is consistent with audit committee

PAGE 137

137 members having few directorships failing to recognize the additional risk associated with extremely high CEO short-term incentives a nd audit committees having many directorships increasing their effort when the ri sk of an irregularity is extremely high. This suggests that directorships enhance audit committee effectiveness. The biasing effect of director ships on audit committee decisions appeared minimal. For all levels of CEO short-term equity incentives except when the CEO has extremely high shortterm equity incentives, the did vary between audit committees with above the median directorships and audit committees with below the median directorships. When the CEO had extremely high incentives, the committees bias is greater when the audit committee had above the median directorships than when the audi t committee had below the median directorships (1.54 vs. 1.18). This is consistent with a udit committees exhibiting a pro-management bias by accommodating management requests when the CEO will benefit most from it. Profit Sample Equity Holdings The influence of audit co mmittee equity holdings on d depended upon the CEOs shortterm equity incentives (Table 8-1, Figures 8-5 an d 8-6). Audit committee equity holdings were positively associated with d, except when the CEO has low incentives and the audit committee has below the median equity holdings (Figure 8-5) This is consistent with audit committees having low incentives exerting low effort in a low risk environment. The biasing effect of audit committee e quity holdings also depended upon the CEOs short-term equity incentives. Specifically, re gardless of the level of equity holdings, the committees bias is highest when the CEO had lower incentives (Table 8-1). This is consistent

PAGE 138

138 with audit committees willing to overlook evidence of opportunistic repor ting if there is not a risk factor to suggest opportunistic reporting. Directorships The effect of directorships on d depended upon the CEOs incentives (Table 8-1 and Figures 8-5 and 8-6). If the CEO had below median incentives and the audit committee had above the median directorships, then d was 2.81, but under the other scenarios, d varied between 3.75 and 3.92. This suggests that wh en members with many directorships reduce their effort when the risk of oppor tunistic reporting is low. The influence of directorships on also depended upon the CEOs incentives. When the CEO had below the median incentives a nd the audit committee had above the median directorships, was 1.45. Under the other scenarios, varied between 2.22 and 2.51 (Table 81). This suggests that when the risk of opport unistic reporting is low, the members with many directorships compensate for their reduced effort by reducing their evidence threshold. Discussion While my results vary slightly between th e growth and profit sample, the results do suggest three conclusions. First, audit committe e equity holdings motivate greater effort and induce a pro-management bias. This suggests th at while equity holdings may compromise the committees judgment, the committee is exerting greater effort to a void irregularities. Unexpectedly, audit committee equity holdings did not motivate the committee to be attentive to the risk associated with high CEO short-term equity incentives. This suggests that educating the members on the risk associated with CEO shor t-term equity incentives can enhance audit committee effectiveness.

PAGE 139

139 Second, while directorships influenced both th e committees effort and bias, the influence depended upon the threshold under consideration a nd the CEOs short-term equity incentives. This indicates that the influence of directorships is contextual. Hence, it is be problematic to design any regulations restricting directorships or to identify an optimal level of directorships. Lastly, audit committees appeared to exert mo re effort and have less bias when earnings were near the profit thre shold than when earnings were near th e growth threshold. This suggests that audit committees are more a ttentive to the risk of opportunist ic reporting when earnings are near zero than when earnings are near the previous years level. This supports the assertion that audit committees adjust their effort and eviden ce demands based upon the perceived risks. This may also indicate that managers are more a ggressive in opposing earnings restatements if restated earnings result in reporting losses.

PAGE 140

140 Table 8-1. Applying signal detection theory Equity incentives Growth sample Profit sample Hit rate False alarm rate Detection ability Bias Hit rate False alarm rate Detection ability Bias Total sample 0.88 0.09 2. 48 1.24 0.94 0.03 3.46 2.04 CEO incentives Extreme high 0.87 0.08 2.51 1.39 Above median 0.88 0.09 2. 57 1.25 0.95 0.02 3.79 2.57 Below median 0.87 0.10 2. 39 1.23 0.92 0.04 3.21 1.78 Extreme low 0.86 0.10 2.37 1.33 Audit committee equity incentives Independent of CEO's short-term incentives Above median 0.89 0.07 2. 68 1.38 0.96 0.02 3.90 2.13 Below median 0.86 0.11 2. 31 1.14 0.91 0.04 3.15 1.95 CEO's short-term equity incenti ves are above the 75th percentile Above median 0.90 0.06 2. 88 1.54 0.92 0.04 3.14 1.81 Below median 0.81 0.14 1. 96 1.19 0.95 0.02 3.70 1.89 CEO's short-term equity in centives are above the median Above median 0.89 0.07 2. 70 1.40 0.95 0.02 3.75 2.29 Below median 0.88 0.11 2. 39 1.09 0.95 0.01 3.90 3.36 CEO's short-term equity incentives are below the median Above median 0.86 0.03 2. 96 3.25 0.98 0.03 3.89 0.94 Below median 0.88 0.14 2. 27 0.87 0.88 0.08 2.58 1.44 CEO's short-term equity incentives are below the 25th percentile Above median 0.87 0.05 2. 76 2.08 0.98 0.02 3.98 1.00 Below median 0.85 0.13 2. 15 1.11 0.86 0.06 2.63 1.77

PAGE 141

141 Table 8-1 Continued Reputation incentives Growth sample Profit sample Hit rate False alarm rate Detection ability Bias Hit rate False alarm rate Detection ability Bias Audit committee reputation incentives Independent of the CEO's short-term incentives Above median 0.87 0.10 2. 42 1.26 0.96 0.02 3.88 2.38 Below median 0.88 0.09 2. 54 1.21 0.92 0.04 3.17 1.66 CEO's short-term equity incenti ves are above the 75th percentile Above median 0.90 0.06 2. 88 1.54 0.94 0.02 3.63 2.50 Below median 0.82 0.14 1. 97 1.18 0.97 0.01 4.08 2.13 CEO's short-term equity in centives are above the median Above median 0.87 0.10 2. 39 1.20 0.95 0.02 3.75 2.49 Below median 0.92 0.06 2. 98 1.30 0.96 0.02 3.93 2.22 CEO's short-term equity incentives are below the median Above median 0.87 0.09 2. 47 1.30 0.90 0.06 2.81 1.45 Below median 0.87 0.07 2. 58 1.50 0.95 0.02 3.83 2.51 CEO's short-term equity incentives are below the 25th percentile Above median 0.86 0.08 2. 48 1.49 0.85 0.07 2.47 1.69 Below median 0.86 0.08 2. 49 1.44 0.95 0.03 3.57 1.52 Note: CEO's shortterm equity incentives Audit committee equity incentives Audit committee reputation incentives Extreme high (75 percentile) 2.30% 1.11% 1.60 Median 0.95% 0.37% 0.67 Extreme low (25th percentile) 0.17% 0.12% 0.00 Audit committee equity incentives (equi ty holdings/total share outstanding) CEO incentives (exercisable op tion/total shares outstanding) Audit committee reputation incentives (outside directorships)

PAGE 142

142 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th Percentiled' Audit Committee Equity Holdings Above Median Audit Committee Equity Holdings Below Median Figure 8-1. Growth Sample: d` Audit committee equity holdings, and CEO short-term equity incentives

PAGE 143

143 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th PercentileBia s Audit Committee Equity Holdings Above Median Audit Committee Equity Holdings Below Median Figure 8-2. Growth Sample: Bias, audit committe e equity holdings, and CEO short-term equity incentives

PAGE 144

144 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th Percentiled' Audit Committee Directorships Above Median Audit Committee Directorships Below Median Figure 8-3. Growth Sample: d`, audit committee directorships, and CEO short-term equity incentives

PAGE 145

145 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th PercentileBia s Audit Committee Directorships Above Median Audit Committee Directorships Below Median Figure 8-4. Growth Sample: Bias, audit committee directorships, and CEO short-term equity incentives

PAGE 146

146 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th Percentiled' Audit Committee Equity Holdings Above Median Audit Committee Equity Holdings Below Median Figure 8-5. Profit Sample: d`, audit committee eq uity holdings, and CEO short-term equity incentives

PAGE 147

147 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th PercentileBia s Audit Committee Equity Holdings Above Median Audit Committee Equity Holdings Below Median Figure 8-6. Growth Sample: Bias, audit committe e equity holdings, and CEO short-term equity incentives

PAGE 148

148 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th Percentiled' Audit Committee Directorships Above Median Audit Committee Directorships Below Median Figure 8-7. Profit Sample: d`, audit committee directorships, and CE O short-term equity incentives

PAGE 149

149 0.00 0.50 1.00 1.50 2.00 2.50 3.00 CEO STEI above the 75th Percentile CEO STEI above the Median CEO STEI below the Median CEO STEI below the 25th PercentileBia s Audit Committee Directorships Above Median Audit Committee Directorships Below Median Figure 8-8. Profit Sample: Bias, audit committee directorships, and CEO short-term equity incentives

PAGE 150

150 CHAPTER 9 CONCLUSIONS AND FUTURE WORK In the prior chapters, I fo cused upon understanding the re lationship between the audit committee decision-making process and audit co mmittees willingness to allow aggressive reporting. In Chapter 2, I presented a mode l of audit committee decision-making based upon signal detection theory. From this model, I iden tified numerous factors that may affect the audit committees decision-making process; many not prev iously documented. In Chapters 4 through 7, I presented extensive empirical evidence to validate the models inference that incentives influence the committees judgment. I found that both audit committee equity and reputation incentives were positively associ ated with the likelihood the firm meets an earning threshold but the influence is moderated by the risk environm ent. Unexpectedly, I found no evidence that audit committee ability was related to the likelihood that a firm met an earnings threshold. This is consistent with miss and meet firms having simila r attributes. In Chapter 8, I re-interpreted the evidence to determine whether incentives bi as the committee judgm ent or motivate the committee to be more diligent. I found that au dit committee incentives both bias and motivate the committee. Collectively, this research provides strong eviden ce that researchers need to consider more thoroughly the audit committees decisi on-making process. I propose three avenues to continue the e xploration of the audit committee decisionmaking process. The first avenue considers th e influence of audit committee incentives in an alternative setting. The second avenue consider s an additional audit committee incentive. The third avenue considers how group dynamics in fluence the decision-making process. Audit Committee Incentives and Seasoned Equity Offerings In the prior chapters, I examined the influen ce of audit committee incentives when it is uncertain whether or not the manager opportunistica lly reported. The uncertainty is a critical

PAGE 151

151 element because it meant that investors ha d to rely upon the audit committee to prevent opportunistic reporting. However, absent this uncertainty, it is unclear how audit committee incentives would influence the committees deci sion and, ultimately, investor wealth. For example, consider the research by Shivakum ar [2003] on earnings management and seasoned equity offerings. Shivakumar [2003] examin ed managerial and i nvestor behavior around seasoned equity offerings. Using a game theory framework, the author found a Nash equilibrium where the manager artificially inflated earni ngs to maximize the yi eld and the investor discounted reported earnings to counter any earnings management This equilibrium existed because managers cannot signal investors whethe r they are truthfully or opportunistically reporting. The author found empirical evidence that supports his model. Absent from the Shivakumar model was an act ive role for the audit committee. In the presence of an audit committee, the equilibriu m condition becomes significantly more complex to predict. Specifically, an investors valua tion decision depends upon th e investors perception of audit committee effectiveness and actual audit committee effect iveness. Assuming that each of these variables has only two states (effective or ineffective), there are four potential outcomes. Further, assume that if investors perceive the audit committee is effective, then they conclude that earnings are not inflated. Lastly, assume that managerial behavior does not depend upon perceived audit committee effectiveness (i.e. the manager always opportunistically report). Under these assumptions, the following outcomes result: Earnings Initially Reported To Audit Committee Audit Committee Response Earnings Reported To Investors Investor Perception of the Audit Committee Investor Response to Reported Earnings Investor Valuation Decision Inflated Reject Aggressive Reporting AccurateEffectiveAcceptAccurate Inflated Reject Aggressive Reporting AccurateIneffectiveDiscount Undervalue Inflated Allow Aggressive Reporting InflatedEffectiveAcceptOvervalue Inflated Allow Aggressive Reporting InflatedIneffectiveDiscount Accurate

PAGE 152

152 As seen by these theoretical outcomes, th e investors response should depend upon their perceptions of the audit committee and the a udit committee willingness to allow aggressive reporting. This raises three criti cal issues. First, what are the attributes of an audit committee that investors perceive as critical to audit co mmittee effectiveness? Second, do audit committee equity incentives motivate the audit committee to allow aggressive reporting in order to maximize the yield of a seasoned equity offering ? Third, does the investor valuation decision depend upon perceived audit committee effectiven ess and/or audit committee incentives? In other words, are effective audit committees ac ting in the shareholde rs best interests? Compensation Does director compensation improve director performance? While there is extensive literature on the relationship be tween CEO compensation and firm performance, I know of no research that explored the relationship between director compensation and audit committee effectiveness. The lack of research may be partially attributable to the less detailed proxy disclosures in the pre-SOX era. However, in re cent years, the details of director compensation included in corporate disclosures have increased dramatically. While director compensation is likely immaterial relative to CEO compensation, directors are potentially paid millions in cash and equity for serving on several boards. Th is raises several important issues including: What is the association between director compensation and aud it committee effectiveness? Does excessive equity compensation cause audit committee members to cease focusing on being an effective monitor and instead focus on a maximizing stock value? Is the annual director compen sation of all directorships a mo re effective proxy for reputation than the number of directorships held? Group Dynamics In my model, I identified group dynamics as a critical element of audit committee potential, yet there is little empirical research on how group dynamics influence audit

PAGE 153

153 committees. Group dynamics refers to the unique effects created by combing the opinions and analyses of several individuals into a committee decision. The potential research topics are extremely diverse, including disc ussion strategies, minor ity influence, group effects, and member diversity. Consider how the interacti on between non-experts and experts may influence audit committee effectiveness. Sarbanes-Oxley required that every audit committee have a financial expert. While it is generally accepted that expert s have a greater ability to understand issues and interpret evidence, it is unclear how group dynami cs influence the impact of experts on audit committee effectiveness. An expert influenc e likely depends upon three factors: recognition, social skills, and non-expert response. It is critical that the member s correctly identify who are th e experts so they can assign appropriate weight to their opin ions (Libby, Trotman, Zimmer [ 1987]). Experts are identified by talkativeness (Littlepage and Mueller [1997]), confidence, (Zarnoth and Sniezek [1997]), and reputation (Yaniv and Kleinberger [2000]). Audit committee members may fail to identify experts because all members of audit committees at least for public companies, likely have distinguished rsums, extraordinary leadership ability, and excellent reputations, but only a fraction may have the technical trai ning. Even if a member meets the standard to be an expert, they may still lack the technical skills, given th e lax SOX definition of expertise (Defond et al. [2005]). While experts may have greater knowledge and e xperience, the influence of experts will be constrained by the members social skills (Ferris, Witt, and Hochwa rter [2001]). Social skill is the ability to understand the feelings, thoughts, and behaviors of persons, including oneself, in interpersonal situations and to act appropriate ly upon that understandi ng (Marlowe [1986]).

PAGE 154

154 Hence, an experts ability to persuade, explain, and influence depends upon their social skills (Argyle [1969]). Ferris, Witt, and Hochwarter [ 2001] found that if a person has low social skills, then ability is irrelevant. Therefore, if an expe rt on the audit committee has the technical skills but lacks the social skills, then they may be una ble to articulate their position effectively and persuade other members. Lastly, the benefits of experts may depe nd upon how non-experts respond to the presence of an expert. In the presence of an expert, non-experts may reduce thei r contribution if they defer to the experts judgment without questi on (i.e. excessive deference) (Taylor and Fiske [1978]), if they believe experts will ignore their contribution (Yan iv [2004]), or if they conform to Sanherderins anti-seniority rule.57 This is potentially harmful to the committees effectiveness because non-experts can still contribute to a discussion by focusing on different issues (McDaniel et al. [2002]). 57 Sanherderins anti-seniority rule claims less qualified me mbers do not disagree with more qualified members.

PAGE 155

155 APPENDIX Variable Measurement Audit Committee Incentives AC direct holdings 100 times the average audit committee member personal stockholding divided by total shares outstanding at the end of fiscal year as specified in the proxy statement AC indirect holdings 100 times the average audit committee member deemed stock ownership not personally owned divided b y total shares outstandin g at the end of fiscal y ear as s p ecified in the p rox y AC total holdings 100 times the average audit committee me mber total stock ownership divided by total shares outstanding at the end of fiscal year as specified in the proxy statement AC directorships Average number of other directorship held by audit committee members at the end of fiscal year Audit Committee Controls Committee Size committee size as sp ecified in the proxy statement Meetings meeting frequency as sp ecified in the proxy statement Charter 1 if the proxy statement discloses the pr esence of an audit co mmittee written charter, 0 otherwise Ind. acc. expert 1 if there is at least one independe nt member has a professional accounting designation or degree, or experience working in an senior le vel financial position based on information in the proxy statement, 0 otherwise Tenure average years of service by audit committee members as specified in proxy statement Wholly independent 1 if the committee is composed entirely of independent me mbers according to the SOX definition of independent based on information in the proxy statement, 0 otherwise AC ability 1 if the audit committee has all independent members, at least 3 members, and an average tenure greater than the sample median, 0 otherwise. Board ability 1 if the board of directors has 3 of follo wing attributes: below sample median board size, 50 percent independent members, member's average at least 1 outside directorship, less than 20 percent of the board have tenure less than 3 years, CEO-Chair of board separation, or nominating committee with only inde pendent members, 0 otherwise. CEO Incentives CEO STEI 100 times the C.E.O.'s exercisable stock options divided by total shares outstanding at the end of fiscal year CEO LTEI 100 times the sum of C.E.O.'s stock holdings and non-exercisable stock options divided by total shares outstanding at the end of fiscal year Control Variables Size The natural log of assets (milli ons) at the end of the fiscal year Market-to-book market value of equity divided by the book value of equity at th e end of the fiscal year Leverage total long-term debt divided by total assets at the end of the fiscal year Litigation risk 1 if the company's industry is litigation risk belong to SIC codes 28336, 357077, 737074, 3600674, and 520061 (Matsumoto 2002)., 0 otherwise Loss For the firms with earnings near the previous year's level or the analyst sample, 1 if the company's EPS is less than $0.00, 0 otherwise. For firms with earnings near zero, 1 if the company's EPS last year was less than $0.00, 0 otherwise. No Growth 1 if the company's current earnings is less than the prior year's earnings, 0 otherwise

PAGE 156

156 LIST OF REFERENCES ABBOTT, L.; Y. PARK; AND S. PARKER. The Effects of Audit Committee Activity and Independence on Corporate Fraud. Managerial Finance 26 (2000): 55-67. ABBOTT, L.; S. PAKER; AND G. PETERS. Audit Committee Characteristics and Restatements. Auditing: A Journal of Practice and Theory 23 (2004): 69-87. AGRAWAL, A., AND S. CHANDHA. Corporate Governance and Accounting Scandals. Journal of Law and Economics 45 (2005): 371-406. AIER, J.; J. COMPRIX; M. GUNL OCK; AND D. LEE. The Financial Expertise of CFOs and Accounting Restatements. Accounting Horizons 19 (2005): 123-135. ANDERSON, R.; S. MANSI; AND D. REEB. Board Characteristics, Accounting Report Integrity, and the Cost of Debt. Journal of Accounting and Economics 37 (2004): 315-342. ARCHAMBEAULT, D., AND DEZOORT, F. Auditor Opinion Shopping and the Audit Committee: An analysis of Su spicious Auditor Switches. International Journal of Auditing 5 (2001): 33-52. ARGYLE, M. Social Interaction. 19 69. New York: Atherton Press. ASHBAUGH, H.; R. LAFOND; AND B. MAYHEW. Do Nonaudit Services Compromise Auditor Independence? Further Evidence. The Accounting Review 78 (2003): 611-639. ASHBY, G., AND W. MADDOX. Complex Deci sion Rules in Categorization: Contrasting Novice and Experienced Performance. Journal of Experiment al Psychology: Human Perception and Performance 18 (1992): 50-71. BARKAN, R.; D. ZOHAR; AND I. EREV. Acc idents and Decision-Making under Uncertainty: A Comparison of Four Models. Organizational Behavior and Human Decision Processes 74 (1998): 118-144. BARTH, M.; J. ELLIOTT; AND M. FINN. Marke t Rewards Associated with Patterns of Increasing Earnings. Journal of Accounting Research 37 (1999): 387-413. BARTOV, E., AND P. MOHANRAM. Private Information, Earnings Manipulations and Executive Stock Option Exercises. The Accounting Review 79 (2004): 889-892. BEASLEY, M. An Empirical Analysis of the Relation between the Board of Director Composition and Financial Statement Fraud. The Accounting Review 71 (1996): 443-466. BEASLEY, J.; T. CARCELLO; D. HERMANSON ; AND P. LAPIDES. Fraudulent Financial Reporting: Consideration of Industry Traits and Corporat e Governance Mechanisms. Accounting Horizons 14 (2000): 441-454.

PAGE 157

157 BDARD, J.; S. MARAKCHI-CHTOUROU; AND L. COURTEAU. The Effect of Audit Committee Expertise, Independence, and Activ ity on Aggressive Earnings Management. Auditing: A Journal of Practice and Theory 23 (2004): 13-35. BLACK, B.; B. CHEFFINS; AND M. KLAUSNER. Liability Risk for Outside Directors: A Cross Border Analysis. European Financial Management 1 (2005): 153-171. BLOCHER, E.; R. MOFFIE; AND R. ZMUD. Report Format and Task Complexity: Interaction in Risk Judgments. Accounting, Organizations and Society 11 (1986): 457-470. BLUE RIBBON COMMITTEE ON IMPROVING THE EFFECTIVENESS OF CORPORATE AUDIT COMMITTEES. Report and R ecommendations of the Blue Ribbon Committee on I mproving the E ffectiveness of C orporate A udit C ommittees. New York: New York Stock Exchange and the National Associat ion of Securities Dealers, 1999. BRADBURY, M.; Y. MAK; AND S. TAN. B oard Characteristics, Audit Committee Characteristics and Abnormal Accruals. Work ing Paper at http:// ssrn.com/abstract=535764, 2004. BRAIOTTA, L. The Audit Committee Handbook, Fourth Edition. Hoboken, NJ: Wiley, 2004. BRICKLEY, J.; J. COLES; AND R. TERRY. Outsi de Directors and th e Adoption of Poison Pills. Journal of Financial Economics 35 (1994): 371-90. BROWN, C. Human Information Processing for Decisions to Investigate Cost Variances. Journal of Accounting Research 19 (1981): 62-85. BURGSTAHLER, D., AND I. DICHEV. Earnings Management to Avoid Earnings Decreases and Losses. Journal of Accounting and Economics 21 (1997): 99-126. BURGSTAHLER, D., AND M. EAMES. Earnings Ma nagement to Avoid Losses and Earnings Decrease: Are Analysts Fooled? Contemporary Accounting Research 20 (1998): 253-294. BURNS, N., AND S. KEDIA. The Impact of Performance-Based Compensation on Misreporting. Journal of Financial Economics 79 (2004): 35-67. CANEGHEM, T. The Impact of Audit Quality on Earnings RoundingUp Behaviour: Some UK Evidence. European Accounting Review 13 (2004): 771-786 CARCELLO, J., AND T. NEAL. Audit Comm ittee Composition and Auditor Reporting. The Accounting Review 75 (2000): 453-467. CARCELLO, J., AND T. NEAL. Audit Committ ee Characteristics and Auditor Dismissals Following New Going-Concern Reports. The Accounting Review 78 (2003): 995-117.

PAGE 158

158 CARCELLO, J.; C. HOLLINGSWORTH; A. K LEIN; AND T. NEAL. Audit Committee Financial Expertise, Competing Corporat e Governance Mechanisms, and Earnings Management. Working Paper at SSRN: http://ssrn.com/abstract=887512 2006. CARCELLO, J.; T. NEAL; Z.-V. PALMROSE; AND S. SCHOLZ. CEO Involvement in Selecting Board Members and Audit Committ ee Effectiveness (August 2, 2006). Working Paper at SSRN: http://ssrn.com/abstract=922301 2006. CHENG, Q., AND T. WARFIELD. Equity In centives and Earnings Management. The Accounting Review 80 (2005): 441-476. CIALDINI, R. Influence: The New Psychol ogy of Modern Persuasion. New York: Quill Publishing, 1985. COLAROS, P., AND L. ANDERSON. Effect of Perceived Expertness upon Creativity of Members of Brainstorming Groups. Journal of Applied Psychology 53 (1969): 1159-1163. CORE, J.; R. HOLTHAUSEN; AND D. LARCKER. Corporate Governance, Chief Executive Officer Compensation, and Firm Performance. Journal of Financial Economics 51 (1999): 371 406. DAVIS, J.; P. ZARNOTH; L. HULBERT; X. CHEN; C. PARKS; AND K. NAM. The Committee Charge, Framing Interpersonal Ag reement, and Consensus Models of Group Quantitative Judgment. Organizational Behavior and Human Decision Processes 72 (1997): 137-157. DAVIS, J. Some Compelling Intuitions about Group Consensus Decisions, Theoretical and Empirical Research. Organizational Behavior and Human Decision Processes 52 (1992): 3-38. DE DREU, C., AND N. DEVRIES. Differentia l Processing and Attitude Change Following Majority and Minority Arguments. British Journal of Social Psychology 35 (1996): 77-90. DE DREU, C., AND M. WEST. Minority Dissent and Team Innovation: The Importance of Participation in Decision-Making. Journal of Applied Psychology 86 (2001): 1191-1201. DECHOW, P.; S. RICHARDS ON; AND I. TUNA. Why Ar e Earnings Kinky?: An Examination of the Earnings Management Explanation. Review of Accounting Studies 8 (2003): 355-384. DECHOW, P.; R. SLOAN; AND A. SWEENEY. Causes and Consequences of Earnings Manipulations: An Analysis of Firms Subj ect to Enforcement Actions by the SEC. Contemporary Accounting Research 13 (1996): 1-36. DEFOND, M.; R. HANN; AND X. HU. Does the Market Value Financial Expertise on the Audit Committees of Boards of Directors?. Journal of Accounting Research 43 (2005): 154194.

PAGE 159

159 DEFOND, M., AND J. JIAMBALVO. Incidence and Circumstances of Accounting Errors. The Accounting Review 66 (1991): 643-655. DEFOND, M., AND J. JIAMBALVO. Debt C ovenant Violation a nd Manipulation of Accruals. Journal of Accounting and Economics 17 (1994): 145-176. DEGEORGE, F.; J. PATEL; AND R. ZECKHAUS ER. Earnings Management to Exceed Thresholds. Journal of Business 72 (1999) 1-33. DENIS, D.; D. DENIS; AND A. SARIN. Ownership Structure and Top Executive Turnover. Journal of Financial Economics 45 (1997): 193-221. DEZOORT, F. An Investigation of Audit Committees' Oversight Responsibilities. Abacus 33 (1997): 208-227. DEZOORT, F. An Analysis of Experience Effects on Audit Committee Members' Oversight Judgments. Accounting, Organizations and Society 23 (1998): 1-22. DEZOORT, F.; D. HERMANSON; D. ARCH AMBEAULT; AND S. RE ED. Audit Committee Effectiveness: A Synthesis of the Em pirical Audit Committee Literature. Journal of Accounting Literature 21 (2002): 38-75. DHALIWAL, D.; V. NAIKER; AND F. NAVISSI. Audit Committee Financial Expertise, Corporate Governance and Accruals Quality: An Empirical Analysis (May 2006). Working Paper at SSRN: http://ssrn.com/abstract=906690 2006. DRISKELL, J., AND E. SALAS. G roup Decision-Making under Stress. Journal of Applied Psychology 76 (1991): 473-478. DUNN, K., AND B. MAYHEW. Audi t Firm Industry Specialization and Client Disclosure Quality. Review of Accounting Studies 9 (2004): 35-58. DURLACH, N.; L. BRAIDA; AND Y. ITO. Towar d a Model for Discrimination of Broadband Signals. Journal of the Acoustical Society of America 80 (1986): 63-72. EFENDI, J.; A, SRIVASTA; AND E. SWANSON. Why Do Corporat e Managers Misstate Financial Statements? The Role of Option Co mpensation, Corporate Governance, and Other Factors. Working paper, Texas A&M University, 2004. EISENBERG, T.; S. SUNDGREN; AND M. WELLS. Larger Boar d Size and Decreasing Firm Value in Small Firms. Journal of Financial Economics 48 (1998): 35. FAMA, E. Agency Problem and the Theory of the Firm. Journal of Political Economy 88 (1980): 288-308.

PAGE 160

160 FAMA, E., AND M. JENSEN. Separat ion of Ownership and Control. Journal of Law and Economics 26 (1983): 301. FARBER, D. Restoring Trust After Fra ud: Does Corporate Governance Matter?. The Accounting Review 80 (2005): 539-561. FELO, A.; S. KRISHNAMURTHY; AND S. SOLIER I. Audit Committee Characteristics and the Perceived Quality of Financial Reporting: An Empirical Analysis. Working Paper at http://ssrn.com/abstract=401240 2003. FERRIS, S.; M. JAGANNATHAN; AND A. PRITCHARD. Too Busy to Mind the Business? Monitoring by Directors with Mu ltiple Board Appointments. The Journal of Finance 58 (2003) 1087-1111. FORKER, J. Corporate Governan ce and Disclosure Quality. Accounting and Business Research 22 (1992):111-124. FRANCIS, J.; R. LAFOND; P. OLSSON; AND K. SCHIPPER. Earnings Quality and the Pricing Effects of Earnings Pa tterns. Working Paper at SSRN: http://ssrn.com/abstract=414142 or DOI: 10.2139/ssrn.414142, 2003. FRANKEL, R.; M. JOHNSON; AND K. NELSON. The Relation between Auditors' Fees for Nonaudit Services and Earnings Management. The Accounting Review 77 (2002): 71-105. GENDRON, Y., AND J. BDARD. G etting Inside the Black Box: A Field Study of Practices in Effective Audit Committees. Auditing: A Journal of Practice and Theory 23 (2004): 153-171. GENDRON, Y., AND BEDARD, J. On the Cons titution of Audit Committee Effectiveness. Accounting, Organizations, and Society 31 (2006): 211-239. GILSON, S. Bankruptcy, Boar ds, Banks, and Blockholders. Journal of Financial Economics 27 (1990): 315-355. GODFREY, J.; P. MATHER; AND A. RAMSAY. Earnings and Impression Management in Financial Reports: The Ca se of CEO Changes. Abacus 39 (2003): 95-123. GOYAL, V., AND C. PARK. Board Leader ship Structure and CEO Turnover. Journal of Corporate Finance 8 (2002): 49. GUAY, W. The Sensitivity of CEO Wealth to E quity Risk: An Analysis of the Magnitude and Determinants. Journal of Financial Economics 53 (1999): 43-71. GUL, F., AND S. LEUNG. Board Leadership, Outside Directors Expertise and Voluntary Corporate Disclosures. Journal of Accounting and Public Policy 23 (2004): 351.

PAGE 161

161 HACKENBRACK, K., AND M. NELSON. Auditors Incentives and their Application to Financial Accounting Standards. The Accounting Review 71 (1996): 43-60. HARRIS, J., AND P. BROMILEY Incentives to Cheat: The Influence of Executive Compensation and Firm Performance on Financ ial Misrepresentation. Working paper, University of Minnesota, 2005. HARVEY, N., AND I. FISCHER. Taking Advi ce: Accepting Help, Improving Judgment and Sharing Responsibility. Organizational Behavior and Human Decision Processes 70 (1997): 117. HEALY, P., AND J. WAHLEN. A Review of th e Earnings Management Literature and its Implications for Standard Setting. Accounting Horizons 13 (1999): 365-383. HERMALIN, B., AND M. WEISBACH. The E ffects of Board Composition and Direct incentives on Firm Performance. Financial Management 20 (1991): 101-12. HOBALLAH, I., AND P. VARSHNEY. Distr ibuted Bayesian Si gnal Detection. IEEE transition on information theory 35 (1989): 995-1000. JACKSON, S. Team Compositi on in Organizations. In S. Worchel, W. Wood, and J. Simpson (eds.): Group Process and Productivity (1992): 1-12. JANIS, I. Groupthink. Boston, MA: Houghton Mifflin Company, 1982. JENSEN, M. The Modern Industrial Revolution, Exit, and the Failure of Internal Control Systems. Journal of Finance 48 (1993): 831-880. JOHNSON, T.; D. BUDESCU; AND T. WALLS TEN. Averaging Probability Judgments: Monte Carlo Analyses of Asym ptotic Diagnostic Values. Journal of Behavioral DecisionMaking 14 (2001): 123-140. KALBERS, P., AND T. FOGARTY. Audit Co mmittee Effectiveness: An Empirical Investigation of the C ontribution of Power. Auditing: A Journal of Practice and Theory 13 (1993): 24-49. KAPLAN, S., AND D. REISHUS. Outside Dir ectorships and Corporate Performance. Journal of Financial Economics 27 (1990): 389-410. KARAMANOU, I., AND N. VAFEAS. The Association between Corporate Boards, Audit Committees and Management Earnings Fo recasts: An Empirical Analysis. Journal of Accounting Research 43 (2005): 453-486. KARPOFF, J.; D. LEE; AND G. MARTIN. The Cost of Cooking Books. Working paper, University of Washington, 2004.

PAGE 162

162 KARPOFF, J.; P. MALATESTA; AND R.WALKING. Corporate Governance and Shareholder Initiatives: Empirical Evidence. Journal of Financial Economics 42 (1996): 365-395. KIRK, D. Experiences with the Public Over sight Board and Corporate Audit Committees. Accounting Horizons 14 (2000): 103. KLEIN, A. Economic Determinants of Audit Committee Independence. The Accounting Review 77 (2002): 435-452. KOSNIK, R. Effects of Board Demography and Directors' Incen tives on Corporate Greenmail Decisions. Academy of Management Journal 33 (1990): 129-150. KRISHNAN, J. Audit Committ ee Quality and Internal Control: An Empirical Analysis. The Accounting Review 80 (2005): 649-675. KRISHNAN, J., AND J. KRISHNAN. Litig ation Risk and Auditor Resignation. The Accounting Review 72 (1997): 539-560. LEE, T., AND M. STONE. Economic Agency and Audit Committees: Responsibilities and Membership Composition. International Journal of Auditing 1 (1997): 97-116. LIBBY, R.; K.T. TROTMAN; AND I. ZIMMER. M ember Variation, Recognition of Expertise, and Group Performance. Journal of Applied Psychology 72 (1987): 81-87. LITTLEPAGE, G., AND A.L. MUELLER. Recogni tion and Utilization of Expertise in Problem-Solving Groups: Expert Char acteristics and Behavior. Group Dynamics 1(1997): 324328. LIU, C. The Influence of Audit Committee Ov ersight on Corporate Disclosures. Working paper, 2004. LODERER, C., AND U. PEYER. Board Over lap, Seat Accumulation and Share Prices. European Financial Management 8 (2002): 165. LORSCH, J., AND E. MACIVER. Pawns or Pote ntates: The Reality of America's Corporate Boards. Boston: Harvard Business School Press, 1989. MARLOWE, H. Social Intelligence: Eviden ce for Multidimensionality and Construct Independence. Journal of Educational Psychology 78 (1986): 52-58. MATSUMOTO, D. Managements Incentives to Avoid Negative Earnings Surprises. The Accounting Review 77 (2002): 483-514. MATZ, D., AND W. WOOD. Cognitive Dissona nce in Groups: The Consequences of Disagreement. Journal of Personalit y and Social Psychology 88 (2005): 22-37.

PAGE 163

163 MCDANIEL, L.; R. MARTIN; AND L. MAINES. E valuating Financial Reporting Quality: The Effects of Financial Expertis e vs. Financial literacy. The Accounting Review 77 (2002): 139167. MCMULLEN, D., AND K. RAGHU NANDAN. Enhancing Audit Committee Effectiveness. Journal of Accountancy 182 (1996): 79. MCMULLEN, D. Audit committee performance: An Investigation of the Consequences Associated with Audit Committees. Auditing: A Journal of Practice and Theory 15 (1996): 87103. MENON, K., AND D. WILLIAMS. Former Aud it Partners and Abnormal Accruals. The Accounting Review 79 (2004): 1095-1118. MENON, K., AND J. WILLIAMS. The Use of Audit Committees for Monitoring. Journal of Accounting and Public Policy 13 (1994): 121-139. MILGROM, P., AND J. ROBERTS. Economics, Organization, and Management. PrenticeHall, Englewood Cliffs: NJ, 1992. MILLSTEIN, I. Oversight Heari ng on Accounting and Investor Protection Issues Raised by Enron and Other Public Companies, US Senate Committee on Banking, Housing, and Urban Affairs, Washington, DC, available at: http://banking.senate.g ov/02_02hrg/022702/millstn.htm 2002. MORCK,R.; A. SHLEIFER; AND R. VISHNY Management Ownership and Market Valuation: An Empirical Analysis. Journal of Financial Economics 20 (1988): 291-315. MOSCOVICI, S., AND M. ZAVALLONI. The Group as a Polarizer of Attitudes. Journal of Personality and Social Psychology 12 (1969): 125-135. MYERS, J.; L. MYERS; Z. PALMROSE; AND S. SCHOLZ. The Length of Auditor-Client Relationships and Financial Statement Restatements. Working Paper at http://ssrn.com/abstract=740525 2005. MYERS, L., AND D. SKINNER. 2000. Earning s Momentum and Earnings Management. Working paper, University of Michigan, Ann Arbor, MI. NABAR, S.; Y. KIM; AND W. G. HENINGER. Earnings Misst atements, Restatements, and Corporate Governance. Working paper, 2006. NELSON, M.; J. ELLIOTT; AND R. TARPLEY. How are Earnings Managed? Examples from Auditors. Accounting Horizons 17 (2003): 17-35. NEMETH, C. Differential Contributions of Majority and Minority Influence. Psychological Review 93 (1986): 23-32.

PAGE 164

164 OLSON, J. How to Make Audit Committees More Effective. The Business Lawyer 54 (1999): 1097. PALMROSE, Z-V.; V. RICHARDSON; AND S. SC HOLZ. Determinants of Market Reactions to Restatement Announcements. Journal of Accounting and Economics 37 (2004): 59-89. PEASNELL, K.; P. POPE; AND S. YOUNG. Accrual Management to Meet Earnings Targets: UK Evidence Pre and Post Cadbury. The British Accounting Review 32 (2000): 415-445. PETRA, S. Do Outside Independent Dire ctors Strengthen Corporate Boards?. Corporate Governance 5 (2005): 55-64. PITT, H. Testimony Concerning th e Corporate and Auditing Acc ountability, Responsibility, and Transparency Act. House of Representatives Committee on Financial Se rvices. 107th Cong.,2nd sess. 20 March. Available at: http: //fina ncialservices.house.gov/media/pdf/032002hp.pdf, 2002. PUNCOCHAR, J., AND P. FOX. Confidence in Individual and Group D ecision-Making: When Two Heads Are Worse Than One. Journal of Educational Psychology 96 (2004): 582. RAGHUNANDAN, K., AND I. MCHUG H. Internal Auditors' Inde pendence and Interactions with Audit Committees: Challenge s of Form and Substance. Advances in Accounting 12 (1994): 313-333. RAGHUNANDAN, K.; D. RAMA; AND D. SCARBROUGH. Accounting and Audit Knowledge Level of Canadian Audit Co mmittee: Some Empirical Evidence. Journal of International Accounting, Auditing, and Taxation 7 (1996): 181-194. RAGHUNANDAN, K.; W. RE AD; AND D. RAMA. Audit Co mmittee Composition: Gray Directors and Interaction with Internal Auditing. Accounting Horizons 15 (2001): 105-118. RAMSAY, R., AND R. TUBBS. A nalysis of Diagnostic Tasks in Accounting Research using Signal Detection Theory. Behavioral Research in Accounting 17 (2005): 149-173. RANGAN, S. Earnings Management and the Pe rformance of Seasoned Equity Offerings. Journal of Financial Economics 56 (1998): 101-122. RICHARDSON, S.; I. TUNA; AND M. WU. Predi cting Earnings Management: The Case of Earnings Restatements. Working paper, Univer sity of Pennsylvania a nd Hong Kong University of Science and Technology, 2003. ROSENSTEIN, S., AND J. WYATT. Inside Directors, Board Effectiveness, and Shareholder Wealth. Journal of Financial Economics 44 (1997): 229-50. RUDER, D. Oversight Hearing on Accounting and Investor Prot ection Issues Raised by Enron and Other Public Companies. Senate Comm ittee on Banking, Housing, and Urban Affairs.

PAGE 165

165 107th Cong., 2nd sess. 12 February. Availabl e at: http: //www.senate.gov/_banking/02 02hrg/021202/ruder.htm., 2002. SAFDAR, I. Executive Stock Option Exercise, Earnings Management, and Stock Returns. Working paper, University of Rochester. New York, NY, 2003. SHIVAKUMAR, L. Do Firms Mislead Investor s by Overstating Earnings before Seasoned Equity Offerings?. Journal of Accounting and Economics 29 (2000): 339-371. SHIVDASANI, A., AND D. YERMACK. CEO I nvolvement in the Selection of New Board Members: An Empirical Analysis. Journal of Finance 54 (1999): 1829-1853. SHIVDASANI, A. Board Composition, Owner ship Structure and Hostile Takeovers. Journal of Accounting and Economics 16 (1993): 167-198. SHORT, H., AND K. KEASEY. Managerial Ow nership and the Performance of Firms: Evidence from the UK. Journal of Corporate Finance 5 (1999): 79-101. SKINNER, D., AND R. SLOAN. E arnings Surprises, Growth Expectations and Stock Returns. Review of Accounting Studies : 7 (2002): 289-312. SONG, J., AND B. WINDRAM. Benchmarking Audit Committee Effectiveness in Financial Reporting. International Journal of Auditing 8 (2004): 195-205. SORKIN, R., AND H. DAI. Signal Det ection Analysis of the Ideal Group. Organizational Behavior and Human Decision Processes 60 (1994): 1-13. SORKIN, R.; J. COHEN; C.J. HAYS; AND R. WE ST. Signal-Detection Analysis of Group Decision-Making. Psychological Review 108 (2001): 183-203. SORKIN, R.; R.WEST; AND D. ROBINSON. G roup Performance Depends on the Majority Rule. Psychological Science 9 (1998): 456-463. SPIRA, L. Audit Committees Begging the Question. Corporate Governance 11 (2002): 180188 SPRINKLE, G., AND R. TUBBS. The Effects of Audit Risk and Information Importance on Auditor Memory During Working Paper Review. The Accounting Review 73 (1998): 475-502. SWETS, J. Indices of Discrimination or Di agnostic Accuracy: Their ROCs and Implied Models. Psychological Bulletin 99 (1986): 100-117. SWETS, J.; W. TANNER; AND T. BIRDSALL. Decision Pr ocesses in Perception. Psychological Review 68 (1961): 301-340.

PAGE 166

166 TANNER, W., AND A. SWETS. The Human Use of Information. Transactions of IRE Professional Group Information Theory 4 (1954): 213-221. TANNER, W., AND T. BIRDSALL Definitions of d' and as Psychophysical Measures. Journal of the Acoustical Society of America 30 (1958): 922-928. TAYLOR, S., AND S. FISKE. Salience, Attent ion, and Attribution: Top of the Head Phenomena. In L. Berkowitz (Ed.): Advances in experimental social psychology 11(1978): 249288. TEOH, S.; I. WELCH; AND T. WONG. Earnings Management and the Underperformance of Seasoned Equity Offerings. Journal of Financial Economics 50 (1998): 63-99. THOMSON FINANCIAL. Gossary I/B/E/S Summary History Version 2.0, 2003. TRICKER, R. 1978. The Independent Director A Study of the N on-Executive Director and of the Audit Committee. London. UK. Tolle y with Deloitte, Haskins and Sells. TSUI, J.; B. JAGGI; AND F. GUL. CEO Domina tion, Growth Opportunities, and their Impact on Audit Fees. Journal of Accounting, Auditing and Finance 16 (2001): 189. UNITED STATES CONGRESS. Sarbanes-Oxley Act of 2002, 107th Congress of the United States of America, H.R. 3763, US Governme nt Printing Office, Wa shington, DC., 2002. VAFEAS, N. Length of Board Tenure and Outside Director Independence. Journal of Business Finance and Accounting 30 (2003): 1043-64. VAFEAS, N. Audit Committees, Boards, and the Quality of Reported Earnings. Contemporary Accounting Research 22 (2005): 1093-1122. VANCIL, R. A Look at CEO Succession. Harvard Business Review 65 (1987): 107-118. VERA-MUNOZ, S. Corporate Governance Refo rms: Redefined Expectations of Audit Committee Responsibilities and Effectiveness. Journal of Business Ethics 62 (2005): 115-127. WEISBACH, S. Outside Directors and CEO Turnover. Journal of Financial Economics 20 (1988): 431-60. WELLS, P. Earnings Manageme nt Surrounding CEO Changes. Accounting and Finance 42 (2002): 169-193. WILD, J. Managerial Accountability to Shareh olders: Audit Committees and the Explanatory Power of Earnings for Returns. The British Accounting Review 26 (1994): 353-374. WILD, J. The Audit Committee and Earnings Quality. Journal of Accounting; Auditing and Finance 11 (1996): 247-276.

PAGE 167

167 WU, M. Earnings Restatements: A Capital Mark ets Perspective. U npublished dissertation, New York University, 2002. XIE, B.; W. DAVIDSON; AND P. DADALT. Earnings Management and Corporate Governance: The Role of the Board and the Audit Committee. Journal of Corporate Finance 9 (2003): 295-316. YANG, J., AND J. KRISHNAN. Audit Committees and Quarterly Earnings Management. International Journal of Auditing 9 (2005): 201-219 YANIV, I. Receiving Other Peoples Advice: Influence and Benefit. Organizational Behavior and Human Decision Processes 93 (2004): 1. YANIV, I., AND R. HOGARTH. Judg mental Versus Statistical Prediction: Information Asymmetry and Combination Rules. Psychological Science 4 (1993) 58-62. YANIV, I., AND E. KLEINBERGE R. Advice Taking in Deci sion-Making: Egocentric Discounting and Reputation Formation. Organizational Behavior and Human Decision Processes 83 (2000): 260. YERMACK, D. Remuneration Retention and Re putation Incentives for Outside Directors. Journal of Finance 59 (2004): 2281-2308. YERMACK, D. Higher Market Valuation of Co mpanies with a Small Board of Directors. Journal of Financial Economics 40 (1996): 185. ZARNOTH, P.; L. HULBERT; X. CHEN; C. PARK S; AND K. NAM. The Committee Charge Framing Interpersonal Agreement and Consen sus Models of Group Quantitative Judgment. Organizational Behavior and Hu man Decision Processes 72 (1997): 137-157.

PAGE 168

168 BIOGRAPHICAL SKETCH Jason MacGregor originally from Sarnia, Ontario, Canada, came to the University of Florida after completing his undergraduate studies in business at the University of Windsor. He is known for his deep concern for colleagues, his willingness to learn to know new things, and his love for his family and His God. In the fa ll, he will begin his professional career as an assistant professor at Baylor University.