Measuring the Impact of Enterprise Resource Planning (Erp) Systems Through the Prism of Accounting Theory

MEASURING THE IMPACT OF ENTERPRISE RESOURCE PLANNING (ERP) SYSTEMS THROUGH THE PRISM OF ACCOUNTING THEORY

A dissertation submitted to the Kent State University Graduate School of Management in partial fulfillment of the requirements for the degree of Doctor of Philosophy

John J. Morris

February 2009

Dissertation written by

John J. Morris

B.S.BA., University of South Dakota, 1976

MBA, Kent State University, 2003

Ph.D., Kent State University, 2009

Approved by

______Chair, Doctoral Dissertation Committee Dr. Ran Barniv ______Members, Doctoral Dissertation Committee Dr. Indrarini Laksmana ______Dr. Alan Brandyberry ______

______Accepted by

______Doctoral Director, Graduate School of Management Dr. Jay Muthuswamy ______Dean, Graduate School of Management Dr. Fredrick Schroath

ACKNOWLEDGEMENTS

I would like to thank the members of my committee, Dr. Ran Barniv, Chair; Dr. Rini

Laksmana; and Dr. Alan Brandyberry for their guidance and support throughout the dissertation process. I would also like to thank participants at various workshops and conferences where one or more of the essays that make up this dissertation have been presented and for which helpful feedback was received. They include: the annual meeting of the American Accounting

Association in 2007, the Journal of Information Systems New Scholars Workshops held in conjunction with the mid-year meeting of the Information Systems Section of the American

Accounting Association in 2006 and 2007, the Kent State University Statistics Seminar in 2007, the Kent State University Graduate Student Senate Colloquium in 2007, and the Kansas State

University Faculty Research Seminars in 2007 and 2008. I would also like to thank Jennifer

Francis for her suggestions and comments about using the concept of earnings quality and the e- loading factor as a way to measure the impact of ERP systems during a breakout session at the

AAA/Deloitte/J. Michael Cook Doctoral Consortium at Tahoe City, California in June 2006.

On a personal level, I would like to thank my wife, Anna, for the sacrifices she made in giving up her home to live in married student housing for four years, and working in social services to supplement my doctoral student’s stipend so I could pursue this transition to an academic career. Her support and inspiration were critical during those difficult times when it would have been so easy to just give up and return to the more lucrative financial profession that had supported us for 25 years.

i TABLE OF CONTENTS

CHAPTER 1 INTRODUCTION ...... 1

1.1 BACKGROUND AND RESEARCH QUESTIONS ...... 1

1.2 AGENCY THEORY AS A FOUNDATION ...... 3

1.3 EARNINGS MANAGEMENT ...... 4

1.4 SHAREHOLDER VALUE ...... 5

1.5 INTERNAL CONTROL ...... 6

1.6 RESEARCH CONTRIBUTION ...... 7

1.7 SUMMARY ...... 8

CHAPTER 2 LITERATURE REVIEW ...... 10

2.1 INTRODUCTION ...... 10

2.2 ERP SYSTEMS RELATED LITERATURE ...... 10

2.2.1 Background on ERP Literature ...... 10

2.2.2 Behavioral Research ...... 11

2.2.3 Archival Research ...... 16

2.2.4 Research on Other ERP Issues ...... 21

2.2.5 Incremental Contribution ...... 26

2.3 AGENCY THEORY LITERATURE ...... 26

2.4 EARNINGS MANAGEMENT LITERATURE ...... 29

2.5 SHAREHOLDER VALUE LITERATURE ...... 33

2.6 INTERNAL CONTROL LITERATURE ...... 38

2.7 SUMMARY ...... 43

CHAPTER 3 EARNINGS MANAGEMENT (ESSAY ONE) ...... 44

3.1 INTRODUCTION ...... 44

3.2 PRIOR RESEARCH AND HYPOTHESES DEVELOPMENT ...... 47

ii 3.3 DATA SELECTION AND METHODOLOGY /M ODELS ...... 50

3.3.1 Sample Data Selection ...... 50

3.3.2 Discretionary Accrual Model ...... 53

3.3.3 Earnings Management Model...... 56

3.3.4 Earnings Quality Model ...... 58

3.3.5 Earnings Distribution Methodology...... 60

3.4 EMPIRICAL RESULTS ...... 61

3.4.1 Estimation of Discretionary Accruals ...... 61

3.4.2 Earnings Management Results ...... 63

3.4.3 Earnings Quality Results ...... 66

3.4.4 Earnings Distribution Results...... 68

3.5 CONCLUSIONS ...... 70

CHAPTER 4 SHAREHOLDER VALUE (ESSAY TWO) ...... 90

4.1 INTRODUCTION ...... 90

4.2 PRIOR RESEARCH AND HYPOTHESES DEVELOPMENT ...... 92

4.3 DATA SELECTION AND METHODOLOGY /M ODELS ...... 98

4.3.1 Sample Data Selection ...... 98

4.3.2 Long-Horizon Returns Model ...... 102

4.3.3 Market to Book (MTB) Model ...... 104

4.3.4 Value-to-Price (V/P) Model...... 105

4.4 EMPIRICAL RESULTS ...... 107

4.4.1 Long-Horizon Returns ...... 107

4.4.2 Market-to-Book (MTB) Analysis ...... 110

4.4.3 Value-to-Price (V/P) Analysis ...... 112

4.5 CONCLUSIONS ...... 114

CHAPTER 5 INTERNAL CONTROL (ESSAY THREE) ...... 134

iii 5.1 INTRODUCTION ...... 134

5.2 PRIOR RESEARCH AND HYPOTHESES DEVELOPMENT ...... 137

5.3 DATA SELECTION AND METHODOLOGY /M ODELS ...... 142

5.3.1 Sample Data Selection ...... 142

5.3.2 Logistic Regression Model ...... 144

5.4 EMPIRICAL RESULTS ...... 146

5.4.1 Frequencies of Internal Control Weaknesses ...... 146

5.4.2 Logistic Regression Analysis ...... 147

5.4.3 Factors Contributing to Material Weakness Determination...... 149

5.5 CONCLUSIONS ...... 153

CHAPTER 6 CONCLUSIONS ...... 174

6.1 CONCLUSIONS ...... 174

6.2 CONTRIBUTIONS ...... 176

6.3 LIMITATIONS ...... 178

6.4 FUTURE RESEARCH ...... 178

BIBLIOGRAPHY ...... 180

iv LIST OF FIGURES

Figure 3-1 Mean Absolute Value of Discretionary Accruals ...... 74

Figure 3-2 Mean Value of e-loading Factors Five Years Before and After Implementation ...... 75

Figure 3-3 Frequency of Earnings Intervals ...... 76

Figure 3-4 Frequency of Earnings Change Intervals ...... 77

Figure 4-1 Long-Horizon Returns ...... 118

Figure 4-2 Mean Market-to-Book Ratios ...... 119

Figure 4-3 Mean Value-to-Price Ratios ...... 120

v LIST OF TABLES

Figure 3-1 Mean Absolute Value of Discretionary Accruals ...... 74

Figure 3-2 Mean Value of e-loading Factors Five Years Before and After Implementation ...... 75

Figure 3-3 Frequency of Earnings Intervals ...... 76

Figure 3-4 Frequency of Earnings Change Intervals ...... 77

Table 3-1 Summary of Sample Selection Process ...... 78

Table 3-2 ERP Implementing Firms by 2 Digit SIC Code and Implementation Year ...... 79

Table 3-3 Descriptive Statistics used with Equations (1) and (4) ...... 80

Table 3-4 Descriptive Statistics for Absolute Value of Discretionary Accrual Variables ...... 81

Table 3-5 Test of Equal Means for Absolute Value of Discretionary Accruals ...... 82

Table 3-6 Pearson Correlation Matrix for Earnings Management Regressions...... 83

Table 3-7 Earnings Management Regression Results ...... 84

Table 3-8 Descriptive Statistics for Equation 12 ...... 85

Table 3-9 Descriptive Statistics for e-Loading Five Years Before and After ERP Implementation.... 86

Table 3-10 Pearson Correlation Matrix for E-Loading Regressions ...... 87

Table 3-11 e-Loading Regression Results ...... 88

Table 3-12 Frequency of Actual vs. Expected Earnings (Change) Just Below Zero...... 89

Figure 4-1 Long-Horizon Returns ...... 118

Figure 4-2 Mean Market-to-Book Ratios ...... 119

Figure 4-3 Mean Value-to-Price Ratios ...... 120

Table 4-1 Summary of Sample Selection Process ...... 121

Table 4-2 ERP Implementing Firms by 2 Digit SIC Code and Implementation Year ...... 122

Table 4-3 Summary of Long-Horizon Return Statistics ...... 123

Table 4-4 Pearson Correlation Matrix for Long-Horizon Returns OLS Regression ...... 124

Table 4-5 Long-Horizon Returns OLS Regression Results ...... 125

Table 4-6 Summary of Mean Market-to-Book (MTB) Values ...... 126

Table 4-7 Pearson Correlation Matrix for Market to Book Regression ...... 127

vi Table 4-8 Market-to-Book (MTB) Value OLS Regression Results ...... 129

Table 4-9 Value-to-Price (V/P) Statistics ...... 130

Table 4-10 Pearson Correlation Matrix for V/P Regression Variables ...... 131

Table 4-10 ...... 132

Table 4-11 Value-to-Price (V/P) Ratio OLS Regression Results ...... 133

Appendix 5-1 Audit Analytics’ Definition of Factors Leading to Internal Control Weakness ...... 156

Appendix 5-2 Example of an Audit Report - Pride International, Inc...... 163

Table 5-1 Summary of Sample Selection Process ...... 166

Table 5-2 ERP Implementing Firms by 2 Digit SIC Code and Implementation Year ...... 167

Table 5-3 Frequency of Internal Control Weaknesses ...... 168

Table 5-4 Logistic Regression Results ...... 169

Table 5-5 Frequency (Proportion) of Factors Contributing to Internal Control Weaknesses ...... 170

Table 5-6 Frequencies of Internal Control Weakness IT vs. Non-IT Factors...... 171

Table 5-7 Logistic Regression Results IT Related Factors ...... 172

Table 5-8 Logistic Regression Results Non-IT Related Factors ...... 173

vii

CHAPTER 1 INTRODUCTION

1.1 Background and Research Questions

Enterprise Resource Planning (ERP) systems emerged in the 1990s as one of the fastest growing and perhaps one of the most important developments in corporate use of information technology (Davenport 1998). These systems, which evolved from Material Requirements

Planning (MRP) and Manufacturing Resource Planning (MRP-II) systems (Umble et al. 2003), consolidate into one common system all the back office information processing needs of a typical business including: accounting/finance, human resources, operations, supply chain, and customer information (Davenport 1998). This common system design allows firms to capture information once and then share it across functional areas enabling “information congruence” (O'Leary 2002).

Vendors often cite this feature in their promotional material as noted in the following quote from

Oracle Corporation’s Web site 1:

Oracle E-Business Suite is your fastest path to high-quality enterprise intelligence, bringing your company a true 360-degree view of your finances, your customers, and your supply chains, so you can make faster, better decisions and grow profitability in a competitive marketplace (Oracle 2007).

ERP systems replaced older legacy systems that had been developed in most organizations along functional boundaries. The evolution of these legacy systems often resulted in duplicate information being collected and stored in various departments, business units, geographic regions, factories, offices, etc. (O'Leary 2002). These so called “silos of information” were inefficient because of the duplicate effort needed to maintain them, and because they created an environment in which data was not easily shared among various members of the management

1 Oracle Corporation is the 2 nd largest supplier of ERP systems, which they refer to as “Oracle E-Business Suite.”

1 2

team. These systems also presented internal control risks because there was no assurance that if data changed in one system, the same data would be updated in other systems on a timely basis.

For instance, if the sales order system does not interact with the production scheduling system, a sales order change will not be reflected in the production system, and therefore, the order could be filled incorrectly leading to numerous management problems. ERP systems address this issue by tightly linking all systems together with a common database. Once data is entered by one functional area, the same data is used by other functional areas to serve their needs. This common design feature has impacted all functional areas of business, especially the accounting function.

Most legacy accounting information systems were designed to automate manual processes developed in the 14 th Century by Luca Pacioli (Hollander et al. 2000). As a result, data from paper transactions (sales invoices, purchase orders, shipping documents, etc.), many of which were computer generated in the first place, were entered into accounting systems by re- keying the same data. With this approach, management was often left at the end of the month waiting for the accounting department to finish posting transactions so they could determine the financial results. In contrast, ERP systems use workflow automation to track information from sales order to cash collection, generating the necessary accounting transactions in the process

(Hunton et al. 2004). The result is timelier and higher quality accounting information available for management decision making at all levels of the organization on an almost continuous basis.

In spite of the significant impact that ERP systems have had on the accounting function, there has been very little ERP related academic research driven by accounting theory. In fact, most of the early ERP research has been more practice oriented, addressing issues such as key success factors and measuring ex ante productivity improvements, without relating them to any specific academic theory (Poston and Grabski 2001; Bradford and Florin 2003). This dissertation

addresses that gap in the literature by relying on a theoretical framework well established in accounting research, specifically agency theory (Lambert 2001), to formulate three research questions: (1) Does implementation 2 of an ERP system have an impact on the level of earnings management activity by the implementing firm? (2) Does implementation of an ERP system have an impact on shareholder value? and (3) Does implementation of an ERP system have an impact on a firm’s ability to provide an adequate system of internal controls as mandated by

Sarbanes-Oxley Section 404? These questions are addressed in three separate essays that use different methodologies from the accounting literature to measure the impact from three different perspectives.

1.2 Agency Theory as a Foundation

Each of these research questions can be framed in the context of agency theory. Lambert

(2001) points out that agency theory, with its roots in information economics literature, provides a framework for examining the link between information systems, incentives, and behavior. Many examples of agency relationships exist in business, including: the shareholder as principal and the manager as agent, the parent corporation as principal and the subsidiary as agent, or even the manager as principal and the employee as agent 3. Lambert (2001) argues that, although agency theory is used extensively in analyzing financial accounting and auditing issues, its largest contributions have been to managerial accounting, because financial measures of performance are calculated at all levels of an organization, not just at the firm-wide level. One could argue that in large firms, multiple agency levels exist, beginning with the shareholders and ending at all levels of employees where decisions are made on behalf of the firm. Eisenhardt (1989) suggests that

2 Implementation, as used throughout this dissertation, is assumed to be initial implementation or adoption, recognizing that the implementation process is often an ongoing continuous process in many firms. 3 Conyon and He (2004) use a three-tier model with shareholders, board of directors, and executives as the three layers which they describe as a principal-supervisor-agent structure.

“when a principal has information to verify agent behavior, the agent is more likely to behave in the interest of the principal.” Therefore, one could argue that ERP systems are designed to address the agency problem because they reduce information asymmetry by providing transparency throughout an organization. Given the impact that agency theory has throughout the organization, it plays a key role in developing testable hypotheses for all three of the research questions addressed in this dissertation. For instance, based on agency theory, increased information transparency should encourage agents at all levels to act in the interest of the principals, which should include limiting earnings management activities (research question #1).

Increased transparency should also provide better quality information for financial reporting to the shareholders (principals) by management (agents), which will ultimately be reflected in the price shareholders are willing to pay for stock thus improving their value (research question #2).

Finally, internal controls have evolved as the primary check on management’s (agent’s) fiduciary responsibility to shareholders (principals) and systems that enhance that control mechanism will aid in addressing the agency problem (research question #3).

1.3 Earnings Management

The first research question relates to earnings management, which has evolved in the accounting literature as a classic example of agency theory. Agency relationships exist at various levels throughout most organizations, with managers at each level acting as agents on behalf of the principal(s) at higher levels. Prior research indicates that earnings management activities take place at all levels of organizations (Nelson et al. 2003) and that greater transparency in income reporting reduces the likelihood that managers will engage in earnings management (Hunton et al. 2006). If ERP systems provide more transparency of information throughout the firm does the increased level of transparency following ERP system implementation discourage earnings

management activity, and thus, improve earnings quality? The first essay addresses this question using various proxies developed in prior accounting research to measure earnings management.

1.4 Shareholder Value

The second research question related to shareholder value has been linked to the explicit and intrinsic value of the firm by prior accounting research. Ohlson (1995) shows that a firm’s intrinsic value is equal to the book value of equity plus the present value of expected future residual income 4. Ohlson (1995), and Feltham and Ohlson (1995) provide a foundation for using the stock price as a proxy for this intrinsic value of the firm. Since ERP systems are designed to provide transparency of information and to provide more accurate and timely information for management decision making purposes, management should be making better decisions, which should lead to increased future residual income. It follows that if the benefits of an ERP system are realized in the form of improved performance, and that improved performance flows into residual income, the impact will be reflected in the firm’s intrinsic value. Also, assuming ERP systems provide timelier, more accurate, and more transparent data for financial reporting and management decision making, shareholders should experience increased confidence in the quality of the financial reports, which should be reflected in the price of the firm’s stock. Essay two addresses this research question using previously developed proxies to measure the impact on shareholder value as measured by return, market-to-book value, and the intrinsic (residual income) value.

4 Residual income, also referred to as abnormal earnings, is defined in the accounting literature as the difference between accounting earnings and a charge for the normal return on equity See Penman (2004) page 142 for a discussion of residual income.

1.5 Internal Control

Internal control and audit processes exist as a means of providing assurances to the shareholder (principal) that management (agent) has complied with all relevant laws and reporting requirements. One of the features included in ERP systems is the ability to control user access and facilitate the separation of duties, which is one of the most common internal control mechanisms used to deter fraud within financial systems. For instance, the ERP system by SAP 5 assigns profiles consisting of role-based access controls to users, which prevents them from performing incompatible activities (Little and Best 2003) 6. This is just one example of many features that are built into ERP systems based on best practices. This design should make it easier for firms to comply with Section 404 of The Sarbanes-Oxley Act of 2002 (SOX).

Motivated by corporate scandals at Enron, WorldCom and others, SOX was enacted to restore public confidence in corporate governance of publicly traded companies. Section 404 of the act addresses reporting on internal controls and requires companies to include in their annual report (Form 10-K) a separate management report on the company’s internal control over financial reporting 7 and an attestation report issued by a registered public accounting firm 8 (SEC

2003). Oracle and other ERP vendors are reporting increased sales of ERP and ERP related

5 SAP is the worldwide market leader in ERP software according to a comprehensive study by Gartner (SAP 2006) 6 As an example, someone who is assigned the role of recording accounts payable transactions would be prevented from issuing a purchase order. 7 The internal control report requires specific language including a statement of management’s assessment of the effectiveness of the company’s internal control over financial reporting, a statement identifying the framework used by management to evaluate the effectiveness, and a statement that the registered public accounting firm that audited the company’s financial statements included in the annual report has issued an attestation report on management’s assessment of the company’s internal control over financial reporting. 8 The attestation report must express an opinion on management’s assessment of the internal controls and on the overall effectiveness of the internal controls themselves.

software as a result of SOX9. Therefore, one would assume that firms that implemented ERP systems prior to SOX would find it easier to comply with the internal control requirements of

Section 404. Essay three addresses this research question by examining compliance reporting for firms that implemented ERP systems compared to firms that did not.

1.6 Research Contribution

This dissertation makes a significant contribution to the academic literature, because it addresses a topic that is timely and of interest not only to academics, but also to the practice community. It is a topic that has generated considerable discussion in both environments, as firms have spent major portions of their IT budgets on these systems. Yet, researchers have not associated this investment in ERP systems with significant benefits. By taking a different approach to measuring the impact of these systems, this dissertation provides new insights into this cost/benefit controversy.

As discussed in more detail in the next section, this dissertation is expected to fill a gap that exists in prior research related to ERP systems. Specifically, most prior research has focused on addressing practitioner driven questions related to implementation methodologies and cost/benefit economic analysis. Very little ERP research is driven by theory based academic discovery methods. This dissertation relies on well established theories and methodologies used in accounting research, to explore alternative methods of measuring the impact that ERP systems have on firms that implement them.

9 Fred Studer, vice president of ERP applications marketing at Oracle was recently quoted as saying “we’ve seen a lot of people upgrading, and one of the biggest drivers is corporate governance. In the U.S., it’s Sarbanes-Oxley, and in Europe and Asia, it’s the International Accounting Standards” (Millman 2004).

1.7 Summary

The first essay uses three different proxies developed in prior accounting research to examine the impact of ERP systems on earnings management: discretionary accruals, earnings quality, and frequency of small earnings losses and small negative earnings changes with mixed results. Short-term discretionary accruals provide some evidence that ERP systems have a positive impact on earnings management; however long-term discretionary accruals and the other two proxies do not.

The second essay uses three different measures of shareholder value: long-horizon buy- and-hold returns, market-to-book (MTB) ratios, and price-to-value (V/P) ratios to examine the impact that ERP systems have on shareholder value. One-year, three-year, and five-year abnormal returns are 15.1%, 7.7%, and 6.6% respectively following implementation for a sample of firms that implemented ERP systems vs. a control group that did not. However, multivariate analysis comparing before and after data suggests that these abnormal returns may not be due to implementation of ERP systems. In fact there is some evidence that self selection bias exists and firms that choose to implement ERP systems may be those that are better performers to begin with. MTB ratio and V/P ratio analysis provides additional support for these conclusions.

The third essay uses Sarbanes-Oxley Section 404 compliance results to measure the impact that ERP systems have on internal controls. The essay provides evidence that ERP systems have a positive impact on internal control. Specifically, the results show that a smaller proportion of ERP implementing firms reported internal control weaknesses (ICW) than a control group of non-ERP implementers during the first three years that SOX reporting requirements were in effect. The essay also examines factors contributing to ICW; finding that a smaller proportion of ERP implementing firms report IT-related factors than the control group, but the differences between the two groups for non-IT-related factors are not significant.

Overall, this dissertation provides evidence that ERP systems have very little if any impact on earnings management, or shareholder value, but a positive impact on internal controls over financial information. These results contribute to academic research in the fields of both accounting and information systems and should be of interest to a wide variety of constituents, including practitioners of financial and managerial accounting, auditing, and information systems.

The remainder of this dissertation is organized as follows: Chapter two provides a summary of relevant prior research related to this dissertation topic. It includes an extensive review of prior research related to enterprise resource planning systems followed by sub-sections that summarize relevant prior research for agency theory and each of the three essays. It includes a discussion of the incremental contribution expected from the dissertation as it relates to prior research for each topic. The next three chapters present the three essays: chapter three (the first essay) examines earnings management, chapter four presents the second essay, which focuses on shareholder value, and chapter five addresses internal control (the third essay). Finally, chapter six summarizes conclusions, contributions, limitations and opportunities for future research.

CHAPTER 2 LITERATURE REVIEW

2.1 Introduction

The purpose of this chapter is to provide a summary of relevant prior research related to this dissertation. Section 2.2 provides an extensive review of prior research related to enterprise resource planning (ERP) systems in general. Section 2.3 presents relevant prior research on agency theory. Sections 2.4 through 2.6 summarize major research related to each of the three essays and Section 2.7 concludes. Each section closes with a brief discussion of the incremental contribution this dissertation adds to the related research.

2.2 ERP Systems Related Literature

2.2.1 Background on ERP Literature

Early academic research on ERP systems generally follows one of two research streams:

(1) behavioral research focused on analysis of implementation success factors or (2) archival research focused on analysis of firm performance. The first stream uses behavioral research methods to measure the impact of critical success factors on ERP implementation projects. These studies, which serve the practice community as “lessons learned,” tend to measure success, not with traditional accounting or financial metrics, but rather based on perception and user satisfaction. By contrast, the second stream uses archival accounting and financial data to measure changes in firm performance associated with implementation of ERP systems. This research has generally yielded mixed results, with respect to improved firm performance as measured by traditional accounting and financial metrics. More recently, a third stream of research has emerged that uses a combination of methods to investigate such issues as: audit and other business risks, financial statement disclosures, intangible benefits, and organizational and

10 11

strategic implications. A summary of major contributions from each of the three streams of prior research follows.

2.2.2 Behavioral Research

This line of research focuses primarily on management related issues. In particular, researchers have tried to identify key factors that lead to successful ERP system implementation projects and conversely, those that may be responsible for implementation failures. Most of these studies tend to use survey or case study methods designed to solicit factors that lead to successful implementations as well as those that lead to implementation failures. One limitation with this approach is defining success. Most researchers have relied on the subjective “perceptions” of those being interviewed or surveyed to determine whether or not a project has been successful, rather than using objective metrics.

Markus, et al. (2000), complicate this issue further by maintaining that “success depends on the point of view from which you measure it.” They conduct an extensive study, under the sponsorship of one of the major ERP vendors, to assess the problems and outcomes of ERP implementation projects. They combine four research methods in their study, including: (1) a meta-analysis type review of published and in-process research studies and teaching cases of ERP implementations, (2) in-depth case studies of the ERP experience in five ERP-adopting organizations, (3) interviews with 11 additional ERP-adopting organizations, and (4) approximately 20 interviews with ERP implementation consultants and members of the ERP vendor company sponsoring the study. From this study, they conclude that success should be measured at three different points in the implementation process: (1) at the project phase, (2) at the shakedown phase, and (3) at the onward and upward phase. They then use three different subjective measures of success for each of these phases, and conclude that none of the companies in their study could be considered “an unqualified success” at all three stages. They further find

that some projects that were considered a failure at one stage could be considered a success at another stage. This held true in both directions. In other words, some projects that were considered a failure at the project phase were rehabilitated and became successes in the later stages. Likewise, some projects that were considered successful in the early stages were later described as failures.

Themistocleous et al. (2001), using an internet based survey, identify four major management problems faced by companies during the implementation process: (1) project delays and cost problems, (2) conflicts with external entities, (3) internal conflicts and (4) conflicts with business strategy. They also identified a number of technical issues including customization problems and integration with other systems.

Umble et al. (2003), using a case study approach identify ten categories of reasons why

ERP implementations fail: (1) strategic goals are not clearly defined, (2) top management is not committed to the system, (3) implementation project management is poor, (4) the organization is not committed to change, (5) a great implementation team is not selected, (6) inadequate education and training, (7) data accuracy is not ensured, (8) performance measures are not adapted to ensure that the organization changes, (9) multi-site issues are not properly resolved, and (10) technical difficulties can lead to implementation failures.

Al-Mashari, et al. (2003), using a literature review approach, develop a theoretical taxonomy that demonstrates the linkages between ERP critical success factors, ERP success and

ERP benefits. The critical success factors in their model are divided into three categories as follows: (1) Setting-up: management/leadership and visioning/planning, (2) Implementation:

ERP package selection, training/education, systems integrations, communications, project management, systems testing, process management, legacy systems management, cultural & structural changes, and (3) Evaluations: performance evaluation & management.

Duplaga and Astani (2003), conduct in-depth face-to-face interviews with representatives of 30 manufacturing firms in the upper Midwest region of the United States, with particular attention paid to identifying the major problems encountered during implementation and assessing user satisfaction with the system. To identify major problems, they use a Likert scale ranging from 1 to 5 to measure the extent to which each of thirteen factors created problems during implementation. The top ten results are as follows: (1) lack of ERP training and education for affected employees, (2) lack of in-house expertise in ERP, (3) lack of clear goals for

ERP effort, (4) lack of companywide support and involvement, (5) lack of data accuracy, (6) lack of top management commitment and support, (7) lack of communications to users, (8) lack of project management strategy to manage processes, (9) lack of software vendor support, and (10) unsuitability of hardware and/or software. With respect to user satisfaction, almost all interviewees express some degree of user satisfaction with the system.

Sarker and Allen (2003), use a case study to test the role of three human factors and organizational issues or social enablers in ERP implementations. The three factors which they discern from the ERP literature include: (1) strong and committed leadership, (2) open and honest communication, and (3) a balanced and empowered ERP implementation team. Their case study validates the first factor and finds that “strong and committed leadership at the top management level, at the project management level and of the IS function must be given significant priority throughout the life of an ERP implementation project.” However, they are unable to support the other two factors, finding that, while possibly helpful, they are not necessary to the success of the project.

A case study by Barker and Frolick (2003), which focuses on a major project failure at an unnamed bottling company, tends to support the strong and committed leadership factor. One quote from the case study sums up management’s attitude as: “some members of the

management team made it clear to team members that they felt the ERP implementation was a waste of both time and money.” They conclude that in addition to management commitment, excellent planning, employee involvement and good communication are critical success factors.

Another case study about a failed ERP project, which was conducted by Paper et al.

(2003), concludes that the primary reason for failure is that the company made no attempt to analyze existing processes and systems prior to implementation, and that management failed to communicate with people in the organization that were knowledgeable and take their ideas into consideration.

Kumar et al. (2003), survey 20 Canadian organizations using mainly open-ended questions around key issues and typical activities in the ERP implementation process. They conclude that “in implementing ERP systems, firms face more behavioral and management related challenges such as: the end user not being ready, resistance to change, lack of training, turnover of key project persons, and lack of project planning, rather than pure technical glitches such as software bugs and configuration difficulties.” Most of their respondents, when asked about lessons learned from the project, resolved to place more emphasis on the behavioral and management issues of implementation and improving the process.

A study by Bradford and Florin (2003) uses Diffusion of Innovation (DOI) and

Information Systems Success (IS) theories to develop an exploratory model stating that “DOI factors (i.e., innovation, organizational characteristics, and environmental characteristics) will influence ERP implementation success both from a firm performance perspective and from a user satisfaction perspective.” To test this model, they administer a survey to a randomly selected sample of members of America’s SAP User Group (ASUG). Their survey instrument uses a 7 point Likert-type scale to collect responses for seven independent variables and two dependent variables. The seven independent variables are developed from the DOI model and grouped into

three characteristic classifications as follows: (1) Innovative Characteristics: technical compatibility, perceived complexity, and business process reengineering, (2) Organizational

Characteristics: top management support, organizational objectives consensus, and training, and

(3) Environmental Characteristics: competitive pressure. The two dependent variables are: (1)

Perceived Organizational Performance, and (2) User Satisfaction. In addition to the survey variables, two control variables, elapsed time and firm size are also used. Their results, which are collected from 51 firms, show that the degree of consensus in organizational objectives and competitive pressure are significantly related to perceived performance, whereas complexity of the system, training, competitive pressure, and top management support are significantly related to the satisfaction of functional managers using the new systems.

Mabert et al. (2003), find a difference between small and large firms in their study of

ERP implementations in the US manufacturing sector. They use a two-phased approach to analyze ERP adoption and implementation experiences. In the first phase they conduct a case study of twelve different manufacturing ERP implementations using structured interviews of key managers, IT professionals, and users. They also interview senior consultants at six consulting firms specializing in ERP implementations. From this process they develop five propositions that are then tested in the second phase with a survey sent to a randomly selected sample of 5000

APICS 10 members employed in manufacturing companies in the US. They received 482 usable responses for a 9.6% response rate, with no follow up. The respondents were a mix of managerial and staff personnel. The biggest finding to come out of their study is that the size of the enterprise plays an important role in ERP implementations. They also conclude that “while early

10 The Association for Operations Management, formerly the American Production and Inventory Control

Society

adopters may have received some competitive advantages, late adopters generally benefited from upgraded systems and a better implementation knowledge base.”

The information in this section related to the behavioral research stream is provided as background information only. This dissertation will not expand on this research stream, but rather will focus on contributing to the archival research stream discussed in the next section.

2.2.3 Archival Research

The concept of using firm performance to measure the impact of information technology

(IT) spending evolved in the literature during the 1980s and 1990s. Initially researchers were unable to find empirical evidence that major investments in IT resulted in increased productivity.

This came to be known as the “productivity paradox.” Brynjolfsson (1993) in reviewing prior work offers four possible explanations for the paradox: (1) measurement error: outputs (and inputs) of information-using industries are not being properly measured by conventional approaches, (2) lags: time lags in the payoffs to IT make analysis of current costs vs. current benefits misleading, (3) redistribution: it is especially likely that IT is used in redistributive activities among firms (i.e. passing cost savings on to customers in the form of lower prices), making it privately beneficial without adding to total output, and (4) mismanagement: the lack of explicit measures of the value of information makes it particularly vulnerable to misallocation and over-consumption by managers. He concludes that measurement error is responsible for the

“lion’s share of the gap,” which is later supported by additional studies using econometric analysis techniques (Brynjolfsson and Hitt 1996; Hitt and Brynjolfsson 1996).

Different researchers have sought to use different approaches to measuring performance.

Some have used stock price as a proxy for firm value, drawing on the efficient market theory of finance that says markets do an efficient job of taking into consideration all relevant information when arriving at stock prices (Eakins 2002). Therefore, as the public becomes aware of major IT

investment projects, the stock price will be adjusted to reflect that fact. Under this approach, if the market perceives the investment will improve firm performance, the stock price should increase. Dos Santos et al. (1993) use event-study methodology to assess the effect of announcements about IT investments on stock price as a measure of firm value. Their research finds no excess returns overall; however in a cross-sectional analysis, they find that the market responds differently to announcements about innovative IT investments vs. non-innovative investments. They conclude that innovative investments increase the value of the firm but non- innovative investments do not.

Hayes et al. (2001) use a similar event-study methodology directed specifically at announcements about ERP systems, and include contextual factors for firm size, health, and ERP vendor size. Their results, using standardized cumulative abnormal returns (SCAR), indicate that there is an overall favorable reaction to such announcements, with the most positive reaction to announcements by small/healthy firms and a negative reaction to small/unhealthy firms.

Reactions to announcements by large/healthy and large/unhealthy firms are positive with mean

SCAR results in between the two extremes of the small firms. Additionally, they conclude that

ERP vendor size has an impact, with reaction to announcements using packages from large ERP vendors significantly more positive than those of small vendors.

Hunton et al. (2002) follow the Hayes et al. study with an experimental study that uses forecasted earnings of financial analysts as the dependent variable, rather than stock price. Their results are supportive of Hayes et al, concluding that forecasted earnings by financial analysts post-announcement are significantly higher than their pre-announcement estimates. They also find congruent results with respect to the two contextual factors related to firm size and financial health. That is, the two studies report at least marginally significant interaction terms, which suggest that the combined effect of size and health moderate the influence of ERP announcements

on stock price and earnings forecasts. Specifically, they both report a significant difference between small/unhealthy and small/healthy firms.

A more recent study by Ho et al. (2008) complements Huton et al. (2002) using archival data from I/B/E/S to show that financial analysts revise their three-year-ahead earnings forecasts significantly upward for firms that announce ERP implementations. They also found that financial analysts react less positively to middle adopters (1998-1999) than to early (1993-1997) and late (2000-2002) adopters.

Other researchers have used more traditional accounting and financial ratios to measure changes in firm performance related to ERP spending. Poston and Grabski (2001), use four traditional accounting measures to compare firm performance post ERP implementation to pre

ERP implementation. They include: (1) selling, general & administrative expense as a percent of revenues, (2) cost of goods sold as a percent of revenue, (3) residual income (defined as net operating income less “imputed” interest), and (4) number of employees divided by revenues.

Their study includes a sample of 50 firms that publicly disclosed ERP adoption through 1997, limited to the top five ERP vendors at the time (SAP, Oracle, PeopleSoft, Baan and J.D.

Edwards). They compare these ratios from one year before implementation to the same ratios one, two and three years after implementation. Their results are mixed, finding a significant decrease in the ratio of employees to revenues in all three years, and a reduction in the ratio of cost of goods sold to revenues in year three. However, they report no significant improvement in the ratio of selling, general and administrative expenses to revenues, or residual income. They conclude that although productivity improvement could be demonstrated with the employee ratios, they were not flowing through to the bottom line in the form of residual income. One explanation they offer, supported by Hitt and Brynjolfsson (1996), is that the benefits of

productivity (residual profits) are passed on to the consumer, rather than being retained within the firm as profits.

Hunton et al. (2003), citing the work of Poston and Grabski (2001), compare four performance measures for firms that adopted ERP systems with firms that did not adopt ERP systems. The four measures are (1) ROA (return on assets), (2) ROS (return on sales), (3) ATO

(asset turnover), and (4) ROI (return on investment). They use the data on ERP-adopting firms from Hayes et al. (2001) and matched them to a control group of non-ERP-adopting firms of similar size and industry, using SIC codes and data from Compustat. They compare the above ratios for three years prior to three years after the announcement. Their final sample contains 126 firms, 63 adopting and 63 non-adopting. Their hypothesis is that ERP-adopting firms will show little or no improvement in these profitability ratios, but that non ERP-adopting firms will show decreases in the ratios, thus resulting in a competitive advantage for ERP-adaptors. Their results find that three of the four metrics (ROA, ATO and ROI) support the hypothesis, with all three results significantly lower for the non-adopting firms the third year after ERP implementation.

Therefore, their results are supportive of Poston and Grabski in that both studies find evidence of efficiency gains but little evidence of gains in profitability. However, since there are significant differences in profitability ratios between non-adopters and adopters, the results also support the suggestion of Poston and Grabski, that profits could be passed on to customers. They also investigated the interactive effect of firm size and financial health on performance of ERP adopters, similar to Hayes et al. (2001) and Hunton et al. (2002), and find significant interaction between size and health for three of the financial measures (ROA, ROI and ROS). Specifically they find that large/unhealthy adopters experience better ROI than large/healthy adopters, and those small/healthy firms that adopt ERP systems experience better ROA, ROI and ROS than small/unhealthy firms.

Hitt et al. (2002) examine results for a group of firms that implemented SAP-R/3, the

ERP system from SAP America, from 1986-1998. They use eight common ratios including: (1) sales per employee, (2) return on assets, (3) inventory turnover, (4) return on equity, (5) profit margin, (6) asset turnover, (7) accounts receivable turnover, and (8) debt to equity ratio, plus

Tobin’s q in their analysis. They find a positive association between all of the ratios and ERP implementation except for return on equity. Further, they find with the Tobin’s q analysis that the market rewards adopters of ERP systems, suggesting that the market expects the improvements will continue into the future. Although this was an extensive study, the authors suggest several limitations including the fact that only one ERP vendor is represented in the sample, even though it is the largest vendor in terms of sales. Other limitations include the fact that the ERP implementers were compared to all other public firms in the Compustat database, and although they did control for size and industry, it is likely that firms in the control sample will have adopted ERP systems from other vendors. A final limitation was the fact that they had very little post implementation data available, which led them to suggest that future research needed to be done to follow-up on the longer-term implications from their study.

Nicolaou (2004) also uses eight performance measures to compare adaptors of ERP systems to non-adaptors, including: (1) ROA (return on assets), (2) ROI (return on investment),

(3) OAI (operating income over assets), (4) ROS (return on sales), (5) OIS (operating income over sales), (6) CGSS (cost of goods sold over sales), (7) SGAS (selling general and administrative expense over sales), and (8) ES (employees over sales). His control sample uses a matched pairs approach based on size and industry. Consistent with prior research, his results were mixed. ROI was significantly higher for ERP adopters two years after implementation, while ROA was significantly higher four years after implementation. Both measures are significantly lower in the year of implementation and the first year after implementation, which is

a pattern also observed for OIA, OIS and ROS, whereas CGSS results are not significant and

SGAS are marginally significantly higher in the year of implementation. OIS shows significant improvement in years three and four, ROS is marginally significantly better in year 4, and OIA is not significantly improved in years two through four.

Although a number of these ex ante empirical studies use archival accounting data, they are not developed around accounting theory, which Tuttle (2005) describes as a disturbing trend, because without theory, academics become indistinguishable from practitioners investigating

“interesting” questions. This dissertation attempts to reverse that trend by using accounting theory as a basis to measure the impact that ERP systems have on firms that adopt them. The academic field of accounting has a rich history of developing theories and using archival methodologies to test these theories. Section 2.2.5 provides further discussion of the incremental contribution of this dissertation to ERP related literature.

2.2.4 Research on Other ERP Issues

In more recent years researchers have begun to investigate other issues related to ERP systems. For the most part, these more recent studies tend to be developed around some theoretical base, although not always related directly to accounting. Examples include: auditing and assurance risk, financial statement disclosures, and strategic & intangible values.

Wright and Wright (2002) use a semi-structured interview study of 30 experienced information systems auditors, from three of the Big Four CPA firms, who specialize in assessing risks for ERP systems. They find that a number of common implementation problems, including business process reengineering and customization, can result in increased risk. The most frequently cited implementation problems were: users not adequately involved in the system design (68.2%), users not adequately trained (40.9%), process reengineering was required

(40.9%), and the ERP system initially lacked adequate controls (31.8%). They also find that

ongoing risks differ across applications and across vendor packages, and that participants overwhelmingly focus on testing the process rather than the system output.

Hunton et al. (2004) use an experimental method to assess the ability of financial auditors versus IT audit specialists to identify risks associated with ERP systems. They administered the experiment at four continuing professional education (CPE) sessions involving 165 auditors with varied financial and IT auditing backgrounds from four of the (then) Big Five CPA firms. They find that IT audit specialists assess significantly higher levels of network, database, and application security risks with ERP systems as compared to non-ERP systems, while the financial auditors record similar security risk assessments in both environments. Both groups of auditors assessed the overall internal control risk higher in the ERP system than the non-ERP system; however, the IT specialists risk differential is rated significantly greater than that of the financial auditors.

Using an interview process with representatives of major ERP vendors, Debreceny et al.

(2005) examine the feasibility of using Embedded Audit Modules (EAMs) in ERP systems.

EAMs are software applications embedded in host systems or linked to host systems that allow continuous monitoring of accounting information systems. They find that many of the ERP systems have the capability through embedded query tools and business alert tools to develop such EAMs, even though they do not specifically provide for them. The authors argue that future efforts should be directed at increasing the rate of EAM adoption as a means to strengthen internal controls and thus better comply with Sarbanes-Oxley Section 302 and 404 requirements.

Financial statement disclosure is another area of research that has received some attention recently. Mauldin and Richtermeyer (2004) examine firms with annual report disclosures related to ERP systems and compare them to firms that choose not to disclose ERP information in their annual report. They find that at least a third of the firms implementing ERP systems do not

disclose anything about the implementation in their annual report. Using a logistic regression analysis, they find that annual report disclosures are increasing in firms issuing debt or stock, firms with higher than median return on assets, and in the manufacturing industry, but decreasing in firm size. They find that for firms that choose to disclose, there is significant diversification of disclosure practice with respect to costs, goals and risks, suggesting that additional standards may be needed to enhance the value of annual report disclosures about technology.

Another issue receiving attention lately is the strategic value that ERP systems may have that is difficult to quantify using traditional metrics. For instance, Chand et al. (2005) use a field study approach to develop an “ERP Scorecard,” which integrates the four categories in the balanced scorecard from Kaplan and Norton (1992) with Zuboff’s (1985) three goals of information systems ( automate, informate, and transformate ) to measure the contribution of ERP systems on the strategic goals of the company. The resulting 4 x 3 matrix provides twelve cells that can be used to capture measures of success by analyzing the strategic goals and operationalizing them at the firm level.

Murphy and Simon (2002) point out that the strategic nature of ERP systems contributes to the problem of how to measure benefits with quantitative data. They argue that, similar to

R&D and advertising expenditures that are expensed when incurred rather than capitalized and amortized, these systems create intangible benefits that are not quantified and reported in the financial records. Using case study methodology they examine an attempt by a large computer manufacturer to incorporate intangibles into traditional cost-benefit analysis for an ERP project.

Ragowsky and Adams (2005) develop a model to identify the value ERP applications add to Porter’s organizational primary activities (Porter and Millar 1985) and the information system applications related to ERP that help deliver added value through organizational characteristics.

Four dependent variables are identified and related to Porter’s value chain as follows: (1)

reductions in inventory holding costs are related to inbound and outbound logistics, (2) reductions in unit production costs are related to operations, (3) reductions in cost of after-sale-services are related to service, and (4) customer retention through differential advantages are related to marketing and sales. Each of these dependent variables are then associated with specific applications within the ERP system and specific organizational characteristics. As an example, reductions in inventory holding costs are associated with the following applications: (1) inventory management, (2) sales management, (3) MRP, and (4) MRP-II along with the following organizational characteristics: (1) number of suppliers, (2) number of production lines, (3) parallel production lines yes/no, (4) production for orders percentage, and (5) length of work order in days. The authors conducted surveys of over 200 manufacturing organizations as part of a longitudinal study that ran from 1992 to 2004, using path analysis to determine the statistical strength of the various relationships. The results indicate that organizational characteristics mediate the relationship between individual applications and the value ERP can add to primary organizational activities. Each primary activity was supported by some, though not necessarily all, of the individual applications in most ERP packages. Thus the authors conclude that an organization’s characteristics are related to the return that may be gained from the use of an ERP system.

In a survey of 111 manufacturing plants, Gattiker and Goodhue (2005) find that interdependence is associated with increased plant-level benefits while differentiation is associated with decreased benefits. They also find that customization of the basic package and the amount of time elapsed since ERP implementation, both have a positive impact on plant-level benefits.

O’Leary (2004) examines a database repository of information about changes in organizations due to the implementation of ERP systems that was obtained from Oracle

Corporation’s web site. He finds that in addition to tangible benefits (i.e. inventory reduction, productivity improvement, financial close cycle reductions, etc.) a number of intangible benefits are identified by customers. The most frequently cited intangible benefit was information visibility (64%), followed by integration (44%), flexibility (40%) and customer relations (40%).

A further analysis by industry finds that tangible benefits are largely industry-independent, while intangible benefits vary across industries, with one exception, information access/visibility was the most frequently cited benefit regardless of industry.

Nicolaou and Bhattacharya (2006) in a follow-up to Nicolaou (2004) find that ERP adapting firms that initiate early enhancements in the form of either add-ons or upgrades, may enjoy superior differential financial performance in comparison to other ERP adopting firms’ differential performance.

Grabski and Leech (2007) use a theoretical foundation from economics, complementarity, to explain why successful ERP implementations are associated with multiple controls, and that these controls are used in a complementary vs. substitutable fashion. They argue that this theoretically grounded framework provides a much-needed foundation for further understanding the assurance service required during implementation and subsequent upgrades of

ERP systems or other complex organization-wide systems.

Wier et al. (2007) use two theoretical perspectives, to tie together two streams of research related to: (1) non-financial performance indicators (cybernetic control theory) and (2) ERP systems (agency theory). Their research supports the hypothesis that joint adoption of ERP and use of non-financial performance indicators (NFPI) results in significantly higher short-term and long-term return on assets and stock returns than either ERP-only or NFPI-only firms.

As discussed in more detail in the next section, this dissertation will continue this recent trend to examine other issues related to ERP systems, including the impact on earnings management, shareholder value, and internal control.

2.2.5 Incremental Contribution

As this review of prior ERP systems related research shows, there is a lack of theory driven research in this field. Most of the early research is driven by practitioner related questions about implementation processes, or by a desire to justify the investment with cost/benefit type analysis. While these are important questions, especially to practitioners that are charged with justifying and implementing these complex systems, they do not necessarily advance any theoretical foundation, which is at the heart of academic discovery research. Because these ERP systems are having a major impact on the field of accounting, a unique opportunity exists to tap into the rich field of theory based research that has been developed by accounting academics to bridge this gap. This dissertation helps to fill that gap, by using a theory driven approach, centered on agency theory, to formulate three research questions and develop hypotheses that can be tested using established accounting research methodology to measure the impact of these systems. The results should be of interest to academics in the fields of accounting, finance, and information systems, as well as practitioners in those fields and regulators charged with oversight of these functions.

2.3 Agency Theory Literature

Agency theory has been the subject of research by academics in many different fields of study. Examples include: accounting (Demski and Feltham 1978), economics (Spence and

Zeckhauser 1971), finance (Fama 1980), marketing (Basu et al. 1985), political science (Mitnick

1982), and organizational behavior (Eisenhardt 1989). Agency theory grew out of “risk sharing”

research by economists in the 1960s and 1970s to include the situation that occurs when cooperating parties have different goals and divisions of labor (Eisenhardt 1989). The basic agency model assumes two actors, the principal and the agent. The agent is hired by the principal to perform work on behalf of the principal. An agency problem arises when the agent works in his or her own self-interest instead of on behalf of the principal or firm (Mahaney and Lederer

2003). Furthermore, the level of risk that each party is willing to assume may be substantially different (Eisenhardt 1989), in which case the principal and agent may prefer different courses of action depending on their level of risk aversion or risk tolerance.

Eisenhardt (1989) offers the following two propositions that summarize the positivist stream of research focused on governance mechanisms to solve the agency problem:

Proposition 1 : When the contract between the principal and agent is outcome based, the agent is more likely to behave in the interests of the principal.

Proposition 2 : When the principal has information to verify agent behavior, the agent is more likely to behave in the interests of the principal.

These propositions represent two approaches that are used in business to control the agency problem. Both approaches have costs associated with them. In the first approach, the principal gives up some portion of the profits of the firm to encourage the agent to act in the best interest of the principal (i.e. bonus, stock options, commissions, etc.). With the second approach, the cost of providing information (i.e. the purchase of an information system) is compounded by the price the principal has to pay to understand the information once it is provided. In fact there are instances where the principal does not have the basic knowledge to comprehend the actions of the agent, even if the information is available. For instance, Kohli and Kettinger (2004) find that even with good information systems, hospital administrators have difficulty establishing legitimized managerial authority in their relation with physicians.

Outcome based contracts are found at all levels of business organizations, from piece rates for factory workers and commissions for sales representatives, to executive compensation contracts loaded with stock options and bonus incentives. Information systems are also in place throughout organizations to provide feedback on agent behavior, from time keeping and production reporting systems, to accounting information systems that track sales, profits, and other information used to measure performance, and most recently, ERP systems that combine all of the information reporting and feedback requirements of an organization into one system.

A number of researchers have examined the impact of employment contracts on the agency problem. For instance, Basu et al. (1985) examine sales force compensation plans, while several others examine various performance based incentive plans (Banker et al. 1996; Ittner and

Larcker 1998; Banker et al. 2000; Chenhall and Langfield-Smith 2003). Christensen, et al.

(2002) examine the impact of accounting policies on incentive plans, Demski and Feltham (1978) look at the impact of incentives in budget control systems, and Duru et al. (2002) examine the impact on strategic expenditures. Task level issues have also been examined as they relate to incentive plans (Feltham and Xie 1994; Fessler 2003). Fogarty et al. (2009), inspired by agency theory, examine the role of: (1) executive compensation, (2) governance structure at the board level, (3) ownership structure, and (3) earnings management in the collapse of Nortel, the biggest meltdown in Canadian financial markets’ history.

This dissertation does not use this approach to address the agency problem, but instead focuses on the second approach, which uses transparency of information provided by information systems to encourage the agent to act in the best interest of the principal. Considerably less research has been devoted to the impact of information systems such as budget and reporting systems (or ERP systems) that provide information revealing the actions of the agent to the principal. Fama (1980) describes the information effects of efficient capital and labor markets on

managerial opportunism, and Eisenhardt (1989) points out that richer information systems control managerial opportunism and therefore lead to less performance-contingent pay schemes.

Ekanayake (2004) addresses the issue of cultural differences on agent behavior in the design of management control systems. This dissertation helps to fill this gap in the literature by using this approach to addressing the agency problem as the basis for formulating three research questions and developing several related hypotheses in the areas of earnings management, shareholder value, and internal control. Since relatively little research has been done in this area, an opportunity exists to provide significant incremental contributions to this field of study.

2.4 Earnings Management Literature

Earnings management has been the subject of academic research for more than a quarter century. Graber and Jarnagin (1979) define three types of earnings management and review early action by the Financial Accounting Standards Board (FASB) in their efforts to “discourage the practice of earnings management.” During the 1980s several analytical models were developed to explain earnings management (Demski et al. 1984; Lambert 1984; Verrecchia 1986; Dye 1988;

Trueman and Titman 1988). Dye (1988), uses analytical models to demonstrate two reasons why shareholders might want to encourage rather than discourage earnings management. Using another analytical model, Trueman and Titman (1988) show that if a manager can choose to recognize income in one of two periods, he/she will prefer the choice that smoothes income, arguing that it will demonstrate less volatility, which is preferred by shareholders.

Empirical research on earnings management also evolved during the 1980s. Healy

(1985), examined the influence of bonus schemes on accounting decisions, focusing on the association between accrual policies of managers with income reporting incentives in their employment contracts. Other empirical research during this time period included examination of the impact of earnings management on: labor union negotiations (Liberty and Zimmerman 1986),

management led buyouts of public stockholders (DeAngelo 1986), proxy battles by dissident shareholders (DeAngelo 1988), and specific mandated accounting changes (Beatty and

Verrecchia 1989). Most of these early studies measured earnings management by comparing reported earnings to reported cash flow, thus inferring that earnings management activity was buried in the accrual entries. McNichols and Wilson (1988) focused on a single account, the provision for bad debts, finding mixed results and calling into question some of the earlier findings on the basis of a correlated omitted variables problem.

Schipper (1989) provides an excellent review of this early research, including discussion of a working paper that was later published in the Journal of Accounting Research, (Jones 1991), and become a seminal model for future earnings management research. Although prior studies had considered accruals as the best approach to measure earnings management, most of them either ignored the difference between the discretionary and non-discretionary portion of accruals, or assumed that the non-discretionary component was constant over time. Jones (1991) uses a regression model to estimate the non-discretionary component of accruals by regressing total accruals on revenue and fixed assets. The difference between total accruals and the amount of accruals predicted by the regression equation (the regression residual) is then considered to be the measure of discretionary accruals. This model was subsequently modified by Dechow et al.

(1995) who adjust revenue for the change in receivables, assuming that credit sales are subject to earnings manipulation. These two models have since been referenced in the literature as the

“Jones Model ” and the “ Modified Jones Model ” respectively.

Following publication of these two papers, numerous academic studies have used these models to measure various aspects of earnings management 11 . These studies cover a wide range

11 A recent query of the EBSCO Host database lists 83 papers that reference Jones (1991) and 140 that reference Dechow et al. (1995).

of topics, including: an analysis of earnings manipulation by firms subject to enforcement action by the SEC (Dechow et al. 1996); an examination of the extent to which the stock market prices discretionary accruals (Subramanyam 1996); earnings management to smooth income (DeFond and Park 1997; Tucker and Zarowin 2006); the relationship between discretionary accruals, auditor changes, and audit quality (Becker et al. 1998; DeFond and Subramanyam 1998;

Krishnan 2003); the impact of earnings management on seasoned equity offerings (Rangan 1998;

Teoh, Welch et al. 1998; Kim and Park 2005); the extent to which earnings are managed in relation to earnings expectations (Bartov et al. 2002; Matsumoto 2002); and the relationship between earnings quality and accounting conservatism (Penman and Zhang 2002), to name just a few.

Over the years, researchers have tried to improve on the Jones and modified Jones models. For instance, Teoh, Welch et al. (1998) segregate the current (working capital) portion of accruals using the argument that managers have more discretion over current accruals than over long-term accruals, therefore, the discretionary component of working capital accruals may be a better proxy for earnings management than total accruals. More recently, Kothari et al.

(2005) show that both the Jones Model and the Modified Jones Model may be misspecified in cases where sample firms demonstrate extreme performance levels. They control for performance by adding return on assets (ROA) to the models, and find that both models perform better with the additional control variable across a wide variety of simulated events. Although it is too early to determine how widespread the acceptance of this alternative version of the Jones models will be, some recent researchers have incorporated it either directly (Tucker and Zarowin

2006) or indirectly in sensitivity analysis (Francis et al. 2005) or with similar uses of ROA in estimating discretionary accruals (Leone and Van Horn 2005). Venkataraman et al. (2008) in

addition to using five different measures of abnormal accruals compute performance-matched and growth-matched versions for a total of fifteen measures.

Davidson et al. (2004) examine earnings management, or “impression management” as they call it, by using methods grounded in ethnostatistics and agency theory. Ekanayake (2004) identifies two distinct types of principal-agent relationships: (1) shareholders (principal) and top management (agent), and (2) top management (principal) and divisional managers (agents).

Nelson et al. (2003) survey auditors to identify specific approaches used by managers when they attempt to manage earnings, many of which can be found at various levels of an organization where accounting transactions are recorded. Therefore, it is reasonable to conclude that earnings management activity takes place throughout an organization, especially in decentralized structures where corporate management assigns decision rights to lower-level managers that also have incentive bonus plans. Abernethy et al. (2004) examine information asymmetries between corporate and divisional managers, finding some evidence that decentralization choice and use of performance measures are complementary. Hunton et al. (2006) show that greater transparency in income reporting reduces the likelihood that managers will engage in earnings management.

Another method used to measure earnings management was developed by Burgstahler and

Dichev (1997) to show that firms manage reported earnings to avoid earnings decreases and losses. They use cross-sectional distributions of earnings and earnings changes, finding unusually low frequencies of small decreased in earnings and small losses and unusually high frequencies of small increases in earnings and small positive earnings. They also find evidence that cash flow from operations and changes in working capital are used to achieve increases in earnings.

Recently a stream of research has emerged that extends the earnings management stream by examining accruals quality as a proxy for overall earnings quality. For instance, Francis et al.

(2005) use a measure of accrual quality (AQ), first developed by Dechow and Dichev (2002) to

show that poorer AQ is associated with higher cost of debt and equity. In a related study, Ecker et al. (2006) use this AQ factor to augment a returns-based representation of earnings quality, estimated from firm-specific asset pricing regressions. They show that the coefficient on the earnings quality factor, referred to as the “e-loading” factor, captures the sensitivity of the firms returns to earnings quality.

Ball and Shivakumar (2005; 2008) question the evidence of Teoh, Welch et al. (1998) that managers opportunistically inflate earnings to influence IPO pricing. They show that IPO firms report more conservatively, and attribute it to the higher quality reporting demanded of public firms. They offer four alternative explanations for the alleged earnings management, which Lo

(2008) examine further from the perspective of a crime scene investigator, noting that the alternatives have merit.

In spite of all this research related to earnings management and earning quality, the author is not aware of anyone who has used this theoretical framework to measure the impact of ERP systems. This dissertation fills that gap by using methodologies developed in prior accounting research to measure changes in proxies for earnings management. Specifically it uses the modified Jones model developed by Kothari et al. (2005) to measure discretionary accruals, the e- loading model developed by Ecker et al. (2006) to measure perceived earnings quality, and the earnings distribution methodology developed by Burgstahler and Dichev (1997) to measure small earnings (changes) near zero. This information should be of interest to directors and officers of corporations charged with reducing earnings management activities, and with academics in the fields of accounting, finance, information systems, and management.

2.5 Shareholder Value Literature

Shareholder value has been the subject of extensive research by academics not only in the field of accounting but also in the fields of finance and economics. Often referred to as capital

markets research, this field traces its roots to Modigliani and Miller (1958), two financial economists who developed a theory of investment, and show how the theory can lead to an operational definition of the cost of capital, which in turn can be used for rational investment decision-making within the firm. Ball and Brown (1968) are among the first to link accounting data, specifically income, to stock prices. Since then, accounting academics have focused considerable research effort on this relationship, which relies to a large extent on the efficient market theory.

Market efficiency is one of the areas of capital markets research identified by Beaver

(2002), as making the greatest contribution to accounting academic knowledge over the previous ten years. He points out that early studies including Ball and Brown (1968) and his own paper

(Beaver 1968) support the theory, but that more recent studies find the markets to be inefficient with respect to at least three areas: post-earnings announcement drift, market-to-book ratios, and contextual accounting issues. Researchers in these areas have been able to show that it is possible to exploit specific instances of market inefficiencies and generate abnormal returns. For instance, post-earnings announcement drift research finds that the market does not fully price the information contained in earnings announcements immediately, which provides a window of opportunity to generate abnormal returns (Bernard and Thomas 1989; Freeman and Tse 1989;

Bernard and Thomas 1990; Abarbanell and Bernard 1992). Other researchers have been able to demonstrate abnormal returns associated with portfolio strategies developed by grouping together stocks with similar market-to-book ratios, and buying long or selling short those stocks that are in the extreme deciles (Fama 1991; Fama and French 1992; Dechow and Sloan 1997; Frankel and

Lee 1998; Dechow et al. 1999). With respect to contextual accounting issues, Sloan (1996) is the first to show that the accrual components of earnings are less persistent than the cash flow components, implying that investors fixate on earnings and ignore the different properties of

accruals and cash flows, thus providing an opportunity for abnormal returns by developing portfolio strategies based on the magnitude of accruals. Xie (2001) shows that this mispricing of accruals is due mostly to abnormal accruals vs. normal accruals and other studies support these findings (Teoh, Welch et al. 1998, 1998; Teoh, Wong et al. 1998; DeFond and Park 2001).

Event study is a research approach that relies on the efficient market theory. It has been used in both the early accounting literature (Brown and Warner 1985; Strong 1992) as well as the information technology literature (Dos Santos et al. 1993; Subramani and Walden 2001). This approach assumes that the market reacts to news as it is released and incorporates the expected impact of the news into the price of stock within a relatively short time frame. This is the basis for the early research on market reaction to ERP systems by Hayes et al. (2001) who measure the abnormal returns during this relatively short window as a proxy for the reaction of shareholders.

Positive (negative) abnormal returns are interpreted as favorable (unfavorable) reaction by the shareholders to the decision to implement an ERP system.

Lev (1989) assesses the usefulness of earnings to investors based on the available returns/earnings research evidence and finds that the correlation between earnings and stock returns is very low, suggesting that the usefulness of quarterly and annual earnings is limited. He suggests that the fault may lie in the low quality or information content of reported earnings and calls for a reexamination of the returns/earnings research paradigm. Easton et al. (1992) use a research design consistent with the Hicksian income theory (Hicks 1939) and a clean surplus relation, finding that as the interval for calculating the association between earnings and returns widens, the strength of that relationship increases. They show that the R 2 associated with market and earnings variables increases from 5% to 15% to 33% and to 63% for one year, two year, five year, and ten year return periods respectively. Kothari and Zimmerman (1995) examine the differences between price and returns models, finding that the earnings response coefficients

(ERC) are less biased in price models but that price models more frequently reject tests of heteroscedasticity and/or misspecification than the returns models. Teets and Wasely (1996) compare two methods of estimating ERCs and find that cross-sectional coefficients are downward biased relative to firm specific coefficients.

Most of the previously discussed research evolved from empirical rather than theoretical foundations. A formal theoretical relationship between accounting information and shareholder value is presented by Ohlson (1995), and Feltham and Ohlson (1995). They show that the value of a firm is equal to its book value, assuming a clean surplus relation, plus the present value of expected future residual income, or abnormal earnings, which is the difference between accounting earnings and a charge for the normal rate of return on equity (Penman 2004). This theoretical framework provides the basis for many subsequent empirical studies that further examine the relationship between accounting information and shareholder value (Barth et al.

1996; Burgstahler and Dichev 1997; Collins et al. 1997; Barth et al. 1998; Barth and Clinch 1998;

Frankel and Lee 1998; Aboody et al. 1999; Barth et al. 1999; Brown et al. 1999; Collins et al.

1999; Francis and Schipper 1999; Myers 1999; Ali et al. 2003; Core et al. 2003).

Additional studies emerged that examine other aspects of this framework, including

Bernard (1995) who argues that the framework can effect thinking in two ways, first a shift from explaining stock price behavior toward predicting future earnings and growth in book value, and second how we structure the relation between accounting data and firm value. He compares a dividend model to an earnings model finding that the earnings model is much more useful in forecasting stock prices (R 2=68%) versus the dividend model (R 2=29%). Frankel and Lee (1998) use analyst forecasts as a means of estimating the expected future residual income component and introduce an assumption about the terminal value as a perpetuity. They find that estimates based on consensus forecasts are highly correlated with contemporaneous stock prices. They develop a

V/P ratio that compares an intrinsic value estimate (V) with stock price (P) and show that firms with higher V/P ratios experience higher long-term cross-sectional returns. Subsequent researchers have expanded on the approach, including Gebhardt et al. (2000) who use a discounted residual income model to generate an implied market cost-of-capital, and Ali et al.

(2003) who replicate Frankel and Lee (1998) finding that the value/price effect is partially concentrated around future earnings announcements, consistent with the mispricing explanation.

Bradshaw (2004) uses a similar approach with different assumptions related to the terminal value to compare different models to analyst recommendations. He finds very little correlation between any of the models and analysts’ recommendations. Other researchers have used the V/P ratio to examine the impact of arbitrage costs and earnings forecast accuracy (Zhou 2002), sell-side analysts’ buy/sell recommendations (Finger and Landsman 2003), the use of fundamental values to predict stock returns (Pillay 2004), and as a comparison to historical models in an international setting using data from 17 developed contries and six accounting regimes (Barniv and Myring

2006). Most recently, Xu (2007) examines the componants of Frankel and Lee’s (1998) model to test whether the abnormal returns are due to the predictive ability of the residual income model or to that of the components used in constructing V/P. She finds that V/P does not provide additional explanatory power for subsequent abnormal returns over its component variables, especially analysts’ earnings forecasts.

Given this wealth of prior research that relates accounting information to stock price it is surprising that the only efforts to date that use stock price to evaluate the impact of ERP systems has been the use of event study methodology (Hayes et al. 2001; Hunton et al. 2002), which only captures the initial market reaction to the announcement that an ERP system is being implemented. Since the benefits of ERP systems are expected to be generated over longer-term horizons, it makes sense that other methods of measuring the impact on shareholder value would

be appropriate. This dissertation fills that gap by using these other methodologies developed in the accounting literature to measure the impact on shareholder value. Specifically it uses long- horizon shareholder return analysis, market to book (MTB) ratio analysis, and the intrinsic value model developed by Fankel and Lee (1998) to measure the impact on shareholder value. This information should be of interest to academics and professionals in the fields of accounting, finance and information systems as well as investors.

2.6 Internal Control Literature

The subject of internal control and its relationship to the agency problem has been around for many years, even though significant academic research on the subject has not taken place until recently. Samson et al. (2006) describe steps taken as early as 1831 by the Baltimore and Ohio

(B&O) Railroad to mitigate the agency problem, including: (1) provisions in the charter that the president issue an annual “statement of affairs” letter to the shareholders (annual report), (2) a requirement by the board of directors that a committee of directors audit the financial records on a quarterly basis (audit committee), (3) provisions requiring that vouchers be authorized by the board of directors in support of all major disbursements (audit trail), and (4) a recommendation early on by the audit committee that an officer (controller) be added to the corporate structure in addition to the treasurer to keep the books and examine and certify all claims or accounts against the company (segregation of duties). These early corporate governance steps were encouraged in part by the financing arrangement that the B&O Railroad had with Baring Brothers & Co., one of

London’s oldest investment banking establishments.

Even though internal controls have played a major role in corporate governance for many years, academic research has been somewhat limited until the Sarbanes-Oxley Act of 2002

(SOX). This is due in part to the fact that internal control was considered an “internal issue” and public companies were generally not required to disclose information about their internal control

procedures, making access to data difficult for academics. Prior to SOX, the only source of formal reporting on internal controls occurred when companies changed auditors. Under prior

SEC rules, companies were required to disclose any internal control problems that were pointed out by their predecessor auditors in a Form 8-K filing. Krishnan (2005) uses these Form 8-K disclosures as a data source to compare internal control quality to audit committee quality, and finds that independent audit committees and audit committees with financial expertise are less likely to be associated with incidences of internal control problems.

In addition to the Form 8-K filing required for a change of auditors, some firms have provided voluntary disclosures in a “report of management’s responsibility” in their annual reports, which El-Gazzar and Fornaro (2004) examine during the five-year period prior to SOX.

They find that audit committee activity and internal control structure along with profitability and risk exposure are important influences on management’s behavior. Specifically they find significant differences between firms that issue such reports and firms that do not. In another study using a sample of voluntary disclosures prior to SOX, Gupta and Nayar (2005) examine whether control deficiency disclosures convey valuation-relevant information to the market.

They find that such disclosures are associated with a negative stock price reaction, on average, indicating that such disclosures do indeed convey value-relevant information. They also note that this reaction is mitigated to some extent if management also discloses that remediation steps have been taken to correct the weaknesses identified in the disclosures. Bronson et al. (2006) examine

397 mid-sized firms that issued voluntary management reports on internal control in 1998, finding that voluntary disclosures are more likely for firms that are: larger, have an audit committee that meets more often, have a greater level of institutional ownership, and have more rapid income growth.

Following implementation of SOX, internal control related research has increased significantly. This added reporting requirement has placed information in the public domain that researchers are using to examine numerous issues related to internal control and corporate governance in general. Section 404 of the act in particular requires companies to include in their annual report (Form 10-K), a separate management report on the company’s internal control over financial reporting. Initially, research related to SOX focused on reaction to the legislation and compliance with sections other than 404. For instance, Li et al. (2008) report results of an event study surrounding the legislative process leading to the final act, finding significantly positive stock returns associated with events that resolve uncertainty about the Act’s final provisions or were informative about its enforcement. They also found evidence of a positive relation between these event based stock returns and the extent of earnings management. They suggest that their results are consistent with investors anticipating that SOX would constrain earnings management and enhance the quality of financial statement information. Ge and McVay (2005) examine a sample of 261 companies that disclosed material weaknesses in internal control in their SEC filings between the effective date of Section 302 12 of SOX, and the effective date of Section 404.

They find that poor internal control is usually associated with inadequate accounting resources, deficient revenue-recognition policies, lack of segregation of duties, deficiencies in the period- end reporting process, and inappropriate account reconciliation. They also find that the most common account-specific material weaknesses occur in current accrual accounts such as accounts receivable and inventory.

12 Section 302 requires the CEO and CFO to certify that their financial statements “present fairly” in all material respects, the financial condition of their company, and that they have evaluated the effectiveness of their internal controls, and disclosed any material weakness and any significant changes in internal control procedures (Ge and McVay 2005).

To date, a relatively small number of research papers directly related to Section 404 have been published. They include Ettredge, Sun et al. (2006) who find that the presence of material weaknesses in internal control is associated with longer delays in the financial audit and thus makes it more difficult for firms to comply with the SEC’s desire to shorten 10-K filling deadlines. They also find that the type of material weakness matters, with weaknesses related to personnel, process and procedure, segregations of duties, and the closing process experiencing longer delays. Raghunandan and Rama (2006) examine the association between audit fees and internal control disclosures made pursuant to Section 404 for a sample of 660 manufacturing firms with December 31, 2004 fiscal year-ends that filed the Section 404 report by May 15, 2005.

They find that audit fees are 43% higher for clients with material weakness disclosures compared to clients without such disclosures. They also find that the association between audit fees and the presence of a material control weakness disclosure does not vary depending on the type of material weakness. Tackett et al. (2006) examine the cost and benefits associated with Section

404, and find that the new requirements have negative net benefits to the securities markets because of excessive cost and ambiguous interpretation. On the other hand, Gupta and Leech

(2006), place the requirements of Section 404 into historical perspective and offer a number of recommendations that they believe have the potential to make compliance a value-adding activity that will cost-effectively achieve the goal of promoting transparency and accountability in the US capital markets. Ashbaugh-Skaife et al. (2008) document that firms reporting internal control deficiencies have lower quality accruals, and that those whose auditors confirm remediation exhibit an increase in accrual quality relative to firms that do not remediate their control problems.

Because the subject is relatively new, most of the academic research directly related to

SOX Section 404 is still in the form of working papers. In a recent search of the Social Science

Research Network (SSRN) the author found 38 working papers based on a key word search using

“Sarbanes-Oxley Section 404” in the title or abstract. The topics cover a wide range of research, including: characteristics of firms with material weaknesses (Bryan and Lilien 2005); the impact on cost of equity (Ogneva et al. 2006); the cost of compliance (Eldridge and Kealey 2005); the role of audit committees and auditors (Krishnan and Visvanathan 2005); earnings quality, earnings management and accruals (Chan et al. 2005; Altamuro and Beatty 2006; Bédard 2006); auditor realignments (Ettredge, Li et al. 2006); restatements of financial reports (Li and Wang

2006); CFO turnover (Ettredge, Li, and Sun 2007); audit pricing (Hoitash et al. 2007) and value relevance (Cheng et al. 2006) 13 . More recent researchers have expanded the research stream to include the cost of Section 404 compliance (Krishnan et al. 2008) and analysis of determinants of

SOX compliance (Webb 2008).

Only a few studies examine the relationship between SOX Section 404 compliance and information systems in general. For instance Li et al. (2007) examine the effects of internal and external governance factors on material weakness related to information technology (IT) controls.

They find that companies with more IT-experienced senior managers and with longer tenured

CIOs are less likely to have IT material weaknesses. They also find that companies with a higher percentage of independent directors and those with more IT-experienced audit committee members are less likely to have IT control related material weaknesses, and that the clients of Big

4 audit firms and IT-experienced auditors have higher IT control quality. Canada et al. (2007) examine the relationship between audit fees and IT and non-IT related internal control weaknesses. They find that audit fees are higher for firms with IT related material control weaknesses as compared to both firms without any material weaknesses and those with only non-

13 Some of these working papers have subsequently been published in peer reviewed journals, including: (Ettredge, Li, and Scholz 2007; Krishnan and Visvanathan 2007; Ogneva et al. 2007; Hoitash et al. 2008).

IT related material weaknesses. Grant et al. (2008) examine 278 companies reporting IT control deficiencies in the first three years of SOX 404 requirements (2004-2006) finding that firms with

IT-control deficiencies report more internal control (IC) deficiencies, are smaller, pay higher audit fees, and are typically audited by smaller accounting firms. Although Walker (2008) and others provide anecdotal information about the relationship between ERP, BPM (business process management), and SOX, to the author’s knowledge no empirical research has been completed, or is underway, examining the relationship between ERP systems and SOX Section 404 compliance.

This dissertation fills that gap by addressing this relationship in the third essay. It examines SOX Section 404 compliance for ERP implementers compared to non-implementers by reviewing Form 10-K filings, as compiled by Audit Analytics, for instances of material weaknesses reported by management and attested to by the auditors. It also examines general internal control compliance and specific IT related control compliance. These findings provide new insights into the impact that ERP systems have on internal controls, and compliance with

SOX Section 404. The results should be of interest to practitioners and academics in several fields, especially finance, accounting, auditing, and information technology.

2.7 Summary

This chapter has summarized prior research that relates to the overall topic of this dissertation and each of its three essays. It begins with an extensive review of literature related to

ERP systems, emphasizing the point that most prior research is not based on any form of accounting related theory, thus providing an opportunity to make a unique contribution. It then reviews significant prior contributions related to agency theory and to each of the areas of accounting research addressed in the three essays: earning management, shareholder value, and internal control. In each of these areas gaps in the current literature are identified, which provides an opportunity for this dissertation to make significant incremental contributions.

CHAPTER 3 EARNINGS MANAGEMENT (ESSAY ONE)

3.1 Introduction

Since the collapse of Enron and Arthur Andersen, the academic community has shown renewed interest in earnings management issues, as demonstrated by the fact that eight of the twelve articles in the January 2006 issue of The Accounting Review relate either directly or indirectly to earnings management. Earnings management is described by Healy and Wahlen

(1999) as follows:

Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.

Prior academic research has provided overwhelming evidence that earnings management takes place 14 . Most of this prior research has focused on examining specific events that provide opportunities for “increased” levels of earnings management activity such as: SEC enforcement actions (Dechow et al. 1996), auditor changes (Becker et al. 1998), equity offerings (Rangan

1998; Teoh, Welch et al. 1998), income smoothing (Tucker and Zarowin 2006), analyst forecasts

(Abarbanell and Lehavy 2003), stock-based compensation expense (Aboody et al. 2004), the auditor-to-client revolving door (Geiger et al. 2005) and transparency (Hunton et al. 2006).

However, very little published research examines management actions that reduce rather than increase the level of earnings management activity. Implementation of ERP systems represents one such management action. These systems have been designed to provide transparency of information throughout an organization, which based on agency theory, should result in

14 See (Healy and Wahlen 1999) for an excellent review of earnings management literature up to that date.

44 45

“decreased” levels of earnings management activity. This is especially true for large organizations that delegate management decision making to the operating unit level, where earnings management activities in the past may have gone undetected because of the lack of transparency provided by older legacy systems with their silos of information.

This essay examines the relationship between earnings management and ERP systems by using three different proxies for earnings management developed in prior accounting research: (1) discretionary accruals, (2) earnings quality, and (3) small earnings (changes) near zero. It begins by examining three levels of discretionary accruals: short-term (working capital) accruals, long- term accruals and total accruals before and after implementation of ERP systems using the Jones

(1991) model as modified and applied by prior research (Dechow et al. 1995; Teoh, Welch et al.

1998; Aboody et al. 2005; Kothari et al. 2005). It provides some evidence that the absolute values of all three discretionary accrual categories decrease following the ERP implementation event for a sample of firms that implemented ERP systems between 1994 and 2003. However, when compared to a control group of similar firms over a ten year period (five before and five after implementation), and adding controls for possible correlated omitted variables, only the short-term working capital accruals reflect a marginally significant difference. This difference is not robust, and loses its significance in smaller tests of specifically matched pairs over two, three, four, and five year periods before and after implementation.

Recent research has expanded the concept of discretionary accruals as a measure of earnings quality (Dechow and Dichev 2002; Francis et al. 2005; Ecker et al. 2006). As a second proxy, this essay uses the “e-loading” methodology developed by Ecker et al. (2006) to examine earnings quality for firms that implement ERP systems. Relative to other proxies for earnings management, such as discretionary accruals, this approach can be used to estimate quality over shorter intervals of time because it relies on daily market prices rather than annual accounting

data. This approach is of particular interest with respect to ERP systems, because a lot of these systems have been implemented in recent years. Therefore, the number of years of accounting data available post-implementation for comparative analysis is limited. The essay provides evidence that the quality of earnings as measured by the “e-loading” factor is better for firms that implement ERP systems than for the control firms, and that earnings quality improved for both groups in the five years after implementation compared to the five years before. However, the improvement was not significantly different for the ERP implementers than for the non- implementers.

The third proxy used in this essay is one developed by Burgstahler and Dichev (1997) that measures the frequency of small earnings (changes) just above and below zero on the theory that companies will manage earnings away from reporting small losses or small decreases in earnings. The essay provides some evidence that earnings management activity decreased, but consistent with the other two proxies, the improvements as measured by the frequency of small earnings (changes) surrounding zero are not significantly different for firms that implement ERP systems than for the control firms.

These results should be of interest to the academic community and the professional communities of accounting, finance, and information technology as evidence that the ERP investments made in the 1990s may not have provided a level of information transparency that is sufficient to modify earnings management behavior.

The remainder of this essay is organized as follows: Section 3.2 reviews prior research and develops the hypotheses, Section 3.3 describes the sample data selection process and research methodology, Section 3.4 presents empirical results, and Section 3.5 concludes the essay by discussing implications, limitations, and future research opportunities.

3.2 Prior Research and Hypotheses Development

Antle (1989) suggests that “earnings management” is really “information management” and the agency model is a useful tool to evaluate the impact, especially given the fact that the choice of information reporting system is often delegated to the agent. Given the design of ERP systems and their emphasis on information transparency throughout an organization combined with Eisenhardt’s (1989) second proposition, which states that “if the principal has information to verify an agent’s behavior, the agent will be more likely to behave in the interest of the principal,” ERP systems should discourage earnings management activity. Since the principal/agent relationship exists at many levels throughout a firm, this new access to data by the various principals should encourage the agents to act in the best interest of the principals at all levels. This should result in more truthful reporting, especially with respect to the operating activities that generate working capital accruals. Since ERP systems track operational activity from sales order to cash receipt on the revenue side, and purchase order to disbursement on the procurement side, including inventory flow, there is less opportunity for income increasing and/or income decreasing accrual entries to be posted or to go un-noticed if they are posted. For instance, Nelson et al. (2003) report earnings management activities in such working capital accounts as bad debt reserves, inventory reserves, cut-off issues for both revenue (accounts receivable) and expenses (accounts payable), all of which become much more visible in an ERP environment.

Since earnings management cannot be measured directly, accounting researchers have developed proxies for earnings management. One of the most common approaches is to examine accrual activity based on the theory that accruals provide the best opportunity to manipulate earnings because they represent the difference between reported earnings and cash flow, which can not be as easily manipulated. One problem with accruals as a proxy is that they contain both

discretionary accruals, which can be easily manipulated and non-discretionary accruals, which can not be as easily manipulated. Jones (1991) developed an approach to separating the two components that has become widely accepted in accounting research, and subsequently modified by others to improve the overall model (Dechow et al. 1995; Kothari et al. 2005), and segregate short-term (working capital) accruals (Teoh, Welch et al. 1998; Aboody et al. 2005) from long- term accruals. These modified Jones models are especially useful in developing a proxy for earnings management in this essay because they are able to measure short-term (working capital) accruals, which are expected to be affected more directly by implementation of ERP systems than long-term accruals. This leads to the following hypothesis stated in the alternative:

H1a: Implementation of Enterprise Resource Planning Systems will have a positive impact on earnings management activity, as measured by changes in discretionary short-term (working capital) accruals.

The impact of ERP systems on long-term discretionary accruals is not as clear. Long- term accrual entries other than depreciation tend to be made at higher levels within a corporation to reflect such things as deferred taxes, amortization of intangibles, impairment of goodwill, etc.

Therefore, it is possible that implementation of an ERP system may have no impact at all on earnings management activities carried out through manipulation of long-term discretionary accruals. It is also possible that by implementing an ERP system, management may shift their earnings management activities from short-term to long-term accounts resulting in no change in total accrual activity. On the other hand, the overall increased transparency of financial transactions may have the same effect on long-term accruals as short-term accruals resulting in an overall decline in earnings management activity. Thus the effect is an empirical question with no clear theory as to what the expected impact will be, so it is stated in the alternative form in the following hypothesis with no directional prediction:

H1b: Implementation of Enterprise Resource Planning Systems will have an impact on earnings management activity, as measured by changes in discretionary long-term and total accruals.

In addition to providing an opportunity for earnings management, accruals provide a signal to the market with respect to future cash flows. This signal provides investors with information about the relationship between earnings and cash flow, which Francis et al. (2005) argue contains information risk that is not diversifiable. To the extent that investor pricing decisions are made based on this information, the quality of the information is important. In fact, they show that firms with poor accrual quality (AQ) experience higher cost of capital in terms of both debt and equity. In a related study, Ecker et al. (2006) use the AQ factor to develop a returns-based measure of earnings quality, which they call an “e-loading” factor. Because this e- loading factor represents the market’s perception of earnings quality, and is developed based on daily stock market information and not on annual accounting data, it can be used to measure trends over relatively short time periods, which is of particular interest in this essay because of the limited amount of annual data available post-ERP implementation. Assuming ERP systems discourage earnings management, the quality of accruals and therefore earnings will also improve, thus making the e-loading factor another proxy for earnings management. Since this proxy applies to overall total accrual activity and is not distinguishable between short-term and long-term, the following hypothesis is stated in the alternative form with no directional prediction:

H1c: Implementation of Enterprise Resource Planning Systems will have an impact on earnings management activity, as measured by changes in e-loading factors.

Another approach to measuring earning management was developed by Burgstahler and

Dichev (1997), who show that firms tend to manage earnings to avoid small losses and small

decreases in earnings. They segregate earnings (changes) into small intervals and show that the intervals around zero have a distinct pattern with lower frequencies of small negative earnings

(changes) than small positive earnings (changes). As with the e-loading proxy discussed above, this approach does not take into account the differences between short-term and long-term earnings management activity, so the following hypothesis is also stated in the alternative form with no directional prediction:

H1d: Implementation of Enterprise Resource Planning Systems will have an impact on earnings management activity, as measured by changes in small intervals of earnings (changes) around zero.

3.3 Data Selection and Methodology/Models

3.3.1 Sample Data Selection

The sample data for this essay is based on 147 firms from 33 industry groups that implemented ERP systems between 1994 and 2003. Panel A of Table 3-1, provides a summary of the sample selection process for these ERP implementing firms. The process starts with the list of 91 ERP announcements made between 1994 and 1998 from Hayes et al. (2001) 15 , from which 36 firms were eliminated because they are no longer listed or data was otherwise not available. The second step in the process involves a search of all available newswire services using the Lexis-Nexis service for years after 1998, searching on key phrases such as: “ERP”,

“Enterprise Resource Planning”, and “Enterprise Systems.” This search found an additional 92 firm announcements yielding a total of 147 firms for which data is available for at least two years before and after implementation in the Research Insight Compustat database.

15 I would like to thank David C. Hayes and Jacqueline L. Reck for providing this list of firms

Following the recommendations of Barber and Lyon (1997), and the approach used by

Nicolaou (2004), this essay uses a matched pairs approach to select a control group. ERP implementing firms are first matched with other firms based on their four digit SIC code, then by total assets at the beginning of the implementation year, then by the availability of Compustat data for the two years before and two years after implementation. If a reasonable match is not available then three or two digit SIC codes are used to obtain a match. Once a matched firm is identified, a further search of Lexis-Nexis Newswires is conducted for all available years using a combination of the firm name, and the following terms: (1) ERP, (2) Enterprise Resource

Planning, (3) Enterprise Systems, (4) SAP, (5) Oracle, (6) QAD, (7) Baan, (8) Peoplesoft, (9) JD

Edwards, and (10) Lawson 16 . If no newswire records are found, the firm is used as a match for this study. If a record is found that indicates an ERP system may be in use, then the next closest firm in terms of total assets is used, and the process repeated. Six matches required the use of three digit SIC codes: (201)-Food and Kindred Products Manufacturing, one match; (252)-

Furniture and Fixture Manufacturing, two matches; (284)-Chemicals and Allied Products

Manufacturing, two matches; and (594)-Miscellaneous Retail, one match. Two digit SIC codes are used for four matches: (28) Chemical Manufacturing, two matches, (35)-Industrial and

Commercial Machinery and Computer Equipment, one match; and (38) Measuring, Analyzing, and Control Instruments, one match. After adding the 147 control firms, the total number of firms available for analysis is 294.

Panel B of Table 3-1 provides a comparison of the mean, median, standard deviation, minimum and maximum values of total assets at the beginning of the implementation year and market capitalization as of the end if the implementation year for each of the two groups. T-tests

16 The specific vendors listed (4-10) represent the seven most common ERP systems identified in Nicolaou (2004).

and Wilcoxon tests indicate that the mean values are statistically different from each other, with the ERP firms larger than the control firms on average. Although it would be preferable for the two samples to be closer to the same size, emphasis in the matching process was place on matching based on industry first, and size second, recognizing that it may not be possible to match the size as close because larger firms tend to be the ones that implement ERP systems and publicize the fact with press releases. This could be a major limitation for this essay, and is discussed further in the conclusions section of this document. However, given the nature of the information available in the public records and the methodology used to search these records, this is the best match available at this time. To control for the size differences, control variables are used in the analysis where appropriate.

[Insert Table 3-1 here ]

Table 3-2 provides a breakdown of ERP implementing firms by two-digit SIC code and implementation year. ERP announcements are of two types, those made before the system was implemented and those made after the system was implemented. For instance, some of the announcements are reporting that the firm had entered into an agreement with a particular vendor or consulting firm to implement an ERP system. In other cases the announcements are made by the vendor or a consultant announcing the “successful” implementation of an ERP system. Or in still other cases the announcements are part of a regular earnings announcement indicating that earnings are different than forecast due to implementation of an ERP system. In some cases the announcements do not contain specific implementation dates or time lines. In those cases, judgment was used to determine which year best represents the implementation year. Generally, for announcements made after implementation, the implementation year is assumed to be the firm’s fiscal year in which the announcement is made or the previous year if the announcement is dated early in the fiscal year. For announcements made prior to implementation, the

implementation year is assumed to be the fiscal year following the announcement. For before and after analysis, the implementation year is considered to be the last year for before implementation data.

[Insert Table 3-2 here ]

3.3.2 Discretionary Accrual Model

Prior research has used a number of methods and models to estimate discretionary accruals, which are, in turn, used as proxies for earnings management. This essay uses the cross- sectional version of the modified Jones (1991) model initially developed by Dechow et al. (1995), with a modification for accounts receivable and ROA incorporated by Kothari et al. (2005) 17 , hereafter referred to as the modified Jones model. The model uses the following OLS regression to estimate total accruals:

TAC it /TA it-1 = a 1(1/TA it-1) + a 2(∆ REV it −∆ REC it )/TA it-1 + a 3(PPE it )/TA it-1

+ a 4(ROA it )/TA it-1 + εit (1)

Where (reference to Compustat data elements are in parenthesis):

18 TAC it = Total Accruals in year t for firm i ( ∆A4 – ∆A1 – ∆Α 5 + ∆A34 – A14)

19 TA it-1 = Total Assets (A6) in year t-1 for firm i

∆REV it = Revenue (A12) in year t less revenues in year t-1 for firm i

∆REC it = Receivables (A2) in year t less receivables in year t-1 for firm i

17 Dechow et al. (1995) exclude the receivable adjustment in the estimation period of the time-series setting, but include it in the event period calculations. The effect is to assume that sales are not managed in the estimation period, but that the entire change in accounts receivable in the event year represents earnings management. Kothari et al. (2005) include the adjustment in both equations, which is more appropriate for the cross-sectional setting. 18 A4=Total Current Assets; A1=Cash & Equivalents; A5=Total Current Liabilities; A34=Current Portion of Long-Term Debt; A14=Depreciation & Amortization 19 Following Jones (1991), total lagged assets are used as a scale factor to reduce heteroscedasticity because lagged total assets are assumed to be positively associated with the variance of the disturbance term. Most subsequent applications of this model and its modified forms use this same approach.

PPE it = Net property plant and equipment (A8) in year t for firm i

ROA it = Return on assets in year t for firm i

εit = Prediction error (residual) in year t for firm i

Coefficients from the regression model (a 1, a 2, a 3, a 4) are then used as estimates of ( α1,

α2, α3, α4) to calculate the nondiscretionary portion of total accruals, with the residuals representing the discretionary portion, using the following equations:

NTAC it /TA it-1 = α1(1/TA it-1) + α2(∆ REV it −∆ REC it )/TA it-1 + α3PPE it /TA it-1

+ α4ROA it /TA it-1 (2)

20 DTAC it /TA it-1 = TAC it /TA it-1 – NTAC it /TA it-1 (3)

Where:

NTAC it = Nondiscretionary Total Accruals in year t for firm i

DTAC it = Discretionary Total Accruals in year t for firm i

And other variables are as previously defined

To differentiate between short-term (working capital) and long-term accruals, this essay follows prior research (Teoh, Welch et al. 1998; Aboody et al. 2005), with the accounts receivable and ROA modification by Kothari et al. (2005), and defines working capital and long- term discretionary accruals with the following set of equations applied in the same manner as the total accrual set of equations above:

WAC it /TA it-1 = a 1(1/TA it-1) + a 2(∆ REV it −∆ REC it )/TA it-1+ a 4(ROA it )/TA it-1 + εit (4)

NWAC it /TA it-1 = α1(1/TA it-1) + α2(∆ REV it −∆ REC it )/TA it-1+ α4ROA it /TA it-1 (5)

DWAC it /TA it-1 = WAC it /TA it-1 – NWAC it /TA it-1 (6)

DLAC it /TA it-1 = DTAC it /TA it-1 – DWAC it /TA it-1 (7)

20 DTAC are also equal to the residuals from regression equation (1)

Where:

WAC it = Working Capital Accruals in year t for firm i ( ∆A4 – ∆A1 – ∆Α 5 + ∆A34)

NWAC it = Nondiscretionary Working Capital Accruals in year t for firm i

DWAC it = Discretionary Working Capital Accruals in year t for firm i

DLAC it = Discretionary Long-Term Accruals in year t for firm i

And other variables are as previously defined

Prior earnings management research has used two different approaches when applying these models to calculate discretionary accruals; time-series and cross-sectional. Jones (1991) uses the time-series approach with multiple years prior to some event, used as an estimation period, for each firm in the regression equation, to estimate the coefficients in the model. These coefficient estimates are then applied to data in the “event year or years” to measure discretionary and non-discretionary accruals. Other studies use a cross-sectional approach by combining all firms in the same industry based on SIC codes to estimate nondiscretionary accruals for specific years (Aboody et al. 2005). Kothari et al. (2005) suggest that the cross-sectional approach measures the level of “abnormal” earnings management compared to a “normal” level of earnings management that exists based on industry averages. Most studies that use the time-series approach require a minimum of 10 years worth of data, which often results in the loss of observations (McNichols 2000). Since most ERP systems were implemented during the latter half of the 1990s and early 2000s, it is difficult to have ten years worth of post-implementation data to compare to ten years worth of pre-implementation data; therefore, this essay uses the cross-sectional approach. The initial regressions are estimated on all firms in the same industry using three-digit SIC codes each year where possible, which ensures a large number of observations in industry specific groups. If three digit codes do not result in a sufficient number of observations, then two digit codes are substituted. Since this is an estimate of the “normal”

level of discretionary and non-discretionary accruals based on industry averages, it is important to differentiate by industry as precisely as possible. The resulting coefficients are applied to individual firm-year observations in the sample.

3.3.3 Earnings Management Model

Once the discretionary accruals have been estimated, Dechow et al. (1995) suggest the following model to measure earnings management:

DAC it = a i + b iPART it + c iCONT it + εit (8)

Where:

DAC it = Discretionary accruals

PART it = Partitioning variables

CONT it = Control variables

Aboody et al. (2005) suggest that the absolute value of discretionary accruals better reflects earnings management, because of the reversing nature of most accrual entries. Therefore the following regression models are used to estimate earnings management for short-term, long-term and total discretionary accruals:

ADTAC it /TA it-1 = a it + biPART it + c iCONT it + εit (9)

ADWAC it /TA it-1 = a it + biPART it + c iCONT it + εit (10)

ADLAC it /TA it-1 = a it + biPART it + c iCONT it + εit (11)

Where:

ADTAC it = absolute value of discretionary total accruals in year t for firm i

ADWAC it = absolute value of discretionary working capital accruals in year t for firm i

ADLAC it = absolute value of discretionary long-term accruals in year t for firm i

PART it = partitioning variables including:

ERP = (1) for firms implementing ERP and (0) for matching control firms.

IMP = (1) for post implementation periods, and (0) for pre-implementation.

ERP*IMP = interaction term based on ERP x IMP. This is the primary variable

of interest and is expected to be negative if ERP implementation is

associated with lower levels of earnings management relative to non-

implementers.

CONT it = control variables, including:

SIZE = natural log of total assets at the beginning of year t for firm i is used as a

control variable for size because prior research shows it has an influence

(Hayes et al. 2001; Hunton et al. 2002).

ROA = return on assets is used as a control variable for financial health as prior

research shows it has an influence (Hayes et al. 2001; Hunton et al. 2002).

Y2K = dummy variable set to (1) for firms that implemented ERP systems (and

the matched control firms) after the year 2000, otherwise (0). This

variable is used to control for the possible differences in efficiencies that

may exist between early adopters and late adopters of technology.

MFG = dummy variable set to (1) for firms that are in manufacturing SIC codes

(2000 to 3999), otherwise (0). This is used to control for the possibility

that manufacturing firms may benefit more from ERP systems because of

the complexity of their operations and the origin of ERP in this industry.

3.3.4 Earnings Quality Model

To test the earnings quality hypothesis (H1c), this essay uses the methodology developed by Ecker et al. (2006) to measure perceived earnings quality by adding a variable for accrual quality ( AQfactor ), to the standard CAPM asset pricing regression model 21 as follows:

Rj,t – RF,t = αj,t + βj,T (R M,t – RF,t ) + e j,T AQfactor t + εj,t (12)

Where:

t = index for the number of trading days in year T.

Rj,t = firm j’s return on day t.

RF,t = the risk free rate on day t.

RM,t = the market return on day t.

AQfactor t = accrual quality factor on day t.

The coefficient on the AQfactor variable, referred to as an e-loading factor by Ecker et al.

(2006), is a returns-based representation of earnings quality. Similar to the way the CAPM beta captures exposure to market risk, the e-loading captures investor perceptions of the firm’s earnings quality exposure in year T. AQfactors, risk free rates, and market rates are made available by Jennifer Francis of Duke University in a file with 8,586 daily portfolio returns from

1970 to 2005 downloaded from the Duke University web site 22 . Following is a detail description of how the AQfactor is constructed, which is provided on the web site:

The construction of AQfactor starts with identifying all firms with the necessary data to estimate the underlying accruals quality metric, developed by Dechow and Dichev (2002) and modified by McNichols (2002). Requiring a minimum of 20 firms per industry-year, we run annual cross-sectional regressions of total current accruals on past, present and future cash flows from operations, as well as on gross property, plant and equipment and

21 In addition to the CAPM model, Ecker et al. (2006) use the 3-factor Fama and French model which includes variables for SMB (small-minus-big) and HML (high-minus-low) portfolios. They report similar results for both models. 22 I would like to thank Jennifer Francis for making this data available.

the change in sales revenues, separately for each of the 48 Fama and French (1997) industries. Accruals quality at the end of Year T (AQ) is the standard deviation of the five firm- and year-specific residuals obtained from the regressions in Years T-5 to T-1. Lagging AQ by one year accounts for the fact that the industry regressions contain the leading cash flow from operations. We assign firms to AQ deciles using a dynamic portfolio technique that allows for differences in firms' fiscal year ends as well as over- time changes in accruals quality. Specifically, we further lag the AQ metric by three months after fiscal year end to ensure public availability of the accounting data and then form deciles on the first day of each month based on the firm's most recent value of AQ. If the AQ signal for the following fiscal year is missing due to insufficient data, the firm is excluded from this portfolio formation after twelve months (but allowed to re-enter the portfolio later). Finally, AQfactor is defined as the equal-weighted daily return of the four deciles of firms with the highest (=poorest) AQ less the equal-weighted daily return of the four deciles of firms with the lowest (=best) AQ.

Ecker et al. (2006) show that e-loadings exhibit predictably positive correlations with most other proxies used to represent earnings quality, including the seven earnings attributes considered by Francis et al. (2004): accruals quality itself (AQ), persistence, predictability, smoothness, value relevance, timeliness and conservatism. They also show that firms with higher e-loadings have lower earnings response coefficients and more dispersed and less accurate analysts’ forecasts. They argue that this characteristic is consistent with market participants perceiving that higher e-loading firms have noisier earnings signals compared to lower e-loading firms. They also find a decline in the magnitude of e-loadings and an increase in autocorrelation over time, as firms mature. In a final test, they find that e-loadings are highest during years containing restatement announcements, lawsuit filings, and bankruptcies, all events that are indicative of poor earnings quality.

Based on this prior work, e-loadings are used as a proxy for earnings quality for this essay. A larger e-loading implies greater sensitivity to poor earnings quality, therefore this coefficient is expected to be lower for ERP firms relative to non-ERP firms following implementation. To test this hypothesis (H1c), firm specific e-loading factors obtained in equation (12) are regressed on the same partitioning and control variables used to test discretionary accruals (equations 9, 10 & 11) as follows:

e-loading it = a it + biPART it + c iCONT it + εit (13)

Where:

e-loading it = e-loading coefficient from equation (12) for year t for firm i

Other variables are as previously defined in equations (9, 10 & 11). The variable of interest (ERP*IMP) is expected to be negative if ERP implementation is associated with improved earnings quality because e-loading values reflect sensitivity to poor earnings quality.

Therefore higher values for e-loadings are associated with poorer quality earnings.

3.3.5 Earnings Distribution Methodology

To test hypothesis (H1d) related to small negative earnings (changes) vs. small positive earnings (changes) this essay uses the methodology developed by Burgstahler and Dichev (1997).

It uses net income (Compustat mnemonic NI) scaled by beginning of the year total assets

(Compustat mnemonic AT) 23 for the small earnings analysis, and net income for year t minus net income for year t-1 scaled by total assets at the beginning of year t-1 for the earnings change analysis. The earnings (change) values are then grouped into interval widths of .005 (.0025) with the resulting frequencies compared to expected frequencies that are based on the average of the number of observations in the two immediately adjacent intervals, which is their definition of smoothness. Burgstahler and Dichev develop a test statistic to test the null hypothesis that the distribution is smooth by dividing the difference between the actual number of observations and the expected number of observations by the estimated standard deviation of all the differences.

This approach is also used in this essay.

23 Burgstahler and Dichev (1997) use beginning of the year market value for their primary analysis, but report that alternative results using total assets obtain qualitatively similar results.

3.4 EMPIRICAL RESULTS

3.4.1 Estimation of Discretionary Accruals

To estimate the cross-sectional discretionary accruals, this essay uses data for fiscal years

1990-2006 for all active and research firms in the Compustat database dated June 21, 2007. The database contains records for 22,880 firms, which equates to 388,960 firm-year observations for the 17 year period. A number of these observations (169,819) contain no data because the firms were not in business or otherwise did not have financial data reported to Compustat in some of the years. Another 52,148 were missing total asset data for year t-1, needed as a scale factor,

42,237 observations were for 2 digit SIC codes not included in the ERP sample, and 17,221 observations were missing data needed to calculate the change in revenue. This leaves a total of

107,735 firm-year observations used in the analysis. Equations (1) and (4) are used to estimate the coefficients each year for total accruals and working capital accruals, respectively, grouped by three-digit SIC codes provided that at least five observations exist for each industry-year regression. If five observations are not available, then a two digit SIC code is substituted, which resulted in 42 industry groups compared to the 33 two-digit groups. All of the variables used in the regressions are first winsorized at the 1% and 99% levels to reduce distortions that may be introduced by outliers. Table 3-3 provides descriptive statistics for the primary variables in Panel

A, coefficient estimates from the 714 industry-year regressions (42 industries x 17 years) for non- discretionary total accruals in Panel B and working capital accruals in Panel C. The results of the first regression (Panel B) reflect a negative mean coefficient on the variable for property, plant and equipment (a 3), which is similar to other studies, reflecting the impact of depreciation. Both of the regressions generate positive coefficients on the other variables, indicating that the non- discretionary component of total accruals and working capital accruals tends to be positive. The

R2 values are relatively strong at 0.45 for the total accruals model and 0.35 for the working

capital model on average, indicating that the models do a reasonable job of explaining the variation in the data.

[Insert Table 3-3 here ]

The estimated coefficients for each industry group and fiscal year combination from the previous regressions are then used in equations (2), (3), (5), (6) and (7) to calculate firm specific, non-discretionary and discretionary accruals as the fitted values and residuals, respectively for the sample firms. Table 3-4 presents descriptive statistics for the absolute value of discretionary working capital, long-term, and total accruals grouped by ERP and non-ERP implementing firms for the five years before and five years after implementation.

[Insert Table 3-4 here ]

Figure 3-1 provides a graphical presentation of the mean absolute value of discretionary accruals from Table 3-4. The first graph shows that the mean absolute value of discretionary working capital accruals decreases for both ERP implementing firms and for the control group following implementation, with the slope of the line for implementers declining more rapidly than for non-implementers. The average for ERP implementers decreased 0.0139 (21%) from 0.0665 to 0.0526, compared to only 0.0017 (2.3%) from .0710 to .0693 for the control firms. Notice also that the line for ERP implementers is lower than the control group both before and after implementation. The absolute value of long-term discretionary accruals in the second graph however shows that the ERP implementers have higher values than the control group both before and after the implementation event. The implementer values decline slightly (0.0008; 2.8%) while the control group line is flat. The total accrual graph shows a pattern with the value before implementation for both groups being about the same (0.0763 vs. 0.0748), however the slope of the line for ERP implementers is once again declining more than the control group, reflecting a

decrease of 0.0150 (20%) to 0.0598 compared to a decrease of only 0.0037 (5%) for the control group.

[Insert Figure 3-1 here ]

Table 3-5 provides a summary of t-tests, which further supports the graphical presentation.

Notice that the decrease in absolute value of working capital accruals for ERP implementers represented by the downward sloping line in the graphical presentation is significant (t=3.1274; p=0.0009) as is the decrease in total accruals (t=3.1376; p=0.0017) while the control group changes are not significantly different from zero. Also notice that the difference between implementers and non-implementers is significant after implementation for both working capital accruals (t=3.2981; p=0.0005) and total accruals (t=2.5624; p=0.0105). By contrast, none of the differences in the long-term accruals are significant.

[Insert Table 3-5 here ]

These results provide initial support for the first hypothesis, indicating that ERP systems may have an impact on earnings management activities as measured by changes in the absolute value of discretionary short-term, working capital accruals following implementation. They also provide mixed support for the null version of the second hypothesis showing no difference for long-term accruals, but a significant difference for total accruals.

3.4.2 Earnings Management Results

While the univariate analysis discussed in the prior section provides initial support for

H1a and mixed support for H1b, it does not take into consideration the impact that other factors may have on the changes taking place. The earnings management regression models expressed in equations 8, 9, & 10 are used to control for some of these factors. Table 3-6 provides a correlation matrix of the variables used in these models, and Table 3-7 presents results from the regressions. The only sign that is predicted is the sign on the coefficient for the interaction term,

which is the primary variable of interest, for the absolute value of discretionary working capital accruals (ADWAC). Based on the first hypothesis, the sign is predicted to be negative indicating that firms that implement ERP systems should expect to see a reduction in the absolute value of working capital accruals relative to matched firms that did not implement ERP systems. The results show only a marginally significant negative coefficient on the interaction term

(ERP*IMP=-0.0120; p=0.0922) for the working capital regression model (ADWAC). The interaction term for the long-term accrual model (ADLAC) is also negative but not significant

(ERP*IMP=-0.0011; p=0.6872), which supports the second hypothesis that was stated in the null form. Likewise, the coefficient on the total accrual model (ADTAC) is negative and not significant (ERP*IMP=-0.0108; p=0.1386), which is also consistent with the second hypothesis, and different from the univariate analysis.

The main effect coefficients reflect results similar to those depicted in Figure 3-1 and the t-tests discussed previously. The main effect on the ERP variable is not significant in any of the models, except ADLAC. This indicates that there is not a significant difference in the overall level of the absolute values of discretionary accruals between firms implementing ERP systems and the matched control firms across the ten year period, except for the long-term accruals, which were actually higher for implementers than non-implementers. This is reflected in the Figure 3-1 graphic display, which shows that the lines for implementers and non-implementers are very close together, especially before implementation for ADWAC and ADTAC, but not for ADLAC.

The main effect on the IMP variable is not significant for any of the three models, which indicates that overall the changes are not significant when the two groups are combined. The marginally significant coefficient on the interaction variable ERP*IMP for ADWAC confirms the graphic presentation that shows the two lines separating in the model for discretionary working capital

accruals, and that the rate of separation for long-term and total accruals is not significant although moving in the same direction.

The control variables are all significant with the exception of Y2K in the ADLAC model, suggesting that they are having an impact on the overall results. The negative coefficients on the

SIZE, ROA, and MFG variables suggest that larger, more profitable, and manufacturing firms contribute more to the decrease in discretionary accruals than smaller, less profitable and non- manufacturing firms for all three models. The positive coefficient on Y2K indicates that early adopters of ERP systems contribute more to the decrease than late adopters, with respect to short- term working capital accruals and total accruals, but not long-term accruals.

[Insert Table 3-6 here ]

[Insert Table 3-7 here ]

Overall, the results provide weak support for the first hypothesis (H1a), indicating that earnings management activities, as measured by changes in discretionary short-term working capital accruals, may be affected by implementation of ERP systems, but that long-term and total discretionary accruals may not be affected. Because the results were weak for the original ten year pooled sample of 2,638 observations, additional regressions were run using various combinations of specifically matched pairs for two, three, four, and five years before and after the implementation event. The original sample of 147 pairs of matched firms all had data available for the two years before and two years after implementation resulting in 1,176 firm-year observations. For the three years before and after analysis, 134 of the pairs had all the data needed; yielding 1,608 firm-year observations. Only 99 pairs had data for the four years before and after for 1,584 observations, and 71 pairs had enough data for the entire ten years, five before and five after. For all of these matched pair regressions (not tabulated) the significance of the variable of interest dropped below generally accepted levels (i.e. p > 0.10). This is an indication

that the changes in working capital accruals may not necessarily be associated with implementation of an ERP system.

3.4.3 Earnings Quality Results

Daily returns were extracted from the Center for Research in Security Prices (CRSP) database for the ERP implementing firms and the matched control firms for the period 1990 through 2005. One of the ERP implementing firms and five of the control firms do not have data available in CRSP because they are traded on the over the counter market. These daily returns are merged with a data file from Jennifer Francis at Duke University with daily risk free rates, market rates, and AQfactors for the same time period 24 . Table 3-8, Panel A provides descriptive statistics for the resulting 925,591 observations that are then used in annual firm specific regressions using equation 12. The resulting coefficients from the firm-year regressions are summarized in Panel B.

[Insert Table 3-8 here ]

The resulting coefficient on the AQfactor, referred to as the e-loading factor by Ecker et al. (2006), is the variable of interest for this analysis. It is merged with firm-year data from the discretionary accruals analysis in the previous section. Only e-loading factors from annual regressions with at least 10 observations are retained based on the assumptions that it takes a minimum of 10 observations to generate reliable coefficients each year. The data is further reduced to include only those observations where the e-loading factor is available for both the

ERP implementing firm and the matched control firm each year. Table 3-9 summarizes descriptive statistics for the e-loading factor for each of the five years before and after

24 The Francis file contains daily data from 1970 – 2005, which is why 2005 is the last year used in this analysis. The AQfactors are computed based on the methodology described in Ecker et al. (2006). I would like to thank Jennifer Francis for making this data available.

implementation. Panel A shows information for implementing firms, Panel B for control firms, and Panel C summarizes all firms in the sample. Panel D summarizes the data for each group before and after implementation.

[Insert Table 3-9 here ]

Figure 3-2 provides a graphical representation of the e-loading means from Table 3-9.

The first graph plots the year by year mean values for ERP implementing and control firms on separate lines. The line for ERP implementing firms is lower than that of the control firms, indicating a better quality of earning, since higher e-loading factors are an indication of lower perceived earnings quality. The trend for both groups appears to be down, although there is no distinct changed at the point of ERP implementation. The second graph shows the mean values grouped by before and after values for each group, again with the ERP firms reflecting lower average e-Loading factors than the control group. The slope of both lines is downward, although the slope for the control firms appears to decrease more rapidly than for ERP implementing firms.

[Insert Figure 3-2 here ]

Equation 13 provides a model for regression of the e-loading factor on the same partitioning and control variables uses in the discretionary accruals model. Table 3-10 provides a correlation matrix for the variables used in the regression, and Table 3-11 summarizes results.

The results support the graphical presentation discussed above in that the coefficient on ERP and

IMP are both negative and at least marginally significant, with p-values of 0.0514 and 0.0024 respectively. This indicates that ERP implementing firms have lower e-loading factors than the control firms, and that the factors are lower for both groups after the ERP implementation event.

The positive coefficient on the interaction term would indicate that the decrease is less for implementers than for the control group as depicted by the faster sloping line in the graph.

However, the p-value of 0.2827 shows that the difference is not significant, which supports the null form of hypothesis (H1c).

[Insert Table 3-10 here ]

[Insert Table 3-11 here ]

3.4.4 Earnings Distribution Results

Two variables are used to examine the earnings distribution hypothesis: earnings and earnings change. Earnings is based on net income (Compustat mnemonic NI) scaled by beginning of the year total assets (Compustat mnemonic AT). Earnings change is based on NI t –

NIt-1 scaled by total assets at the beginning of year t-1. The required data was extracted from

Compustat for the five years before and five years after ERP implementation for all firms.

Following Burgstahler and Dichev (1997) the resulting earnings (changes) are then grouped into interval widths of 0.005 (0.0025).

Figure 3-3 shows graphical presentations of the frequency of earnings between -0.2500 and +0.2500 grouped by ERP implementers and non-implementers pre and post implementation.

The dotted line represents zero reported earnings with the first bar to the left of the dotted line equal to the frequency of small negative earnings (between -0.005 and 0.000) and the first bar to the right of the dotted line equal to the frequency of small positive earnings (between 0.000 and

0.005) . A pattern similar to that reported by Burgstahler and Dichev (1997) is obvious in some of the graphs, especially those before the implementation event, in that the size of the first bar to the left of zero is smaller than the first bar to the right of zero. The pattern is not as obvious in the post implementation graphs, especially for the control firms.

[Insert Figure 3-3 here ]

Figure 3-4 provides similar graphical displays of the frequency of earnings changes between -0.1500 and +0.1500. The change in patterns between pre and post implementation

charts is more apparent in these graphs, with the post implementation graphs for both ERP firms and the control group reflecting smoother earnings around zero.

[Insert Figure 3-4 here ]

Although the graphs in Figures 3-3 and 3-4 provide a visual image of the activity near zero, the differences need to be measured and compared to test (H1d). Table 3-12 provides a summary of information for the small losses and small negative changes in earnings, those in the first interval below zero. Panel A has information for small earnings losses, and Panel B has information for small negative changes in earnings. The actual frequency of earnings (changes) for each category of firm pre and post implementation is compared to an expected frequency, which is developed based on the average of the interval just above and below. The difference between the actual and expected values is then tested using the same test statistic developed by

Burgstahler and Dichev (1997), which is the difference divided by the standard deviation of all the differences in the sample.

The results show the difference in small earnings losses increasing for ERP implementing firms (-2 to -6) but decreasing for control firms (-5 to +1), whereas the small negative earnings changes show both groups decreasing (ERP firms from -4.5 to 0; control firms from -1 to-0.5).

The test statistic however suggests that none of the differences are significant, in that all of the test statistics are less than the level in most tables that would be considered significant at the 0.10 level, given the sample sizes. These results tend to support the null hypothesis (H1d) that there is no difference in earnings management activity as measured by small earnings losses and negative changes in earnings.

[Insert Table 3-12 here ]

3.5 CONCLUSIONS

This essay uses three different proxies developed in prior accounting research to examine the relationship between ERP systems and earnings management for a sample of 147 firms that implemented ERP systems between 1994 and 2002. The first proxy is discretionary accruals using a cross-sectional version of the modified Jones (1991) model to estimate the absolute value of discretionary short-term, long-term, and total accruals for a five year period before and after implementation. Univariate results show that the proxy for earnings management decreases for all firms in the sample following the implementation event, with ERP implementing firms decreasing more than a control sample of firms that did not implement such systems. This is especially true with respect to short-term (working capital) accruals, which are hypothesized to be affected more by an ERP system than long-term accruals. A multivariate analysis that controls for potential correlated omitted variables retains a marginally significant difference with respect to the change in working capital accruals, but not for long-term or total accruals. Additional robustness tests that match specific firms over two, three, four, and five year periods of before and after data do not hold up, with the differences losing their significance at generally accepted levels (i.e. p<0.10). These results suggest that any association between implementation of an

ERP system and the level of earnings management as measured by discretionary accruals is weak at best, and then only for short-term working capital accruals. It is possible that the changes identified in the univariate analysis are being driven more by the other control variables, and that the relationship to ERP implementation is either a matter of coincidence or possibly due to self selection bias. It is possible for instance, that firms that implement ERP systems are the firms that also possess characteristics that motivate a change in discretionary accrual patterns.

The second proxy is the Ecker et al. (2006) e-loading factor, which measures the markets perception of earnings quality. This proxy, which can only measure the total level of earnings

management (not short-term vs. long-term), has an advantage over discretionary accruals in that it can be applied to shorter time periods. A year by year analysis of the e-loading factors shows a pattern of better quality earnings for ERP implementing firms than for the control firms in each of the five years before and after implementation. A multivariate regression of the e-loading factor using the same control variables as the discretionary accruals analysis suggests that ERP implementing firms have better quality earnings overall, and that both groups experience an improvement in earnings quality after the implementation event. However, the improvement is not significantly better for the ERP implementing firms than it is for the control firms.

The third proxy is the frequency of small earnings losses and small negative changes in earnings developed by Burgstahler and Dichev (1997). The results suggest some improvement in the pattern of avoiding small negative earnings changes for both groups and small earnings losses for the control firms, but none of the differences are significant using traditional levels. The results, consistent with the null hypothesis, suggest that there is no association between earnings management and implementation of ERP systems as measured by small earnings losses or small negative earnings changes.

Taken as a whole, the analysis of these three proxies suggests that the overall level of earnings management is not significantly associated with implementation of ERP systems. The results do suggest that short-term working capital accruals may be marginally associated, and that earnings quality is perceived to be better for firms that implement ERP systems, although any changes as a result of implementing the system are not significantly different from non-ERP firms. It is possible that these results may imply a self-selection bias in that firms that implement

ERP systems may be the ones more likely to have better quality earnings and better manage their working capital without regard to the system implementation itself.

This essay has implications for both practitioners and researchers, considering the added emphasis on earnings management following the recent accounting scandals, and the amount of money that has been spent on these systems. It suggests that the added transparency of these systems by itself may not affect earnings management behavior in the manner that agency theory would suggest. Specifically, the results suggest that either the amount of transparency added by these systems is insufficient to deter earnings management behavior, or the methods used to measure earnings management are unable to detect the changes in behavior that may result from such transparency. It is also possible that because of the effects of consolidation accounting, the type of working capital accrual information used to measure earnings management at the division level is getting lost in the noisy proxy.

This essay has several limitations that should be taken into consideration when applying the results. First, like other earnings management studies, the use of discretionary accruals is a noisy proxy for earnings management. McNichols (2000) argues that traditional aggregate approaches to measuring earnings management (i.e. Jones and modified Jones models) can be biased if there are correlated omitted variables. Also, because the sample was self selected, based on firms that implemented ERP systems, and because the matched firms were selected based on specific criteria, the sample does not represent an independent random sample that is required for most statistical interpretations and for generalization of the results. Perhaps the biggest limitation is the methodology used to select the control sample, which relied on first matching by industry based on SIC codes, with size used as a secondary factor. The result is a sample in which the

ERP implementers are significantly larger than the control firms. This is explained in part by the high cost of these systems and the fact that most of the largest corporations have implemented them, leaving only smaller firms in each industry to use as control firms. A related issue is the use of public announcements to determine if a firm qualifies as a control firm, which may bias the

results even further. It is possible that firms selected as control firms have indeed implemented

ERP systems, but have not made any public announcement about it. Future research in this area should explore alternative methods for developing a control sample.

Future researchers may want to address some of the above limitations. For instance,

McNichols (2000) argues that alternative approaches to earnings management, may have merit, especially measurement of specific accruals such as receivables, inventory, and accounts payable, similar to the approach used by Beneish (1997). This would require access to more detail levels of data than are generally available in public databases, but could provide valuable insight into the make up of discretionary accruals. Another area for future research consideration is to further define the extent of the ERP implementation projects. In particular, it would be interesting to see if there are differences that could be identified based on the number of ERP modules implemented (scope), the level of ERP system usage throughout the firm (diffusion), the specific

ERP vendor software being used, and the implementation process that was used (big bang vs. sequential roll out). These kinds of factors may contribute to further understanding the impact that ERP systems may or may not have on earnings management. However, access to the detail level of information needed will require a survey or case study type of methodology that is beyond the scope of this essay.

Figure 3-1 Mean Absolute Value of Discretionary Accruals ADWAC = Absolute Value of Discretionary Working Capital Accruals Scaled by Lagged Total Assets ADLAC = Absolute Value of Discretionary Long -Term Accruals Scaled by Lagged Total Assets ADTAC = Absolute Value of Discretionary Total Ac cruals Scaled by Lagged Total Assets ERP = 0 Matching Non -ERP Implementing Firms ERP = 1 ERP Implementing Firms IMP = 0 Five Years Before ERP Implementation IMP = 1 Five Years After E RP Implementation

ADWAC

0.075

0.070

0.065 ERP=0 0.060 ERP=1 0.055

0.050 IMP=0 IMP=1

ADLAC

0.030

0.029

0.028 ERP=0 0.027 ERP=1 0.026

0.025 IMP=0 IMP=1

ADTAC

0.080 0.075

0.070 ERP=0 0.065 ERP=1

0.060 0.055 IMP=0 IMP=1

Figure 3-2 Mean Value of e-loading Factors Five Years Before and After Implementation ERP=1: ERP Implementing Firms ERP=0: Matching Control Fi rms IMP=1: Five Years After Implementation IMP=0: Five Years Before Implementation

e-Loading Means by YRS

0.400 0.350 0.300 0.250 0.200 0.150 0.100 -5 -4 -3 -2 -1 1 2 3 4 5 YRS

ERP=1 ERP=0

e-Loading Means Years -5 to +5

0.350 0.300 0.250 0.200 0.150 0.100 IMP=0 IMP=1

ERP=1 ERP=0

Figure 3-3 Frequency of Earnings Intervals

Earnings: ERP Firms Pre-Implementation

30 25 20 15 10

Frequency 5 0 0.0050 0.0300 0.0550 0.0800 0.1050 0.1300 0.1550 0.1800 0.2050 0.2350 -0.2250 -0.1150 -0.0750 -0.0500 -0.0250 Intervals of .005

Earnings: Control Firms Pre-Implementation

25 20 15 10 5 Frequency 0 0.0100 0.0350 0.0600 0.0850 0.1100 0.1350 0.1600 0.1900 0.2250 -0.2200 -0.1450 -0.0950 -0.0700 -0.0450 -0.0200 Intervals of .005

Earnings: ERP Firms Post-Implementation

30 25 20 15 10 5 Frequency 0 0.0100 0.0350 0.0600 0.0850 0.1100 0.1350 0.1600 0.1850 0.2100 -0.2300 -0.1600 -0.1250 -0.1000 -0.0700 -0.0450 -0.0200 Intervals of .005

Earnings: Control Firms Post-Implementation

30 25 20 15 10

Frequency 5 0 0.0050 0.0300 0.0550 0.0800 0.1050 0.1300 0.1550 0.1950 -0.2450 -0.2150 -0.1600 -0.1350 -0.1050 -0.0750 -0.0500 -0.0250 Intervals of .005

Figure 3-4 Frequency of Earnings Change Intervals

Earnings Change: ERP Firms Pre-Implementation

25 20 15 10 5 Frequency 0 0.0150 0.0300 0.0450 0.0625 0.0775 0.0925 0.1075 0.1225 0.1375 -0.1500 -0.1175 -0.0975 -0.0825 -0.0650 -0.0500 -0.0325 -0.0175 -0.0025 Intervals of .0025

Earnings Change: Control Firms Pre-Implementation

20 15 10 5 Frequency 0 0.0125 0.0275 0.0425 0.0575 0.0725 0.0875 0.1075 0.1225 0.1450 -0.1500 -0.1300 -0.1050 -0.0850 -0.0675 -0.0500 -0.0350 -0.0200 -0.0050 Intervals of .0025

Earnings Change: ERP Firms Post-Implementation

25 20 15 10 5 Frequency 0 0.0075 0.0225 0.0375 0.0525 0.0675 0.0825 0.1000 0.1175 0.1425 -0.1475 -0.1225 -0.1025 -0.0875 -0.0700 -0.0550 -0.0400 -0.0250 -0.0100 Intervals of .0025

Earnings Change: Control Firms Post-Implementation

20 15 10 5 Frequency 0 0.0100 0.0275 0.0450 0.0625 0.0850 0.1025 0.1200 0.1400 -0.1475 -0.1200 -0.0975 -0.0800 -0.0625 -0.0450 -0.0275 -0.0100 Intervals of .0025

Table 3-1 Summary of Sample Selection Process

Panel A – Sample Selection Sequence: Firms Initial ERP announcements from Hayes et al. (2001) 91 Less firms no longer listed or data otherwise not available (36) Remaining firms from Hayes et al. (2001) 55 Additional ERP announcements collected from Lexis -Nexis search 92 Firms implementing ERP systems used in this study 147 Control firms not implementing ERP systems 147 Total firms used in this study 294

Panel B – Comparison of Firm Sizes: Market Capitalization Total Assets ($Millions) Control ERP Control ERP Firms Firms Firms Firms N 147 147 147 147 Mean 1,691.3 5,799.7 1,934.8 3,584. 7 Median 606.3 1,268.6 587.2 915.6 Standard Deviation 2,896.5 13,091.1 3,584.7 7,067.9 Minimum 3.5 3.7 1.6 1.0 Maximum 19,705.1 83,683.7 34,369.00 36,147.0

Test for differences: t stat = 3.72 p = .0002 t stat = 2.48 p = 0.0139 Wilcoxon p = .00 11 Wilcoxon p = 0.0341

Note: Market Capitalization is as of the end of the implementation year and total Assets are as of the beginning of the implementation year.

Table 3-2 ERP Implementing Firms by 2 Digit SIC Code and Implementation Year

2 Digit SIC Codes 94 95 96 97 98 99 00 01 02 03 Total 01 -Agricultural Production Crops 1 1 13 -Oil and Gas Extraction 1 1 3 2 1 8 16 -Heavy Construction 1 1 20 -Mfg: Food and Kindre d Products 2 1 2 1 1 1 8 23 -Mfg: Apparel 1 1 1 3 24 -Mfg: Lumber & Wood Products 1 1 25 -Mfg: Furniture & Fixtures 1 2 1 4 26 -Mfg: Paper & Allied Products 1 1 1 1 4 27 -Mfg: Printing & Publishing 1 1 1 1 4 28 -Mfg: Chemicals 1 1 2 2 1 4 2 3 2 18 29 -Mfg: Petroleum Refining 1 1 30 -Mfg: Rubber & Misc. Plastic 1 1 33 -Mfg: Primary Metal Industries 1 1 1 3 34 -Mfg: Fabricated Metal Products 1 1 1 3 35 -Mfg: Ind. & Com. Machinery 1 5 8 2 4 1 1 22 36 -Mfg: Electronic & Elect. Equip. 1 2 5 3 3 4 18 37 -Mfg: Transportation Equipment 1 1 1 1 1 5 38 -Mfg: Measuring & Control Instr. 1 2 1 1 1 2 1 1 10 39 -Mfg: Misc. Manufacturing 1 1 42 -Motor Freight Transportation 1 1 45 -Transportation by Air 2 2 48 -Communications 1 1 2 49 -Electric, Gas & Sanitary Services 2 2 50 -Wholesale: Durable Goods 2 1 3 51 -Wholesale: Non -Durable Goods 1 1 52 -Retail: Bldg Materials, Hardwar e 1 1 54 -Retail: Food Stores 1 1 59 -Retail: Miscellaneous 1 3 2 6 63 -Insurance Carriers 1 1 2 67 -Holding & Other Investment 1 1 2 73 -Automotive Repair & Service 1 2 3 6 80 -Health Services 1 1 87 -Engineering, Actg., R&D, Mgt 1 1 Totals 1 8 8 15 28 34 18 20 9 6 147

Table 3-3 Descriptive Statistics used with Equations (1) and (4) Panel A: Variables used in the models based on 107,735 observations after winsorizing at 1% and 99%

Panel B: TAC it /TA it-1 = a 1(1/TA it-1) + a 2(∆ REV it −∆ REC it )/TA it-1 + a 3(PPE it )/TA it-1+a 4(ROA it ) + εit (1) Panel C: WAC it /TA it-1 = a 1(1/TA it-1) + a 2(∆ REV it −∆ REC it )/TA it-1 +a 3(ROA it ) + εit (4)

Panel A – Model Variables (N=107,735): TAC WAC TA/1 ∆∆∆Rev-∆∆∆Rec PPE ROA Mean -0.0549 0.0036 0.1667 0.1358 0.5837 -18.9418 Median -0.0403 0.0034 0.0100 0.0506 0.4345 1.8630 Std Dev 0.2763 0.2594 0.6759 0.4725 0.5551 76.0112 Minimum -1.6576 -1.4090 <.00 01 -1.1889 0.0000 -560.3570 Maximum 1.0501 1.1733 5.6180 2.7407 3.1887 31.7030

Panel B – Total Non-Discretionary Accrual Coefficients (N=714): 2 a1 . a2 . a3 . a4 . R . Mean 0.0514 0.0189 -0.0652 0.0017 0.4491 Median 0.0276 0.0253 -0.0660 0.0013 0.4131 Std Dev 1.6252 0.1652 0.0816 0.0039 0.2463 Minimum -18.3123 -0.7291 -0.3924 -0.0319 0.0151 Maximum 8.4885 0.8543 0.4321 0.0320 0.9967

Panel C – Working Capital Non-Discretionary Accrual Coefficients (N=714): 2 a1 . a2 . a3 . R . Mean 0.0956 0.0397 0.0016 0.3535 Median 0.0396 0.0422 0.0010 0.2831 Std Dev 1.8223 0.1599 0.0031 0.2606 Minimum -29.3816 -0.5604 -0.0197 0.0069 Maximum 14.4610 0.7 435 0.0274 0.9918

This table presents the summary statistics for 107,735 observations and 714 industry-year regressions from 1990-2006, where industries are classified by the first three digits of the SIC code, except for those that have less than 5 observations, where two digit SIC codes are used . Variables are defined as follows, with Compustat reference in parenthesis:

TAC it = Total Accruals in year t for firm i ( ∆A4 – ∆A1 – ∆Α 5 + ∆A34 – A14)* WAC it = Working Capital Accruals in year t for firm i ( ∆A4 – ∆A1 – ∆Α 5 + ∆A34) TA it-1 = Total Assets (A6) in year t-1 for firm i ∆REV it = Revenue (A12) in year t less revenues in year t-1 for firm i ∆RECit = Receivables (A2) in year t less receivables in year t-1 for firm i PPE it = Net property plant and equipment (A8) in year t for firm i ROA it = Return on Assets (Computstat mnemonic ROA) in year t for firm i εit = Prediction error (residual) in year t for firm i

* A4=Total Current Assets; A1=Cash & Equivalents; A5=Total Current Liabilities; A34=Current Portion of Long-Term Debt; A14=Depreciation & Amortization

Table 3-4 Descriptive Statistics for Absolute Value of Discretionary Accrual Variables Absolute value of discretionary accruals computed after applying coefficients from the modified Jones model to the following equations for the five years before and five years after ERP implementation: NTAC it = α1(1/TA it-1) + α2(∆ REV it −∆ REC it )/TA it-1 + α3PPE it /TA it-1 + α4ROA it /TA it-1 (2) DTAC it = TAC it – NTAC it (3) NWAC it = α1(1/TA it-1) + α2(∆ REV it −∆ REC it )/TA it-1 + α3ROA it /TA it-1 (5) DWAC it = WAC it – NWAC it (6) DLAC it = DTAC it – WAC it (7) ADLAC it = |DLAC it | ADWAC it = |DWAC it | ADTAC it = |DTAC it | N Mean Median Std Dev Minimum Maximum ADWAC; ERP=0; IMP=0 654 0.0710 0.0411 0.0947 <0.0001 1.1260 ADWAC; ERP=0; IMP=1 665 0.0693 0.0381 0.1166 <0.0001 1.3435 ADWAC; ERP=1; IMP=0 654 0.0665 0.0402 0.0988 <0.0001 1.2721 ADWAC; ERP=1; IMP=1 665 0.0526 0.0334 0.0586 0.0001 0.4256 ADLAC; ERP=0; IMP=0 654 0.0260 0.0189 0.0404 <0.0001 0.8134 ADLAC; ERP=0; IMP=1 665 0. 0260 0.0171 0.0342 <0.0001 0.4721 ADLAC; ERP=1; IMP=0 654 0.0290 0.0205 0.0308 <0.0001 0.3565 ADLAC; ERP=1; IMP=1 665 0.0282 0.0189 0.0308 <0.0001 0.3244 ADTAC; ERP=0; IMP=0 654 0.0763 0.0485 0.0958 0.0002 1.0523 ADTAC; ERP=0; IMP=1 665 0.0726 0.0421 0.1143 <0.0001 1.5627 ADTAC; ERP=1; IMP=0 654 0.0748 0.0502 0.1078 <0.0001 1.6286 ADTAC; ERP=1; IMP=1 665 0.0598 0.0448 0.0592 <0.0001 0.4022 ERP=0 Matched Firms that did not implement ERP system ERP=1 Firms that Implemented ERP systems IMP=0 Five Fis cal Years Before Implementation IMP=1 Five Fiscal Years After Implementation TAC = Total Accruals WAC = Working Capital Accruals NTAC = Non -Discretionary Total Accruals NWAC = Non -Discretionary Working Capital Accruals DTAC = Disc retionary Total Accruals DWAC = Discretionary Working Capital Accruals DLAC = Discretionary Long -Term Accruals ADTAC = Absolute Value of Discretionary Total Accruals ADWAC = Absolute Value of Discretionary Working Capital Accruals ADLAC = Absolute Value of Discretionary Long -Term Accruals

Table 3-5 Test of Equal Means for Absolute Value of Discretionary Accruals

Absolute value of discretionary accruals computed after applying coefficients from the modified Jones model for the five years before and five years after ERP implementation:

Panel A – Absolute Value of Discretionary Working Capital Accruals (ADWAC) ERP = 0 ERP = 1 Difference t-statistic p-value IMP = 0 0.0710 0.0665 0.0045 0.8320 0.2028 IMP = 1 0.0693 0.0526 0.0167 3.2981 0.0005 Difference 0.0017 0.0139 t-statistic 0.2953 3.1274 p-value 0.3839 0.0009

Panel B – Absolute Value of Discretionary Long-Term Accruals (ADLAC) ERP = 0 ERP = 1 Difference t-statistic p-value IMP = 0 0.0260 0.0290 -0.0030 -1.4975 0.1345 IMP = 1 0.0260 0.0282 -0.0022 -1.1885 0.2340 Difference 0.0000 0.0008 t-statistic -0.0177 0.4818 p-value 0.9859 0.6300

Panel C – Absolute Value of Discretionary Total Accruals (ADTAC) ERP = 0 ERP = 1 Difference t-statistic p-value IMP = 0 0.0763 0.0748 0.0015 0.2769 0.7819 IMP = 1 0.0726 0.0598 0.0128 2.5624 0.0105 Difference 0.0037 0.0150 t-statistic 0.6479 3.13 76 p-value 0.5172 0.0017

ERP=0 Matched Firms that did not implement ERP system ERP=1 Firms that Implemented ERP systems IMP=0 Five Fiscal Years Before Implementation IMP=1 Five Fiscal Years After Implementation

Table 3-6 Pearson Correlation Matrix for Earnings Management Regressions

ADWAC ADLAC ADTAC ERP IMP ERPIMP SIZE ROA Y2K

ADWAC 1.0000

ADLAC .2070 1.0000 (<.0001) ADTAC .9251 .3117 1.0000 (<.0001) (<.0001) ERP -.0561 .0371 -.0373 1.0000 (.0040) (.0565) (.0553) IMP -.0414 -.0057 -.0485 .0000 1.0000 (.0334) (.7698) (.0128) (1.0000) ERPIMP -.0750 .0147 -.0663 .5806 .5757 1.0000 (.0001) (.4508) (.0007) (<.0001) (<.0001) SIZE -.2425 -0.871 -.2224 .1556 .1243 .1691 1.0000 (<.0001) (<.0001) (<.0001) (<.0001) (<.0001) (<.0001) ROA -.1702 .1783 -.1315 .0819 -.0738 -.0039 .1232 1.0000 (<.0001) (<.0001) (<.0001) (<.0001) (<.0001) (.8409) (<.0001) Y2K .0280 -.0023 .0347 .0000 -.0332 -.0191 .0705 -.0678 1.0000 (.1500) (.9070) (.0748) (1.0000) (.0885) (.3269) (.0003) (.0005) MFG -.0108 -.0729 -.0163 .0000 .0103 .0059 -.1650 .0049 .0945 (.5776) (.0002) (.4031) (1.0000) (.5974) (.7611) (<.0001) (.8002) (<.0001)

ADTAC = Absolute value of discretionary total accruals ADWAC = Absolute value of discretionary working capital accruals ADLAC = Absolute value of discretionary long-term accruals ERP = (1) for firms implementing ERP and (0) for matching control firms. IMP = (1) for post implementation periods, and (0) for pre implementation periods. (This analysis uses five years before and five years after implementation) ERP*IMP = interaction term based on ERP x IMP. SIZE = Natural log of total assets at beginning of year t for firm i used as control variable. ROA = Return on Assets. Y2K = (1) for ERP implementations after 2000 and (0) for before 2000. MFG = (1) for firms with SIC codes (2000 to 3999) and (0) for others.

P-values in parentheses. Note that th e dependent variables (ADWAC, ADLAC, & ADTAC) are not used in the same regressions.

Table 3-7 Earnings Management Regression Results

Results from the following earnings management regression models: ADTAC it = a it + b 1ERP it + b 2IMP it + b 3ERP*IMP it + b 4SIZE it + b 5ROA it + b 6Y2K it + b 7MFG it + εit (9) ADWAC it = a it + b 1ERP it + b 2IMP it + b 3ERP*IMP it + b 4SIZE it + b 5ROA it + b 6Y2K it + b 7MFG it + εit (10) ADLAC it = a it + b 1ERP it + b 2IMP it + b 3ERP*IMP it + b 4SIZE it + b 5ROA it + b 6Y2K it + b 7MFG it + εit (11)

ADWAC ADLAC ADTAC Exp Variable Sign Coeff p-value Coeff p-value Coeff p-value

Intercept ? 0.1454 <.0001 0.0403 <.0001 0.1485 <.0001 ERP ? 0.0044 0.4213 0.0050 0.0073 0.0065 0.2631 IMP ? 0. 0020 0.7219 <.0001 0.9593 0.0002 0.9684 ERP*IMP (a) -0.0120 0.0922 -0.0011 0.6872 -0.0108 0.1386 SIZE ? -0.0116 <.0001 -0.0016 0.0154 -0.0112 <.0001 ROA ? -0.0005 0.0410 -0.0002 0.0353 -0.0004 0.0203 Y2K ? 0.0078 0.0535 <.0001 0.9895 0.0097 0.0234 MFG ? -0.0109 0.0170 -0.0066 0.0019 -0.0121 0.0104

F-statistic 34.53 <0.0001 18.70 <0.0001 26.94 <0.0001 Adjusted R 2 0.0817 0.0449 0.0644 N 2638 2638 2638

(a) Expected sign for ADWAC model is (-), the other two models are (?) P-values are based on White adjusted t-statistics, and two-tail test. Analysis of Variance Inflation Factors (VIF) indicates there is not a multicollinearity problem.

ADTAC it = Absolute value of discretionary total accruals in year t for firm i ADWAC it = Absolute value of discretionary working capital accruals in year t for firm i ADLAC it = Absolute value of discretionary long-term accruals in year t for firm i ERP = (1) for firms implementing ERP and (0) for matching control firms. IMP = (1) for post implementation periods, and (0) for pre implementation periods. (This analysis uses five years before and five years after implementation) ERP*IMP = interaction term based on ERP x IMP. SIZE = Natural log of total assets at beginning of year t for firm i used as control variable. ROA = Return on Assets. Y2K = (1) for ERP implementations after 2000 and (0) for before 2000. MFG = (1) for firms with SIC codes (2000 to 3999) and (0) for others.

Table 3-8 Descriptive Statistics for Equation 12 Descriptive statistics used in the following regression model :

Rj,t – RF,t = αj,t + βj,T (R M,t – RF,t ) + e j,T AQfactor t + εj,t (12)

N Mean Median Std Dev Minimum Maximum Panel A – Raw Data from 1990 to 2005: Variables Rj 925,591 0.0913% 0.0000% 3.7174% -71.7510% 172.0000% Rf 925,591 0.0153% 0.0180% 0.7130% 0.3000% 3.4000% Rj – Rf 925,591 0.0760% -0.0190% 3.7175% -71.7660% 171.9800% Rm – Rf 925,591 0.0289% 0.0600% 1.0145% -6.6500% 5.3100% AQ factor 925,591 0.0013% 0.0015% 0.9582% -7.5719% 7.9256%

Panel B – Results from Firm-Year Regression: Coefficients Intercept 3,762 0.0002 0.0002 0.0026 -0.0569 0.0263 RmRf 3,762 0.8945 0.7984 0.7566 -12.2280 11.3553 Aqfactor 3,762 0.2224 0.1188 0.5228 -2.4997 4.7773 R-Squared 3,762 0.1300 0.0901 0.1220 <0.0001 0.8596

Rj = Daily returns from CRSP database for ERP implementers and control firms Rf = Risk free daily returns from Francis website Rm – Rf = Daily market returns in excess of risk free returns from Francis website AQfactor = Daily AQfactor from Francis website

Table 3-9 Descriptive Statistics for e-Loading Five Years Before and After ERP Implementation

Year N Mean Median Std Dev Minimum Maximum Panel A – ERP Implementing Firms: -5 84 0.1756 0.1043 0.4168 -1.0456 1.5550 -4 96 0.2170 0.1677 0.4097 -0.6549 1.3750 -3 111 0.2632 0.1932 0.5296 -0.5913 2.8270 -2 123 0.1910 0.0874 0.4266 -0.5550 1.8262 -1 128 0.1379 0.0359 0.4751 -0.8816 1.3842 1 127 0.1786 0.0573 0.5456 -0.7836 2.2272 2 130 0.1777 0.0206 0.5689 -0.5878 2.6171 3 121 0.1545 -0.0302 0.5996 -0.7791 2.9464 4 106 0.1558 0.0893 0.4903 -0.6659 1.6531 5 88 0.1432 0.0113 0.5587 -1.5863 2.0345 Panel B – Matching Control Firms: -5 84 0.2881 0.2422 0.3565 -0.5358 1.3224 -4 96 0.3587 0.2765 0.4979 -0.9752 1.9386 -3 111 0.2849 0.1986 0.4481 -0.5520 1.5102 -2 123 0.3008 0.1961 0.5026 -0.7708 1.7110 -1 128 0.2978 0.1769 0.5499 -1.7678 1.99 71 1 127 0.2800 0.1423 0.5744 -0.7050 3.3364 2 130 0.1945 0.1031 0.4475 -0.9286 2.2616 3 121 0.1887 0.1250 0.4586 -0.7167 1.4430 4 106 0.2203 0.0540 0.5778 -0.6044 2.4755 5 88 0.2209 0.1187 0.5021 -0.6420 2.0047 Panel C – All Firms: -5 168 0.231 8 0.1848 0.3907 -1.0456 1.5550 -4 192 0.2878 0.2553 0.4603 -0.9752 1.9386 -3 222 0.2740 0.1945 0.4895 -0.5913 2.8270 -2 246 0.2459 0.1588 0.4685 -0.7708 1.8262 -1 256 0.2179 0.0821 0.5191 -1.7678 1.9971 1 254 0.2293 0.0992 0.5614 -0.7836 3.3364 2 260 0.1861 0.0693 0.5109 -0.9286 2.6171 3 242 0.1716 0.0450 0.5330 -0.7791 2.9464 4 212 0.1881 0.0734 0.5355 -0.6659 2.4755 5 176 0.1821 0.0572 0.5311 -1.5863 2.0345 Panel D –Before and After: ERP Implementing Firms: Before 542 0.1955 0.1008 0.4 572 -1.0456 2.8270 After 572 0.1636 0.0287 0.5535 -1.5863 2.9464 Control Firms: Before 542 0.3051 0.2189 0.4821 -1.7678 1.9971 After 572 0.2211 0.1186 0.5130 -0.9286 3.3364

Table 3-10 Pearson Correlation Matrix for E-Loading Regressions

ELOAD1 ERP IMP ERPIMP SIZE ROA Y2K MFG

ELOAD1 1.0000

ERP -0.8191 1.0000 (0.0001) IMP -0.0572 0.0000 1.0000 (0.0069) (1.0000) ERPIMP -0.0661 0.5877 0.572 1 1.0000 (0.0018) (<.0001) (<.0001) SIZE -0.4166 0.1276 -0.0076 0.0639 1.0000 (<.0001) (<.0001) (0.7200) (0.0026) ROA -0.0721 0.0455 -0.0700 -0.0208 0.0638 1.0000 (0.0007) (0.0316) (0.0009) (0.3259) (0.0026) Y2K 0.0167 0.0000 -0.0441 -0.0252 0.0770 -0.0334 1.0000 (0.4311) (1.0000) (0.0375) (0.3441) (0.0002) (0.1153) MFG 0.1416 0.0000 0.0103 0.0059 -0.2154 -0.0169 0.0872 1.0000 (<.0001) (1.0000) (0.6262) (0.7805) (<.0001) (0.4242) (<.0001)

ELoad = coefficient on AQfactor ERP = (1) for firms implementing ERP and (0) for matching control firms. IMP = (1) for post implementation periods, and (0) for pre implementation periods. (This analysis uses five years before and five years after implementation) ERP*IMP = interaction term based on ERP x IMP. SIZE = Natural log of total assets at beginning of year t for firm i used as control variable. ROA = Return on Assets. Y2K = (1) for ERP implementations after 2000 and (0) for before 2000. MFG = (1) for firms with SIC codes (2000 to 3999) and (0) for others.

P-values in parentheses

Table 3-11 e-Loading Regression Results

Results from the following regression model: e-Load it = a it + b 1ERP it + b 2IMP it + b 3ERP*IMP it + b 4SIZE it + b 5ROA it + b 6Y2K it + b7MFG it + εit (13)

Variable Coefficient t-value p-value

Intercept 0.9826 20.84 <.0001 ERP -0.0500 -1.79 0.0514 IMP -0.0838 -3.05 0.0024 ERP*IMP 0.0413 1.07 0.2827 SIZE -0.1165 -20.19 <.0001 ROA -0.0005 -2.44 0.0595 Y2K 0.0426 2.06 0. 0376 MFG 0.0584 2.60 0.0074

F-statistic 72.16 <.0001 Adjusted R 2 0.1828 N 2,228

P-values are based on White adjusted t-statistics, two-tail test. Analysis of Variance Inflation Factors (VIF) indicates there is not a multicollinearity problem.

e-Load = coefficient on AQfactor for firm i, year t. ERP = (1) for firms implementing ERP and (0) for matching control firms. IMP = (1) for post implementation periods, and (0) for pre implementation periods. (This analysis uses five years before and five years after implementation) ERP*IMP = interaction term based on ERP x IMP. SIZE = Natural log of total assets at beginning of year t for firm i used as control variable. ROA = Return on Assets. Y2K = (1) for ERP implementations after 2000 and (0) for before 2000. MFG = (1) for firms with SIC codes (2000 to 3999) and (0) for others.

Table 3-12 Frequency of Actual vs. Expected Earnings (Change) Just Below Zero

Standard Test N Actual Expected Difference Deviation Statistic Panel A – Earnings: Pre Implementation: ERP Firms 71 4 6.0 -2.0 3.1623 -0.6325 Control Firms 71 4 9.0 -5.0 3.1870 -1.5689 Post Implementation: ERP Firms 78 6 12.0 -6.0 6. 2102 -0.9661 Control Firms 78 12 11.0 1.0 2.6726 0.3742

Panel B – Earnings Change: Pre Implementation: ERP Firms 104 10 14.5 -4.5 2.7791 -1.6192 Control Firms 103 7 8.0 -1.0 1.7843 -0.5604 Post Implementation: ERP Firms 102 12 12.0 0.0 2.6649 0.0000 Control Firms 105 11 11.5 -0.5 2.8372 -0.1762

Pre Implementation = five years before ERP system implementation Post Implementation = five years after ERP system implementation N = Number of observations between -0.2500 and 0.2500 for earnings and between -0.1500 and 0.1500 for changes Actual = number of observations with earnings (change) in first increment below zero (-0.005 to 0.0) for earnings and (-0.0025 to 0) for change. Expected = expected number of observations with earnings (change) in first increment below zero based on average of increment just above and just below Std Dev = Standard deviation of differences for all observations between -0.2500 and 0.2500 for earnings and between -0.1500 and 0.1500 for changes Test Statistic = Difference / Std Dev

CHAPTER 4 SHAREHOLDER VALUE (ESSAY TWO)

4.1 Introduction

Since Enterprise Resource Planning (ERP) systems emerged in the 1990s as a major tool to integrate business processes and improve productivity, researchers have used different approaches to measure their impact. Some have used survey data or field studies to assess user satisfaction (Themistocleous et al. 2001; Bradford and Florin 2003; Umble et al. 2003). Others have used traditional accounting and financial metrics such as ROA, ROI, inventory turnover, etc.

(Poston and Grabski 2001; Hitt et al. 2002; Hunton et al. 2003; Nicolaou 2004). Still others have used stock market and financial analyst reactions to the announcements made disclosing the technology investment (Hayes et al. 2001; Hunton et al. 2002). One could argue that these approaches use proxies to measure the impact of the investment in ERP on shareholder value.

Generally, management makes investments on behalf of shareholders, based on an assumption that the investment will benefit the shareholders. Therefore, the ultimate measure of any corporate investment should be the impact it has on shareholder value. This essay extends this prior research by using alternative direct approaches to measure the impact of ERP systems on shareholder value.

Software vendors promote ERP systems based on their ability to improve productivity by reducing redundant processes, streamlining operations, breaking down silos of information, and providing fast and accurate information throughout the organization so that management can make faster, better decisions. To the extent that these factors lead to improved productivity and firm performance, they should manifest themselves ultimately in the form of increased profitability. This increased profitability should lead to higher dividends and/or re-investment of

90 91

retained earnings, which will increase the intrinsic value of the firm (Ohlson 1995). Furthermore, the improved quality and timeliness of information should increase shareholders confidence in management by reducing uncertainty and the associated risk. Therefore, one would expect these events to affect the company’s stock price, which will directly impact shareholder value. This essay examines the impact of ERP systems on shareholder value using three different metrics: long-horizon buy-and-hold returns, market-to-book ratios, and value-to-price ratios.

The first analysis uses long-horizon buy-and-hold returns as a way to measure the impact on shareholder value because ERP systems are long-term investments. The long-term buy-and- hold investors better represent the beneficiaries of the ERP system than short-horizon traders that exploit short-term gaps in information asymmetry to generate returns. The returns analysis uses a sample of 147 firms from 33 industry groups that implemented ERP systems between 1994 and

2003 and a control sample of firms matched by size and industry. The results show that average returns for ERP implementers are significantly higher than the control firms over three different buy-and-hold periods: one-year, three-years, and five-years 25 following implementation.

However, when these results are compared to one-year, three-year, and five-year returns before implementation, the change in returns for ERP implementers relative to the control group is not significant, suggesting that ERP systems have no significant impact on long-horizon returns.

To supplement this analysis, market-to-book (MTB) ratios are used, because they are often cited by both academics and financial professionals as explanatory variables for differences in firm and industry characteristics. It is generally assumed for instance, that firms with higher

MTB ratios represent faster growing firms and/or firms with more intangible assets because the market is pricing the stock higher than the underlying book value. The MTB analysis compares

25 The sample size for the five-year analysis is reduced to 132 ERP implementing firms because 15 firms do not have data available for the fifth year.

average ratios for three, four and five years before implementation to three, four, and five years after implementation for ERP implementers and the control firms. The results consistently show that ERP implementers have higher MTB ratios for all three time periods. The ratios do not change significantly from before to after implementation for either group, indicating a possible self selection bias with higher MTB firms perhaps being more inclined to implement ERP systems.

A third analysis uses the model developed by Frankel and Lee (1998) to estimate the value-to-price (V/P) ratio of the sample firms. The analysis is limited to 103 of the ERP implementing firms (and their matched control firms) that have analyst forecast data available from I/B/E/S. The results show that ERP implementing firms have significantly higher V/P ratios than their control firms at the time of implementation. This finding is consistent with Frankel and

Lee’s (1998) finding that higher V/P ratios are indicators of higher future long-horizon returns.

The results also show that this difference begins to fade after one year, indicating that the market may be catching up with the information and pricing the stock accordingly.

The remainder of this essay is organized as follows: Section 4.2 summarizes prior research and develops the hypotheses, Section 4.3 describes the data selection process and the methodology/models used, Section 4.4 presents empirical results, and Section 4.5 concludes.

4.2 Prior Research and Hypotheses Development

Ohlson (1995) shows that a firm’s intrinsic value is equal to the book value of equity plus the present value of expected future residual income. Ohlson (1995), and Feltham and Ohlson

(1995) provide a foundation for using the stock market as a proxy for this intrinsic value of the firm. The efficient market hypothesis holds that the price of a security is an accurate estimate by the market of its true value, fully reflecting all information that is publicly available (Eakins

2002). Although there is debate in the academic community as to which form of efficiency the

market reflects in the short-term (i.e. strong, semi-strong, or weak), there is evidence that in the long-run, the market is efficient (Fama 1998). Therefore, the market price of a firm’s stock should include anticipated future residual income based on information that becomes publicly available. It follows that as benefits from an ERP system are realized, and the market becomes aware of these benefits, the value of the improvements will be included in the stock price.

Measuring this change in price is one way to assess the impact that ERP systems have on shareholder value. However, a more complete measure may be the total returns experienced by shareholders based on the change in price (capital appreciation) as well as dividends earned over a period of time. From the shareholder’s perspective, this is the ultimate measure of value since investors will generally not maintain their investment in firms that do not provide a rate of return competitive with the market. Accounting researchers use both price and returns to measure shareholder value. Kothari and Zimmerman (1995) examine price models (stock price regressed on earnings per share) and return models (returns regressed on scaled earnings variables), finding that earnings response coefficients are less biased in price models, but return models have less serious econometric problems.

Many accounting and finance researchers using return models focus on abnormal returns, measuring cumulative abnormal returns (CAR), usually over a three day window surrounding some event. A number of these studies seek to exploit timing differences and short-term market inefficiencies to generate abnormal returns (Bernard and Thomas 1989; Freeman and Tse 1989;

Bernard and Thomas 1990; Fama 1991; Abarbanell and Bernard 1992; Fama and French 1992).

This methodology has also been popular with early researchers in the information technology (IT) field (e.g. Dos Santos et al. 1993; Subramani and Walden 2001) and was used by Hayes et al.

(2001) to measure the reaction of the market to announcements made by firms that were

implementing ERP systems. They use standardized cumulative abnormal returns (SCAR) over a three day window, and find an overall favorable reaction to the announcements.

This concept of abnormal returns is generally considered to be a short-term phenomena because, based on the efficient market hypothesis, the stock price adjusts to information fully within a narrow window of a few days (Ang and Zhang 2004). This may be true for announcement type events, but in the case of an ERP system, the actual event takes much longer to occur than one day and the benefits are realized over many future periods. Since firms make public disclosures of earnings at least quarterly, and analysts provide forecast information even more frequently, it is reasonable to assume that each of these disclosures provide implicit additional information about the impact of the ERP investment. Therefore, one could argue that the appropriate measurement period for this type of event will last for many months, or even years, rather than just a few days.

An alternative method of measuring shareholder returns is to measure long-horizon buy- and-hold returns. Easton et al. (1992) argue that the association between returns and earnings are stronger over longer intervals, because value relevant events occurring before and during the return interval have a better chance of being incorporated in the explanatory earnings variable as the window is expanded 26 .

This longer-horizon approach is especially appropriate for ERP systems, which represent major investments expected to improve earnings over a long-term horizon, with an understanding that short-term earnings may suffer during the implementation process. Assuming ERP systems provide a competitive advantage, as advertised by ERP vendors, one would expect that this competitive advantage would manifest itself in the form of increased residual income. This

26 Using R 2 as the measure of association, they find it increases from 5% for one-year return periods to 15% for two-year periods, 33% for five-year periods, and 63% for ten-year periods.

increased residual income will lead to either increased dividend payouts or increased stock prices reflecting the reinvestment of earnings back into the firm, both of which will impact shareholder returns over a long-horizon, which leads to the following hypothesis:

H2a: Implementation of Enterprise Resource Planning (ERP) systems will have a positive impact on shareholder value as measured by long-horizon returns.

A significant part of the returns analysis is dependent on the change in price of the stock over the selected time horizon. Therefore a separate analysis using the price model would be beneficial in further understanding the impact of ERP systems on shareholder value. A number of models have been developed using variations of the Ohlson (1995) model, with price as the dependent variable and book value, earnings, and other accounting information as explanatory variables. Brown et al. (1999) show that variation in the scale factor can impact results and suggest that researchers control for these differences by deflating individual observations by a proxy for scale. Core et al. (2003) control for this by scaling all variables in their model by the book value of equity, which has the effect of creating a model of the market-to-book ratio. The relationship between the market value of a firm and its recorded book value, expressed as a market-to-book (MTB) ratio, or its inverse book-to-market (BTM) ratio, is commonly used by both academics and financial professionals to compare and contrast firms and industries. Fama and French (1992; 1995) find that firms with high BTM (low MTB) ratios tend to be persistently distressed and, conversely, firms with low BTM (high MTB) ratios are associated with sustained strong profitability. Assuming ERP systems contribute to increased residual income as discussed above, one would expect the market to price this increase, which would then be reflected in the

MTB ratios. This leads to the following hypothesis:

H2b: Implementation of Enterprise Resource Planning (ERP) systems will have a positive impact on shareholder value as measured by market-to-book (MTB) ratios.

Accounting and finance researchers have used the residual income model (RIM) first proposed by Ohlson (1995) in a number of different settings to empirically analyze stock price and its relationship to accounting information. Dechow et al. (1999) conclude that the model provides a unifying framework for earnings-based valuation research, even though they found only modest improvements in explanatory power over methods using the dividend-discounting model. Fankel and Lee (1998) operationalize the model by using analyst earnings forecasts to estimate future residual income and compare the resulting intrinsic value (V) with stock price (P) forming a value-to-price (V/P) ratio that they show is a good predictor of long-term cross- sectional returns. Zhou (2002) finds that firms with small arbitrage costs can benefit from a strategy based on the V/P ratio, but those with high arbitrage costs may not. He also documents that forecast accuracy has an impact on the predictive power of the V/P ratio. Ali et al. (2003) replicate Frankel and Lee (1998) finding the V/P effect is partially concentrated around future earnings announcements. Finger and Landsman (2003) find, consistent with theory, that more optimistic analyst recommendations are associated with higher mean forecast errors, forecast revisions, and forecasted earnings-to-price ratios. However, contrary to expectations they also have higher market-to-book ratios, higher market values, and lower value-to-price ratios.

Bradshaw (2004) later refined this model by developing alternative methods of projecting the terminal value component, one that persists, and another that fades. He also examined a price- earnings-growth (PEG) model and a long-term earnings growth model, finding very little correlation between any of the four models and analysts’ recommendations. He argues that buy- and-hold investors would earn higher returns relying on present value models that incorporate analysts’ earnings forecasts than by relying on analysts’ recommendations. Xu (2007) decomposes the Frankel and Lee model to examine the incremental effects of each of the

components that make up the model. She finds that the predictive power of V/P is fully attributable to to an ex post linear combination of its components, especially analyst forecasts of earnings in a regression setting. She argues that these findings suggest that the abnormal returns related to V/P are due to the predictive ability of its components, not the specific transformation form of the residual income model. Therefore, she concludes, V/P is a repackaging of other known anomalies and not a distinct market anomaly.

Hunton et al. (2002) find that forecasted earnings by financial analysts increase significantly following announcements that ERP systems are being implemented, which would be reflected in the intrinsic value of the firm calculated using the Frankel and Lee (1998) model.

The resulting V/P ratio would depend on how fast the market incorporates this information into the stock price. If the information is fully incorporated at the same time the ratio would be near

1.0. Frankel and Lee (1998) show that firms with higher V/P ratios experience higher long-term returns, especially over 24 to 36 months. If implementation of an ERP system is expected to yield higher long-term returns, then one would expect this to be reflected in the V/P ratio leading to the following hypothesis:

H2c: Implementation of Enterprise Resource Planning (ERP) systems will have a positive impact on shareholder value as measured by value-to-price (V/P) ratios.

To test these hypotheses, initial univariate comparisons are supported by multivariate archival analysis using: a long-horizon buy-and-hold return model; a market-to-book value model; and a value-to-price model. Each model is described in more detail in the following section.

4.3 Data Selection and Methodology/Models

4.3.1 Sample Data Selection

The initial sample consists of 147 firms representing 33 industries that implemented ERP systems between 1994 and 2003. Table 4-1 Panel A, provides a summary of the sample selection process for ERP implementing firms used to test long-term buy-and-hold returns in H2a. The sample selection process starts with the list of 91 ERP announcements made between 1994 and

1998 from Hayes et al. (2001) 27 , from which 36 firms were eliminated because they are no longer listed or data was otherwise not available. The second step in the process involves a search of all available newswire services using the Lexis-Nexis service for years after 1998, searching on key phrases such as: “ERP”, “Enterprise Resource Planning”, and “Enterprise Systems.” This search found an additional 92 firm announcements yielding a total of 147 firms for which returns data is available for at least three years following implementation in the Research Insight Compustat database. To examine H2a based on one year before and after returns, seven of these firms were eliminated because they (or their matched control firm) did not have data available one year prior to implementation, leaving 140 firms. Another 25 firms were illuminated for the three year analysis and 34 more for the five year analysis due to missing data, leaving 115 firm with three year before and after returns and 81 with five year before and after returns.

Panel B summarizes the sample selection process for ERP implementing firms used to examine H2b, related to changes in the market-to-book ratio. This analysis compares three different periods before and after implementation to assess the effect of changes in the ratio for

ERP-implementers vs. non-implementers. The three time periods are: (1) three years before and three years after, (2) four years before and four years after, and (3) five years before and five

27 I would like to thank David C. Hayes and Jacqueline L. Reck for providing this list of firms

years after. The selection process begins with the 147 firms from Panel A, and eliminates 22 firms that did not have data for three years before and three years after implementation for the first analysis, another 34 that did not have data four years before and four years after for the second analysis, and another 28 that did not have data available for the five years before and five years after implementation. The result is 125 ERP implementing firms used in the first analysis,

91 in the second and 63 in the third.

Panel C summarizes the selection process for ERP implementing firms used to examine

H2c, related to valuation using the V/P ratio model. It also begins with the initial 147 firms selected in Panel A, and eliminates 44 firms that do not have I/B/E/S forecast data available, resulting in 103 firms used for this analysis.

Following the recommendations of Barber and Lyon (1997), and the approach used by

Nicolaou (2004) 28 , this essay uses a matched pairs approach to selecting a control group. ERP implementing firms are first matched with other firms based on their four digit SIC code, then by total assets at the beginning of the implementation year, then by the availability of Compustat return data for the one, three, or five years following implementation. If a reasonable match is not available then three or two digit SIC codes are used to obtain a match. Once a matched firm is identified, a further search of Lexis-Nexis Newswires is conducted for all available years using a combination of the firm name, and the following terms: (1) ERP, (2) Enterprise Resource

Planning, (3) Enterprise Systems, (4) SAP, (5) Oracle, (6) QAD, (7) Baan, (8) Peoplesoft, (9) JD

Edwards, and (10) Lawson 29 . If no newswire records are found, the firm is used as a match for

28 Firms were matched by the four-digit Standard Industrial Classification (SIC) code first. Second, companies were matched by size, using total assets. The selection rule was to form firm pairs within the same four-digit SIC with the closest asset size match. There were some instances (n = 44 out of 247) were it was necessary to go beyond the four-digit SIC and select a firm using a three-digit match. 29 The specific vendors listed (4-10) represent the seven most common ERP systems identified in Nicolaou (2004).

100

this study. If a record is found that indicates an ERP system may be used, then the next closest firm in terms of total assets is used, and the process repeated.30 Panel D provides a comparison of the mean, median, standard deviation, minimum and maximum values of total assets at the beginning of the implementation year and market capitalization as of the end if the implementation year for each of the two groups. T-tests and Wilcoxon tests indicate that the mean and median values are statistically different from each other, with the ERP firms larger than the control firms on average. Although it would be preferable for the two samples to be closer to the same size, emphasis in the matching process was place on matching based on industry first, and size second, recognizing that it may not be possible to match the size as close because larger firms tend to be the ones that implement ERP systems and publicize the fact with press releases.

This could be a major limitation for this essay, and is discussed further in the conclusions section of this document. However, given the nature of the information available in the public records and the methodology used to search these records, this is the best match available at this time. To control for the size differences, control variables are used in the analysis where appropriate.

[Insert Table 4-1 here ]

Table 4-2 provides a breakdown of firms by two-digit SIC code and implementation year.

ERP announcements are of two types, those made before the system was implemented and those made after the system was implemented. For instance, some of the announcements report that the firm had entered into an agreement with a particular vendor or consulting firm to implement an

ERP system. In other cases the announcements are made by the vendor or a consultant announcing the “successful” implementation of an ERP system. In other cases the

30 Six matches required the use of three digit SIC codes: (201)-Food and Kindred Products Manufacturing, one match; (252)-Furniture and Fixture Manufacturing, two matches; (284)-Chemicals and Allied Products Manufacturing, two matches; and (594)-Miscellaneous Retail, one match. Two digit SIC codes are used for four matches: (28) Chemical Manufacturing, two matches, (35)-Industrial and Commercial Machinery and Computer Equipment, one match; and (38) Measuring, Analyzing, and Control Instruments, one match.

101

announcements are part of a regular earnings announcement indicating that earnings are different than forecast due to implementation of an ERP system. In some cases the announcements do not contain specific implementation dates or time lines. In those cases, judgment was used to determine which year best represents the implementation year 31 . Generally, for announcements made after implementation, the implementation year is assumed to be the firm’s fiscal year in which the announcement is made or the previous year if the announcement is dated early in the fiscal year. For announcements made prior to implementation, the implementation year is assumed to be the fiscal year following the announcement 32 .

[Insert Table 4-2 here ]

Firm specific data selection is based on each firm’s fiscal year, with accounting data as of the end of the fiscal year, and market data as of the end of the fourth month following the end of the fiscal year 33 . This approach, which is common in accounting research, helps to ensure that the market has the latest fiscal year-end accounting information available to incorporate in the price.

For the returns analysis related to H2a, the baseline is the fiscal year that corresponds to the implementation year discussed above, with one, three, and five year buy-and-hold returns calculated from that point. The market-to-book (MTB) analysis related to H2b uses the implementation year as the last year of the baseline and the first year following implementation as the first year of the analysis period. The value-to-price (V/P) analysis related to H2c uses a

31 As an example, if the announcement is made early in the fiscal year indicating that an implementation has been recently completed, then the prior year is used, or if the announcement is made in the middle of the year indicating that an implementation will be taking place, then the following year is used. 32 This method is consistent with Nicolaou (2004) who documents an average implantation of 9.92 months.

33 For instance firms with a calendar fiscal year would have accounting data as of December 31 st and market data as of April 30 th .

102

similar approach for the IBES forecast data, which is based on the last forecast issued in the fourth month following each firm’s fiscal year end 34 .

4.3.2 Long-Horizon Returns Model

Previous researchers have used a number of different techniques to measure long-horizon returns. Some have extended the concept of cumulative abnormal returns beyond the narrow three-day window used for short-horizon studies. Kothari and Warner (1997) find that many of the techniques used to measure long-horizon returns are severely misspecified, including those that use variations of the capital asset pricing model. Barber and Lyon (1997) find similar results for models based on reference portfolios and application of the three-factor model of Fama and

French (1993). Specifically, they find that cumulative abnormal returns calculated using reference portfolios yield positively biased test statistics, and that buy-and-hold abnormal returns calculated using reference portfolios yield negatively biased test statistics. They attribute these contradictory results to the impact of new listings, rebalancing, and skewness biases resulting from the use of reference portfolios. As an alternative, they document that a control firm approach yields test statistics that are well specified in virtually all sampling situations they considered, because it alleviates these new listing, rebalancing, and skewness biases. They also argue that researchers should calculate abnormal returns as the simple buy-and-hold return on a sample firm less the simple buy-and-hold return on a reference portfolio or control firm 35 .

Following Barber and Lyon (1997), a simple long-term buy-and-hold model is used to calculate abnormal returns using the control firm approach with the following equation:

34 Similar to the approach used for market data, the assumption is that the I/B/E/S forecast issued in the fourth month following the fiscal year end will be based on information contained in the firm’s actual audited financial statements. This provides for consistency of information used in the valuation models such that book value and forecasted earnings are based on information as of the same point in time. 35 They document that biases are induced by summing daily or monthly abnormal returns referred to as cumulative abnormal returns.

103

AR i = Ri − BR i (1)

Where: AR i is abnormal return for the ERP implementing firm i, Ri is the long-term buy-and-hold return of the ERP implementing firm i, and BR i is the long-term buy-and-hold return for a benchmark control firm i that did not implement an ERP system. This approach assumes that an investment is made in the stock of the subject firm, and its control firm, as of the end of the fiscal year in which an ERP system is implemented and held through the end of the analysis period.

The returns are taken from Standard & Poor’s Research Insight Compustat database as follows 36 :

1 Year Total Return (Compustat mnemonic TRT1Y)

3 Year Total Return (Compustat mnemonic TRT3Y)

5 Year Total Return (Compustat mnemonic TRT5Y)

To perform an initial univariate test of the first hypothesis, abnormal returns are examined over each of the three time periods using a t-statistic to test the null hypothesis that the mean buy-and-hold abnormal returns are equal to zero vs. the alternative hypothesis that they are greater than zero. Although this provides an initial assessment of the impact on shareholder value, other factors may also be influencing the results. To address this concern, returns for implementers and non-implementers are compared using the following multivariate regression model, which is a modified form of the Ohlson (1995) model for each of the long-horizon time periods of one, three and five years:

RET i = αi + β1ERP i + β2IMP i + β3ERP*IMP i + β4BVPS i + β5EPS i + β6Y2K i + εit (2)

Where:

36 The returns are defined by Compustat as annualized rates of return reflecting price appreciation plus reinvestment of monthly dividends and the compounding effect of dividends paid on reinvested dividends.

104

RET i = total returns for firm i for each of the one, three and five year periods following

implementation (Compustat mnemonic: TRT1Y, TRT3Y, TRT5Y)

ERP = dummy variable set to (1) for ERP implementers, (0) for control firms

IMP = dummy variable set to (1) for post- and (0) for pre-implementation periods

ERP*IMP = interaction term based on ERP x IMP, expected to be positive if ERP

implementation is associated with increased returns relative to non-

implementers.

BVPS = book value per share for firm i at the end of the implementation fiscal year

(Compustat mnemonic: BKVLPS)

EPS = earnings per share change for firm i based on the change in EPS over each of

the one, three and five year periods (Compustat mnemonic: EPSCHG1,

EPSCHG3, EPSCHG5)

Y2K = dummy variable set to (1) for firms that implemented ERP systems (and the

matched control firms) after the year 2000, otherwise, (0) 37 .

4.3.3 Market to Book (MTB) Model

To examine the second hypothesis, this essay uses a cross-sectional price model adopted from Core et al. (2003), which uses MTB as a dependent variable, and several accounting measures as explanatory variables. The model is used to measure MTB over three different time periods: (1) three years before and three years after implementation, (2) four before and four years after implementation, and (3) five years before and five years after implementation.

Following is the model:

37 This control variable is used for two reasons: (1) it addresses the possible differences in efficiencies that may exist between early adopters and late adopters of technology (the learning curve effect), and (2) it address the impact that may be driven by the fact that the stock market changed from a bull market to a bear market in the year 2000.

105

MVE it /BVE it = αi + β1ERP it + β2IMP it + β3ERP*IMP it + β41/BVE it + β5IBX it /BVE it +

β6RND it /BVE it + β7ADVERT it /BVE it + β8CAP_EX it /BVE it + εit (3)

Where:

38 MVE it = market value of equity for firm i period t (Compustat mnemonic MKVAL)

BVE it = book value of equity for firm i period t (Compustat mnemonic CEQ)

ERP = dummy variable set to (1) for ERP implementers, (0) for control firms

IMP = dummy variable set to (1) for post- and (0) for pre-implementation periods

ERP*IMP = interaction term based on ERP x IMP. This variable is expected to be

positive if ERP implementation is associated with increased market-to-book

ratios following implementation relative to non-implementers.

IBX it = income before extraordinary items for firm i period t (Compustat mnemonic IB)

RND it = research and development expenses for firm i period t (Compustat mnemonic

XRD)

ADVERT it = advertising expenses for firm i period t (Compustat mnemonic XAD)

CAP_EX it = capital expenditures for firm i period t (Compustat mnemonic CAPXV) .

4.3.4 Value-to-Price (V/P) Model

To examine the third hypothesis, this essay uses the valuation model operationalized by

Frankel and Lee (1998) 39 , which they refer to as the Edwards-Bell-Ohlson (EBO) valuation technique. This model assumes that intrinsic value is equal to the book value of a firm plus the

38 Compustat uses a three month lag between book value and market value to allow the market to incorporate the results of the prior period financial reports (quarterly and annual). 39 Although Bradshaw (2004) provides a more sophisticated model, with alternative methods of calculating the terminal value, the Frankel and Lee (1998) model with a single terminal value approach is sufficient for this essay because the emphasis is on the difference between the matched pairs not the accuracy of the model itself.

106

present value of expected residual income, assuming clean surplus accounting. Following is their primary model:

(FROE t − re ) (FROE t+1 − re ) (FROE t+2 − re ) Vt = Bt + Bt + 2 Bt+1 + 2 Bt+2 (4) 1( + re ) 1( + re ) 1( + re ) re

Where:

FROE t = forecasted return on equity in year t = FY1/[(B t-1 + B t-2)/2]

FROE t+1 = forecasted return on equity in year t+1 = FY2/[(B t + B t-1)/2]

FROE t+2 = forecasted return on equity in year t+2 = [FY2(1 + Ltg)]/[(B t+1 + B t)/2]

40 Bt = the estimated book value of common stockholders’ equity at the end of year t

41 Re = the estimated cost of equity capital .

FY1 = I/B/E/S one-year consensus earning estimate

FY2 = I/B/E/S two-year consensus earning estimate

Ltg = I/B/E/S consensus growth rate

Once (V) has been estimated, the resulting V/P ratio is computed for ERP implementers and compared to the ratio for non-implementers to measure the differences. Although this provides an initial assessment of the impact of the V/P ratio, other factors may also be influencing the results. To address this concern, the V/P ratio for implementers and non-implementers are compared using the following regression model:

V/P i = αi + β1ERP i + β2NUM i + β3ACC i + β4STD i + εit (5)

Where:

40 A sequential process is used to estimate FROE and B (see Frankel and Lee (1998) Appendix A) using the Compustat mnemonics BKVLPS, for book value per share and DVPOR5 for dividend payout ratio (5 year). 41 The cost of equity capital is estimated for each firm using the Easton (2004) PEG methodology, which is based on the square root of projected EPS growth divided by price.

107

V/P i = the ratio derived above

Pi = stock price (Compustat mnemonic PRCC)

ERP = dummy variable set to (1) for firms implementing ERP else (0)

NUM = average number of financial analysts providing 1- and 2-year-ahead forecasts

ACC = average forecast accuracy for 1- and 2-year ahead forecasts based on the

absolute value of differences between actual EPS and forecasted EPS

STD = average standard deviation of 1- and 2-year ahead forecasts.

Note: All continuous variables are scaled by share price 42 .

4.4 Empirical Results

4.4.1 Long-Horizon Returns

4.4.1.1 Univariate Analysis

An initial analysis (not tabulated) shows that the 147 firms implementing ERP systems experienced statistically significant abnormal returns of 15.1%, 7.7%, and 6.6% over the one, three, and five year periods following implementation using the matched control firms as the benchmark. Although this initial finding provides some support for H2a, it doesn’t address how well these firms would have done without implementing an ERP system. To address that issue, return data for the one, three, and five year periods prior to implementation is also examined.

One problem with this approach is that some of the firms (or their matched pairs) do not have return data available for the period of time prior to implementation, which reduces the sample size. Table 4-3 provides a summary of the long-horizon buy-and-hold return statistics for the

42 This reduced model is being used because it includes the more common variables associated with forecasting.

108

three different time horizons: (1) one-year returns, (2) three-year returns, and (3) five-year returns both before and after implementation. To avoid distortions from outliers, the return data has been winsorized at the .01 and .99 levels. Panel A, shows that mean one-year returns for ERP implementers are 12.0% before and 18.7% after implementation, compared to 10.5% before and

0.840% after for the benchmark non-implementers. The resulting abnormal return of 17.8% is significant after implementation (t=2.40; p=0.017), but the 1.5% abnormal return before implementation is not. The changes that take place from before to after are not significant for either group. Panel B shows a similar pattern for the three-year returns, with an abnormal return of 6.9% after implementation significant (t=1.79; p=0.075), but not the 5.3% abnormal return before implementation. The change from before to after is also not significant for either group.

The five-year returns in Panel C have no significant abnormal returns, however, both groups reflect marginally significant decreases in returns (t=1.64; p=0.103 and t=1.85; p=0.066). Figure

4-1 provides a graphical representation of the data in Table 4-3.

[Insert Table 4-3 here ]

[Insert Figure 4-1 here ]

4.4.1.2 Multivariate Analysis

Table 4-4 provides a correlation matrix of the variables used in the multivariate regressions of long-horizon buy-and-hold returns for the three different time horizons and Table

4-5 summarizes the results. The one year model is not very robust as indicated by the low F- statistic (1.41; p=0.2099) and Adjusted R 2 of only 0.4%. The three and five year models however both have significant F-statistics (17.28, p<0.0001; and 2.61, p0.0174) and Adjusted R 2 of 7.6% and 2.9% respectively. The coefficient on the interaction term ERP*IMP is not significant in any of the models indicating that ERP implementation has no impact on change in returns relative to non implementers after controlling for book value, EPS, and Y2K. Consistent with the univariate

109

analysis, the coefficient on the IMP variable in the five year model is negative and marginally significant (-6.118; p=0.0962) indicating that the five year returns are lower on average for both groups. None of the models reflect significant differences in the ERP coefficient, indicating that when averaging the before and after returns, there is no difference between the two groups. This is also consistent with the univariate analysis which finds significant differences only in the after period for the one and three year returns. As for control variables, the coefficient on book value is highly significant in the three and five year models, and the coefficient on EPS is highly significant in the three year model. The coefficient on the Y2K variable is not significant in any of the models.

[Insert Table 4-4 here ]

[Insert Table 4-5 here ]

Taken as a whole, the results reflected in tables 4-3 and 4-5 do not provide strong support for hypothesis H2a. Although the returns are higher for ERP implementers after implementation, the evidence does not support the argument that implementation of ERP systems has an impact on shareholder value as measured by long-horizon returns. These results raise the question of self selection bias. In other words, are the firms that self select to implement ERP systems those that would have generated better returns with or without the system. To explore this issue further, an analysis of EPS before and after implementation was conducted on the assumptions that EPS may be an alternative indicator of future firm performance. The results (not tabulated) show that EPS for one and three years before implementation are significantly higher for ERP implementing firms (p=0.036; p=0.029) than for control firms, with no significant differences in the one and three year periods after implementation. There are no significant differences between control groups for either of the five year periods before or after implementation. These results provide some support for the possibility of self selection bias in that the firms that self selected to

110

implement ERP systems were those that were already generating better earnings than the control group during the one and three year time periods leading up to the decision. This issue should be explored further in future research.

4.4.2 Market-to-Book (MTB) Analysis

4.4.2.1 Univariate Analysis

Table 4-6 provides a summary of the market-to-book (MTB) ratios covering three different blocks of time surrounding ERP implementation: (1) three years before and after, (2) four years before and after, and (3) five years before and after. To avoid distortions from outliers, the MTB ratios have been winsorized at the .01 and .99 levels. All three time periods present similar results, namely that the mean MTB ratio is higher for ERP implementers than for non implementers both before and after implementation, and that the change in ratios from before to after is not statistically significant for either group. Panel A provides results for 1,500 firm-year observations from 125 ERP implementers and 125 control firms for three years before and three years after implementation (125 x 2 x 3 x 2). The mean ratio for ERP-implementers is higher than non-implementers both before (0.885; p=0.0049) and after (0.997; p=0.0013) implementation. The change in ratio is not significant for either the ERP implementing firms or the control group. Panel B provides similar results for 1,456 firm-year observations from 91 implementing firms and 91 control firms for four years before and four years after implementation (91 x 2 x 4 x 2). As with Panel A, the mean MTB ratio for implementing firms is higher than the control firms both before (0.862; p=0.0035) and after (0.750; p=0.0173) implementation. The change in ratio is again not significant for either the ERP implementing firms or the control group. Panel C provides results for 1,260 firm-year observations for 63 firms that implemented ERP systems and 63 control firms for five years before and after

111

implementation (63 x 2 x 5 x 2). Similar to the other two panels, the MTB ratio is higher for implementers both before (0.964; p=0.0003) and after (0.750; p=0.0191) implementation. The change from before to after is again not significant for either implementers or the control firms.

[Insert Table 4-6 here ]

Figure 4-2 provides a graphical view of these MTB values. For all three time periods, the

MTB ratio is higher for ERP implementers than for non-implementers, and in all three cases the ratios decline following implementation as reflected by the downward sloping lines. However the decline is not significant and the distance between them does not change significantly. These results do not provide support for hypothesis H2b because the MTB ratios are higher for ERP implementers both before and after implementation, with no significant change for either group.

[Insert Figure 4-2 here ]

4.4.2.2 Multivariate Analysis

Although Table 4-6 does not provide support for hypothesis H2b, other factors may be contributing to the results. To address this, the multivariate regression model adapted from Core et al. (2003), as describe in equation (3), is used to further test the hypothesis. Table 4-7 provides a correlation matrix of the variables used in the regression and Table 4-8 summarizes the results.

Three models are used, one for each of the three different blocks of time. The first model uses three years of data before implementation and three years of data after implementation for 125

ERP implementers and 125 control firms for a total of 1,500 firm year observations. The second model uses four years of data before and after implementation for 91 ERP implementers and 91 control firms for 1,456 firm year observations. The third model uses five years of data before and after implementation for 63 ERP implementers and 63 control firms for a total of 1,260 firm year observations.

112

All three models reflect positive significant coefficients on the ERP variable (0.6176; p=0.0331), (0.5883; p=0.0254), and (0.7642; p=0.0014), which supports the univariate analysis where the MTB ratio was higher for ERP firms both before and after implementation. However, neither the implementation variable (IMP), nor the interaction term (ERP*IMP), have significant coefficients, indicating that the implementation event itself does not have a significant impact on the ratios. This would tend to support a self-selection argument that firms with higher MTB ratios are the ones that implement ERP systems.

[Insert Table 4-7 here ]

[Insert Table 4-8 here ]

4.4.3 Value-to-Price (V/P) Analysis

4.4.3.1 Univariate Analysis

Table 4-9 summarizes key statistics from the value-to-price (V/P) analysis surrounding

ERP implementations based on the Frankel and Lee (1998) model. To avoid distortions from outliers, V/P ratios have been winsorized at the .05 and .95 levels. Panel A results are based on data from the year before implementation, Panel B is based on data from the implementation year and Panel C, is based on data from the year after implementation. The year before implementation shows mean (median) V/P ratios of 0.772 (0.770) for ERP implementing firms and 0.642 (0.635) for control firms, which is a significant difference based on a Wilcoxon

Nonparmetric test at the p=0.0058 (one-tail) level. The mean (median) V/P ratios for the implementation year are 0.780 (0.766) for ERP implementing firms vs. 0.654 (0.602) for control firms. The differences are significant at the p=0.0138 (one-tail) level based on the Wilcoxon

Nonparametric test. The year after implementation however provides a different result, with the

113

mean (median) V/P ratios of 0.668 (0.637) for ERP firms, and 0.596 (0.625) for control firms not significantly different (p=0.2318)43 .

[Insert Table 4-9 here ]

Figure 4-3 provides a graphical view of these V/P ratios. For all three years, the V/P ratio is higher for ERP implementers than for non-implementers. However in the year after implementation, the differences are no longer significant.

[Insert Figure 4-2 here ]

4.4.3.2 Multivariate Analysis

As with other analysis in this essay, these results could be influenced by other factors, which is why the multivariate regression model (equation 5) is used to further test the hypothesis.

Table 4-10 provides a correlation matrix of the variables used in the regression, and Table 4-11 summarizes the results. The primary variable of interest (ERP) is highly significant in the year before implementation (0.1312; p=0.0092), marginally significant in the implementation year

(0.1167; p=0.0762), and not significant in the year after implementation (0.0452; p=0.4350).

These results indicate that just prior to and during implementation of ERP systems, the V/P ratio is higher for implementing firms than for control firms which, consistent with Frankel and Lee

(1998), would predict higher future long-horizon returns. However, in the year after implementation, the difference is not significant, which would indicate that the market does not anticipate higher returns. The control variables all have negative coefficients, although only the number of analysts is significant in all three years, and the standard deviation is significant in the

43 Some of the observations (14 in the implementation year) result in negative V/P ratios, which would imply a negative future value since price can not be negative. Although this is conceptually difficult, the observations were retained in order to maximize the sample size. A sensitivity analysis was run with two alternative assumptions, one that the value would only drop to zero, and the other with the negative observations excluded. The results are substantially the same, with V/P ratios for ERP implementers still significantly higher than control firm V/P ratios.

114

year after implementation. These results are consistent with the univariate analysis in Table 4-9, indicating that after controlling for factors thought to influence V/P ratios, implementation of

ERP systems may not have an impact on shareholder value as measured by V/P ratios.

[Insert Table 4-10 here ]

[Insert Table 4-11 here ]

4.5 Conclusions

This essay examines the impact that ERP systems have on shareholder value using three different measures of shareholder value: long-horizon returns, market-to-book ratios, and value- to-price ratios. The primary analysis examines long-horizon buy-and-hold returns over one-, three-, and five-year time horizons for a sample of firms that implemented ERP systems between

1994 and 2003. It uses a sample of control firms matched by size and industry to compare returns before and after implementation. Univariate results show that ERP firms have higher returns after implementation for the one and three year periods, than their matched control firms, but not before implementation. The results also show no significant change in returns for the one and three year periods from before to after implementation for either group. The five year returns show no significant differences either before or after implementation between the two groups, however both groups reflect marginally significant reductions in returns 44 . A multivariate OLS regression supports the initial results after adding control variables for book value, earnings, size, and the bull vs. bear market timing. Although there is evidence that ERP firms have higher returns following implementation than the control firms, the results do not support the hypothesis that implementation of ERP systems has an impact on the results. In fact there is some evidence

44 A larger sample of firms does have higher five year returns for ERP implementers after implementation.

115

that self selection bias may exist in that firms that implemented ERP systems experienced higher earnings during the one and three year periods prior to implementation.

To further investigate the impact using other measures, this essay examines the market- to-book (MTB) ratio for the sample and control firms using three different blocks of time based on three-, four-, and five-years of data before and after implementation of the ERP system.

Although the sample size gets smaller as more years are included in the analysis, the results are similar for all three groups, with ERP implementers experiencing higher average MTB ratios than the non-implementer control group. Both groups experience decreases in their respective ratios following implementation, although the decreases are not statistically significant.

A multivariate OLS regression analysis is used to examine the effect of implementing

ERP on MTB after controlling for other factors that may be influencing the results. The results are consistent with the results for long-horizon returns, indicating that implementation of ERP systems may not have an impact on shareholder value as measured by MTB ratios. However, these results are consistent with a self selection bias in that firms with higher MTB ratios may be those that choose to implement ERP systems because the ratios were higher both before and after implementation.

Another measure used in this essay is the value-to-price (V/P) ratio, which Frankel and

Lee (1998) show is a good indicator of future long-horizon returns. This study shows that ERP implementing firms have higher V/P ratios immediately before and during the implementation process than the control firms, which would indicate higher future long-horizon returns.

However, the ratio one year after implementation is not significantly different, indicating that higher returns are no longer predicted. Multivariate OLS regressions that control for other factors thought to impact the V/P ratio support these initial findings, indicating that implementation of

ERP systems may not have a favorable impact on shareholder value as measured by V/P ratios.

116

Major investments, such as ERP systems, are usually made with the expectation that they will benefit a business over a long-horizon time frame, which should translate into benefits for the shareholder owners of that business as well. Taken as a whole, the results of this study fail to provide sufficient evidence that implementation of ERP systems has a significant positive impact on shareholder value as measured by long-horizon returns. This finding is supported by supplemental analysis of market-to-book ratios and value-to-price ratios that are fully consistent with the primary results. The results provide some support for the argument that self selection bias may have occurred. In other words, firms that provide better shareholder value as measured by long-horizon returns, market-to-book ratio, and value-to-price ratios may be more inclined to implement ERP systems.

As with most research, this essay has several limitations that restrict the ability to draw broad conclusions. Perhaps the biggest limitation is the methodology used to select the control sample, which relied on first matching by industry based on SIC codes, with size used as a secondary factor. The result is a sample in which the ERP implementers are significantly larger than the control firms. This is explained in part by the high cost of these systems and the fact that most of the largest corporations have implemented them, leaving only smaller firms in each industry to use as control firms. A related issue is the use of public announcements to determine if a firm qualifies as a control firm, which may bias the results even further. It is possible that firms selected as control firms have indeed implemented ERP systems, but have not made any public announcement about it. Future research in this area should explore alternative methods for developing a control sample.

These results should be of interest to investors, ERP vendors, and management personnel responsible for major IT investment decisions. The results also contribute to the literature on

117

capital markets and information systems by applying capital markets research methods to information systems research events.

118

Figure 4-1 Long-Horizon Returns ERP = 0 Matching Non -ERP Implementing Firms ERP = 1 ERP Implementing Firms IMP = 0 Before ERP Implementation IMP = 1 After ERP Implementation

119

Figure 4-2 Mean Market-to-Book Ratios ERP = 0 Matching Non -ERP Implementing Firms ERP = 1 ERP Implementing Firms IMP = 0 Before ERP Implementation IMP = 1 After ERP Implementation

Three Years Before & After Implementation

4.000

3.500

3.000

MTB MTB Ratio 2.500

2.000 IMP=0 IMP=1

ERP=0 ERP=1

Four Years Before & After Implementation

4.000

3.500

3.000

MTB MTB Ratio 2.500

2.000 IMP=0 IMP=1

ERP=0 ERP=1

Five Years Before & After Implementation

4.000

3.500

3.000

MTB Ratio 2.500

2.000 IMP=0 IMP=1

ERP=0 ERP=1

120

Figure 4-3 Mean Value-to-Price Ratios ERP = 0 Matching Non -ERP Implementing Firms ERP = 1 ERP Implementing Firms

Value-to-Price (V/P) Ratios

0.8

0.7

0.6

0.5 1 Yr Before Imp Year 1 Yr After

ERP=1 ERP=0

V/P = Ratio of Value -to -Price P = Stock Price (Compustat mnemonic PRCC) V is from the following Frankel and Lee (1998) model:

(FROE t − re ) (FROE t+1 − re ) (FROE t+2 − re ) Vt = Bt + Bt + 2 Bt+1 + 2 Bt+2 1( + re ) 1( + re ) 1( + re ) re Where: FROE t = forecasted return on equity in year t = FY1/[(B t-1 + B t-2)/2] FROE t+1 = forecasted return on equity in year t+1 = FY2/[(B t + B t-1)/2] FROE t+2 = forecasted return on equity in year t+2 = [FY2(1 + Ltg)]/[(B t+1 + B t)/2] Bt = the estimated book value of common stockholders’ equity at the end of year t, Re = the estimated cost of capital. FY1 = I/B/E/S one-year consensus earning estimate FY2 = I/B/E/S two-year consensus earning estimate Ltg = I/B/E/S consensus growth rate

121

Table 4-1 Summary of Sample Selection Process Panel A – Analysis of Returns (H2a): Firms Initial ERP anno uncements from Hayes et al. (2001) 91 Less firms no longer listed or data otherwise not available (36) Remaining firms from Hayes et al. (2001) 55 Additional ERP announcements collected from Lexis -Nexis search 92 Total firms implementing ERP systems with one & three year return data 147 Less firms without one year return data prior to implementation * ( 7 ) Firms used to examine H2a one year before and after return s 140 Less firms without three year return data prior to implementatio n* ( 25) Firms used to examine H2a three year before and after returns 115 Less firms without five year return data before or after implementation * ( 34) Firms used to examine H2a five year before and after returns 81

Panel B – Analysis of Market-to-Book Ratio (H2b): Total firms implementing ERP systems from above 147 Less firms without 3 years of data before and after implementation * (22) Firms used to examine H2b 3 years before and after 125 Less firms without 4 years of d ata before and after implementation * (34) Firms used to examine H2b 4 years before and after 91 Less firms without 5 years of data before and after implementation * (28) Firms used to examine H2b 5 years before and after 63

Panel C – Analysis of Residual Income Valuation Model (H2c): Total firms implementing ERP systems from above 147 Less firms without IBES earnings forecast data * (44) Firms used to examine H2c 103

Panel D – Comparison of Firm Sizes: Market Capitalization Total Assets ($Millions) Control ERP Control ERP Firms Firms Firms Firms N 147 147 147 147 Mean 1,691.3 5,799.7 1,934.8 3,584.7 Median 606.3 1,268.6 587.2 915.6 Standard Deviation 2,896.5 13,091.1 3,584.7 7,067.9 Minimum 3.5 3.7 1.6 1.0 Ma ximum 19,705.1 83,683.7 34,369.00 36,147.0

Test for differences: t stat = 3.72 p = .0002 t stat = 2.48 p = 0.0139 Wilcoxon p = .0011 Wilcoxon p = 0.0341 * Firms eliminated may have been due to lack of available data for their matched control firm Note: Market Capitalization is as of the end of the implementation year and total Assets are as of the beginning of the implementation year.

122

Table 4-2 ERP Implementing Firms by 2 Digit SIC Code and Implementation Year

2 Digit SIC Codes 94 95 96 97 98 99 00 01 02 03 Total 01 -Agricultural Production Crops 1 1 13 -Oil and Gas Extraction 1 1 3 2 1 8 16 -Heavy Construction 1 1 20 -Mfg: Food and Kindred Products 2 1 2 1 1 1 8 23 -Mfg: Apparel 1 1 1 3 24 -Mfg: Lumber & Wood Products 1 1 25 -Mfg: Furniture & Fixtures 1 2 1 4 26 -Mfg: Paper & Allied Products 1 1 1 1 4 27 -Mfg: Printing & Publishing 1 1 1 1 4 28 -Mfg: Chemicals 1 1 2 2 1 4 2 3 2 18 29 -Mfg: Petroleum Refining 1 1 30 -Mfg: Rubber & Misc. Plastic 1 1 33 -Mfg: Primary Metal Industries 1 1 1 3 34 -Mfg: Fabricated Metal Products 1 1 1 3 35 -Mfg: Ind. & Com. Machinery 1 5 8 2 4 1 1 22 36 -Mfg: El ectronic & Elect. Equip. 1 2 5 3 3 4 18 37 -Mfg: Transportation Equipment 1 1 1 1 1 5 38 -Mfg: Measuring & Control Instr. 1 2 1 1 1 2 1 1 10 39 -Mfg: Misc. Manufacturing 1 1 42 -Motor Freight Transportation 1 1 45 -Transportat ion by Air 2 2 48 -Communications 1 1 2 49 -Electric, Gas & Sanitary Services 2 2 50 -Wholesale: Durable Goods 2 1 3 51 -Wholesale: Non -Durable Goods 1 1 52 -Retail: Bldg Materials, Hardware 1 1 54 -Retai l: Food Stores 1 1 59 -Retail: Miscellaneous 1 3 2 6 63 -Insurance Carriers 1 1 2 67 -Holding & Other Investment 1 1 2 73 -Automotive Repair & Service 1 2 3 6 80 -Health Services 1 1 87 -Engineering, Actg., R &D, Mgt 1 1 Totals 1 8 8 15 28 34 18 20 9 6 147

123

Table 4-3 Summary of Long-Horizon Return Statistics Panel A – One Year Returns (TRT1Y): ERP = 0 ERP = 1 Difference t-statistic p-value IMP = 0 10.529 12.006 1.477 .20 0.8437 IMP = 1 .840 18.653 17.813 2.40 0.0170 Difference 9.6899 6.647 t-statistic 1.31 0.89 p-value 0.1916 0.3770 n 280 280

Panel B – Three Year Returns (TRT3Y): ERP = 0 ERP = 1 Difference t-statistic p-value IMP = 0 3.834 9.174 5.340 1.24 0.2146 IMP = 1 0.520 6.341 6.861 1.79 0.0753 Difference 4.354 2.833 t-statistic 1.03 0.73 p-value 0.3054 0.4681 n 230 230

Panel C – Panel C Five Year Returns (TRT5Y): ERP = 0 ERP = 1 Difference t-statistic p-value IMP = 0 9.212 13.100 3.888 0.95 0.3439 IMP = 1 3.226 5.413 2.187 0.59 0.5557 Difference 5.986 7.687 t-statistic 1.64 1.85 p-value 0.1026 0.0656 n 162 162

ERP=0 Matched Firms that did not implement ERP system ERP=1 Firms that Implemented ERP systems IMP=0 Five Fiscal Years Before Implementation IMP=1 Five Fiscal Years After Implementation

124

Table 4-4 Pearson Correlation Matrix for Long-Horizon Returns OLS Regression Panel A – One Year Returns (N=560 ): RET ERP IMP ERP*IMP BVPS EPS Y2K RET 1.0000 ERP 0.0772 1.0000 (0.0678) IMP -0.0122 0.0000 1.0000 (0.7736) (1.0000) ERP*IMP 0.0753 0.5774 0.5774 1.0000 (0.0749 (<.0001) (<.0001) BVPS -0.0644 -0.0768 0.0177 -0.0139 1.0000 (0.1283) (0.0695) (0.6766) (0.7422) EPS 0.0254 0.0201 0.0255 -0.0129 -0.1371 1.0000 (0.5483) (0.6346) (0.5465) (0.7607) (0.0011) Y2K -0.0267 0.0000 0.0000 0.0000 0.0575 -0.024 8 1.0000 (0.5286) (1.0000) (1.0000) (1.0000) (0.1742) (.5578) Pa nel B – Three Year Returns (N=460 ): RET ERP IMP ERP*IMP BVPS EPS Y2K RET 1.0000 ERP 0.0986 1.0000 (0.0346) IMP -0.0581 0.0000 1.0000 (0.2139) (1.0000) ERP*IMP 0.0305 0.5774 0.5774 1.0000 (0.5143) (<.0001) (<1.000) BVPS -0.1792 -0.0897 0.0060 -0.0117 1.0000 (0.0001) (0.0546) (0.8983) (0.8018) EPS 0.2169 0.0520 0.0211 0.0320 -0.0538 1.0000 (<.0001) (0.2654) (0.6520) (0.4930) (0.2493) Y2K 0.0666 0.0000 0.0000 0.0000 0.0019 0.0597 1.0000 (0.1540) (1.0000) (1.0000) (1.0000) (0.9669) (0.2012) Pa nel C – Five Year Returns (N=32 4): RET ERP IMP ERP*IMP BVPS EPS Y2K RET 1.0000 ERP 0.0608 1.0000 (0.2752) IMP -0.1369 0.0000 1.0000 (0.0137) (1.0000) ERP*IMP -0.0537 0.5774 0.5774 1.0000 (0.3349) (<.0001) (<.0001) BVPS 0.1499 -0.0642 -0.0504 -0.0979 1.0000 (0.0069) (0.2492) (0.3663) (0.0784) EPS 0.0009 -0.0497 0.1396 0.0936 -0.0054 1.0000 (0.9867) (0.3722) (0.0119) (0.0926) (0.9227) Y2K 0.0510 0.0000 0.0000 0.0000 0.0370 -0.0851 1.0000 (0.3601) (1.0000) (1.0000) (1.0000) (0.5070) (0.1265) RET i = total returns for firm i for each of the one, three and five year periods ERP = dummy variable set to (1) for ERP implementers, (0) for control firms IMP = dummy variable set to (1) for post- and (0) for pre-implementation periods ERP*IMP = interaction term based on ERP x IMP, BVPS = book value per share for firm i at the end of the implementation fiscal year EPS = earnings per share change for firm i Y2K = dummy variable set to (1) for firms that implemented ERP systems (and the matched control firms) after the year 2000, otherwise, (0) P-values in parentheses

125

Table 4-5 Long-Horizon Returns OLS Regression Results Results from the following long-horizon returns regression models:

RET i = αi + β1ERP i + β2IMP i + β2ERP*IMP i + β4BVPS i + β5EPS i + β6Y2K i + εit (2)

Exp TRT1Y TRT3Y TRT5Y Variable Sign Coeff p-value Coeff p-value Coeff p-value

Intercept ? 13.8462 0.0295 6.2529 0.0655 8.2825 0.0028 ERP + 0.4573 0.9516 2.9204 0.4880 4.1301 0.2850 IMP + -10.0531 0.1756 -5.4192 0.1770 -6.1180 0.0962 ERP*IMP + 17.1830 0.1038 3.2429 0.5567 -1.1080 0.8319 BVPS + -0.1451 0.1452 -0.1618 0.0030 0.0148 0.0002 EPS + 0.0018 0.5394 0.0088 0.0002 0.0002 0.8561 Y2K + -2.9719 0.5769 3.5213 0.2209 2.5994 0.3547

F-statistic 1.41 0.2099 7.28 <0.0001 2.61 0.0174 2 Adjusted R 0.004 0.076 0.029 N 560 460 324

White test for heteroskedastisity yields Chi-Square results that are not significant. Analysis of Variance Inflation Factors (VIF) indicates there is not a multicollinearity problem.

RET i = total returns for firm i for each of the one, three and five year periods ERP = dummy variable set to (1) for ERP implementers, (0) for control firms IMP = dummy variable set to (1) for post- and (0) for pre-implementation periods ERP*IMP = interaction term based on ERP x IMP, BVPS = book value per share for firm i at the end of the implementation fiscal year EPS = earnings per share change for firm i Y2K = dummy variable set to (1) for firms that implemented ERP systems (and the matched control firms) after the year 2000, otherwise, (0)

126

Table 4-6 Summary of Mean Market-to-Book (MTB) Values Panel A – Three Years Before and Three Years After Implementation (a) ERP = No ERP = Yes Difference t-statistic p-value Before Implementation 2.694 3.579 0.885 2.82 0.0049 After Implementation 2.229 3. 206 0.977 3.22 0.0013 Change (After - Before) -0.465 -0.373 t-statistic -1.42 -1.29 p-value 0.1555 0.1966 N 750 750

Panel B – Four Years Before and Four Years After Implementation (a) ERP = No ERP = Yes Difference t-stati stic p-value Before Implementation 2.632 3.494 0.862 2.93 0.0035 After Implementation 2.437 3.187 0.750 2.39 0.0173 Change (After - Before) -0.195 -0.307 t-statistic -0.62 -1.03 p-value 0.5325 0.3027 N 728 728

Panel C – Five Years Before and Five Years After Implementation (a) ERP = No ERP = Yes Difference t-statistic p-value Before Implementation 2.458 3.422 0.964 3.68 0.0003 After Implementation 2.290 3.040 0.750 2.35 0.0191 Change (After - Before) -0.168 -0.382 t-statistic -0.56 -1.34 p-value 0.5731 0.1809 N 630 630

(a): Data for the implementation year is included in the period before implementation.

127

Table 4-7 Pearson Correlation Matrix for Market to Book Regression Panel A: Three Years Before and After ERP Implementation (N=1,500): MVE ERP IMP ERPIMP BVE/1 IBX RND ADVERT MVE 1.00000

ERP 0.10952 1.00000 (<.0001) IMP -0.04931 0.00000 1.0 0000 (0.0562) (1.0000) ERPIMP 0.03790 0.57735 0.57735 1.00000 (0.1423) (<.0001) (<.0001) BVE/1 0.21236 0.01290 -0.02035 -0.02208 1.00000 (<.0001) (0.6176) (0.4309) (0.3929) IBX 0.18561 0.06000 -0.04895 -0.00098 0.51452 1.00000 (<.0001) (0.0201) (0.0580) (0.9698) (<.0001) RND 0.29262 0.05946 -0.02793 0.01833 0.12281 0.24883 1.00000 (<.0001) (0.0213) (0.2796) (0.4780) (<.0001) (<.0001) ADVERT 0.21943 0.05678 -0.00868 0.04011 0.06295 -0.01859 0.023 25 1.00000 (<.0001) (0.0279) (0.7369) (0.1205) (0.0147) (0.4718) (0.3682) CAP_EX 0.25683 0.04161 -0.01665 0.00740 0.29253 0.51933 0.53378 0.07297 (<.0001) (0.1072) (0.5194) (0.7747) (<.0001) (<.0001) (<.0001) (0.0047)

Panel B: Four Years Before and After ERP Implementation (N=1,456): MVE ERP IMP ERPIMP BVE/1 IBX RND ADVERT MVE 1.00000

ERP 0.09770 1.00000 (0.0002) IMP -0.03039 0.00000 1.00000 (0.2465) (1.0000) ERPIMP 0.03493 0.57735 0.57735 1.00000 (0.1828) (<.0001) (<.0001) BVE/1 0.24748 0.01825 -0.01984 -0.01587 1.00000 (<.0001) (0.4865) (0.4494) (0.5452) IBX 0.20629 0.05773 -0.03394 0.01097 0.51804 1.00000 (<.0001) (0.0276) (0.1956) (0.6757) (<.0001) RND 0.31871 0.03952 -0.01263 0.00994 0.13826 0.19690 1.000 00 (<.0001) (0.1318) (0.6300) (0.7046) (<.0001) (<.0001) ADVERT 0.25451 0.04552 0.00737 0.03388 0.04168 -0.00714 0.06389 1.00000 (<.0001) (0.0825) (0.7789) (0.1963) (0.1119) (0.7854) (0.0147) CAP_EX 0.27513 0.04496 0.00467 0.02930 0.25775 0.54954 0. 38506 0.05521 (<.0001) (0.0864) (0.8588) (0.2639) (<.0001) (<.0001) (<.0001) (0.0352)

Continued on next page

128

Table 4-7 Pearson Correlation Matrix for Market to Book Regression (continued)

Panel C: Five Years Before and After ERP Implementation (N=1,260): MVE ERP IMP ERPIMP BVE/1 IBX RND ADVERT MVE 1.00 000

ERP 0.11624 1.00000 (<.0001) IMP -0.03733 0.00000 1.00000 (0.1854) (1.0000) ERPIMP 0.03717 0.57735 0.57735 1.00000 (0.1873) (<.0001) (<.0001) BVE/1 0.36441 -0.00068 0.00659 -0.00071 1.0 0000 (<.0001) (0.9809) (0.8152) (0.9798) IBX 0.16147 0.05231 -0.01861 0.02316 0.16521 1.00000 (<.0001) (0.0634) (0.5092) (0.4115) (<.0001) RND 0.34853 0.06602 -0.01607 0.02609 0.53453 0.32422 1.00000 (<.0001) (0.0191) (0.5688) (0.3548) (<.0001) (<.0001) ADVERT 0.29665 0.03699 0.01949 0.03247 0.21350 -0.00705 0.06195 1.00000 (<.0001) (0.1894) (0.4895) (0.2495) (<.0001) (0.8027) (0.0279) CAP_EX 0.28640 0.04096 0.00976 0.03357 0.34516 0.54100 0.42547 0.05677 (<.0001) (0.1462) (0.7292) (0.2337) (<.0001) (<.0001) (<.0001) (0.0439)

MVE = market value of equity ERP = dummy variable set to (1) for firms implementing ERP else (0); IMP = dummy variable set to (1) for post implementation periods else (0); ERP*IMP = interaction term based on ERP x IMP; BVE = book value of equity IBX = income before extraordinary items; RND = research and development expenses; ADVERT = advertising expenses; CAP_EX = capital expenditures

P-values are in parenthesis

129

Table 4-8 Market-to-Book (MTB) Value OLS Regression Results Results from the following market to book regression model:

MVE it /BVE it = αi + β1ERP it + β2IMP it + β3ERP*IMP it + β41/BVE it + β5IBX it /BVE it + β6RND it /BVE it + β7ADVERT it /BVE it + β8CAP_EX it /BVE it + εit

Exp 3 Yrs Bef & Aft 4 Yrs Bef & Aft 5 Yrs Bef & Aft Variabl e Sign Coeff p-value Coeff p-value Coeff p-value

Intercept ? 2.5489 <.0001 2.4790 <.0001 2.2885 <.0001 ERP + 0.6176 0.0331 0.5883 0.0254 0.7642 0.0014 IMP + -0.3658 0.2147 -0.2046 0.4576 -0.2081 0.4014 ERP*IMP + 0.1126 0.7786 -0.0261 0.9457 -0.1972 0.5841 1/BVE + 2.8874 <.0001 3.4522 0.0002 13.8358 0.0001 IBX + 0.0402 0.7111 -0.0043 0.9654 -0.0033 0.9697 RND + 1.7549 0.0069 2.0218 <.0001 1.4250 0.0002 ADVERT + 4.1876 <.0001 4.2603 <.0001 3.7285 <.0001 CAP_EX + 0.1298 0.2782 0.1879 0.0931 0.1681 0.0175

F-statistic 38.17 <.0001 49.47 <.0001 51.10 <.0001 Adjusted R 2 0.166 0.210 0.242 N 1,500 1,456 1,260

P-values are based on White adjusted t-statistics, two-tailed test. Analysis of Variance Inflation Factors (VIF) indicates there is not a multicollinearity problem.

MVE i,t = market value of equity for firm i period t (Compustat mnemonic MKVAL) ERP = dummy variable set to (1) for firms implementing ERP else (0) IMP = dummy variable set to (1) for post implementation periods else (0) ERP*IMP = interaction term based on ERP x IMP. This is the primary variable of interest and is expected to be positive if ERP implementation is associated with higher market to book ratios. BVE = book value of equity (Compustat mnemonic CEQ) IBX = income before extraordinary items (Compustat mnemonic IB) RND = research and development expenses (Compustat mnemonic XRD) ADVERT = advertising expenses (Compustat mnemonic XAD) CAP_EX = capital expenditures (Compustat mnemonic CAPXV)

130

Table 4-9 Value-to-Price (V/P) Statistics N Mean Std Dev Median Minimum Maximum Panel A, Year Before Implementation: ERP Implementing Firms 102 0.772 0.373 0.770 0.052 1.460 Matching Control Firms 102 0.642 0.362 0.635 0.052 1.460 Difference 0.130 0.368 Wilcoxon Nonparametric Test for Difference; p -value = 0.0058 (one tail)

Panel B, Implementation Year: ERP Implementing Firms 103 0.780 0.436 0.766 -0.118 1.735 Matching Control Firms 103 0.654 0.504 0.602 -0.118 1.7 35 Differences 0.126 0.471 Wilcoxon Nonparametric Test for Difference; p -value = 0.0138 (one tail)

Panel C, Year After Implementation: ERP Implementing Firms 100 0.668 0.396 0.637 -0.192 1.464 Matching Control Firms 100 0.596 0.440 0.625 -0.192 1.464 Differences 100 0.072 0.418 Wilcoxon Nonparametric Test for Difference; p -value = 0.2318 (one tail)

V/P = Ratio of Value-to-Price P = Stock Price (Compustat mnemonic PRCC) V is from the following Frankel and Lee (1998) model:

131

Table 4-10 Pearson Correlation Matrix for V/P Regression Variables

Panel A – One Year Before Implementation (N=204): V/P ERP NUM ACC STD V/P 1.00000

ERP 0.17504 1.00000 (0.0123) NUM -0.18636 0.07839 1.00000 (0.0076) (0.2651) ACC -0.10426 -0. 12335 -0.05154 1.00000 (0.1378) (0.0788) (0.4641) STD -0.16614 -0.10374 0.20237 0.55072 1.00000 (0.0176) (0.1398) (0.0037) (<.0001)

Panel B – Implementation Year (N=206): V/P ERP NUM ACC STD V/P 1.00000

ERP 0.13349 1.00000 (0.0558) NUM -0.20875 -0.00553 1.00000 (0.0026) (0.9371) ACC -0.08915 -0.11759 0.14072 1.00000 (0.2026) (0.0923) (0.0437) STD -0.144843 -0.10024 0.35406 0.34418 1.00000 (0.0378) (0.1517) (<.0001) (<.0001)

Panel C – One Year After Implementation (N=200): V/P ERP NUM ACC STD V/P 1.00000

ERP 0.08615 1.00000 (0.2251) NUM -0.22080 -0.00830 1.00000 (0.0017) (0.9071) ACC -0.14063 -0.08509 0.05675 1.00000 (0.0470) (0.2309) (0.4248) STD -0.25227 -0.13549 0.2 4034 0.30725 1.0000 (0.0003) (0.0557) (0.0006) (<.0001)

Continued on next page

132

Table 4-10 Pearson Correlation Matrix for V/P Regression Variables (Continued) V/P = Ratio of Value -to -Price P = Stock Price (Compustat mnemonic PRCC) V is from the following Frankel and Lee (1998) model:

(FROE t − re ) (FROE t+1 − re ) (FROE t+2 − re ) Vt = Bt + Bt + 2 Bt+1 + 2 Bt+2 1( + re ) 1( + re ) 1( + re ) re Where: FROEt = forecasted return on equity in year t = FY1/[(B t-1 + B t-2)/2] FROE t+1 = forecasted return on equity in year t+1 = FY2/[(B t + B t-1)/2] FROE t+2 = forecasted return on equity in year t+2 = [FY2(1 + Ltg)]/[(B t+1 + B t)/2] Bt = the estimated book value of common stockholders’ equity at the end of year t, Re = the estimated cost of capital. FY1 = I/B/E/S one-year consensus earning estimate FY2 = I/B/E/S two-year consensus earning estimate Ltg = I/B/E/S consensus growth rate

ERP = Dummy variable set to (1) for firms implementing ERP else (0). NUM = Average number of financial analysts providing one- and two-year-ahead forecasts. ACC = Average forecast accuracy for one- and two-year-ahead forecast based on absolute value of difference between actual EPS and forecasted EPS. STD = Average standard deviation of one- and two-year-ahead forecasts. All continuous independent variables are scaled by price. P-values in parentheses

133

Table 4-11 Value-to-Price (V/P) Ratio OLS Regression Results Results from the following value-to-price regression model:

V/P = αi + β1ERP it + β2NUM it + β3ACC it + β4STD it + εit Exp One Year Before Implementation Yr One Year After Variable Sign Coeff p-value Coeff p-value Coeff p-value Intercept ? 0.7473 <.0001 0.7864 <.0001 0.7651 <.0001 ERP + 0.1312 0.0092 0.1167 0.0762 0.0452 0.4350 NUM ? -0.1508 0.0078 -0.1909 0.0118 -0.1490 0.0052 ACC ? -0.1549 0.8320 -0.2638 0.7571 -0.4675 0.4363 STD ? -4.2147 0.3075 -4.2027 0.5844 -8.1261 0.0325

F-statistic 4.50 0.0017 3.53 0.0082 5.33 0.0004 Adjusted R 2 0.065 0.047 0.080 N 204 206 200 P-values are based on White adjusted t -statistics, two -tailed test. Analysis of Variance Inflation Factors (VIF) indicates that multicollinearity is not a major problem.

V/P = Ratio of Value -to -Price P = Stock Price (Compustat mnemonic PRCC) V is from the following Frankel and Lee (1998) model:

CHAPTER 5 INTERNAL CONTROL (ESSAY THREE)

5.1 Introduction

Internal control has played a major role in moderating the agency problem in corporations for a long time. Samson et al. (2006) document several internal control procedures used by the Baltimore and Ohio Railroad as early as 1831. In more recent times, internal control has become the subject of attention every time there is a prominent scandal in the corporate world. For instance, during the 1970s more than 400 corporations admitted making questionable or illegal payments in excess of $300 million in corporate funds to foreign government officials, politicians, and political parties, which led to enactment of the Foreign Corrupt Practices Act

(FCPA) of 1977 (Staggers 1977). Among other things, the FCPA requires publicly traded companies to devise and maintain a system of internal accounting controls sufficient to provide reasonable assurances that: (1) transactions are properly authorized, (2) transactions are recorded to permit preparation of financial statements in conformity with GAAP, (3) access to assets is authorized, and (4) that recorded asset accountability is compared to existing assets at reasonable intervals and appropriate action is taken with respect to any differences (USC 1998).

During the 1980s, several high profile audit failures led to creation of the Committee of

Sponsoring Organizations of the Treadway Commission (COSO),45 organized for the purpose of redefining internal control and the criteria for determining the effectiveness of an internal control system (Simmons 1997). They studied the causal factors that can lead to fraudulent financial reporting and developed recommendations for public companies, their independent auditor,

45 COSO is a voluntary private sector organization formed in 1985 to sponsor the National Commission on Fraudulent Financial Reporting. It is jointly sponsored by five organizations: the American Accounting Association, the American Institute of Certified Public Accountants, Financial Executives International, the Institute of Internal Auditors, and the National Association of Accountants (now the Institute of Management Accountants) (COSO 1985).

134 135

educational institutions, the SEC, and other regulators (COSO 1985). The product of their work is known as the COSO Internal Control – Integrated Framework (the framework) 46 , which was first published in 1992 (Simmons 1997).

This framework has become even more important following the corporate scandals at

Enron, WorldCom, HealthSouth, etc. and the subsequent passage of the Sarbanes-Oxley Act of

2002 (SOX). SOX includes specific provisions, in Sections 302 and 404, requiring corporations to report on the effectiveness of their internal control over financial reporting (the second of the

COSO categories). Section 302 requires the CEO and CFO to certify that their financial statements “present fairly” in all material respects, the financial condition of their company, and that they have evaluated the effectiveness of their internal controls, and disclosed any material weakness and any significant changes in internal control procedures (Ge and McVay 2005).

Section 404 is more specific, requiring companies to include in their annual report (Form 10-K), a separate management report on the company’s internal control over financial reporting and an attestation report issued by a registered public accounting firm. The internal control report requires specific language including: (1) a statement of management’s assessment of the effectiveness of the company’s internal control over financial reporting; (2) a statement identifying the framework used by management to evaluate the effectiveness; and (3) a statement that the registered public accounting firm that audited the company’s financial statements included in the annual report has issued an attestation report on management’s assessment of the company’s internal control over financial reporting (SEC 2003). Although other frameworks may be accepted, the SEC has specifically stated that the COSO Framework satisfies SEC criteria

46 The framework broadly defines internal control as “a process, effected by an entity’s board of directors, management and other personnel, designed to provide reasonable assurance regarding the achievement of objectives in the following categories: effectiveness and efficiency of operations, reliability of financial reporting, and compliance with applicable laws and regulations (COSO 1992)”.

136

and “may be used as an evaluation framework for purposes of management’s annual internal control evaluation and disclosure requirement” by companies listed on U.S. stock exchanges

(Gupta and Thomson 2006).

The COSO framework identifies five interrelated components of internal controls: (1) the control environment, (2) risk assessment, (3) control activities, (4) information & communication, and (5) monitoring. It states that “there is synergy and linkage among these components, forming an integrated system that reacts dynamically to changing conditions.” The framework also points out that controls are most effective when they are “built into” the entity’s infrastructure. It further states that “built in controls support quality and empowerment initiatives, avoid unnecessary costs and enable quick response to changing conditions” (COSO 1992).

A key feature of enterprise resource planning (ERP) systems is their use of “built-in” control features that mirror the firm’s infrastructure. In fact, vendors that sell these systems emphasize these features in their marketing literature, especially since the enactment of SOX 47 .

Although ERP vendors assert in their marketing literature that these systems will help firms comply with SOX, to my knowledge no independent empirical evidence has been provided to support these claims. This essay addresses that gap in the research by examining annual report data of firms that have implemented ERP systems and comparing their reported internal control weaknesses (ICW) to similar firms that have not implemented ERP systems.

47 SAP AG, the largest provider of ERP software in the world (Eschinger 2006), states the following in one of their brochures: “Embedded system controls within mySAP ERP financials include edit checks and tolerances for document accuracy, required and system-populated fields for document completeness, and checks to prevent duplication of accounting postings…mySAP ERP Financials can help you reduce risk related to compliance with the U.S. Sarbanes-Oxley Act (SAP 2005).” Also, Oracle Corporation, the world’s 2 nd largest provider of ERP software (Brunelli 2006), in one of their brochures, states the following: “Each application in the Oracle Financials product family uses embedded controls to automate process flows and enforce compliance across the organization, such as cross-validation rules for master data, 2-, 3-, and 4-way purchase-order matching, sequential numbering, and the ability to centrally set quantity and price-tolerance limits during invoice processing. This automated approach reduces risk by enforcing business rules and simplifies auditing activities by making it easier to test controls (Oracle 2005).”

137

The essay uses a sample of 147 firms that implemented ERP systems between 1994 and

2003 and a control sample of 147 firms that have not implemented such systems matched by industry and size. From this population, firms that were required to comply with SOX in the first three years (2004-2006) were evaluated based on the frequency of reported ICW. The resulting data from 536 firm-year observations provides evidence that ERP implementing firms are less likely to report ICW than non-ERP firms. The results are robust to a smaller sub-set of 372 observations where firms are specifically matched on a year by year basis. The essay further examines factors contributing to ICW and finds that IT related factors are less likely to exist in

ERP firms than in non-ERP firms, while non-IT related factors are not.

These findings are important because SOX has been such a controversial issue in both the academic and professional communities. It provides support for the argument that ERP vendors put forward that these systems can help companies comply with SOX. These findings should be of interest to a number of constituencies, including: software vendors, purchasers of software, auditors (both internal and external), financial executives and academics.

The remainder of this essay is organized as follow: Section 5.2 summarizes prior research and develops the hypotheses, Section 5.3 describes the data selection process and the methodology/models used, Section 5.4 presents empirical results, and Section 5.5 concludes.

5.2 Prior Research and Hypotheses Development

Section 404 of SOX requires companies to include in their annual report (Form 10-K), a separate management report on the company’s internal control over financial reporting and an attestation report issued by a registered public accounting firm. These requirements became effective December 23, 2004 under a two phase schedule. In the first phase, compliance is

138

required for companies, known as accelerated filers 48 in annual reports for fiscal years ending on or after November 15, 2004 (SEC 2003). The second phase, which includes compliance for all other companies has been extended several times and is effective for fiscal years ending on or after December 15, 2007 for the management report and December 15, 2008 for the auditor attestation report (Cross et al. 2006). Although the exact form and language of the management report on internal controls may vary from one firm to another, the report must disclose if there are any material weaknesses in the internal controls over financial reporting. Therefore it is possible to measure the effectiveness of internal controls based on this report.

Because this legislation is relatively new, the amount of prior research related to SOX is limited. Raghunandan and Dasaratha (2006), for instance, find that audit fees increased in 2004, the first year of SOX compliance over the prior year, that firms reporting ICW had higher audit fees, and that the association between fees and ICW does not vary with respect to the type of

ICW. Ettredge et al. (2007) find that clients paying higher audit fees are more likely to dismiss their auditors in the post SOX period and that dismissals are associated with smaller companies, companies with going concern reports, and companies that later reported material weaknesses in their internal controls. Ogneva et al. (2007) find that on average, ICW is not directly associated with higher cost of equity. Ashbaugh-Skaife et al. (2008) document that firms reporting internal control deficiencies have lower quality accruals, and that those whose auditors confirm remediation exhibit an increase in accrual quality relative to firms that do not remediate their control problems. Grant et al. (2008) examine 278 companies reporting IT control deficiencies in the first three years of SOX 404 requirements (2004-2006) finding that firms with IT control

48 Accelerated filers are firms that have at least $75 million in market capitalization, have been subject to SEC reporting requirements for at least 12 calendar months, that previously have filed at least one annual report, and that are not eligible to file their quarterly and annual reports on Form 10-QSB and 10-KSB. Also excluded are mutual funds and foreign firms that file Form 20-F.

139

deficiencies report more internal control (IC) deficiencies, are smaller, pay higher audit fees, and are typically audited by smaller accounting firms. Although Walker (2008) and others provide anecdotal information about the relationship between ERP, BPM (business process management), and SOX, to the author’s knowledge no empirical research has been completed, or is underway, examining the relationship between ERP systems and SOX Section 404 compliance.

SOX legislation also mandated creation of the Public Company Accounting Oversight

Board (PCAOB), to oversee the auditors of public companies (PCAOB 2007). This private- sector, non-profit corporation has issued Auditing Standard No. 2 as the authoritative guidance for audits of internal control over financial reporting performed in conjunction with an audit of financial statements (PCAOB 2006). This guidance provides some insight into the relationship between information technology and internal controls. For instance paragraph 105 states: “The auditor should subject manual controls to more extensive testing than automated controls. In some circumstances, testing a single operation of an automated control may be sufficient to obtain a high level of assurance that the control operated effectively, provided that information technology general controls also are operating effectively.”

Further, in paragraph 126, the standard indicates that the nature of information technology controls might permit the auditor to rely on the work of others, for instance in the area of change control, which most software vendors provide. The document also provides guidance which may be considered supportive of ERP vendor claims about easier compliance with SOX. For instance, in Appendix B, Example B-4, the document addresses the situation where a company’s computer system performs a three-way match of the receiver, purchase order, and invoice and if there are any exceptions, produces a list of unmatched items that employees review and follow up with on a weekly basis. This is cited as an example of a computerized application control, and a manual follow-up detective control. In this situation, testing of the computerized component would be

140

minimal, provided that the auditor is satisfied that the program functions the same way each time, and that the company does not modify the core functionality of the application. By contrast, a firm that relies on a manual process to perform the three-way match would require additional testing to satisfy the auditor that the controls are effective. Assuming these guidelines are representative of what would be expected from an audit perspective; by extrapolation one could expect that these conditions would also represent the underlying condition of the controls. Since no prior research has addressed this issue, based on these guidelines, and the assertions made by

ERP vendors, the following hypothesis is tested:

H3a: Implementation of Enterprise Resource Planning (ERP) systems will have a positive impact on overall internal controls over financial reporting as measured by SOX Section 404 compliance.

A number of researchers are differentiating between information technology (IT) related controls and general controls. For instance, Canada et al. (2007) examine the relationship between audit fees and material weaknesses in IT related controls, and Li et al. (2007) examine the relationship between internal and external governance factors and material weaknesses in IT controls. The PCAOB does not divide controls into IT and Non-IT categories. Instead, IT controls are assumed to be interspersed within the five COSO categories: (1) control environment, (2) risk assessment, (3) control activities, (4) information and communications, and

(5) monitoring. Furthermore, the PCAOB identifies controls as either company-level controls or specific controls, with company-level controls including IT general controls over program development, program changes, computer operations, and access to programs and data that may help ensure that specific controls over the processing of transactions are operating effectively. To the extent that most of these type of IT controls are designed into the standard ERP systems using so called “best practices,” one could assume that these controls would be better than an

141

alternative environment where home grown legacy software application are in place. This leads to the following hypothesis:

H3b: Implementation of Enterprise Resource Planning (ERP) systems will have a positive impact on information technology (IT) related internal controls as measured by SOX Section 404 compliance.

Even those controls that are not considered to be IT related controls may be impacted by the use of an ERP system that is set-up and implemented using “best practices.” In reviewing each of the COSO categories, one can find instances where the use of a standardized ERP system can contribute to the effectiveness of both company-level and specific controls. For instance, just the act of acquiring an ERP system contributes to the control environment , which sets the tone of internal control in an organization. Knowing that standard controls will be required in order to implement the system successfully sends a message to the organization that internal control is important. Risk assessment is made easier when standardized programs such as those in an ERP system are used, because benchmarks are available for comparisons. Policies and procedures can be programmed into the system, which makes enforcement of control activities easier.

Information and communication is at the heart of any ERP system with a goal of delivering better quality information to decision makers faster. And finally, monitoring is made easier because the system can be programmed to generate exception reports that by-pass potential conflicts of interest. All of these factors taken as a whole lead to the following hypothesis:

H3c: Implementation of Enterprise Resource Planning (ERP) systems will have a positive impact on non-information technology (non-IT) related internal controls as measured by SOX Section 404 compliance.

142

5.3 Data Selection and Methodology/Models

5.3.1 Sample Data Selection

The sample data is based on 147 firms from 33 industry groups that implemented ERP systems between 1994 and 2003. Panel A of Table 5-1, provides a summary of the sample selection process for these ERP implementing firms. The process starts with the list of 91 ERP announcements made between 1994 and 1998 from Hayes et al. (2001) 49 , from which 36 firms were eliminated because they are no longer listed or data was otherwise not available. The second step in the process involves a search of all available newswire services using the Lexis-

Nexis service for years after 1998, searching on key phrases such as: “ERP”, “Enterprise

Resource Planning”, and “Enterprise Systems.” This search found an additional 92 firm announcements yielding a total of 147 firms for which accounting data is available for at least three years following implementation in the Research Insight Compustat database.

Following the recommendations of Barber and Lyon (1997), and the approach used by

Nicolaou (2004), this essay uses a matched pairs approach to select a control group. ERP implementing firms are first matched with other firms based on their four digit SIC code, then by total assets at the beginning of the implementation year, then by the availability of Compustat accounting data for at least three years following implementation. If a reasonable match is not available then three or two digit SIC codes are used to obtain a match. Once a matched firm is identified, a further search of Lexis-Nexis Newswires is conducted for all available years using a combination of the firm name, and the following terms: (1) ERP, (2) Enterprise Resource

Planning, (3) Enterprise Systems, (4) SAP, (5) Oracle, (6) QAD, (7) Baan, (8) Peoplesoft, (9) JD

49 I would like to thank David C. Hayes and Jacqueline L. Reck for providing this list of firms

143

Edwards, and (10) Lawson 50 . If no newswire records are found, the firm is used as a match for this study. If a record is found that indicates an ERP system may be in use, then the next closest firm in terms of total assets is used, and the process repeated. Six matches required the use of three digit SIC codes: (201)-Food and Kindred Products Manufacturing, one match; (252)-

Furniture and Fixture Manufacturing, two matches; (284)-Chemicals and Allied Products

Manufacturing, two matches; and (594)-Miscellaneous Retail, one match. Two digit SIC codes are used for four matches: (28) Chemical Manufacturing, two matches, (35)-Industrial and

This essay examines the first three years that SOX 404 reporting was required, 2004-2006, which would provide for a maximum of 882 firm-year observations if all firms were required to file during the first three years. Not all firms were required to file in all three years for a number of reasons, including the fact that the effective date was for years beginning after November 15,

2004, therefore only firms with November and December year-ends would be included for 2004.

Also some of the sample firms were not classified as accelerated filers due to market capitalization levels and/or filing status. Eliminating these observations leaves 536 firm-year observations for which SOX 404 reporting data is available in the Audit Analytics database.

Because this sample of 536 firm-year observations does not have an exact match of data for each of the firms in each year, a smaller sub-set of 372 firm-year observations has been created where data for both matched pairs are available in each individual year. These two samples are hereafter referred to as the “primary” sample and the “matched pair” sample.

50 The specific vendors listed (4-10) represent the seven most common ERP systems identified in Nicolaou (2004).

144

Panel B of Table 5-1 provides a comparison of the mean, median, standard deviation, minimum and maximum values of total assets at the beginning of the implementation year for each of the two groups. Although the ERP firms are on average larger than the control firms, t- tests indicate that the differences are not statistically significant for the three years at the traditional .05 level.

[Insert Table 5-1 here ]

Table 5-2 provides a breakdown of ERP implementing firms by two-digit SIC code and implementation year.

[Insert Table 5-2 here ]

5.3.2 Logistic Regression Model

This essay uses a logistic regression model adopted from Ogneva et al. (2007) with internal control weakness (ICW) as the dependent variable, and an indicator variable for ERP implementers as the primary variable of interest. It includes control variables that prior research indicates is associated with ICW, including those related to: (1) complexity of operations, (2) organizational change, (3) accounting application measurement risk, and (4) resource constraint indicators (Ge and McVay 2005; Doyle et al. 2007; Ashbaugh-Skaife et al. 2008) plus one control variable related to the age of the ERP systems 51 .

The resulting model is as follows:

Prob (WEAK it ) = f (β + β1ERP it + β2LOGSEG it + β3FOREIGN it + β4M&A it + β5RESTR it

+ β6SALEGRW it + β7INVTAT it + β8LOGMKTV it + β9LOSS it + β10 ZSCORE it +

β11 LOGAGE it + β12 Y2K it + β13 BIG4 it + β14 LOGERPAGE it ) (1)

51 The age of the ERP system is thought to be important in that firms can be expected to implement more of the built-in control features of the system as time goes on.

145

Where:

WEAK = Indicator variable equal to 1 for firms reporting an ICW and 0 otherwise.

ERP = Indicator variable equal to 1 for firms that have implemented ERP systems and

0 otherwise.

The control variables are:

LOGSEG = Natural log of number of business segments (Compustat mnemonic SEGNUM)

FOREIGN = Indicator variable equal to 1 if the firm has a non-zero foreign currency

translation (Compustat mnemonic FCA) and 0 otherwise.

M&A = Indicator variable equal to 1 if acquisitions are reported on the Statement of

Cash Flows (Compustat mnemonic AQC) and 0 otherwise.

RESTR = Indicator variable equal to 1 if at least one of the following restructuring

variables is not equal to zero (Compustat mnemonics RCP, RCA, RCEPS,

RCD) and 0 otherwise.

SALEGRW = Percentage change in sales (Compustat mnemonic SALE)

INVTAT = Ratio of inventory over total assets (Compustat mnemonic INVT / AT)

LOGMKTV = Natural log of market value of equity (Compustat mnemonic MKVAL)

LOSS = Indicator variable equal to 1 if earnings before extraordinary items (Compustat

mnemonic IB) are less than zero and 0 otherwise.

ZSCORE = Altman’s Z-Score (Compustat mnemonic ZSCORE).

LOGAGE = Natural log of years the firm exists in the CRSP database 52 .

BIG4 = Indicator variable equal to 1 if the firm’s auditor is one of the Big 4 firms and 0

otherwise.

52 If firm not found in CRSP (Center for Research in Security Prices at the University of Chicago Graduate School of Business) database, then age was based on information from the company’s web site.

146

LOGERPAGE = Natural log of the number of years since the ERP system was implemented 53 .

The control variables from prior research that are hypothesized to be associated with the discovery of ICW are categorized as: (1) complexity of operations (LOGSEG and FOREIGN);

(2) organizational change (M&A and RESTR); (3) accounting applications measurement risk

(SALEGRW and INVTAT); (4) indicators of resource constraint (LOGMKTV and LOSS); and

(5) other factors including potential for bankruptcy (ZSCORE), age of the company (LOGAGE), and size of the audit firm (BIG4). One additional control variable is added specific to ERP systems, (LOGERPAGE) is used to control for the possibility that ERP systems that have been in place longer are more likely to have implemented more of the built-in internal control features.

5.4 Empirical Results

5.4.1 Frequencies of Internal Control Weaknesses

Table 5-3 summarizes the frequency of internal control weaknesses (ICW) reported by firms in their 10K annual reports for the years 2004-2006 extracted from the Audit Analytics database. The first three columns report results for the primary sample of 536 firm-year observations and the last three columns report results for the matched pair sample. Each sample shows results for each year and a total for the three years. The primary sample shows that 73 firm-years out of the 536 (14%) reported at least one ICW, with 33 out of 311 (11%) ERP implementing firm-years compared to 40 out of 225 (18%) of the control firms. The 7% proportional difference between ERP implementers and control firms has a proportional Z- statistic of 2.39 (p=0.0085), which indicates a significant difference by traditional standards. The matched pairs sample has similar results, with 10% of the ERP implementing firm-years,

53 For control firms, the year in which the matching firm implemented ERP was used to determine the age.

147

compared to 19% of the control firm-years reporting at least one material weakness, with the 9% difference highly significant (Z=2.52; p=0.0058).

The results for individual years reflect significant differences between ERP implementers and non-implementers for 2004 and 2006, but no significant differences in 2005. In 2004, the primary sample is 14% lower for ERP implementers (15% vs. 29%; p=0.0198) and 21% lower for the matched pairs sample (11% vs. 32%; p=0.0099). In 2006, the differences are 8% lower (7% vs. 15%; p=0.0355) and 10% lower (6% vs. 16%; p=0.0279) for the primary sample and matched pairs sample respectively. These results are in line with the overall results reported by Audit

Analytics (Cheffers and Whalen 2007) for the first three years that SOX Section 404 compliance was required: Year 1 = 16.9% of 3,700 opinions filed; Year 2 = 10.5% of 3,791 opinions filed; and Year 3 = 9.6% of 4,051 opinions filed.54

[Insert Table 5-3 here ]

5.4.2 Logistic Regression Analysis

Although the univariate results provide initial support for the first hypothesis (H3a) suggesting that a higher proportion of control firms reported internal control weaknesses than

ERP implementing firms, other omitted variables may be contributing to the differences. To control for these omitted variables, a logistic regression analysis is run using the model described in the prior section that includes variables thought to be associated with internal control weaknesses. The results are presented in Table 5-4 for both the primary sample and the matched pair sample. The probability modeled is WEAK=1 (by using the SAS DESCEND function), which enables interpretation of parameter estimates based on the likelihood they are associated

54 Audit Analytics defines Years 1 through 3 as November 15, 2004 through November 14, 2007, which corresponds to the effective date of the SOX Section 404 reporting requirements. Year 3 data includes 43 registrants that failed to file by the deadline, on the assumption that they are expected to file an adverse opinion.

148

with weaknesses. Positive estimates indicate that dichotomous variables with a value of zero and lower continuous variables are more likely to be associated with weaknesses. By contrast, negative estimates indicate that dichotomous variables with a value of zero and lower continuous variables are less likely to be associated with material weaknesses. Therefore, it is expected that the primary variable of interest (ERP) will be positive, indicating that non-ERP implementers

(ERP=0) are more likely to report material weaknesses than ERP implementers (ERP=1). As for the control variables, firms with fewer segments, no foreign operations, no merger and acquisition activity, no restructuring activity, lower sales growth, lower inventory, lower market value, and no losses are expected to have negative estimates indicating that they are less likely to report weaknesses than their counter parts. By contrast, firms that have lower Z-Scores (indicating a higher risk of bankruptcy), firms that are younger, firms that use non Big 4 audit firms, and firms that have newer ERP systems are expected to have positive estimates indicating that they are more likely to report weaknesses.

The variable of interest (ERP) is positive and significant (p=0.0396) in the primary sample and marginally significant (p=0.0947) in the matched pair sample, as expected. The odds ratio provides for interpretation that non-ERP implementing firms are 1.810 (1.819) times more likely to report material weaknesses in internal controls than ERP implementing firms. A number of the control variables are not significant in either sample. In the primary sample, smaller firms

(LOGMKTV = -0.4481; p<.0001) and those without losses (LOSS=-0.4684; p=.008) were less likely to report weaknesses, as expected. However, sales growth was marginally significant in the opposite direction expected, indicating that firms with lower growth were more likely rather than less likely to report weaknesses (SALEGRW=0.7773; p=0.0829). Similarly, firms with lower Z Scores, which is considered to be an indicator of potential bankruptcy, have a marginally significant negative estimate (ZSCORE=-0.0563; p=0775), indicating that they are less likely to

149

report weaknesses. For the matched pair sample, only the size estimate was significant and in the direction expected (LOGMKTV=-0.5298; P<.0001), indicating that smaller firms are less likely to report weaknesses. Firms with no M&A activity had marginally significant estimates in the expected direction (M&A=-0.3000; p=0.0802) indicating that they were less likely to report weaknesses. Similar to the primary sample, firms with lower Z scores were also marginally less likely to report weaknesses (ZSCORE—0.0658; p=0.0669).

The model is a relatively good fit, with a Likelihood ratio of 64.9 (p < 0.001) for the primary sample and 43.5 (p < 0.001) for the matched pair sample. The Max-Rescaled pseudo R 2 is 0.2076 and 0.1974 for the primary and matched pair sample respectively. Overall, these results tend to support the first hypothesis in that the control firms in both samples are more likely to report material weaknesses than firms that implemented ERP systems, therefore ERP systems appear to have a positive impact on internal controls over financial reporting.

[Insert Table 5-4 here ]

5.4.3 Factors Contributing to Material Weakness Determination

To test the second and third hypotheses, it is necessary to examine the factors contributing to material weakness determination. Audit Analytics includes in their database a listing of these factors which they group into three primary categories: (1) internal controls over financial reporting, (2) accounting rule (GAAP/FASB) application failures, and (3) financial fraud, irregularities and misrepresentations. Appendix 5-1 summarizes the Audit Analytics’ definition of each of the factors included in the first two categories 55 . There is no specific standardized requirement for companies or their auditors to categorize the factors that lead to a determination of internal control weakness. These factors are classified by Audit Analytics by

55 Factors included in the third category are the same factors as the second.

150

reading the compliance document. Appendix 5-2 is an example of an audit report issued by

PricewaterhouseCoopers, LLP for Pride International, Inc. for December 31, 2004, which concurs with managements’ determination that a material weakness has been identified. The appropriate section of the audit report has been underlined to illustrate the point. Based on this language,

Audit Analytics identified four contributing factors related to internal control weaknesses and seven factors related to accounting rules application failures, which are listed at the end of the sample audit report.

Table 5-5 provides a listing of the internal control weaknesses in Panel A and accounting rule application failures in Panel B for the primary sample of 536 firm-year observations and the matched pair sample of 372 matched pair firm-year observations 56 . The table shows the frequency with which each of the factors have been cited and the proportion of the total represented by that frequency in parenthesis. Statistical significance of the difference in proportions is designated with bold italics underline for p-values less than 0.01, bold italics for p-values less than 0.05, and bold face for p-values less than 0.10 based on a one-tailed z-test that assumes the proportion of ERP firms will be less than the proportion of non-implementing control firms. For instance, the most often cited factor in both samples, “accounting documentation, policy and/or procedures” was cited by 39 (17.3%) control firms and 32 (10.3%) ERP firms in the primary sample and 35 (18.8%) control firms and 17 (9.1%) ERP firms in the matched pair sample. The differences are both statistically significant, at a p-value < 0.01 level. This result supports the expectation that the control firms will report more factors than the ERP implementing firms.

56 The frequency of observations in the third category (financial fraud, irregularities and misrepresentations) is small, with only two factors for ERP implementers and two factors for non- implementers. Therefore this category has been excluded from the analysis.

151

Although some of the factors are not statistically different between the ERP implementers and control firms, 4 (6) of the internal control factors and 6 (5) of the accounting rule application failures are at least marginally significant in the primary (matched pair) sample. In the internal control group, they include (in addition to the one discussed above): “accounting personnel resources, competency/training” for the primary sample; “restatement or nonreliance of company filings”; and “restatement of previous 404 disclosures” for both samples; “segregation of duties/design of controls personnel”, and “senior management competency, tone, reliability” for the matched pairs sub-sample. The differences in the accounting rule application failure category that are at least marginally significant include: “capitalization of expenditure issues”; “deferred, stock-based or executive compensation issues”; “intercompany/investment with subsidiaries/affiliates issues”; and “tax expense/benefits/deferral/other (FAS 109) issues” for both samples. “Consolidation, (Fin46r/Off BS) & foreign currency translation issues” and “financial derivatives/hedging (FAS 133) accounting issues” are at least marginally significant in the primary sample, as is “acquisition, merger, disposal or reorganization” in the matched pair sample.

Only one of these contributing factors specifically mentions IT controls; “information technology, software, security, access.” The frequency with which this factor is cited does not provide support for the second hypothesis since ERP implementing firms actually cite this factor more often than the control firms, although the differences are not statistically significant. In the primary sample 10 (3.2%) ERP firms and 6 (2.7%) control firms cite this factor while in the matched pair sample 5 (2.7%) ERP firms and 4 (2.2%) control firms cite it. However, this category addresses only “direct IT” issues, and is not representative of the “IT related” issues that are the subject of H2b. Remember that IT controls are assumed to be interspersed within the five

COSO categories and are integrated into ERP systems based on “best practices.” Therefore, in

152

order to test H2b and H2c, a panel of experts was used to review the various factors defined by

Audit Analytics and identify those that would be directly or indirectly impacted by use of an ERP system. The factors identified by the panel are identified in Table 5-5 with an asterisk (*). They include 9 out 15 in the internal control weakness category and 13 of 22 in the accounting rule application failure category. The panel consisted of three faculty members from a prominent

Midwestern state university, with teaching and research interests in: auditing, managerial accounting, and accounting information systems.

[Insert Table 5-5 here ]

Table 5-6 summarizes frequencies of internal control weaknesses as divided by the expert panel. Panel A shows results for the IT related factors, which relate to hypothesis H3b, while

Panel B shows results for the non-IT related factors which relate to hypothesis H3c. In the primary sample, 11% of ERP implementers report IT related factors contributing to reported weaknesses compared to 17% for the control firms (Z=2.25; p=0.0121). The difference in the matched pairs sub-sample is even more, with only 10% of ERP implementing firms compared to

19% of the control firms citing IT related factors as contributing to weaknesses (Z=2.52; p=0.0058). The non-IT follow a similar pattern, with 14% vs. 10% (Z=1.36; p=.0867) for the full sample and 15% vs. 10% (Z=1.43; .0762) for the matched pairs sample. These results provide initial support for both H3b and H3c before taking into consideration the possible impact of omitted variables.

[Insert Table 5-6 here ]

Table 5-7 summarizes results of logistic regressions of the IT related factors from Table

5-6, using the equation (1) model, which includes control variables thought to be associated with internal control weakness. The primary variable of interest (ERP) is positive and marginally significant in both the primary sample (p=0.0672), and the matched pair sample (p=0.0947). The

153

odds ratio indicates that control firms are 1.703 (1.819) times more likely to report IT related factors contributing to internal control weaknesses than ERP implementing firms, thus further supporting hypothesis H3b. As with the previous logistic regression, only a few of the control variables are significant with the expected sign, indicating they are impacting the results similar to prior research. They include firms with no M&A activity and lower market value, in both samples, and firms with no losses in the primary sample all being less likely to report IT related factors. Contrary to expectations, firms with lower sales growth in the primary sample are more likely to report IT related factors and firms with lower Z-scores in both samples are less likely to report IT related factors. Even though the control variables do not meet expectations, the overall model is reasonable, with likelihood ratios of 66.4841; p<.0001 (43.5915; p<.001) and Max- rescaled pseudo R 2s of .2138 (.1974) for the primary (matched pair) samples.

[Insert Table 5-7 here ]

Table 5-8, summarizes results of a similar logistic regression of the non-IT related factors to test the third hypothesis (H3c). The primary variable of interest (ERP) remains positive, but is no longer significant either in the primary sample (p=0.2948), or the matched pair sample

(p=0.5536). Therefore, H3c is not supported. Results for control variables are generally the same as the regression of IT related factors.

[Insert Table 5-8 here ]

5.5 Conclusions

This essay examines the impact that enterprise resource planning (ERP) systems have on compliance with Sarbanes-Oxley Section 404 (SOX). It uses a sample of firms that implemented

ERP systems from 1994 to 2003 matched with control firms based on industry and size. It uses internal control weaknesses (ICW) to measure the impact by comparing the proportion of firms in each group that have reported such weaknesses. The results show that a smaller proportion of

154

ERP implementing firms have reported ICW than the control firms in the first three years that

SOX Section 404 has been effective, 2004-2006. These results hold up in a logistic regression analysis that includes several control variables that prior research has found to be associated with

ICW. The regression suggests that firms that implemented ERP systems are less likely to report

ICW than the control firms, after controlling for the other variables. These results are also robust in a smaller sub-sample that specifically matches each of the ERP firms and control firms on a year by year basis.

The essay also examines factors contributing to ICW by dividing them into those that are

IT related and therefore would be expected to be either directly or indirectly impacted by the use of an ERP system, and those that are not IT related. The initial examination shows that the proportion of ERP implementing firms reporting both groups of factors is less than the control firms. However, when control variables are added in a logistic regression, only the IT related factors are significantly different, with ERP firms less likely to report IT related factors than the non ERP firms, but no difference for the non-IT related factors.

These results are important because they provide support for the argument that ERP software vendors have been making that implementation of their systems will help firms comply with Sarbanes-Oxley. Also, since internal control is considered as a moderator for the agency problem, these results contribute to a better understanding of that role. To my knowledge, this is the first empirical test of those claims and should be of interest to those responsible for purchasing such systems, auditors, software vendors, and the academic community.

This essay does have some limitations, including the fact that the sample suffers from a large-firm bias that results from the phased-in timing of compliance with SOX Section 404.

Future research should expand this study to include the smaller non-accelerated filers after sufficient data has accumulated. The essay is also relying on the Audit Analytics database as a

155

primary source of compliance information. This is a relatively new source of data for archival research, since its primary focus has been on gathering compliance data that has only recently been required. However other accounting researchers are using this data source in peer reviewed publications (Raghunandan and Rama 2006; Ogneva et al. 2007), which should strengthen its credibility. There is also the problem of human judgment related to the factors that contribute to internal control weaknesses. The first judgment is made by the staff at Audit Analytics who determine which factors apply by reading the management and audit reports that are filed. The second judgment is in determining which of those factors are IT-related; and which are not. The first could be validated with additional reading of the source documents, but would probably not add significant strength to the overall results of the essay. The second is addressed by using an expert panel to spread the judgment to multiple parties from different perspectives. Another limitation is that a number of the firms in the original sample are foreign based and as such were not required to comply with SOX Section 404 in the first two years of the study. Future research should re-visit these firms after addition years of data are available.

A final limitation relates to the fact that that there is no way to know whether a self selection bias may be involved. In other words, it is possible that firms that implemented ERP systems are those that would have had better internal controls prior to implementation. Because the requirement for SOX Section 404 compliance did not begin until after the firms involved in this analysis had already implemented ERP systems, there is no way to perform any kind of an before and after analysis. Therefore these results must be considered with this caveat in mind.

156

Appendix 5-1 Audit Analytics’ Definition of Factors Leading to Internal Control Weakness Internal Control Weaknesses If the assessment of internal controls over financial reporting identified material weaknesses, one or more of the following types of weaknesses are listed.

Accounting documentation, policy and/or procedures Represents material weaknesses deriving from internal control systems that do not contain adequate documentation, policies or other means of justifying account balances. These issues may also include failures to ensure that accounts are recorded based on GAAP, SAB, FASB and/or the appropriate accounting methodology are followed. They may also include failures in policies or procedures designed to gather the correct information on a timely basis or problems with the y/e close process. It also includes failures to employ proper procedures over journal entries, non- routine transactions and other common procedural failures. This is a catch all category. Almost by definition this item will be checked whenever a company indicates an ineffective 404 situation.

Accounting personnel resources, competency/training Consists of problems with accounting personnel resources, competency, training, experience and/or adequacy in any way. To meet these criteria, such an indication would have to be contained in the filing or in the remediation plan.

Ethical or compliance issues with personnel Consists of problems with personnel in the areas of compliance with policies, maintenance of ethical standards, fraud and intentional acts that lead to (or could lead to) misstated account balances or financial reports.

Inadequate disclosure controls (timely, accuracy, complete) Represents material weaknesses related to the adequacy of information flow that should result in a required disclosure.

Ineffective regulatory compliance issues Consists of internal control deficiencies associated with failures to meet regulatory requirements other than taxes.

Ineffective, non-existent or understaffed audit committee Represents circumstances where an audit committee may not exist, have the personnel, expert, experience and/or resources to perform their duties to the extent required by Sarbanes Oxley or their charter. This item can also be checked when a board has not independent directors or other oversight mechanism.

Information technology, software, security & access issue Deficiencies in this category include deficient program controls, software programs/implementation, segregation of duties associated with personnel having access to computer accounting or financial reporting records and related problems with oversight/access to electronic data/programs

Insufficient or non-existent internal audit function

157

Indicates circumstances where a company has stated that its internal audit function was insufficient in identifying and/or advising in the correction of internal control deficiencies. It cannot also identify circumstances where a registrant has identified a failure to have an internal audit department at all, as a ICFR failure.

Journal entry control issues This category is checked whenever the description given by the audit firm or company refers to deficiencies or issues associated with the journal entry process. This category is not checked when there is a journal entry error that originates from control deficiencies in other areas.

Management/Board/Audit Committee investigation(s) Consists of internal control reports indicating that an internal investigation is underway relative to accounting and/or financial reporting matters. This item is demographic in nature.

Material and/or numerous auditor /YE adjustments Represents circumstances where one of the explanations for a material weakness opinion was the number and/or size of year-end adjustments including those proposed by the auditor. These adjustments also consider footnote and related errors that need to be corrected by the auditor at year-end. Too many, or auditor initiated year-end adjustments are consider prima facie evidence of a potential material weakness in financial reporting.

Non-routine transaction control issues This category is checked whenever a registrant specifically describes one of their control deficiencies as emanating from non-routine types of transactions. These could include acquisitions, asset sales, establishment of new systems and other.

Remediation of material weakness identified Refers to disclosures that indicate that material weakness or internal control weaknesses have been remediated.

Restatement or nonreliance of company filings Consists of material weakness opinions deriving from problems that led to restatements. Restatements are often evidentiary of primi-facie internal control deficiencies.

Restatement of previous 404 disclosures Represents circumstances where a company has had to restate its 404 opinion because of some event (most likely a restatement of financials) that has occurred subsequently to filing

SAB 108 adjustments noted This item is checked when the ICFR disclosure identifies that a SAB 108, as opposed to a financial restatement, process is used to correct the beginning retained earnings balances associated with previous period accounting errors.

Scope (disclaimer of opinion) or other limitations A material weakness opinion may derive from assertions from the company or auditor that the company had not completed its own review of internal controls and therefore these controls could not be audited. These limitations could come about for any number of reasons.

158

SEC or other regulatory investigations and/or inquiries An SEC or related investigation into the company affairs is often evidentiary of accounting or financial reporting issues that point to internal control deficiencies. This category seeks to identify circumstances where registrants have indicated in their 404 assertion that an SEC investigation or inquiry is underway.

Segregations of duties/ design of controls (personnel) This category covers internal control deficiencies associated with the design and use of personnel within an organization. It primarily deals with segregation of duty issues, such as clerks having access to both the cash receipts and the bank reconciliation. It may also deal with more sophisticated design of control issues relating to executives having the ability to change customer records ,etc.

Senior management competency, tone, reliability issues This category has been established to identify circumstances where internal control weaknesses are attributed directly to potentially improper or negligent conduct of the current or former senior management of the company. This does not necessarily mean that the assertion is correct, just that such language exists in the filing.

Untimely or inadequate account reconciliations In reviewing internal control assertions or opinions it is often the case that inadequate account reconciliations are identified as the reason for material or numerous adjustments. This category seeks to specifically identify such circumstances.

Accounting Rule (GAAP/FASB) Application Failures If the assessment of internal controls over financial reporting identified accounting rule application failures, one or more of the following types of failures are listed.

Accounts/loans receivable, investments & cash issues Consists of internal control deficiencies in approach, theory or calculations with respect to cash, cash equivalents, accounts receivable, short term investments, certain long term investments, notes, loans collectible, allowance for uncollectibles, notes receivables and/or related reserves.

Acquisition, merger, disposal or reorganization issues Consists primarily of internal control deficiencies in approach, theory or calculation associated with the merger, acquisitions, reorganization or disposal issues for registrants. The internal control issues in this area can vary from incorrect application of GAAP to calculate the proper intangible assets levels associated with acquisitions to failure to record the proper reserves for disposal or reorganization. Accounting rules in this area are considered complex and non-routine. This category is often attributed to failures by personnel in understanding certain issues associated with acquisitions or disposals.

Balance sheet classification of asset issues Consists of internal control deficiencies in approach, theory or calculation associated with how assets were classified on the balance sheet. Primarily this category is made up of misclassified

159

assets as short term versus long term or whether certain assets were properly considered cash equivalents versus short-term investments.

Capitalization of expenditures issues Consists of internal control deficiencies in approach, theory or calculation associated with the capitalization of expenditures. These can include expenditures capitalized for inventory, construction, intangible asset, R&D, software or product development and other purposes. Whether capitalizing expenditures in inventory, leaseholds, buildings or product/software development, the proper methodology can be difficult and demanding on an internal control system.

Cash flow statement (FAS 95) classification errors Consists of internal control deficiencies in approach, theory or calculation that manifested themselves in cash flow statements (FAS 95) that are not consistent with GAAP. These misclassifications can affect cash flow from operations, financing, investment, non-cash and other areas. Difficulties with respect to internal control systems over proper disclosure associated with cash flow statements typically occur with non-routine transactions.

Consolidation, (Fin46r/Off BS) & foreign currency translation issues Consists of internal control deficiencies in approach, theory or calculation with respect to the consolidation of subsidiaries including variable interest entities and off balance sheet arrangements. This can include mistakes in how joint ventures, off balance sheet entities were recorded or disclosed. This category also identifies issues associated with foreign currency translations, minority interests, eliminations or other issues associated with consolidations.

Debt ,quasi-debt, warrants & equity ( BCF) security issues Consists of internal control deficiencies in approach, theory or calculation associated with the recording of financing/bank/securities debt or equity section accounts. Control issues in this area often arise because of incorrect recording of beneficial conversion features in debt/quasi debt or equity securities. They can also occur with the calculation of premiums/discounts on debt securities or the proper valuation of certain non-traded equity securities.

Debt and/or equity classification issues Consists mainly of internal control deficiencies in approach, theory or calculation associated with the proper classification of debt instruments as short term or long term. This area can also refer to reclassifications between equity and debt accounts or within equity accounts.

Deferred, stock-based or executive comp issues Consists of internal control deficiencies in approach, theory or calculation associated with the recording of deferred or executive compensation. The majority of these errors are associated with the valuation of options or similar derivative securities and their recording on the books. Sometimes this issue arises when personnel are paid with shares or options instead of cash. This category also includes other forms of internal control deficiencies associated with executive compensation arrangements.

Depreciation, depletion or amortization issues Consists of internal control deficiencies in approach, theory or calculation associated with depreciation of assets, amortization of assets and/or amortization of debt premiums or discounts.

160

This category can also include deficiencies associated with depletion of reserves or amortization of other fixed assets.

Expense recording (payroll, SG&A) issues Consists of internal control deficiencies in approach, theory or calculation associated with the expensing of assets or understatement of liabilities. These issues can arise from any number areas including failure to record certain expenses, write off certain assets or acknowledge certain liabilities. This category is used primarily for miscellaneous occurrences of expensible items including payroll and SG&A issues.

Fin Stmt, footnote, US GAAP , segment disclosure issues This represents failures or inadequacies in internal controls related to review of preparation of financial statements, footnotes and/or related additions to financial statements. This can also include issues with conversion of foreign company financial statements to US SEC/US GAAP/FASB Standards. It also includes internal control deficiencies associated with segment recording and related annual report disclosures.

Financial derivatives/hedging (FAS 133) accounting issues Consists of internal control deficiencies in approach, theory or calculation of derivative instruments. These can include the valuation of financial instruments such as hedges on currency swings, interest rate swaps, purchases of foreign goods, guarantees and other. Often this category is checked when registrants fail to follow the FAS 133 rules for proper documentation or application of its principles.

Foreign, related party, affiliated and/or subsidiary issues Consists primarily of internal control deficiencies associated with disclosures about related, alliance, affiliated and/or subsidiary entities. This can also refer to accounting issues detected at foreign subsidiaries. This box is checked mostly in conjunction with other categories to indicate that an issue has been raised in association with a failure at a subsidiary (often foreign sub) that has been deemed to be material to the overall financial condition of the company.

Gain or loss recognition issues Consists of internal control deficiencies in approach, theory or calculation with respect to the recording of gains or losses from the sales of assets, interests, entities or liabilities. Mistakes in these areas often result from problems with calculating the proper basis for disposing of an asset or the proper amount to record as sales revenue. Generally, this category relates to issues associated with non-routine or significant transactions.

Income statement classification, margin and EPS issues Consists primarily of internal control deficiencies associated with a registrants disclosure of financial/operational ratios or margins and earnings per share calculation issues. Also included are circumstances where income statement items are misclassified between say gross margin and selling general and administrative expenses. This may also deal with issues associated with exceptional items.

Intercompany/Investment w/ subsidiary/affiliate issues Consists primarily of internal control deficiencies in approach, theory or calculation related to intercompany or affiliate balances, investment valuations or transactions. It is often the case that

161

problems arise when intercompany balances are not reconciled and accounted for on a timely basis.

Inventory, vendor and cost of sales issues Consists of internal control deficiencies in approach, theory or calculation associated with transactions affecting inventory, vendor relationships (including rebates) and/or cost of sales. The proper recording of inventory can be a complex area of accounting requiring many estimates. The issues can range from simple valuation calculations to estimates of completion on construction projects.

Lease, FAS 5, legal, contingency & commit issues Consists primarily of internal control deficiencies associated with FAS 5 type contingencies and commitments. This description also deals with issues associated with the disclosure or accrual of legal exposures by registrants and issues associated leases and lease commitments. One significant area of impact has been internal control deficiencies associated with determining the proper accounting or determination of operating vs. capitalized leases.

Lease, leasehold & FAS 13 (98) (subcategory) issues The category is checked when a lease, leasehold or related issue has been identified with internal or financial reporting controls. This represents a subcategory of the Lease, FAS 5 category.

Liabilities, payables, reserves and accrual estimation failures Consists of internal control deficiencies associated with the accrual or identification of liabilities on the balance sheet. These could range from failures to record pension obligations, to problems with establishing the correct amount of payables, accruals or other reserves. From an internal control perspective, issues in this area most often occur because of cut-off failures in recording liabilities and matching them to related revenue or inventory accounts.

PPE , intangible or fixed asset (value/diminution) issues Consists of internal control deficiencies in calculation, approach or theory that have taken place in the recording of PPE, fixed, intangible, goodwill or long term assets. It also applies to contra liabilities that are required to be valued or assessed for diminution. Generally issues associated with long term development projects and goodwill associated with acquisitions are included in this category.

Revenue recognition issues Consists of internal control deficiencies in approach, understanding or calculation associated with the recognition of revenue. Many of these restatements originate from a failure to properly interpret sales contracts for hidden rebates, returns, barter or resale arrangements. They can also occur because of misapplied credits or debits associated with customer accounts. This account is generally checked without regard to other accounts they impact, such as accounts receivable.

Tax expense/benefit/deferral/other (FAS 109) issues Consists of internal control deficiencies in approach, understanding or calculation associated with various forms of tax obligations or benefits. Many of these restatements relate to foreign tax, local taxes or tax planning issues. Some deal with failures associated with sales taxes, etc. The accounts impacted can include expense, deferral or allowances. With the change in goodwill

162

accounting, a number of issues have arisen with the failure of companies to change the level of permanent differences in their FAS 109 calculations.

Unspecified/unidentified/inapplicable FASB/GAAP issues This flag is identified when the 404 or 302 disclosures are lacking in sufficient information to identify what accounts or areas of financial reporting are being impacted by disclosure controls or internal control deficiencies. It may also indicate that a GAAP/FASB effect is not applicable. This flag may not be checked in circumstances where a recent section 404 report or restatement can provide the missing information.

Unspecified disclosure control deficiencies This category is checked when a company simply states that it's disclosure controls are ineffective but does not detail the actual control deficiencies.

Defective or unreliable accounting/reporting records Consists of disclosures by a registrant that a scope limitation exists with respect to the company's ability to rely on accounting or internal control records. Typically no restatement is announced because the amount, if any, cannot be determined.

163

Appendix 5-2 Example of an Audit Report - Pride International, Inc.

REPORT OF INDEPENDENT REGISTERED PUBLIC ACCOUNTING FIRM 57

To the Stockholders and Board of Directors of Pride International, Inc.: We have completed an integrated audit of Pride International, Inc.'s 2004 consolidated financial statements and of its internal control over financial reporting as of December 31, 2004 and audits of its 2003 and 2002 consolidated financial statements in accordance with the standards of the Public Company Accounting Oversight Board (United States). Our opinions, based on our audits, are presented below.

Consolidated financial statements In our opinion, the accompanying consolidated balance sheet and the related consolidated statements of operations, stockholders' equity and cash flows present fairly, in all material respects, the financial position of Pride International, Inc. and its subsidiaries (the Company) as of December 31, 2004 and 2003, and the results of their operations and their cash flows for each of the three years in the period ended December 31, 2004 in conformity with accounting principles generally accepted in the United States of America. These financial statements are the responsibility of the Company's management. Our responsibility is to express an opinion on these financial statements based on our audits. We conducted our audits of these statements in accordance with the standards of the Public Company Accounting Oversight Board (United States). Those standards require that we plan and perform the audit to obtain reasonable assurance about whether the financial statements are free of material misstatement. An audit of financial statements includes examining, on a test basis, evidence supporting the amounts and disclosures in the financial statements, assessing the accounting principles used and significant estimates made by management, and evaluating the overall financial statement presentation. We believe that our audits provide a reasonable basis for our opinion. As discussed in Note 2 to the consolidated financial statements, the Company has restated its 2003 and 2002 consolidated financial statements.

Internal control over financial reporting Also, we have audited management's assessment, included in Management's Report on Internal Control Over Financial Reporting appearing under Item 9A, that Pride International, Inc. did not maintain effective internal control over financial reporting as of December 31, 2004 because the Company did not maintain effective controls over the communication among operating, functional and accounting departments of financial and other business information that is important to the period-end financial reporting process, including the specifics of non-routine and non-systematic transactions, based on criteria established in Internal Control -- Integrated Framework issued by the Committee of Sponsoring Organizations of the Treadway Commission (COSO). The Company's management is responsible for maintaining effective internal control over financial reporting and for its assessment of the effectiveness of internal control over financial reporting. Our responsibility is to express opinions on management's assessment and on the effectiveness of the Company's internal control over financial reporting based on our audit. We conducted our audit of internal control over financial reporting in accordance with the

57 Source of Audit Report is from Form 10K Filing in EDGAR

164

standards of the Public Company Accounting Oversight Board (United States). Those standards require that we plan and perform the audit to obtain reasonable assurance about whether effective internal control over financial reporting was maintained in all material respects. An audit of internal control over financial reporting includes obtaining an understanding of internal control over financial reporting, evaluating management's assessment, testing and evaluating the design and operating effectiveness of internal control, and performing such other procedures as we consider necessary in the circumstances. We believe that our audit provides a reasonable basis for our opinions. A company's internal control over financial reporting is a process designed to provide reasonable assurance regarding the reliability of financial reporting and the preparation of financial statements for external purposes in accordance with generally accepted accounting principles. A company's internal control over financial reporting includes those policies and procedures that (i) pertain to the maintenance of records that, in reasonable detail, accurately and fairly reflect the transactions and dispositions of the assets of the company; (ii) provide reasonable assurance that transactions are recorded as necessary to permit preparation of financial statements in accordance with generally accepted accounting principles, and that receipts and expenditures of the company are being made only in accordance with authorizations of management and directors of the company; and (iii) provide reasonable assurance regarding prevention or timely detection of unauthorized acquisition, use, or disposition of the company's assets that could have a material effect on the financial statements. Because of its inherent limitations, internal control over financial reporting may not prevent or detect misstatements. Also, projections of any evaluation of effectiveness to future periods are subject to the risk that controls may become inadequate because of changes in conditions, or that the degree of compliance with the policies or procedures may deteriorate. A material weakness is a control deficiency, or combination of control deficiencies, that results in more than a remote likelihood that a material misstatement of the annual or interim financial statements will not be prevented or detected. The following material weakness has been identified and included in management's assessment. As of December 31, 2004, the Company did not maintain effective controls over the communication among operating, functional and accounting departments of financial and other business information that is important to the period-end financial reporting process, including the specifics of non-routine and non-systematic transactions. Contributing factors included the large number of manual processes utilized during the period-end financial reporting process and an insufficient number of accounting and finance personnel to, in a timely manner: (1) implement extensive structural and procedural system and process initiatives during 2004, (2) perform the necessary manual processes, and (3) analyze non-routine and non-systematic transactions. This control deficiency resulted in errors that required the restatement of the Company's consolidated financial statements for 2003 and 2002, the first three quarterly periods in 2004 and all quarterly periods in 2003, as discussed in Note 2 to the consolidated financial statements. The errors primarily affected property and equipment and the related depreciation expense, debt and the related interest and financing costs, minority interest balances and activity and income tax balance sheet accounts and the related provisions. This deficiency also resulted in audit adjustments to the 2004 consolidated financial statements primarily affecting accrued employee benefits and interest and the corresponding expense accounts. Additionally, this control deficiency could result in a material misstatement to the Company's annual or interim consolidated financial statements that would not be prevented or detected. Accordingly, management has determined that this control deficiency constitutes a material weakness . This material weakness was considered in determining the nature, timing, and extent of audit tests applied in our audit of the 2004 consolidated financial statements, and our opinion regarding the effectiveness of the Company's internal control over financial reporting does not affect our

165

opinion on those consolidated financial statements. In our opinion, management's assessment that Pride International, Inc. did not maintain effective internal control over financial reporting as of December 31, 2004, is fairly stated, in all material respects, based on criteria established in Internal Control -- Integrated Framework issued by the COSO. Also, in our opinion, because of the effect of the material weakness described above on the achievement of the objectives of the control criteria, Pride International, Inc. has not maintained effective internal control over financial reporting as of December 31, 2004, based on criteria established in Internal Control -- Integrated Framework issued by the COSO.

PricewaterhouseCoopers LLP Houston, Texas March 25, 2005

The Audit Analytics Database lists the following contributing factors based on the above opinion. The section of the audit report that discusses the internal control weaknesses is underlined.

Internal Control Weaknesses (4): Accounting documentation, policy and/or procedures Non-routine transaction control issues Restatement or nonreliance of company filings Restatement of previous 404 disclosures

Accounting Rules (GAAP/FASB) Application Failures (7): Accounts/loans receivable, investments & cash issues Consolidation, (Fin46r/Off BS) & foreign currency translation issues Debt, quasi-debt, warrants & equity (BCF) security issues Depreciation, depletion or amortization issues Liabilities, payables, reserves and accrual estimate failures PPE, intangible or fixed asset (value/diminution) issues Tax expense/benefit/deferral/other (FAS 109) issues

166

Table 5-1 Summary of Sample Selection Process

Panel A – Sample Selection Process: Initial ERP announcements from Hayes et al. (2001) 91 Less firms no longer listed or data otherwise not available - 36 Remain ing firms from Hayes et al. (2001) 55 Additional ERP announcements collected from Lexis -Nexis search 92 Total firms implementing ERP systems 147 Matched control firms based on SIC and total assets 147 Total firms available 294 Number of ye ars SOX 404 reporting required (2004, 2005 & 2006) x 3 Maximum potential firm -year observations available 882 Less firm -years when reporting was not required due to phase in -346 Total firm -years with SOX 404 reporting in Audit Analytics database 536 Less firm -years where both members of matched pair not available -164 Net firm -year observations for matched pairs analysis 372

Panel B – Comparison of Total Assets ($ millions) by Year of SOX 404 Report: 2004 2005 2006 Contro l ERP Control ERP Control ERP Firms Firms Firms Firms Firms Firms N (total = 372) 44 44 72 72 70 70 Mean 2,552 3,971 1,912 3,214 2,067 3,792 Median 1,113 1,348 714 1,040 714 1,120 Std. Deviation 5,262 7,308 4,304 6,209 4,407 7,079 Minimum 110 150 36 34 11 17 Maximum 34,369 34,778 34,369 34,778 34,369 34,778 Test for mean differences: t = 1.045 p = .2988 t = 1.463 P = .1457 t = 1.731 p = .0856

Note: Total Assets are as of the beginning of the implementation year.

167

Table 5-2 ERP Implementing Firms by 2 Digit SIC Code and Implementation Year

2 Digit SIC Codes 94 95 96 97 98 99 00 01 02 03 Total 01 -Agricultural Production Crops 1 1 13 -Oil and Gas Extraction 1 1 3 2 1 8 16 -Heavy Construction 1 1 20 -Mfg: Food and Kindred Products 2 1 2 1 1 1 8 23 -Mfg: Apparel 1 1 1 3 24 -Mfg: Lumber & Wood Products 1 1 25 -Mfg: Furniture & Fixtures 1 2 1 4 26 -Mfg: Paper & Allied Products 1 1 1 1 4 27 -Mfg: Printing & Publishing 1 1 1 1 4 28 -Mfg: Chemicals 1 1 2 2 1 4 2 3 2 18 29 -Mfg: Petroleum Refining 1 1 30 -Mfg: Rubber & Misc. Plastic 1 1 33 -Mfg: Primary Metal Industries 1 1 1 3 34 -Mfg: Fabricated Metal Products 1 1 1 3 35 -Mfg: Ind. & Com. Machinery 1 5 8 2 4 1 1 22 36 -Mfg: Electronic & Elect. Equip. 1 2 5 3 3 4 18 37 -Mfg: Transportation Equipment 1 1 1 1 1 5 38 -Mfg: Measuring & Control Instr. 1 2 1 1 1 2 1 1 10 39 -Mfg: Misc. Man ufacturing 1 1 42 -Motor Freight Transportation 1 1 45 -Transportation by Air 2 2 48 -Communications 1 1 2 49 -Electric, Gas & Sanitary Services 2 2 50 -Wholesale: Durable Goods 2 1 3 51 -Wholesale: Non -Durable Goods 1 1 52 -Retail: Bldg Materials, Hardware 1 1 54 -Retail: Food Stores 1 1 59 -Retail: Miscellaneous 1 3 2 6 63 -Insurance Carriers 1 1 2 67 -Holding & Other Investment 1 1 2 73 -Automotive Repa ir & Service 1 2 3 6 80 -Health Services 1 1 87 -Engineering, Actg., R&D, Mgt 1 1 Totals 1 8 8 15 28 34 18 20 9 6 147

168

Table 5-3 Frequency of Internal Control Weaknesses

Primary Sample (N=536) Matched Pair Sample (N=372) Weak = 0 Weak = 1 Total Weak = 0 Weak = 1 Total 2004: ERP=0 41 71% 17 29% 58 30 68% 14 32% 44 ERP=1 73 85% 13 15% 86 39 89% 5 11% 44 Totals 114 79% 30 21% 144 69 78% 19 22% 88 Z-sta tistic 2.06 Z-statistic 2.33 Prob > Z 0.0198 Prob > Z 0.0099 2005: ERP=0 71 88% 10 12% 81 62 86% 10 14% 72 ERP=1 101 89% 12 11% 113 63 88% 9 12% 72 Totals 172 89% 22 11% 194 125 87% 19 13% 144 Z-statistic 0.37 Z-statistic 0.25 Prob > Z 0.3542 Prob > Z 0.4028 2006: ERP=0 73 85% 13 15% 86 59 84% 11 16% 70 ERP=1 104 93% 8 7% 112 66 94% 4 6% 70 Totals 177 89% 21 11% 198 125 89% 15 11% 140 Z-statistic 1.81 Z-statistic 1. 91 Prob > Z 0.0355 Prob > Z 0.0279 Total : ERP=0 185 82% 40 18% 225 151 81% 35 19% 186 ERP=1 278 89% 33 11% 311 168 90% 18 10% 186 Totals 463 86% 73 14% 536 319 86% 53 14% 372 Z-statistic 2.39 Z-statistic 2.52 Prob > Z 0.0085 Prob > Z 0.0058

Weak = 0: Firms with no Internal Control Weakness es in the Audit Analytics database Weak = 1: Firms with at least one Internal Control Weakness in the Audit Analytics database ERP=1: Firms that implemented ERP systems between 1994 and 2003 ERP=0: Control firms that have not reported implementation of an ERP system Prob > Z: One-tail Z-test based on proportion of ERP=0 > ERP=1

169

Table 5-4 Logistic Regression Results

Primary Sample Matched Pair Sample Exp. Odds Odds Parameter Sign Estimate Pr> ΧΧΧ2 Ratio Estimate Pr> ΧΧΧ2 Ratio Intercept ? 1.2283 0. 2311 2.0169 0. 1134 ERP (0 vs. 1) + 0.29 68 0.03 96 1.81 0 0. 2990 0.0 947 1.81 9 LOGSEG - 0.2322 0.3113 1.261 0.3401 0.2053 1.405 FOREIGN (0 vs. 1) - -0.0087 0. 9513 0.9 83 -0.0 535 0. 7522 0.8 98 M&A (0 vs. 1) - -0. 2267 0. 1201 0. 635 -0.3 000 0.0 802 0.5 49 RESTR (0 vs. 1) - -0.04 53 0. 7634 0. 913 0.0 839 0.6 368 1.1 83 SALEGRW - 0.77 73 0.0 829 2.1 76 0.96 26 0.17 64 2.6 18 INV TAT - 0. 8448 0. 5253 2.328 -0. 5947 0. 7023 0. 552 LOGMKTV - -0.4 481 <.0001 0.63 9 -0. 5298 <.0001 0.5 89 LOSS (0 vs. 1) - -0.4 684 0.0 080 0. 392 -0. 2510 0. 2365 0. 936 ZSCORE + -0.0 563 0.0 775 0.9 45 -0.0 658 0.0 669 0.9 36 LOGAGE + 0.1 634 0.3 373 1.17 7 0.00 27 0.98 88 1.00 3 BIG4 (0 vs. 1) + -0.0 12 5 0.9 62 1 0.9 75 0.1 118 0. 7248 1. 251 LOGERPAGE + -0. 2476 0. 4248 0. 781 -0. 0292 0. 9381 0. 971 Likelihood Ratio 64 .8577 <.0001 43 .4915 <.0001 Max -rescaled R2 0.2 076 0. 1974 Observations 536 372

Prob (WEAK it ) = f (β + β1ERP it + β2LOGSEG it + β3FOREIGN it + β4M&A it + β5RESTR it + β6SALEGRW it + β7INVTAT it + β8LOGMKTV it + β9LOSS it + β10 ZSCORE it + β11 LOGAGE it + β13 BIG4 it + β14 LOGERPAGE it ) Where: WEAK = Indicator variable equal to 1 for firms reporting an ICW else 0. ERP = Indicator variable equal to 1 for firms that have implemented ERP systems else 0. LOGSEG = Natural log of number of business segments (Compustat mnemonic SEGNUM) FOREIGN = Indicator variable equal to 1 if the firm has a non-zero foreign currency translation (Compustat mnemonic FCA) else 0. M&A = Indicator variable equal to 1 if acquisitions reported on the Statement of Cash Flows (Compustat mnemonic AQC) else 0. RESTR = Indicator variable equal to 1 if at least one of the following is not equal to zero (Compustat mnemonics RCP, RCA, RCEPS, RCD) else 0. SALEGRW = Percentage change in sales (Compustat mnemonic SALE) INVTAT = Ratio of inventory over total assets (Compustat mnemonic INVT / AT) LOGMKTV = Natural log of market value of equity (Compustat mnemonic MKVAL/MKVALM) LOSS = Indicator variable equal to 1 if earnings before extraordinary items (Compustat mnemonic IB) are less than zero else 0. ZSCORE = Altman’s Z-Score (Compustat mnemonic ZSCORE). LOGAGE = Natural log of years the firm exists in the CRSP database. BIG4 = Indicator variable equal to 1 if the firm’s auditor is one of the Big 4 firms else 0. LOGERPAGE = Natural log of the number of years since the ERP system was implemented.

170

Table 5-5 Frequency (Proportion) of Factors Contributing to Internal Control Weaknesses Primary Sample Matched Pair Panel A – Internal Control Weakness: ERP = 0 ERP = 1 ERP = 0 ERP = 1 *Acco unting documentation, policy & /or procedure 39(.173) 32(.103) 35(.188) 17(.091) Accounting personnel resources, compet. /training 20(.089) 18(.058) 17(.091) 11(.059) Ethical or compliance issues with personnel 4(.018) 5(.016) 2(.011) 2(.011) Ineffective regulatory compliance issues 1(.004) 0(.000) 0(.000) 0(.000) *Information technology , so ftware, security, access 6(.027) 10(.032) 4(.022) 5(.027) *Journal entry control issues 5(.022) 7(.023) 4(.022) 4(.022) Management/Board/Audit Committee investigatio n 1(.004) 0(.000) 1(.005) 0(.000) Material and/or numerous auditor /YE adjustments 18(.080) 20(.064) 16(.086) 12(.065) *Non -routine transaction control issues 6(.027) 6(.019) 6(.032) 3(.016) *Restatement or nonreliance of company filings 22(.098) 15(.048) 20(.108) 7(.038) *Restatement of previous 404 disclosures 13(.058) 4(.013) 12(.065) 1(.005) *Scope (disclaimer of opinion) or other limitations 1(.004) 0(.000) 1(.005) 0(.000) *Segregati ons of duties/design of control/ personnel 5(.022) 3(.010) 4(.02 2) 1.(005) Senior management competency, tone, reliability 2(.009) 2(.006) 2(.011) 0(.000) *Untimely or inadequate account reconciliations 4(.018) 4(.013) 4(.022) 2(.011) Panel B – Accounting Rule (GAAP/FASB) Application Failures: *Accounts/loans rec eivable, investments & cash 6(.027) 11(.035) 3(.016) 6(.032) Acquisition, merger, d isposal or reorganization 6(.027) 4(.013) 6(.032) 2(.011) *Capitalization of expenditures issues 3(.013) 0(.000) 2(.011) 0(.000) Cash flow statement (FAS 95) classifi cation errors 0(.000) 1(.003) 0(.000) 1(.005) *Consolidation, (Fin46r/Off BS)& foreign currency 5(.022) 2(.006) 5(.027) 2(.011) Debt , quasi -debt, warrants & equity ( BCF) security 0(.000) 1(.003) 0(.000) 1(.005) *Deferred, stock -based or executive comp issues 7(.031) 3(.010) 6(.032) 2(.011) *Depreciation, depletion or amortization issues 5(.022) 4(.013) 5(.027) 2(.011) *Expense recording (payroll, SG&A) issues 2(.009) 1(.003) 2(.011) 1(.005) Fin Stmt, footnote, US GAAP , segment disclosure 1(.004) 4( .013) 1(.005) 3(.016) *Financial derivatives/hedging (FAS 133) acc tg 2(.009) 0(.000) 1(.005) 0(.000) Foreign, related party, affiliated and/or subsidiary 6(.027) 7(.023) 6(.032) 4(.022) Income statement classification, margin and EPS 1(.004) 0(.000) 1( .005) 0(.000) *Intercompany/Investment w/ sub/affil issues 3(.013) 0(.000) 3(.016) 0(.000) *Inventory, vendor and cost of sales issues 13(.058) 14(045) 10(.054) 8(.043) Lease, FAS 5, legal, contingency & commit issues 5(.022) 4(.013) 4(.022) 3(.016) *Lease, leasehold & FAS 13 (98) (subcategory) 5(.022) 3(.010) 4(.022) 2(.011) *Liabilities, payables, reserves and accrual estimat e 6(.027) 12(.039) 5(.027) 7(.038) *PPE , intangible or fixed asset (value/diminution) 8(.036) 7(.023) 7(.038) 4(.022) *Re venue recognition issues 11(.049) 17(.055) 9(.048) 9(.048) Tax expense/benefit /deferral/other (FAS 109) 15(.067) 8(.026) 13(.070) 5(.027) Unspecified/unidentified/inapplicable FASB/GAAP 2(.009) 1(.003) 1(.005) 1(.005) Data from Audit Analytics Databas e (See Exhibit 5 -1 for definition s) * = IT related factors ERP=1: ERP Implementing Firms, ERP=0: Control Firms Whole numbers = Frequency (proportion of respective ERP category in parenthesis) Bold Italics Underlined = p<0.01; One-tail Z-test based on proportion of ERP=0 > ERP=1 Bold Italics = p<0.05; One-tail Z-test based on proportion of ERP=0 > ERP=1 Bold Face = p<0.10; One-tail Z-test based on proportion of ERP=0 > ERP=1

171

Table 5-6 Frequencies of Internal Control Weakness IT vs. Non-IT Factors

Primary Sample (N=536) Matched Pair Sample (N=372) Weak = 0 Weak = 1 Total Weak = 0 Weak = 1 Total

Panel A – IT Related Factors (H3b): ERP=0 186 83% 39 17% 225 151 81% 35 19 % 186 ERP=1 278 89% 33 11% 311 168 90% 18 10 % 186 Totals 467 87% 69 13% 536 319 78% 53 22% 372 Z-statistic 2.25 Z-statistic 2.52 Prob > Z 0.0121 Prob > Z 0.0058

Panel B – Non IT Related Factors (H3c): ERP=0 194 86 % 31 14 % 225 159 85 % 27 15 % 186 ERP=1 280 90 % 31 10 % 311 168 90 % 18 10 % 186 Totals 474 88 % 62 12% 536 327 88 % 45 12 % 372 Z-statistic 1. 36 Z-statistic 1.43 Prob > Z 0.0867 Prob > Z 0.0762

Weak = 0: Firms with no Internal Control Weaknesses in the Audit Analytics database Weak = 1: Firms with at least one Internal Control Weakness in the Audit Analytics database ERP=1: Firms that implemented ERP systems between 1994 and 2003 ERP=0: Control firms that have not reported implementation of an ERP system Prob > Z: One-tail Z-test based on proportion of ERP=0 > ERP=1

172

Table 5-7 Logistic Regression Results IT Related Factors

Primary Sample Matched Pair Sample Exp. Odds Odds Parameter Sign Estimate Pr> ΧΧΧ2 Ratio Estimate Pr> ΧΧΧ2 Ratio Intercept ? 1.2376 0.2333 2.0169 0.1134 ERP (0 vs. 1) + 0.2661 0.0672 1.703 0.2990 0.0947 1.819 LOGSEG - 0.2986 0.2001 1.348 0.3401 0.2053 1.405 FOREIGN (0 vs. 1) - -0.0146 0.9189 0.971 -0.0535 0.7522 0.898 M& A (0 vs. 1) - -0.2432 0.0983 0.615 -0.3000 0.0802 0.549 RESTR (0 vs. 1) - -0.0532 0.7255 0.899 0.0839 0.6368 1.183 SALEGRW - 0.8134 0.0809 2.256 0.9626 0.1764 2.618 INVTAT - 0.7357 0.5860 2.087 -0.5947 0.7023 0.552 LOGMKTV - -0.4660 <.0001 0.628 -0.5298 0.0001 0.589 LOSS (0 vs. 1) - -0.4797 0.0070 0.383 -0.2510 0.2365 0.605 ZSCORE + -0.0558 0.0838 0.946 -0.0658 0.0669 0.936 LOGAGE + 0.1381 0.4185 1.148 0.0027 0.9888 1.003 BIG4 (0 vs. 1) + -0.1558 0.5777 0.73 2 0.1118 0.7248 1.251 LOGERPAGE + -0.2489 0.4244 0.780 -0.0292 0.9381 0.971 Likelihood Ratio 66.4841 <.0001 43.4915 <.0001 Max -rescaled R 2 0.2138 0.1974 Observations 536 372

Prob (ITRELF it ) = f (β + β1ERP it + β2LOGSEG it + β3FOREIGN it + β4M&A it + β5RESTR it + β6SALEGRW it + β7INVTAT it + β8LOGMKTV it + β9LOSS it + β10 ZSCORE it + β11 LOGAGE it + β13 BIG4 it + β14 LOGERPAGE it ) Where: ITRELF = Indicator variable equal to 1 for firms reporting an IT related factor else 0. ERP = Indicator variable equal to 1 for firms that have implemented ERP systems else 0. LOGSEG = Natural log of number of business segments (Compustat mnemonic SEGNUM) FOREIGN = Indicator variable equal to 1 if the firm has a non-zero foreign currency translation (Compustat mnemonic FCA) else 0. M&A = Indicator variable equal to 1 if acquisitions reported on the Statement of Cash Flows (Compustat mnemonic AQC) else 0. RESTR = Indicator variable equal to 1 if at least one of the following is not equal to zero (Compustat mnemonics RCP, RCA, RCEPS, RCD) else 0. SALEGRW = Percentage change in sales (Compustat mnemonic SALE) INVTAT = Ratio of inventory over total assets (Compustat mnemonic INVT / AT) LOGMKTV = Natural log of market value of equity (Compustat mnemonic MKVAL/MKVALM) LOSS = Indicator variable equal to 1 if earnings before extraordinary items (Compustat mnemonic IB) are less than zero else 0. ZSCORE = Altman’s Z-Score (Compustat mnemonic ZSCORE). LOGAGE = Natural log of years the firm exists in the CRSP database. BIG4 = Indicator variable equal to 1 if the firm’s auditor is one of the Big 4 firms else 0. LOGERPAGE = Natural log of the number of years since the ERP system was implemented.

173

Table 5-8 Logistic Regression Results Non-IT Related Factors

Primary Sample Matched Pair Sample Exp. Odds Odds Parameter Sign Estimate Pr> ΧΧΧ2 Ratio Estimate Pr> ΧΧΧ2 Ratio Intercept ? 0.9036 0.4045 1.5480 0.2460 ERP (0 vs. 1) + 0.1621 0.2948 1.383 0.1117 0.5536 1.250 LOGSEG - 0.4218 0.0910 1.525 0.5660 0.0542 1.761 FOREIGN (0 vs. 1) - -0.0747 0.6222 0.861 -0.09 50 0.5987 0.827 M&A (0 vs. 1) - -0.2256 0.1500 0.637 -0.3675 0.0490 0.480 RESTR (0 vs. 1) - -0.0012 0.9943 0.998 0.1489 0.4386 1.347 SALEGRW - 0.7451 0.0938 2.107 0.2537 0.7805 1.276 INVTAT - 1.1244 0.4391 3.078 -0.4233 0.8063 0.655 LOGMKTV - -0.4402 0.0003 0.644 -0.4836 0.0008 0.617 LOSS (0 vs. 1) - -0.5447 0.0041 0.336 -0.3994 0.0757 0.450 ZSCORE + -0.0607 0.0797 0.941 -0.0646 0.0947 0.937 LOGAGE + 0.1877 0.2963 1.206 -0.0204 0.9189 1.021 BIG4 (0 vs. 1) + -0.0798 0.7795 0.852 -0.0323 0.9273 0.937 LOGERPAGE + -9.3884 0.2286 0.678 -0.1988 0.6098 0.820 Likelihood Ratio 61.9544 <.0001 40.7160 0.0001 Max -rescaled R 2 0.2134 0.1987 Observations 536 372

Prob (ITNONF it ) = f (β + β1ERP it + β2LOGSEG it + β3FOREIGN it + β4M&A it + β5RESTR it + β6SALEGRW it + β7INVTAT it + β8LOGMKTV it + β9LOSS it + β10 ZSCORE it + β11 LOGAGE it + β13 BIG4 it + β14 LOGERPAGE it ) Where: ITNONF = Indicator variable equal to 1 for firms reporting a Non-IT related factor else 0. ERP = Indicator variable equal to 1 for firms that have implemented ERP systems else 0. LOGSEG = Natural log of number of business segments (Compustat mnemonic SEGNUM) FOREIGN = Indicator variable equal to 1 if the firm has a non-zero foreign currency translation (Compustat mnemonic FCA) else 0. M&A = Indicator variable equal to 1 if acquisitions reported on the Statement of Cash Flows (Compustat mnemonic AQC) else 0. RESTR = Indicator variable equal to 1 if at least one of the following is not equal to zero (Compustat mnemonics RCP, RCA, RCEPS, RCD) else 0. SALEGRW = Percentage change in sales (Compustat mnemonic SALE) INVTAT = Ratio of inventory over total assets (Compustat mnemonic INVT / AT) LOGMKTV = Natural log of market value of equity (Compustat mnemonic MKVAL/MKVALM) LOSS = Indicator variable equal to 1 if earnings before extraordinary items (Compustat mnemonic IB) are less than zero else 0. ZSCORE = Altman’s Z-Score (Compustat mnemonic ZSCORE). LOGAGE = Natural log of years the firm exists in the CRSP database. BIG4 = Indicator variable equal to 1 if the firm’s auditor is one of the Big 4 firms else 0. LOGERPAGE = Natural log of the number of years since the ERP system was implemented.

CHAPTER 6 CONCLUSIONS

6.1 Conclusions

This dissertation examines whether enterprise resource planning (ERP) systems have had an impact on: earnings management, shareholder value, and internal controls. These three potential effects are stated as research questions. The dissertation uses a theoretical framework well established in accounting research, specifically agency theory, to develop several testable hypotheses for each of these research questions, and tests them with well established methods and models from the accounting literature. The results provide a unique contribution to the literature by measuring the impact of these systems through the prism of accounting theory. Each research question is addressed in a separate essay, with key results summarized in the following paragraphs.

The first essay uses three different proxies to measure the impact of ERP systems on earnings management: discretionary accruals, earnings quality, and the distribution of earnings and earnings changes just above and below zero. Data for a sample of firms that implemented

ERP systems between 1994 and 2003 are compared to data for a control group of firms matched on size and industry. The overall results are mixed and provide some support for improved earnings management levels for ERP implementers relative to the control group of firms based on short-term discretionary accruals, but do not for long-term discretionary accruals. Initial univariate data provides evidence that ERP implementers have lower levels of discretionary short-term and total accruals than the control firms, and that those levels decrease following implementation. However, multivariate regressions that include control variables for size, profitability, implementation timing, and manufacturing vs. non-manufacturing industries show

174 175

that the significance of these differences may be due more to the control variables than to the

ERP implementation. Similar mixed results are found for earnings quality, where the proxy is significantly better for ERP implementers than for control firms, and it improves following implementation of the system. However, the amount of improvement is not significantly different from improved earnings quality of the control firms over the same time-frame. The last proxy also provides mixed results, with some evidence that both the ERP implementing firms and the control firms have smoother earnings (changes) around zero after the implementation event than they do before, indicating less earnings management activity, but the differences are not significant. Overall, this essay provides only limited support for the argument that implementation of ERP systems will have a positive impact on earnings management as measured by discretionary accruals, earnings quality, and the distribution of earnings (and earnings changes) near zero.

The second essay uses three different metrics to measure the impact of ERP systems on shareholder value: long-horizon buy-and-hold returns, market-to-book ratios, and value-to-price ratios. The results provide some evidence that shareholder returns for one, three, and five year horizons are significantly higher for firms that implement ERP systems than for the matched control firms following implementation. However, when these results are compared to similar results for the one, three, and five year periods prior to implementations in a multivariate analysis, there is evidence that these results are not being influenced by implementation of the ERP system.

In fact there is some support for self selection bias indicating that firms that implement ERP systems are firms that were already more profitable than their matching control firms. The results also show that market-to-book ratios for ERP implementers are significantly higher than ratios for non-implementers both before and after the implementation event. Since higher ratios are considered to be an indication that the market anticipates higher future earnings relative to current

176

book value, these results suggest that the market may have signaled the higher future returns prior to implementation and that subsequent monitoring of information has continued that belief. The results also show that the value-to-price ratio was significantly higher for ERP implementers than for non-implementers before and during implementation, but not after. Overall, this essay does not provide evidence that implementation of ERP systems has a positive impact on shareholder value as measured by long-horizon returns, market-to-book ratios, and value-to-price ratios.

The third essay uses Sarbanes-Oxley Section 404 compliance data to measure the impact of ERP systems on internal control. The essay provides evidence that during the first three years of reporting required under SOX a smaller proportion of ERP implementing firms have reported internal control weaknesses than the non-implementing control firms. The findings are further supported by a logistic regression analysis that includes several control variables found in prior research to contribute to internal control weaknesses. The results suggest that firms with ERP systems are less likely to report material control weaknesses than the control firms. A further analysis of factors that contribute to material weaknesses shows that IT-related factors are less likely to occur in reports for ERP implementing firms than for the control firms. This suggests that the built in control features promoted by ERP vendors may have an impact on compliance.

By contrast, the factors that are not IT-related are no more likely to occur in reports for ERP implementers than for non-implementers. Overall, this essay provides strong evidence that implementation of ERP systems has a positive impact on internal controls as measured by SOX

Section 404 compliance.

6.2 Contributions

Taken as a whole, these results suggest that ERP systems do not have a significant impact on earnings management but that they do have an impact on shareholder value and internal control. These results will contribute to the accounting and information systems academic

177

research literature and should also be of interest to the practitioner communities of financial and managerial accounting, auditing, and management information systems.

The contribution to the academic community is important because it combines two streams of research, accounting and information systems, in a unique way. It does so by tapping into the rich tradition of accounting research theory that has developed over the past forty years and by using the methods and models from that stream of research to measure the impact of information systems. The results show that these theories, methods and models have relevance beyond the confines of stand-alone accounting research. The results also show that from the perspective of the information systems academic community, the use of theoretically based accounting approaches to measuring the impact of information system investments can extend the literature in new ways.

From the practitioner community perspective, this dissertation provides evidence both in support of and against some of the marketing hype that ERP vendors use to sell their product.

The results do not support the arguments that information transparency is assumed to be good for corporate governance and that increased transparency should help to reduce earnings management. Nor do the results support the arguments that the improved timeliness and quality of information will contribute to improved productivity, profitability, and ultimately shareholder value. On the other hand, the results support recent promotional literature arguing that the built- in control features of ERP systems using best practices will help firms comply with Sarbanes-

Oxley. Although it is important to point out that the nature of this type of research does not lend itself to support causation. It can only be used to conclude that there is or is not an association between implementation of ERP systems and the three research questions.

178

6.3 Limitations

Like most research, this dissertation has several limitations that should be taken into consideration before applying the results. First, the results are based on a limited sample of firms that have publicly announced their ERP implementation plans, and the control group is based on a sample of firms that have not made such public announcements. This process could bias the results in that firms that did not publicly announce their ERP implementation plans may have been excluded from the first group and incorrectly included in the control group. The sample is also limited to those firms that are required to publicly disclose financial and other information, which may bias the results in favor of larger firms. Another limitation is survivor bias, which is evident by the fact that 36 of the 91 firms in the initial sample from Hayes et al. (2001) are no longer listed on public exchanges. Some of these firms have been successful and been bought out by larger firms, while some have gone bankrupt and gone out of business. The impact on these firms will never be known because information subsequent to the implementation event is not available. The sample for internal control analysis is further biased by the fact that SOX has a two stage implementation requirement that limited initial compliance to early adapters, which are generally larger firms. There was also a delay in mandated reporting for certain foreign firms that by necessity reduces the sample to larger domestic firms. Another limitation is the use of proxies to measure factors that cannot be measured directly, such as earnings management and earnings quality. This is a common limitation in the accounting literature, which is mitigated by the use of proxies that have had general acceptance in prior research.

6.4 Future Research

Future research could address some of the limitation listed above, that are beyond the scope of this dissertation. Examples include the use of survey methodologies to better divide the

179

sample between implementers and non-implementers. This approach could also better define the non-implementers by gathering data as to the degree to which they have developed internal systems that mirror the features of ERP systems, thus making them “de facto” implementers.

This methodology could also gather data from smaller non-public companies and even larger companies that are privately held and not required to file public financials. The survey method could also be used to gather additional information about how the ERP system is being used in the organization. For instance, are all of the modules being used or are only some of them, which would defeat some of the benefits of a fully integrated system. It could also find out how widely diffused the system is throughout the organization, which could also impact the level of benefits that would be expected. A case study approach could be used to follow up on firms that implemented ERP systems and are no longer listed on public exchanges. This approach may provide additional information about the impact of the system, and whether or not the delisting is due to favorable results, such as profitability that led to a buy-out, or unfavorable results that led to bankruptcy. Future research could also follow up on the second phase of SOX Section 404 implementation after more of the smaller firms are required to comply. Related research could also examine the relationship between ERP systems and: audit fees, audit opinions, and restatements of financial reports. Future research may address the issue of implementation timing. There is a wide range of estimates as to how long it takes to completely implement one of these systems. The assumption used in this dissertation uses an “initial implementation” concept which is assumed to take only one year. An alternative approach may be to drop the data for the implementation year and one year before and after to insure that data used for the before and after type analysis are not compromised with activity during the implementation phase.

BIBLIOGRAPHY

Abarbanell, J., and V. L. Bernard. 1992. Tests of Analysts' Overreaction/Underreaction to Earnings Information as an Explanation for Anomalous Stock Price Behavior. The Journal of Finance 47 (July):1181-1207.

Abarbanell, J., and R. Lehavy. 2003. Can Stock Recommendations Predict Earnings Management and Analysts' Earnings Forecast Errors? Journal of Accounting Research 41 (1):1-31.

Abernethy, M. A., J. Bouwens, and L. Van Lent. 2004. Determinants of Control System Design in Divisionalized Firms. The Accounting Review 79 (3):545-570.

Aboody, D., M. E. Barth, and R. Kasznik. 1999. Revaluations of Fixed Assets and Future Firm Performance: Evidence from the UK. Journal of Accounting & Economics 26 (January):149-178.

———. 2004. Firms' Voluntary Recognition of Stock-Based Compensation Expense. Journal of Accounting Research 42 (2):123-150.

Aboody, D., J. Hughes, and J. Liu. 2005. Earnings Quality, Insider Trading, and Cost of Capital. Journal of Accounting Research 43 (5):651-673.

Al-Mashari, M., A. Al-Mudimigh, and M. Zairi. 2003. Enterprise Resource Planning: A Taxonomy of Critical Factors. European Journal of Operations Research 146 (2):352- 364.

Ali, A., L. S. Hwang, and M. A. Trombley. 2003. Residual-Income-Based Valuation Predicts Future Stock Returns: Evidence on Mispricing vs. Risk Explanation. The Accounting Review 78 (2):377-396.

Altamuro, J. L. M., and A. L. Beatty. 2006. Do Internal Control Reforms Improve Earnings Quality? In Working Paper Series : SSRN.

Ang, J. S., and S. Zhang. 2004. An Evaluation of Testing Procedures for Long Horizon Event Studies. Review of Quantitative Finance and Accounting 23:251-274.

Antle, R. 1989. Intellectual Boundaries in Accounting Research. Accounting Horizons 3 (2):103- 109.

Ashbaugh-Skaife, H., D. W. Collins, W. R. Kinney, and R. LaFond. 2008. The Effect of SOX Internal Control Deficiencies and Their Remediation on Accrual Quality. The Accounting Review 83 (1):217-250.

Ball, R., and P. Brown. 1968. An Empirical Evaluation of Accounting Income Numbers. Journal of Accounting Research 6 (2):159-178.

180 181

Ball, R., and L. Shivakumar. 2005. Earnings Quality in UK Private Firms: Comparative Loss Recognition Timelines. Journal of Accounting & Economics 39:83-128.

———. 2008. Earnings Quality at Initial Public Offerings. Journal of Accounting & Economics 45:324-349.

Banker, R. D., S. Y. Lee, and G. Potter. 1996. A Field Study of the Impact of a Performance- Based Incentive Plan. Journal of Accounting & Economics 21:195-226.

Banker, R. D., G. Potter, and S. Dhinu. 2000. An Empirical Investigation of an Incentive Plan that Includes Nonfinancial Performance Measures. The Accounting Review 75 (1):65-92.

Barber, B. M., and J. D. Lyon. 1997. Detecting long-run abnormal stock returns; The empirical power and specification of test statistics. Journal of Financial Economics 43:341-372.

Barker, T., and M. Frolick. 2003. ERP Implementation Failure: A Case Study. Information Systems Management 20 (4):43-49.

Barniv, R., and M. Myring. 2006. An International Analysis of Historical and Forecast Earnings in Accounting-Based Valuation Models. Journal of Business Finance & Accounting 33 (7&8):1087-1109.

Barth, M. E., W. H. Beaver, J. R. M. Hand, and W. R. Landsman. 1999. Accruals, Cash Flows, and Equity Values. Review of Accounting Studies 4 (December):205-229.

Barth, M. E., W. H. Beaver, and W. R. Landsman. 1996. Value-Relevance of Banks' Fair Value Disclosures Under SFAS 107. The Accounting Review 71 (October):513-537.

———. 1998. Relative Valuation Roles of Equity Book Value and Net Income as a Function of Financial Health. Journal of Accounting & Economics 25 (February):1-34.

Barth, M. E., and G. Clinch. 1998. Revalued Financial, Tangible, and Intangivle Assets: Associaitons with Share Prices and Non Market-Based Value Estimates. Journal of Accounting Research 36 (Supplement):199-233.

Bartov, E., D. Givoly, and C. Hayn. 2002. The Rewards to Meeting or Beating Eanrings Expectations. Journal of Accounting & Economics 33:173-204.

Basu, A. K., R. Lal, V. Spinivasan, and R. Staelin. 1985. Salesforce Compensation Plans: An Agency theoretic Perspective. Marketing Science 4:267-291.

Beatty, R., and R. Verrecchia. 1989. The Effect of a Mandated Accounting Change on the Capitalization Process. Contemporary Accounting Research 5 (2):472-493.

Beaver, W. H. 1968. The Information Content of Earnings. Journal of Accounting Research 6 (Supplement):67-92.

———. 2002. Perspectives on Recent Capital Market Research. The Accounting Review 77 (2):453-474.

182

Becker, C. L., M. L. DeFond, J. Jiambalvo, and K. R. Subramanyam. 1998. The Effect of Audit Quality on Earnings Management. Contemporary Accounting Research 15 (1):1-24.

Bédard, J. 2006. Sarbanes Oxley Internal Control Requirements and Earnings Quality : SSRN.

Beneish, M. D. 1997. Detecting GAAP Violation: Implications for Assessing Earnings Management Among Firms with Extreme Financial Performance. Journal of Accounting and Public Policy 16:271-309.

Bernard, V. L. 1995. The Feltham-Ohlson Framework: Implications for Empiricists. Contemporary Accounting Research 11 (2):733-747.

Bernard, V. L., and J. Thomas. 1989. Post-Earnings Announcement Drift: Delayed Price Response or Risk Premium? Journal of Accounting Research 27 (Supplement):1-48.

———. 1990. Evidence that Stock Prices do not Fully Reflect the Implications of Current Earnings for Future Earnings. Journal of Accounting & Economics 13 (December):305- 340.

Bradford, M., and J. Florin. 2003. Examining the Role of Innovation Diffusion Factors on the Implementation Success of Enterprise Resource Planning Systems. International Journal of Accounting Information Systems 4:205-225.

Bradshaw, M. T. 2004. How do analysts use their earnings forecasts in generating stock recommendations? The Accounting Review 79 (1):25-50.

Bronson, S. N., J. V. Carcello, and K. Raghunandan. 2006. Firm Characteristics and Voluntary Management Reprots on Internal Control. Auditing: A Journal of Practice & Theory 25 (2):25-39.

Brown, S., K. Lo, and T. Lys. 1999. Use of R 2 in Accounting Research: Measuring Changes in Value Relevance over the Last Four Decades. Journal of Accounting & Economics 28:83-115.

Brown, S., and J. Warner. 1985. Using Daily Stock Returns: The Case of Event Studies. Journal of Financial Economics 14:3-31.

Brunelli, M. Smaller Vendors Challenge SAP, Oracle in ERP Market 2006 [cited. Available from http://searchoracle.techtarget.com/originalContent/0,289142,sid41_gci1229162,00.html .

Bryan, S. H., and S. B. Lilien. 2005. Characteristics of Firms with Material Weaknesses in Internal Control: An Assessment of Section 404 of Sarbanes Oxley. In Working Paper Series : SSRN.

Brynjolfsson, E. 1993. The Productivity Paradox of Information Technology. Communications of the ACM 36 (12):67-77.

183

Brynjolfsson, E., and L. M. Hitt. 1996. Paradox Lost? Firm-level Evidence on the Return so Information Systems Spending. Management Science 42 (4):541-558.

Burgstahler, D. C., and I. Dichev. 1997. Earnings, Adaptations and Equity Values. The Accounting Review 72 (April):187-215.

Canada, J., J. R. J. Kuhn, and S. G. Sutton. 2007. The Pervasive Nature of IT Congtrols: An Examination of Material Weaknesses in IT Controls and Audit Fees. Paper read at AAA- IS Section Mid-Year Conference, at Savannah, Georgia.

Chan, K. C., B. R. Farrell, and P. Lee. 2005. Earnings Management and Return-Earnings Association of Firms Reporting Material Internal Control Weaknesses Under Section 404 of the Sarbanes-Oxley Act. In SSRN Working Paper Series .

Chand, D., G. Hachey, J. Hunton, V. Owhoso, and S. Vasudevan. 2005. A Balanced Scorecard Based Framework for Assessing the Strategic Impacts of ERP Systems. Computers in Industry 56:558-572.

Cheffers, M., and D. Whalen. 2007. 404 Dashboard: Year 3 Update: Audit Analytics.

Cheng, C. S. A., J. L. Y. Ho, and F. Tian. 2006. An Empirical Analysis of Value-Relevance of Disclosure of Material Weaknesses Under Section 404. In Working Paper Series : SSRN.

Chenhall, R. H., and K. Langfield-Smith. 2003. Performance Measurement and Reward Systems, Trust, and Strategic Change. Journal of Management Accounting Research 15:117-143.

Christensen, P. O., J. S. Demski, and H. Frimore. 2002. Accounting Policies in Agencies with Moral Hazard and Renegotiation. Journal of Accounting Research 40 (4):1071-1090.

Collins, D. W., E. L. Maydew, and I. S. Wiess. 1997. Changes in the Value-Relevance of Earnings and Equity Book Values over the Past Forty Years. Journal of Accounting & Economics 24 (December):39-67.

Collins, D. W., M. Pincus, and H. Xie. 1999. Equity Valuation and Negative Earnings: The Role of Negative Earnings. The Accounting Review 74 (January):29-61.

Conyon, M. J., and L. He. 2004. Compensation Committees and CEO Compensation Incentives in U.S. Entrepreneurial Firms. Journal of Management Accounting Research 16:35-56.

Core, J. E., W. R. Guay, and A. Van Buskirk. 2003. Market valuations in the New Economy: an investigation of what has changed. Journal of Accounting & Economics 34:43-67.

COSO. Organization Background Information 1985 [cited. Available from http://www.coso.org/ .

———. 2007. Internal Control - Integrated Framework: Executive Summary . The Committee of Sponsoring Organizations of the Treadway Commission 1992 [cited February 14, 2007 2007]. Available from http://www.coso.org/publications/executive_summary_integrated_framework.htm .

184

Cross, M. B., C. L. Mo, E. L. Robnison, P. J. Rondeau, K. J. Salhus, T. W. White, and J. Wolfman. SEC and PCAOB Propose SOX 303 Relief Measures, Extend Compliance Dates . WilmerHale 2006 [cited February 13, 2007. Available from www.wilmerhale.com/publications/whPubsDetail.aspx?publication=3533.

Davenport, T. H. 1998. Putting the Enterprise into the Enterprise System. Harvard Business Review .

Davidson III, W. N., P. Jiraporn, Y. S. Kim, and C. Nemec. 2004. Earnings Management Following Duality-Creating Succession: Ethnostatistics, Impression Managment, and Agency Theory. Academy of Management Journal 47 (2):267-275.

DeAngelo, L. 1986. Accounting Numbers as market Valuation Substitutes: A Study of Management Buyouts of Public Stockholders. The Accounting Review LXI (3):400-420.

———. 1988. Managerial Competition, Information Costs, and Corporate Governance. Journal of Accounting & Economics 10 (1):3-36.

Debreceny, R. S., G. L. Gray, J. J.-J. Ng, K. S.-P. Lee, and W.-F. Yau. 2005. Embedded Audit Modules in Enterprise Resource Planning Systesm: Implementation and Functionality. Journal of Information Systems 19 (2):7-27.

Dechow, P., and I. Dichev. 2002. The Quality of Accruals and Earnings: The Role of Accural Estimation Errors. The Accounting Review 77 (Supplement):35-59.

Dechow, P., A. P. Hutton, and R. Sloan. 1999. An Empirical Assessment of the Residual Income Valuation Model. Journal of Accounting & Economics 43 (January):1-34.

Dechow, P., and R. Sloan. 1997. Returns to Contrarian Investment Strategies: Tests of Naive Expectations Hypothesis. Journal of Financial Economics 43 (January):3-27.

Dechow, P., R. Sloan, and A. Sweeney. 1995. Detecting Earnings Management. The Accounting Review 70:193-225.

———. 1996. Causes and Consequences of Earnings Manipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC. Contemporary Accounting Research 13 (1):1-36.

DeFond, M. L., and C. W. Park. 1997. Smoothing Income in Anticipation of Future Earnings. Journal of Accounting & Economics 23:115-139.

———. 2001. The Reversal of Abnormal Accruals and the Market Valuation of Earnings Surprises. The Accounting Review 76 (July):375-404.

DeFond, M. L., and K. R. Subramanyam. 1998. Auditor Changes and Discretionary Accruals. Journal of Accounting & Economics 25 (1):35-67.

Demski, J. S., and G. A. Feltham. 1978. Economic Incentives in Budgetary Control Systems. The Accounting Review LIII (2):336-359.

185

Demski, J. S., J. Patell, and M. Wolfson. 1984. Decentralized Choice of Monitoring Systems. Accounting Review 59:16-34.

Dos Santos, B. L., K. Peffers, and D. C. Mauer. 1993. The Impact of Information Technology Investment Announcements on the Market Value of the Firm. Information Systems Research 4 (1):1-23.

Doyle, J. T., W. Ge, and S. McVay. 2007. Determinants of Weaknesses in Internal Control Over Financial Reporting. Journal of Accounting & Economics 44 (1-2):193-223.

Duplaga, E., and M. Astani. 2003. Implementing ERP in Manufacturing. Information Systems Management 20 (3):68-75.

Duru, A., R. J. Iyenhar, and A. Thevaranjan. 2002. The Shielding of CEO Compensation from the Effects of Strategic Expenditures. Contemporary Accounting Research 19 (2):175-193.

Dye, R. 1988. Earnings Management in an Overlapping Generations Model. Journal of Accounting Research 26 (2):195-235.

Eakins, S. G. 2002. Finance, Investment, Institutions, and Management . Edited by D. Clinton. 2nd ed. New York: Addison Wesley.

Easton, P. 2004. PE Ratios, PEG Ratios, and Estimating the Implied Expected Rate of REturn on Equity Capital. The Accounting Review 79 (1):73-95.

Easton, P., T. Harris, and J. A. Ohlson. 1992. Aggregate Accounting Earnings Can Explain Most of Security Returns: The Case of Long Return Intervals. Journal of Accounting & Economics 15 (2/3):119-142.

Ecker, F., J. Francisa, I. Kim, P. Olsson, and K. Schipper. 2006. A Returns-Based Representation of Earnings Quality. The Accounting Review 81 (4):749-780.

Eisenhardt, K. M. 1989. Agency Theory: An Assessment and Review. Academy of Management Review 14 (1):57-74.

Ekanayake, S. 2004. Agency Theory, National Culture and Management Control Systems. Journal of American Academy of Business 4 (1):49-54.

El-Gazzar, S. M., and J. M. Fornaro. 2004. The Influences of Monitoring Mchanisms and Firm Specific Factors on Management's Voluntary Disclosure of Responsibility for Financil Information. Journal of Accounting and Finance Research 2004 (Winter):22-34.

Eldridge, S. W., and B. T. Kealey. 2005. SOX Costs: Auditor Attestation under Section 404. In Working Paper Series : SSRN.

Eschinger, C. 2006. Market Share: ERP Software, Worldwide, 2005 (Executive Summary). In ID Number: G00141080 , edited by Gartner: Gartner, Inc.

186

Ettredge, M. L., C. Li, and S. W. Scholz. 2006. Auditor Realignments Accompanying Implementation of SOX 404 Reporting Requirements. In Working Paper Series : SSRN.

———. 2007. Audit Fees and Auditor Dismissals in the Sarbanes-Oxley Era. Accounting Horizons 21 (4):371-386.

Ettredge, M. L., C. Li, and L. Sun. 2007. CFO Turnover Following Adverse SOX 404 Opinions. In Working Paper Series : SSRN.

Ettredge, M. L., L. Sun, and C. Li. 2006. The Impact of SOX Section 404 Internal Control Quality Assessment on Audit Delay in the SOX Era. Auditing: A Journal of Practice & Theory 25 (2):1-23.

Fama, E. F. 1980. Agency Problems and the Theory of the Firm. Journal of Political Economy 88:288-307.

———. 1991. Efficient Capital Markets II. The Journal of Finance 46 (December):1575-1617.

———. 1998. Market Efficiency, Long-Term Returns, and Behavioral Finance. Journal of Financial Economics 49:283-306.

Fama, E. F., and K. French. 1993. Common risk factors in returns on stock and bonds. Journal of Financial Economics 33:3-56.

———. 1995. Size and Book-to-Market Factors in Earnings and Returns. The Journal of Finance L (1):131-155.

Fama, E. F., and K. R. French. 1992. The Cross-Section of Expected Stock Returns. The Journal of Finance 47 (July):427-465.

———. 1997. Industry Costs of Equity. Journal of Financial Economics 43 (2):153-193.

Feltham, G. A., and J. A. Ohlson. 1995. Valuation and Clean Surplus Accounting for Operating and Financial Activities. Contemporary Accounting Research 11 (2):689-735.

Feltham, G. A., and J. Xie. 1994. Performance Measure Congruity and Diversity in Multi-Task Principal/Agent Relations. The Accounting Review 69 (3):429-453.

Fessler, N. J. 2003. Experimental Evidence on the Links Among Monetary Incentives, Task Attractiveness, and Task Performance. Journal of Management Accounting Research 15:161-176.

Finger, C. A., and W. R. Landsman. 2003. What do Analysts' Stock Recommendations Really Mean? Review of Accounting and Finance 2 (2):67-88.

Fogarty, T., M. L. Magnan, G. Markarian, and S. Bohdjalian. 2009. Inside Agency: The Rise and Fall of Nortel. Journal of Business Ethics 84 (2):165-187.

187

Francis, J., R. LaFond, P. Olsson, and K. Schipper. 2004. Costs of Equity and Earnings Attributes. The Accounting Review 79 (4):967-1010.

———. 2005. The Market Pricing of Accruals Quality. Journal of Accounting & Economics 39 (2):295-327.

Francis, J., and K. Schipper. 1999. Have Financial Statemetns Lost Their Relevance? Journal of Accounting & Economics 39:295-327.

Frankel, R. M., and C. M. C. Lee. 1998. Accounting Valuation, Market Expectation, and the Cross-Sectionsl Stock Returns. Journal of Accounting & Economics 25 (June):283-319.

Freeman, R., and S. Tse. 1989. The Multiperiod Information Content of Accounting Earnings: Confirmations and Contradictions of Previous Earnings Reports. Journal of Accounting Research 27 (Supplement):49-79.

Gattiker, T. F., and D. L. Goodhue. 2005. What Happens After ERP Implementation: Understanding the Impact of Inter-Dependence and Differentiation on Plant-Level Outcomes. MIS Quarterly 29 (3):559-585.

Ge, W., and S. McVay. 2005. The Disclosure of Material Weaknesses in Internal Control after the Sarbanes-Oxley Act. Accounting Horizons 19 (3):137-158.

Gebhardt, W. R., C. M. C. Lee, and B. Swaminathan. 2000. Toward an Implied Cost of Capital. Journal of Accounting Research 39 (Spring):135-176.

Geiger, M. A., D. S. North, and B. T. O'Connell. 2005. The Auditor-to-Client Revolving Door and Earnings Management. Journal of Accounting, Auditing and Finance 20 (1):1-26.

Graber, D. E., and B. D. Jarnagin. 1979. The FASB - Eliminator of "Managed Earnings"? Financial Analysts Journal March-April:72-78.

Grabski, S., and S. A. Leech. 2007. Complementary Controls and ERP Implementation Success. International Journal of Accounting Information Systems 8 (1):17-39.

Grant, G. H., K. C. Miller, and F. Alali. 2008. The Effect of IT Controls on Financial Reporting. Managerial Auditing Journal 23 (8):803-823.

Gupta, P., and T. Leech. 2006. Making Sarbanes-Oxley 404 Work: Reducing Cost, Increasing Effectiveness. International Journal of Disclosure and Governance 3 (1):27-48.

Gupta, P., and N. Nayar. 2005. Market Reaction to Control Deficiency Disclosures under the Sarbanes-Oxley Act: The Early Evidence. In Lehigh University Working Paper .

Gupta, P., and J. C. Thomson. 2006. Use of COSO 1992 in Management Reporting on Internal Control. Strategic Finance 2006 (September):27-33.

Hayes, D. C., J. E. Hunton, and J. L. Reck. 2001. Market Reaction to ERP Implementation Announcements. Journal of Information Systems 15 (1):3-18.

188

Healy, P. M. 1985. The Effect of Bonus Schemes on Accounting Decisions. Journal of Accounting & Economics 7:85-107.

Healy, P. M., and J. M. Wahlen. 1999. A Review of the Earnings Management Literature and its Implications for Standard Setting. Accounting Horizons 12 (4):365-383.

Hicks, J. 1939. Value and Capital . Oxford: Oxford University Press.

Hitt, L. M., and E. Brynjolfsson. 1996. Productivity, Business Profitability, and Consumer Surplus: three Different Measures of Information Technology Value. MIS Quarterly :121- 142.

Hitt, L. M., D. J. Wu, and X. Zhou. 2002. Investment in Enterprise Resource Planning: Business Impact and Productivity Measures. Journal of Management Information Systems 19 (1):71-98.

Ho, L.-C. J., C.-S. Liu, and J. Tsay. 2008. Further Evidence on Financial Analysts' Reaction to Enterprise Resource Planning Implementation Announcements. Review of Accounting and Finance 7 (3):213-235.

Hoitash, R., U. Hoitash, and J. Bedard. 2008. Internal Control Quality and Audit Pricing under the Sarbanes-Oxley Act. Auditing: A Journal of Practice & Theory 27 (1):105-127.

Hoitash, R., U. Hoitash, and J. C. Bedard. 2007. Internal Control Quality and Audit Pricing Under the Sarbanes-Oxley Act. In Working Paper Series : SSRN.

Hollander, A. S., E. L. Denna, and J. O. Cherrington. 2000. Accounting, Information Technology, and Business Solutions . 2nd ed. Boston, MA: Irwin McGraw-Hill.

Hunton, J. E., R. Libby, and C. L. Mazza. 2006. Financial Reporting Transparency and Earnings Management. The Accounting Review 81 (1):135-157.

Hunton, J. E., B. Lippincott, and J. L. Reck. 2003. Enterprise Resource Planning: Comparing Firm Performance of Adopters and Nonadopters. International Journal of Accounting Information Systems 4 (3):165-184.

Hunton, J. E., R. A. McEwen, and B. Wier. 2002. The Reaction of Financial Analysts to Enterprise Resource Planning (ERP) Implementation Plans. Journal of Information Systems 16 (1):31-40.

Hunton, J. E., A. M. Wright, and S. Wright. 2004. Are Financial Auditors Overconfident in Their Ability to Assess Risks Associated with Enterprise Resource Planning Systems? Journal of Information Systems 18 (2):7-28.

Ittner, C. D., and D. F. Larcker. 1998. Innovations in Performance Measurement: Trends and Research Implications. Journal of Management Accounting Research 10:205-239.

Jones, J. 1991. Earnings Management During Import Relief Investigations. Journal of Accounting Research 29:193-228.

189

Kaplan, R. S., and D. P. Norton. 1992. The Balanced Scorecard--Measures That Drive Performance. Harvard Business Review 70 (1):71-80.

Kim, Y., and M. S. Park. 2005. Pricing of Seasoned Equity Offers and Earnings Management. Journal of Financial and Quantitive Analysis 40 (2):435-463.

Kohli, R., and W. J. Kettinger. 2004. Informating the Clan: Controlling Physicians' Costs and Outcomes. MIS Quarterly 28 (3):363-394.

Kothari, S. P., A. J. Leone, and C. E. Wasley. 2005. Performance Matched Discretionary Accruals Measures. Journal of Accounting & Economics 39:163-197.

Kothari, S. P., and J. B. Warner. 1997. Measuring long-horizon security price performance. Journal of Financial Economics 43:301-339.

Kothari, S. P., and J. Zimmerman. 1995. Price and Return Models. Journal of Accounting & Economics 20:155-192.

Krishnan, G., and G. Visvanathan. 2005. Reporting Internal Control Deficiencies in the Post- Sarbanes-Oxley Era: The Role of Auditors and Corporate Governance. In Working Paper Series : SSRN.

Krishnan, G. V. 2003. Audit Quality and the Pricing of Discretionary Accruals. Auditing: A Journal of Practice & Theory 22 (1):109-126.

Krishnan, G. V., and G. Visvanathan. 2007. Reporting Internal Control Deficienceis in the Post- Sarbanes-Oxley Era: The Role of Auditors and Corporate Governance. International Journal of Auiditing 11 (2):73.

Krishnan, J. 2005. Audit Committee Quality an Internal Control: An Empirical Analysis. The Accounting Review 80 (2):749-675.

Krishnan, J., D. V. Rama, and S. Zhang. 2008. Costs to Comply with SOX Section 404. Auditing: A Journal of Practice & Theory 27 (1):169-187.

Kumar, V., B. Maheshwari, and U. Kumar. 2003. An Investigation of Critical Management Issues in ERP Implementation: Empirical Evidence from Canadian Organizations. Technovation 23 (10):793-807.

Lambert, R. A. 1984. Income Smoothing as Rational Equilibrium Behavior. Accounting Review 59:604-618.

———. 2001. Contracting Theory and Accounting. Journal of Accounting & Economics 32:3-87.

Leone, A. J., and R. L. Van Horn. 2005. How Do Nonprofit Hospitals Manage Earnings? Journal of Health Economics 24 (4):815-837.

190

Lev, B. 1989. On the Usefulness of Earnings and Earnings Research: Lessons and Directions from Two Decades of Empirical Research. Journal of Accounting & Economics 27:153- 192.

Li, C., J.-H. Lim, and Q. Wang. 2007. The Effects of IT Governance on IT Control Material Weakness. Paper read at AAA-IS Section Mid-Year Conference, at Savannah, Georgia.

Li, C., and Q. Wang. 2006. SOX 4O4 Assessments and Financial Reporting Errors. In Working Paper Series : SSRN.

Li, H., M. Pincus, and S. O. Rego. 2008. Market Reaction to Events Surrounding the Sarbanes- Oxley Act of 2002. Journal of Law and Economics February 2008 (Forthcoming).

Liberty, S., and J. Zimmerman. 1986. Labor Union Contract Negotiations and Accounting Choices. The Accounting Review LXI (4):692-712.

Little, A., and P. J. Best. 2003. A Framework for Separation of Duties in an SAP R/3 Environment. Managerial Auditing Journal 18 (5):419-430.

Lo, K. 2008. Earnings Management and Earnings Quality. Journal of Accounting & Economics 45:350-357.

Mabert, V., A. Soni, and M. Venkataramanan. 2003. The Impact of Organization Size on Enterprise Resource Planning (ERP) Implementations in the US Manufacturing Sector. International Journal of Management Science 31 (3):235-246.

Mahaney, R. C., and A. L. Lederer. 2003. Information Systems Project Management: An Agency Theory Interpretation. The Journal of Systems and Software 68:1-9.

Markus, M., S. Axline, D. Petrie, and S. Tanis. 2000. Learning from Adopter's Experience with ERP: Problems Encountered and Success Achieved. Journal of Information Technology 15 (4):245-265.

Matsumoto, D. A. 2002. Management's Incentives to Avoid Negative Earnings Surprises. The Accounting Review 77 (3):483-524.

Mauldin, E. G., and S. B. Richtermeyer. 2004. An Analysis of ERP Annual Report Disclosures. International Journal of Accounting Information Systems 5:395-416.

McNichols, M. 2000. Research Design Issues in Earnings Management Studies. Journal of Accounting and Public Policy 19:313-345.

McNichols, M., and G. P. Wilson. 1988. Evidence of Earnings Management from the Provision for Bad Debts. Journal of Accounting Research 26:1-31.

Millman, G. J. 2004. What Did You Get From ERP and What Can You Get? Financial Executive May:38-42.

Mitnick, B. M. 1982. Regulation and the Theory of Agency. Policy Studies Review 1 (3):442-453.

191

Modigliani, F., and M. H. Miller. 1958. The Cost of Capital, Corporation Finance and the Theory of Investment. The American Economic Review XLVIII:261-297.

Murphy, K. E., and S. J. Simon. 2002. Intangible Benefits Valuation in ERP Projects. Information Systems Journal 12:301-320.

Myers, J. 1999. Implementing Residual Income Valuation with Linear Information Dynamics. The Accounting Review 74 (January):1-28.

Nelson, M. W., J. A. Elliott, and R. L. Tarpley. 2003. How are Earnings Managed? Examples from Auditors. Accounting Horizons Supplement:17-35.

Nicolaou, A. I. 2004. Firm Performance Effects in Relation to the Implementation and Use of Enterprise Resource Planning Systems. Journal of Information Systems 18 (2):79-105.

Nicolaou, A. I., and S. Bhattacharya. 2006. Organizational Performance Effects of ERP Systems Usage: The Impact of Post-Implementation Changes. International Journal of Accounting Information Systems 7 (1):18-35.

O'Leary, D. E. 2002. Knowledge Management Across the Enterprise Resource Planning Systems Life Cycle. International Journal of Accounting Information Systems 3:99-110.

———. 2004. Enterprise Resource Planning (ERP) Systems: An Emprical Analysis of Benefits. Journal of Emerging Technologies in Accounting 1:63-72.

Ogneva, M., K. Raghunandan, and K. R. Subramanyam. 2006. Internal Control Weakness and Cost of Equity: Evidence from SOX Section 404 Disclosures. In Working Paper Series : SSRN.

Ogneva, M., K. R. Subramanyam, and K. Raghunandan. 2007. Internal Control Weakness and Cost of Equity: Evidence from SOX Section 404 Disclosures. The Accounting Review 82 (5):1255-1297.

Ohlson, J. A. 1995. Earnings, Book Values, and Dividends in Equity Valuation. Contemporary Accounting Research 11:661-687.

Oracle. 2007. Information Generates Value: The Cornerstone for Sustainable Compliance and Growth 2005 [cited February 19 2007]. Available from http://www.oracle.com/applications/financials/finance.pdf .

———. 2007. Oracle E-Business Suite 2007 [cited 2007]. Available from http://www.oracle.com/applications/e-business-suite.html .

Paper, D., K. Tingery, and W. Mok. 2003. The Relation Between BPR and ERP Systems: A Failed Project . Edited by M. Kosrow-Pour. Vol. 5, Annals of Cases on Information Technology . Hershey, PA: Idea Group Publishing.

PCAOB. 2006. Auditing Standard No. 2: An Audit of Internal Control Over Financial Reporting Performed in Conjunction with An Audit of Financial Statements: PCAOB.

192

———. 2007. Our Mission 2007 [cited February 27 2007]. Available from http://www.pcaobus.org .

Penman, S. H. 2004. Financial Statement Analysis and Security Valuation . 2nd ed. New York: McGraw-Hill/Irwin.

Penman, S. H., and X. J. Zhang. 2002. Accounting Conservatism, the Quality of Earnings, and Stock Returns. The Accounting Review 77 (2):237-254.

Pillay, C. A. 2004. Can Fundamental Value Predict Stock Returns? An Empirical Assessment of the Feltham-Ohlson Model. Dissertation, Louisiana Tech University.

Porter, M. E., and V. E. Millar. 1985. How Information Gives You Competitve Advantage. Harvard Business Review 63 (4):149-160.

Poston, R., and S. Grabski. 2001. Financial Impacts of Enterprise Resource Planning Implementations. International Journal of Accounting Information Systems 2 (4):271- 294.

Raghunandan, K., and D. V. Rama. 2006. SOX Section 404 Material Weakness Disclosures and Audit Fees. Auditing: A Journal of Practice & Theory 25 (1):99-114.

Ragowsky, A., and D. A. Adams. 2005. Assessing the Value Provided by ERP Applications Through Organizational Activities. Communications of the Associaiton for Information Systems 16:381-406.

Rangan, S. 1998. Earnings Management and the Performance of Seasoned Equity Offerings. Journal of Financial Economics 50:101-122.

Samson, W. D., D. L. Flesher, and G. J. Previts. 2006. Corporate Governance and External and Internal Controls: The Case of teh Baltimore and Ohio Railroad, Circa 1831. Issues in Accounting Education 21 (1):45-62.

SAP. 2007. Enterprise Governance and Sarbanes-Oxley Compliance with mySAP ERP Financials 2005 [cited February 19 2007]. Available from http://www.sap.com/usa/solutions/business- suite/erp/financials/pdf/Entpr_Gov_SO_Compliance.pdf .

———. 2006. SAP Ranked #1 in ERP Software http://www.sap.com/usa/index.epx .

Sarker, S., and S. Allen. 2003. Using a Case Study to Test the Role of Three key Social Enablers in ERP Implementation. Information & Management 40 (8):813-829.

Schipper, K. 1989. Commentary on Earnings Management. Accounting Horizons December:91- 102.

SEC. 2003. Final Rule: Management's Reports on Internal Control Over Financial Reporting and Certification of Disclosure in Exchange Act Periodic Reports, edited by U. S. S. a. E. Commission: http://www.sec.gov/rules/final/33-8238.htm .

193

Simmons, M. R. 1997. COSO Based Auditing. The Internal Auditor 54 (6):68-73.

Sloan, R. 1996. Do Stock Prices Fully Reflect Information in Accruals and Cash Flows About Future Earnings? The Accounting Review 71 (July):289-315.

Spence, M., and R. Zeckhauser. 1971. Insurance, Information and Individual Action. American Economic Review 61:380-387.

Staggers. 1977. Unlawful Corporate Payments Act of 1977 Legislative History - House Report: House of Representatives.

Strong, N. 1992. Modelling Abnormal Returns: A Review Article. Journal of Business Finance & Accounting 19 (4):533-553.

Subramani, M., and E. Walden. 2001. The Impact of E-Commerce Announcements on the Market Value of Firms. Information Systems Research 12 (2):135-154.

Subramanyam, K. R. 1996. The Pricing of Discretionary Accruals. Journal of Accounting & Economics 22:249-281.

Tackett, J. A., F. Wolf, and G. A. Claypool. 2006. Internal Control Under Sarbanes-Oxley: A Critcal Examination. Managerial Auditing Journal 21 (3):317-323.

Teets, W. R., and C. E. Wasely. 1996. Estimating Earnings Response Coefficients: Pooled Versus Firm-Specific Model. Journal of Accounting & Economics 21:279-295.

Teoh, S. H., I. Welch, and T. J. Wong. 1998. Earnings Management and the Subsequent Market Performance of Initial Public Offerings. Journal of Finance 53 (December):1935-1974.

———. 1998. Earnings Management and the Underperformance of Seasoned Equity Offering. Journal of Financial Economics 50:63-99.

Teoh, S. H., T. J. Wong, and G. Rao. 1998. Are Accruals During Initial Public Offerings Opportunistic? Review of Accounting Studies 3:175-208.

Themistocleous, M., I. Zahir, R. O'Keefe, and R. Paul. 2001. ERP Problems and Application Integration Issues: An Empirical Survey. Paper read at 34th Hawaii International Conference on System Sciences, at Hawaii.

Trueman, B., and S. Titman. 1988. An Explanation for Accounting Income Smoothing. Journal of Accounting Research 26 (Supplement):127-139.

Tucker, J. W., and P. A. Zarowin. 2006. Does Income Smoothing Improve Earnings Informativeness? The Accounting Review 81 (1):251-270.

Tuttle, B. M. 2005. Editor's Comments. Journal of Information Systems 19 (2):1-5.

194

Umble, E., R. Haft, and M. Umble. 2003. Enterprise Resource Planning: Implementation Procedures and Critical Success Factors. European Journal of Operations Research 146 (2):241-258.

USC. 1998. Anti-Bribery and Books & Records Provisions of the Foreign Corrupt Practices Act (as ammended November 10, 1998): USDOJ.

Venkataraman, R., J. P. Weber, and M. Willenborg. 2008. Litigation Risk, Audit Quality, and Audit Fees: Evidence from Initial Public Offerings. The Accounting Review 83 (5):1315- 1345.

Verrecchia, R. 1986. Managerial Discretion in the Choice Among Financial Reporting Alternatives. Journal of Accounting & Economics 8:175-195.

Walker, K. B. 2008. SOX, ERP, and BPM: A Trifecta that can Make Your Business Run Better. Strategic Finance 90 (6):47-54.

Webb, E. 2008. Sarbanes-Oxley Compliance and Vilation: An Emprical Study. Review of Accounting and Finance 7 (1):5-.

Wier, B., J. Hunton, and H. R. HassabElnaby. 2007. Enterprise Resource Planning Systems and Non-Financial Performance Incentives: The Joint Impact on Corporate Performance. International Journal of Accounting Information Systems 8 (3):165-190.

Wright, S., and A. M. Wright. 2002. Information System Assurance for Enterprise Resource Planning Systems: Unique Risk Considerations. Journal of Information Systems 16:99- 113.

Xie, H. 2001. The Mispricing of Abnbormal Accruals. The Accounting Review 76 (July):357-373.

Xu, L. E. 2007. Is V/P a Distinct Anomaly? Review of Accounting and Finance 6 (4):404-418.

Zhou, P. 2002. The Impact of Arbitrage Costs and Earnings Forecast Accuracy on the Predictive Power of Value-to-Price Ratios. Dissertation, Georgia State University.

Zuboff, S. 1985. Automate/Informate: The Two Faces of Intelligent Technology. Organizational Dynamics 14 (2):5-18.