Amsterdam Business School
The association between the detection of financial reporting fraud and
subsequent improvements in the quality of governance mechanisms
Name: Lisa Sarian
Student number: 10867945 Master Thesis, draft version
Thesis supervisor: Alexandros Sikalidis Date: 15 August 2016
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam Word count: 16888
Statement of Originality
This document is written by Universiteit van Amsterdam student Lisa Samantha Sarian who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
In this study, we examine the association between the detection of financial reporting fraud and subsequent improvements in the quality of governance mechanisms in the U.S. for the period 2004-2015. The sample size of this study is 41 fraud companies with a total of 82 companies. Prior research with a database that consists of the years before the implementation of the SOX has found that fraud firms have weaker corporate governance mechanisms compared to control firms. This study clarifies the changes in corporate governance mechanisms of fraud firms and control firms, as well as the changes between those firms. This study shows that three years after the detection of fraud, there is a greater increase in the percentage of outside directors for fraud firms. The results also show that fraud firms tend to have less audit committee meetings three years after the detection compared to the year prior the detection. Further, comparing the results of Farber (2005), which are the pre-SOX results, with the results of this study, which are the post-SOX results conclude that the corporate governance mechanisms, percentage of outside board members and CEO duality give the same results for the pre- and the post-SOX fraud companies. For the corporate governance mechanisms, which are audit committee characteristics the results of the post-SOX are different compared to the results of the pre-SOX. The most interesting finding of this study is for CEO duality, where fraud firms choose to split the position CEO and COB, control firms rather want to choose for the combined position and adds a Lead Independent Director to their board.
T
ABLE OFC
ONTENTS1 INTRODUCTION ... 1
1.1 BACKGROUND ... 1
1.2 MOTIVATION ... 1
2 LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT ... 4
2.1 FRAUD ... 4
2.2 REGULATING FRAUD ... 6
2.2.1 U.S. Securities and Exchange Commission ... 6
2.2.2 Sarbanes- Oxley Act 2002 ... 6
2.2.3 Dodd-Frank Wall Street Reform and Consumer Protection Act ... 7
2.3 FRAUD AND CORPORATE GOVERNANCE MECHANISMS ... 8
2.3.1 Board characteristics ... 8
2.3.2 Audit committee characteristics ... 9
2.3.3 CEO characteristics ... 10
3 DATA AND METHOD ... 12
3.1 SAMPLE SELECTION ... 12
3.1.1 Collecting fraud firms ... 12
3.1.2 Search of matching pair ... 13
3.1.3 Measurement and variables used ... 14
3.1.4 Quantitative description of variables ... 17
3.1.5 Research design ... 19 4 RESULTS ... 20 4.1 DESCRIPTIVE STATISTICS ... 20 4.1.1 Comparing means ... 21 4.2 CORRELATIONS ... 25 4.3 REGRESSIONS ... 27
4.3.1 Regression model with audit committee members ... 27
4.3.2 Regression model with outside audit committee members ... 31
5 LIMITATIONS AND FURTHER RESEARCH ... 35
6 CONCLUSION ... 37
6.1 OUTSIDE BOARD MEMBERS ... 37
6.2 OUTSIDE AUDIT COMMITTEE MEMBERS ... 37
6.3 AUDIT COMMITTEE MEETINGS ... 38
6.5 POSITION CEO AND COB ... 39
6.6 IMPLICATIONS ... 40
7 REFERENCES ... 41
1 Introduction
1.1 Background
This study builds on Farber’s (2005) paper on the association between the credibility of the financial reporting system and the quality of governance mechanisms.
Financial reporting fraud has become a major issue in the financial world in the early 2000’s with cases like Enron, Worldcom and Xerox. But even nowadays when new regulations have been applied to combat fraud, there still remains a lot of financial reporting fraud. Companies like Ballast Nedam, Logitech International, Volkswagen, Toshiba, Olympus, Thornton Precision Components and Carter’s Inc. have committed fraud in recent years. (NRC, 2012; U.S. Securities and Exchange commission, 2016; Agrawal & Chadha, 2005; Rezaee, 2005).
As a response to all these accounting scandals in the late 1990’s and the early 2000’s, major changes have been made following these scandals. The United States implemented a new federal law in 2002 called ‘Sarbanes Oxley-Act’. Also the New York Stock Exchange (NYSE), the National Association of Securities Dealers Automated Quotations (NASDAQ) and the American Stock Exchange (AMEX) adopted a new set of corporate governance rules (Agrawal & Chadha, 2005; Clark, 2005; Uzun, Szewczyk, & Varma, 2004). All these new changes have been made to increase the corporate governance applied by companies with stock listings on the markets of NYSE, NASDAQ and AMEX. These changes have also been made to increase the accuracy and reliability of corporate disclosures (Hostak, Lys, & Yang, 2013; Agrawal & Chadha, 2005; Uzun, Szewczyk, & Varma, 2004). Besides these changes, institutions like the Securities and Exchange Commission (SEC), the Securities and Exchange Board of India (SEBI), and the European Securities Commission (ESC) were created to provide compliance and transparency with financial reporting.
In previous studies, associations have been found between weaknesses in corporate governance and poor financial reporting quality, financial reporting fraud, and weaker internal controls (Beasley, 1996; Dechow, Sloan, & Sweeney, 1996; Klein, 2002; Lenard, Petruska, Alam, & Yu, 2012). These studies found that when a company has weak corporate governance mechanisms it is often the case that such a company has dealings with fraud.
1.2 Motivation
Financial reporting fraud is not the most common type of fraud but the amounts associated with financial reporting fraud are much higher than in any other type of fraud. Of all types of fraud,
financial reporting fraud has the highest financial impact on the company and sometimes even on the society (ACFE, 2014). In the past years financial reporting fraud has become a hot issue since a lot of major scandals have happened all over the world (U.S. Securities and Exchange commission, 2016; Agrawal & Chadha, 2005; Rezaee, 2005). The financial world has applied several rules to eliminate and/or reduce the risk of fraud. The United States implemented the Sarbanes Oxley Act in 2002, and also the Stock Exchanges in the United States implemented new corporate governance rules. A lot has been written about these major scandals and new rules, but there have not been many studies that examined the impact of the differences in corporate governance mechanisms between companies. Due to those major financial reporting frauds around 2000 (companies such as, Enron, WorldCom, Parmalat, Ahold), corporate governance mechanisms have been receiving much more emphasis in the academic literature, but also in practice (Blue Ribbon Committee, 1999; Sarbanes & Oxley, 2002; Bebchuck & Cohen, 2005; Larcker, Richardson, & Tuna, 2004).
A key theme in these studies is the relation between corporate governance mechanisms and financial reporting fraud. Now, about a decade after the major changes in regulation of the financial sector and governance mechanisms more specifically, we would like to test the findings of these studies. This leads to the research question that will be answered in this study and that is:
“What is the association between the detection of financial reporting fraud and subsequent improvements in the quality of governance mechanisms?”
By answering this question, this study makes a contribution to the existing literature about financial reporting fraud. We do so by examining the association between the detection of financial reporting fraud and the subsequent improvements in the quality of governance mechanisms. Farber (2005) has examined the association between the credibility of the financial reporting system and the quality of governance mechanisms. Farber (2005) also investigated whether the improved governance influences informed capital market participants. This study is a follow up on the paper of Farber (2005). Farber (2005) used U.S. data from 1982 up to and including 2000. This paper adds value to the existing literature by analysing the period after 2000. This is especially relevant as major financial reporting scandals happened all over the world post-2000, that have led to changes in legislation (e.g. the Sarbanes Oxley Act in 2002) and changes in the policies of stock exchanges in the United States. By evaluating the success of these changes,
that are aimed at reducing or eliminating the risk of such scandals happening again, we also contribute to the community of practice.
In this study, we have used U.S. data from 2003 up to and including 2015, which has not been researched in previous studies yet. In doing so, this study we will expand the existing literature by relating our results to those of Farber (2005). This research could offer contributions to the knowledge in the auditing field. Furthermore, this research will also be interesting for society at large, including business owners, shareholders and investors as financial statement fraud still occurs and a better understanding of the phenomenon might contribute to a decrease of financial statement fraud.
This paper is organized as follows: Chapter two provides an overview of the existing theories and existing literature to provide background information on the topics fraud and regulation (e.g. SEC, Sarbanes Oxley-Act, and the Dodd-Frank Act). This chapter also provides the formulation of the hypotheses for this research. In chapter three we describe the sample selection and the research design used in this study. In this chapter we also will provide an explanation of the used corporate governance mechanisms. The purpose of chapter four is to discuss the results of the analyses. Chapter five provides the limitations of this study and some directions for further research. Lastly, chapter six will provide a conclusion of this study and a discussion of the results that are given in chapter four.
2 Literature review and Hypothesis development
This chapter contains the outline for the literature review to explain the underlying mechanism that relates the theoretical constructs.
2.1 Fraud
In order to get a better understanding of the term fraud and to understand the research question, an explanation of the perception of fraud is presented. There are different definitions of the term fraud. ISA 240 defines fraud as “An intentional act by one or more individuals among management, those charged with governance, employees, or third parties, involving the use of deception to obtain an unjust or illegal advantage” (IFAC, 2009). The Association of Certified Fraud Examiners, henceforth ACFE, explains fraud in the broadest sense as “Fraud can encompass any crime for gain that uses deception as its principal modus operandi” (Association of Certified Fraud Examiners). The ACFE refers for a more specific explanation to the Black’s Law Dictionary1, which defines fraud as “A knowing misrepresentation of the truth or
concealment of a material fact to induce another to act to his or her detriment” (Garner, 2014). There is internal and external fraud, internal fraud is also called occupational fraud and abuse. External fraud against an organization covers a broad range of schemes. External threats to the organization are security breaches and thefts of intellectual property perpetrated by unknown third parties like hacking, theft of proprietary information, tax fraud, bankruptcy fraud, or loan fraud (Association of Certified Fraud Examiners, 2014). In this research the focus is on internal fraud.
Internal fraud can be classified into three categories (Kranacher, 2010; Association of Certified Fraud Examiners, 2014), asset misappropriation, corruption, and financial statement fraud. Those three categories are also used by the ACFE in their classification system referred to as ‘the fraud tree’. An asset misappropriation scheme involves the misuse or theft by an employee of an organization’s resources. Examples of asset misappropriation are theft of cash, theft of inventory, false billing schemes, and inflated expense reports. Corruption schemes involve the misuse by an employee of his or her influence in a business transaction in a way that violates his or her duty to the employer in order to gain a direct or indirect benefit. Examples of corruption are bribery, conflicts of interest, economic extortion, and illegal gratuities. Financial statement fraud schemes involve an employee intentionally causing a misstatement or an
omission of material information in the company’s financial reports. Examples of Financial statement fraud are recording false revenues, and asset and revenue overstatements or understatements (Association of Certified Fraud Examiners, 2014; Rezaee, 2005; Albrecht, Holland, Malagueño, Dolan, & Tzafrir, 2015).
This study focuses in particular on financial statement fraud. The reason for focusing on financial statement fraud is that in all the cases of the ACFE, financial statement fraud2 is the
least common type of fraud but those cases of financial statement fraud had the greatest financial impact (Association of Certified Fraud Examiners, 2014), which we find interesting.
The experienced pressure or motivation for initial perpetrators to commit financial statement fraud often relates to the possible negative outcomes of reporting the firm’s true financial performances (Albrecht, Holland, Malagueño, Dolan, & Tzafrir, 2015). The financial performance of firms is a key measurement, it has a big influence on a firm’s stock price. Shareholders use a firm’s financial performance to measure if expectations are met. Further, the security of the jobs of executives and their financial compensation rest on the financial performance of the firm, because the executives are hired to maintain a strong financial performance and also to make sure the company’s stock prices increases. Therefore, top management feels an enormous pressure to measure up to the investor’s expectations and may even consider using fraudulent means to do so (Albrecht, Holland, Malagueño, Dolan, & Tzafrir, 2015).
In order to commit financial statement fraud without getting caught, there has to be an opportunity for potential perpetrators. The higher the position held by an employee in a company the greater the opportunity to manipulate others and thereby reduce the likelihood of detection. Although most people can expect to be honest and have the right intentions to be honourable to their employer, they still commit fraud (Tsang, 2002). They will try to justify it for themselves and to others, in a way to search for excuses in which they believe they are not violating their moral standards (Tsang, 2002). (Albrecht et al., 2015, p. 805) wrote that typical excuses that have been used for financial statement fraud are “This is our only option; everybody is doing it; it will only be short-term; it is in the best interest of the company, shareholders, or employees”. Within financial statement fraud it often is not one person that commits fraud, but several persons working as a group together (Association of Certified Fraud Examiners, 2014; Zyglidopoulos & Fleming, 2008).
2.2 Regulating fraud
2.2.1 U.S. Securities and Exchange Commission
The U.S. Securities and Exchange Commission (henceforth, SEC) is an agency of the U.S. federal government. The SEC has a three-part mission, which is to protect investors, maintain fair, orderly, and efficient markets, and facilitate capital formation (U.S. Securities and Exchange Commission, 2016). The SEC was designed to restore investor confidence in the capital markets of the United States by providing investors and the markets with more reliable information and clear rules of honest dealing. The functional responsibilities of the agency are organized into five divisions and 23 offices (U.S. Securities and Exchange Commission, 2016). The five divisions are Corporation Finance; Enforcement; Economic and Risk Analysis; Investment Management; Trading and Markets. Of all those divisions, the division Enforcement will be the most important division for this study, since this division undertakes the detecting fraud companies, which relevant for this study.
The Enforcement division assists the commission in executing its law enforcement function by recommending the commencement of investigations of securities law violations. The Enforcement division does so by recommending the commission to bring civil actions in federal court or as administrative proceedings before an administrative law judge, and by prosecuting these cases on behalf of the commission. The federal court actions are executed by the subdivision called Accounting and Auditing Enforcement Releases (AAER) (U.S. Securities and Exchange Commission, 2016).
The SEC is important for this study because we will be making use of their database called, EDGAR (Electronic Data Gathering, Analysis and Retrieval system), where all the filings of the Enforcement division is being held and is publicly available.
2.2.2 Sarbanes- Oxley Act 2002
In 2002 a new federal law, the Sarbanes Oxley Act 2002 (henceforth, SOX) was introduced in the United States. The reason for designing and implementing a new federal law was because of the high profile of scandals that happened in the late 1990’s and early 2000’s. The fraud scandals of Enron, WorldCom and Xerox were the most important reason for implementing SOX in the United States, aimed at reducing the number of fraud scandals in the future (Zhang, 2007). Prior to SOX, there was self-regulation in the US audit profession. This type of regulation proved insufficient to rectify the strong economic motives these days, like the financial scandals mentioned above (Ribstein, 2002; Hayes, Wallage, & Görtemaker, 2014). With the self-regulation
for instances external auditors were allowed to combine internal audit services and external audit services to the same client, but under SOX it was no longer allowed to combine these services for the same client.
With SOX, the United States wanted to improve the audit quality of public companies in the United States (Coates & Srinivasan, 2014; DeFond & Lennox, 2011). SOX introduced major changes in the regulation of financial practice and corporate governance3. SOX is made up of
eleven sections, and as far as compliance concerned the most important sections are often considered to be the sections 302, 401, 404, 409, 802, and 9064. There is especially written a lot
in the literature about section 404, this section is about financial disclosures. It mainly requires issuers to publish information in their annual reports that concerns the scope and adequacy of the internal control structure and also the procedures for financial reporting.
Coates and Srinivasan (2014) stated in their study that SOX has two core goals concerning the relationship between public companies and their auditors, which are (1) to create a quasi-public institution to oversee and regulate auditing. This institution is called the Public Company Accounting Oversight Board (PCAOB). This institution is responsible for the regulation, registration, and inspection of external auditors in the United States (Gunny & Zhang, 2013; Hayes, Wallage, & Görtemaker, 2014) (2) to enlist auditors more extensively in the enforcement of existing laws against theft and fraud by corporate officers, pursuant to regulations from and enforcement by the PCAOB.
2.2.3 Dodd-Frank Wall Street Reform and Consumer Protection Act
Another federal law that has been implemented in the United States is the Dodd–Frank Wall Street Reform and Consumer Protection Act (also known by Dodd-Frank Act). This happened in 2010, after the global financial crisis that started in 2007. The Dodd-Frank Act was, amongst others, implemented in the United States to increase the protection of shareholders. The new federal law also gives the SEC the ability to compensate whistleblowers if their information leads to a sanction of valued at more than 1 million US dollar for the fraudulent company. The whistleblower may receive up to thirty percent of the penalty imposed on the company for compensation. This Act also increases the protection of whistleblowers against reprisals. With this Act, the aim is to give an incentive to top management to report fraud (Rapp, 2012). While the Dodd-Frank Act is relevant for the analysis of fraud firms we choose not to take its impact
3http://www.soxlaw.com/ May 3rd 2016
to account in our analysis as the Dodd-Frank Act was introduced half way through our period of analysis.
2.3 Fraud and corporate governance mechanisms
In general, we conclude from existing literature that fraud companies have weak corporate governance mechanisms (Beasley, 1996; Uzun, Szewczyk, & Varma, 2004; Dechow, Sloan, & Sweeney, 1996). The relation between corporate governance mechanisms and the occurrence of fraud, more specifically financial statement fraud, is at the core of this research. In order to answer our research question we have developed five hypotheses based on the hypotheses used by Farber (2005) in his research. We use a division into three different subjects: board characteristics, audit committee characteristics and CEO characteristics.
2.3.1 Board characteristics
The board of directors is a group of individual people that have been elected as the representatives of the shareholders in a company. The main goal of the board of directors is to determine if corporate management related policies and decisions on major company issues are made in the interest of the company. U.S. stock exchanges require companies to have a board of directors in their company. A board is composed of inside and/or outside board members. Inside board members are people that have a position inside the company, such as Chief Executive Officer, Chief Financial Officer or Chief Operating Offer. Outside board members are members that are not involved in the inner workings of the company. Outside board members are only bringing their experience and network with him, which adds contribution to the company. Stock exchanges require that the majority of a board consist outside board members.
Beasley (1996) researched whether the proportion of outside board members and certain characteristics of outside board members change the likelihood of financial statement fraud. Beasley found that the more outside board members a company has, the smaller the chance that a firm will commit financial reporting fraud. Uzun et al (2004) examined to what degree the characteristics of the board of directors affect the existence of corporate fraud in a company. The result of the Uzun et al study is that companies with a higher percentage of outside board members are less likely to commit fraud. Chen et al (2006) studied whether board of director characteristics have an effect on financial reporting fraud. The outcome of their study shows that a way to decrease financial reporting fraud is to increase the percentages of outside board
members. Chen et al (2006) also discover that the shorter the tenure of the chairman is, the less able he is to deter fraud.
All of these studies are based on information before the implementation of the Sarbanes-Oxley Act (SOX). We seek to verify the validity of these findings in the period post the introduction of SOX. Therefore, for our H1, we examine the percentage of outside director between fraud firms three years after their detection and their corresponding control firms in, therefore we state in the alternate:
Hypothesis 1. Fraud firms have a greater increase in outside director percentage after fraud detection than their corresponding control firms.
2.3.2 Audit committee characteristics
An audit committee is an operating committee of the board of directors. The audit committee helps the board of directors to monitor the company on a lot of characteristics. Examples of those components are processes like the accounting and the financial reporting processes, the integrity of the financial statement, and also the qualifications and independency of the hired accountants. In an audit committee there has to be at least one financial expert according to the stock exchanges of the United States and the Sarbanes-Oxley Act (Sarbanes & Oxley, 2002). A financial expert is a member of the audit committee that has experiences in preparing and auditing financial statements of generally comparable issuers. Furthermore, the audit committee may only exist of members which are independent from management. Also, the NYSE has a requirement that the audit committee should be consisted of at least three members (Klein, 2002).
Uzun et al (2004) find an association between the existence of an audit committee and the likelihood of corporate fraud. Uzun et al (2004) find that when an audit committee is present within a company, the likelihood of committing corporate fraud is lower than when a company does not have an audit committee in their board. Also, Uzun et al (2004) find a negative, but not significant result, which indicates that a lower degree of independence in the audit committee will contribute to a higher likelihood of fraud. Chen et al (2004) discover that firms that are committing fraud are more likely to have much more audit committee meetings than firms that are not committing fraud. Farber (2005) finds the opposite of the results of Chen et al (2004) about audit committee meetings. Farber (2005) finds that fraud firms have fewer audit committee meetings compared to a firm that does not commit fraud. Another finding of Farber
(2005) is that fraud firms also tend to have fewer financial experts on the audit committee than firms that does not commit fraud.
For our H2, we examine the audit committee independence between fraud firms three years after their detection and their corresponding control firms in, therefore we state in the alternate:
Hypothesis 2. Fraud firms have a greater increase in audit committee independence after fraud detection than their corresponding control firms.
For our H3, we examine the audit committee activity5 between fraud firms three years after their
detection and their corresponding control firms in, therefore we state in the alternate:
Hypothesis 3. Fraud firms have a greater increase in audit committee activity after fraud detection than their corresponding control firms.
For our H4, we examine the number of financial experts in the audit committee between fraud firms three years after their detection and their corresponding control firms in, therefore we state in the alternate:
Hypothesis 4. Fraud firms have a greater increase in financial experts6 after fraud detection
than their corresponding control firms 2.3.3 CEO characteristics
The CEO (Chief Executive Officer) is responsible for overseeing operational aspects within the company, as such the CEO is the highest-ranking executive in a company. The CEO is the person within a company that reports all the information to the board of directors. One of the main goals of a CEO is to provide overall leadership to the company by establishing the day-to-day operations of the company. Another goal of a CEO is to maximize the value of the company.
Jensen (1993) claims that CEOs that also has the position Chairman of the Board (COB), cannot execute all of their duties very well. The reason for this claim is that Jensen (1993) assumes there will be conflicts of interest, by being a CEO (an inside board member) that oversees the operational aspects of the company, but at the same time functioning as the chairman of the board, which function is to oversee the company as a whole. Abdullah (2004)
5 Audit committee activity is another word for audit committee meetings.
6 Financial expert is a member of the audit committee that has experiences in preparing and auditing financial statements of generally comparable issuers.
finds that CEO duality or separated positions are not related to firm performance. Farber (2005) believes that fraudulent firms tend to have a combined position of the chairman of the board and the CEO, whereas combined positions occur much less frequent at non-fraudulent firms.
For our H5, we examine the independence between the positions CEO and Chairman of the Board between fraud firms three years after their detection and their corresponding control firms in, therefore we state in the alternate:
Hypothesis 5: Fraud firms have a greater decrease in the proportion of firms with the combined CEO/COB position after fraud detection than their corresponding control firms.
3 Data and Method
In this chapter the choices for the empirical approach, sample and methodology are outlined 3.1 Sample selection
The data used for this study is limited to data from the United States. The reason for using U.S. data is because Farber (2005) also used U.S. data, which provides for a better comparison of results since this study will follow the same sample selection methodology of Farber (2005). In addition, the US data is more readily and publically available. The period that is used for this study is from 2004 up to and including 2015.
We examine whether there is a difference in corporate governance mechanisms between fraud firms and control firms. Therefore the dependent variable has to be dichotomous in other words, the variable has to have a ‘one value’ that stands for fraud firms and a ‘zero value’ for control firms (no fraud firms). All the information that is obtained regarding the variables, control variables and other company information were hand-collected. The steps that have been taken for collecting the sample selection are described below.
3.1.1 Collecting fraud firms
The database that has been used in this study for collecting firms that have committed fraud is the EDGAR (Electronic Data Gathering, Analysis and Retrieval system) database, which belongs to the SEC. The SEC uses the Enforcement division to issue against firms, audit firms, managers, and directors for violations of the SEC rules (Farber, 2005). This study only consists of publicly held companies listed in the United States because the SEC publishes only data of these companies listed in their database. Using the EDGAR database comes also with limitations, since the SEC is only selecting those cases where they have the best chance of winning a judgment. Therefore the SEC selects probably only the most extreme fallacious reports (Feroz, Park, & Pastena, 1991). That is why the result of this study cannot be generalized to the complete population of firms that report fraudulently.
In the EDGAR database all the firms that are issued have been given a number to be identified, called Accounting and Auditing Enforcement Releases (AAERs)7. The total AAERs
that were issued from 2005 up to and including 2012 are 819. 762 of these did not involve financial reporting fraud, were duplicates, or were beyond the scope of the research period. In
order to determine whether the committed fraud of the AAERs is financial reporting fraud, each case has to be read to find if the fraud firm violated rule 10b-5. Rule 10b-5 falls under Section 10b of the Exchange Act 1934. (Farber, 2005, p. 542) describes Section 10b and Rule 10b-5 as:
“Section 10(b) of the antifraud provisions of the Exchange Act (of 1934) and Rule 10b-5 thereunder proscribes the making of materially false and misleading statements in connection with the purchase or sale of any security. Violations of Section 10(b) and Rule 10b-5 occur when an issuer makes material misstatements in registration statements, prospectuses or periodic reports filed with the Commission and trading thereafter occurs in the issuer’s securities. The filing of false and misleading reports is also a violation of Section 10(b) of the Exchange Act (of 1934) and Rule 10b-5 thereunder because reports of publicly traded companies affect the markets for offer, sale and purchase of their securities.”
The remaining 45 AAERs were researched for further information. Information has been gathered on: board characteristics; audit committee characteristics; other governance variables; and additional company information. In the subsection 3.1.3 more will be explained about these variables. All of the information is obtained through the EDGAR database by searching in the annual reports (10-k) of the fraudulently companies and the proxy statements (DEF 14). If one of these variables could not be obtained through the EDGAR database and also not via the website of the specific firm then the firm was excluded from the sample. After obtaining all the data of each fraud firm there were 45 left for which all the information could be gathered in the EDGAR database. Table 1 shows the process of the sample selection of fraud firms.
TABLE 1–PROCESS OF THE SAMPLE SELECTION
Years of AAER detected Total
Total fraud firms between 2005 and 2012 (#2158 and # 3434) 1276
Fraud that didn’t consist financial reporting fraud -931
Total financial reporting fraud companies 345
Duplicates, no information available -300
Total fraud companies 45
No matching pair found -4
Fraud companies for which matches were found 41
3.1.2 Search of matching pair
Another major step in the sample selection is the search for a matching firm. The control firm that has to match with a fraud firm has been obtained through COMPUSTAT. In order to create a matching group, the control firms had to be of similar in the net sales, industry, stock exchange listing, and time period. The year prior to the detection of the fraud firm was taken as the reference year. Each control firm was matched with a fraud firms based on the following conditions:
(1) Net sales: The net sales of the control firm had to be similar to the net sales of a fraud firm. Therefore the net sales of the control firm had to be within ± 25% of the net sales of the fraud firm. In a very few cases the net sales was a little bit more or less than 25% of the net sales.
(2) Standard Industry Code: The SIC (Standard Industry Code) contains four-digits8. To
compare a control firm with a fraud firm, the control firm had to have the same four-digits SIC code. If there was no control firm with the same four-digit code identified, a three-digit or at least a two-digit code was searched for to match with the fraud firm. (3) Stock exchange: Each fraud firms has been listed at the New York Stock Exchange
(NYSE) or the NASDAQ. In matching a control firm to a fraud firm it had to be listed on the same stock exchange. In a very few cases the stock exchange of the control firm was different than the matching fraud firm.
(4) Time period: The annual report and the proxy statement of the control firm had to be available for the same time period of the matching fraud firm.
(5) Fraudulently activities: Each matching control firm had to be fraud free in a period 5 years. The control firm had to be fraud free one year prior to the detection of the fraud firm and the upcoming four years including the detection year.
The net sales, Standard Industry Code and the stock exchange were obtained through COMPUSTAT. The time period and the check for fraudulently activities were obtained through EDGAR database. If the control firm meet all the above requirements, the annual report and proxy statement were obtained in the EDGAR database.
The same steps have been taken for collecting all the information on the variables and the additional company information as for the fraud firms. A sample of 41 control firms was obtained that matches with the fraud firms. The total sample consist 82 firms, 41 fraud firms and 41 control firms.
3.1.3 Measurement and variables used
The variable of importance in this study is the determination of fraud or control firm, which is the dependent variable. The independent variables are: the number of outside board members; the board size; the audit committee size; the number of outside audit committee members; the number of financial experts in the audit committee; and if the CEO also represents the function of Chairman of the Board. The control variables are: if the independent auditor is a Big-4
company; the percentages of blockholders; the percentages of institutional ownership; the percentages of inside ownership and additional information of the company like market value; net sales or revenue; Stock exchange listing; the SIC and the CIK code of the companies. For both fraud and control firms all these variables and the additional information had to be collected for the year prior to the detection of the fraud firm and three years later after the detection year. The variables are described more in depth below (figure 1 gives a brief overview of all the variable).
3.1.3.1 Dependent variable
The dependent variable is firm type: fraud or control. This variable can only take two values, fraud firm or control firm and therefore this variable is dichotomous. It takes the value one if the firm committed financial reporting fraud, and it takes the value zero if it is a control firm.
3.1.3.2 Independent variables
Board Size: Is a variable that takes a nominal value of the total board members.
Outside board members: Is a variable that takes a nominal value of the total outside board members. Farber (2005) defines an outside director as a director who is not a present or former employee of the firm and whose only formal connection to the firm is his/her duty as a director.
Percentage of outside board members: Is the difference of outside board members divided by the total board members expressed in percentages.
Audit committee members: Is the amount of total audit committee members.
Outside audit committee members: Are members of the audit committee that have no relationship with the company. For example, they do not have a position within the company as an employee.
Financial experts: Gives the amount of financial experts that are present in the audit committee. 3.1.3.3 Control variables
CEODuality: and if the CEO also represents the function of Chairman of the Board. This variable is made into a dummy where the value one consist CEODuality and the value zero consist no CEODuality.
Big4: Is a dummy variable which takes the value one if the firm has a Big-4 auditor as there independent auditor and takes the value zero if otherwise.
Percentages of blockholder ownership: Is a variable in percentages. It shows the proportion of blockholders.
Percentages of institutional ownership: Is a variable in percentages. It shows the proportion of institutional ownership.
Percentages of inside ownership: Is a variable in percentages. It shows the proportion of inside ownership.
Market value: Is the amount of the aggregate market value stated in the annual report in numbers. 3.1.3.4 Additional information of the company
Net sales or revenue: Each firm has to state its net sales or revenue in their annual report.
Stock exchange listing: Each firm had to be listed in the U.S. For this study a dummy variable is made, it takes the value one if the firm is listed at the New York Stock Exchange and it takes the value zero if the firm is listed at the NASDAQ.
Standard Industry Classification (SIC): Is a classification code, which takes four-digit numerical codes. The SIC code is assigned by the U.S. government in order to identify the primary business of the company. The SIC code has been used inter alia to find a matching control firm.
Central Index Key (CIK): Every firm that has been listed in the U.S. has a CIK code. It is a number that is given to each company in order to identify the company. For this study the CIK code has been used to identify the companies in the EDGAR database.
FIGURE 1–VARIABLES Dependent variable
Fraud/Control firm = Firm type; it could be a fraud firm or control firm Independent variables
OutsideDir% = Percentage of outside directors #OutsideDir = Number of outside directors
#Directors = Number of directors
#AudComMeet = Number of audit committee meetings
#AudComMbrs = Number of audit committee members
#AudComOutsideDir = Number of outside directors on audit committee #FinExpert = Number of financial experts on audit committee Control variables
CEODuality = One person that holds the position CEO and Chairman of the Board Block% = Percentage of shares held by > 5 percent blockholders
InstOwn% = The percentage of shares held by institutions that file SEC Form 13 F InsideOwn% = The percentage of shares held by management and directors
Big4% = Percentage with Big 4 audit firms
Market value = The market value of the company of the initial or final year
Additional information
Net Sales/Revenue = The net sales/revenue of the company of the initial or final year
SIC = Standard Industry Classification
CIK = Central Index Key
Initial Year = The year prior to fraud detection
3.1.4 Quantitative description of variables
Table 2A provides an overview of general information of the initial year. This table shows that the p-value of net sales/revenue is 0.814, which implies that there is no significant difference in the net sales/revenue between the fraud firms and control firms in the initial year. This could also be seen at the amounts that are shown under the mean, there is a difference. The same conclusion we derive for the market value between the fraud firms and control firms. With a p-value of 0.439 we can conclude that there is also no significant difference in the market p-value between fraud firms and control firms. Because the dependent variable is dichotomous, we have heteroskedasticity, which can lead to biased and misleading estimations.
Further table 2B shows that the number of stock exchanges is almost equal, the reason for an unequal number has to do with the fact that for two firms a matching control firm could not be found with the same stock exchange. Below the stock exchange an overview of the SIC code is given. It shows that the majority of matching pairs has been found with a four-digit SIC code. The last section of table 2B provides an overview of the amount of firms that has been found divided in year. The majority of fraud firms have been found in 2008. This may have to do with the financial crisis, which took place around the years of 2008. Table 2B also shows that in this study the industry where most of the financial reporting fraud occurred are the industries: industrial machinery and equipment, electronic and other equipment, chemical and allied products.
TABLE 2A-MATCHING STATISTICS FOR FRAUD AND CONTROL (INITIAL YEAR)
Variables Fraud firms Control firms
Net sales/ Revenue
Mean $3,676,170,455 $4,121,324,928 Median $1,016,742,000 $990,807,000 Standard deviation $7,852,895,089 $9,162,669,017 T-statistic 0.236 P-value 0.814 Number of observations 41 41 Market value Mean $4,105,859,379 $7,571,923,959 Median $726,000,000 $886,000,000 Standard deviation $12,166,266,848 $25,831,494,754 T-statistic 0.777 P-value 0.439 Number of observations 41 41
TABLE 2B-MATCHING STATISTICS FOR FRAUD AND CONTROL (INITIAL YEAR)
Stock Exchange
NYSE 22 19
NASDAQ 19 22
Total 41 41
Matched based on SIC codes
Four-digit SIC code 25
Three-digit SIC code 5
Two-digit SIC code 11
Sum of the firms per year
2004 - 4 2007 - 8 2010 - 6
2005 - 10 2008 - 22 2011 - 4
2006 - 16 2009 - 12
Total of 82 firms 2-Digit SIC
Code Name of the industry Number of fraud firms that occurred in this study
20 Food & Kindred Products 1
28 Chemical & Allied Products 3
35 Industrial Machinery & Equipment 8
36 Electronic & Other Electric Equipment 4
44 Water transportations 1
45 Transportation by air 2
48 Communications 1
49 Electric, Gas, & Sanitary Services 1
50 Wholesale-Durable goods 1
51 Wholesale Trade – Nondurable Goods 1
53 General Merchandise Stores 1
54 Food Stores 1
55 Automative Dealers & Service Stations 1
59 Miscellaneous Retail 2
60 Depository Institutions 2
61 Nondepository Institutions 1
62 Security and Commodity brokers, Dealers, Exchanges and Services 1
63 Insurance Carriers 2
67 Holding & Other Investment Offices 2
72 Personal Services 1
73 Business Services 2
80 Health Services 1
3.1.5 Research design
As discussed above, we have collected information about the fraud and control firms through the EDGAR database and the COMPUSTAT, WRDS database. In order to test the changes in corporate governance mechanisms between fraud and control firms and those over time from the year prior to detection and three years afterwards, we have made use of a matched pairs design. We have also tested the variables through a bivariate correlation matrix, to see whether there are variables that correlate with each other. The correlation matrix also tells us if we have to separate variables in the logistic regression. Before having prepared a logistic regression we first did an analysis with the linear regression to test on multicollinearity. In the linear regression we have looked at the Variance Inflation Factor (VIF) and Tolerance to test whether there is any multicollinearity between the variables. This enabled us to perform a logistic regression, whereby, we have been able to classify firms as fraud or non-fraud firms by testing the likelihood of fraud indicators.
4 Results
4.1 Descriptive Statistics
Figure 2 and table 3 set out the descriptive statistics of the dependent variable, independent variables, and the control variables. It provides an overview of the differences between fraud and control firms and also between the initial year and the final year.
FIGURE 2,PANEL AFRAUD FIRMS -MEANS OF GOVERNANCE VARIABLES IN PERCENTAGE
FIGURE 2,PANEL B-FRAUD FIRMS -MEANS OF GOVERNANCE VARIABLES IN LEVELS
Figure 2, Panel A, shows the percentages of variables of fraud firms in the initial year and the final year. The percentages of outside directors show the greatest increase of the variables in panel A. Other variables that have been increasing from the initial year to the final year are the
OutsideDir% Big4% CEO=CoB% Block% InstOwn% InsideOwn% Ini9al Year 69.891 87.805 41.463 27.71 0.834 16.14 Final Year 77.713 82.927 39.024 30.41 1.788 15.98 0 10 20 30 40 50 60 70 80 90 100 Percen ta ge Ini9al Year Final Year
#OutsideDir #Directors #AuditComMeet #AuditComMbrs #AuditComOut #FinExperts Ini9al Year 6.52 9.10 9.15 3.67 3.62 1.67 Final Year 7.21 9.20 8.13 3.79 3.77 1.99 0 1 2 3 4 5 6 7 8 9 10 Le ve l Ini9al Year Final Year
blockholders and institutional ownership. The other variables in this panel show a decrease like the variable ‘CEODuality’. Three years after the detection fraud firms have less Chief Executive Officers with a combined function as Chairman of the Board. The highest decrease in this panel is the variable ‘Big4’, where in the initial year 87,81% of the fraud firms had a Big-4-audit firm as their independent auditor, the final year shows a decrease of 4,88%, which leads to a percentage of 82,93.
Figure 2, Panel B, shows the changes of the variables of the initial year and the final year of fraud firms in levels. The only decreasing variable in this panel is the ‘AuditComMeet’. Fraud firms have more audit committee meetings in the year prior to the detection than three years after the detection. The variables ‘OutsideDir’ and ‘FinExpert’ have the greatest increase. Prior research also mentioned that fraud firms have less oustide directors in their board (Beasley, 1996; Chen, Firth, Gao, & Rui, 2006; Uzun, Szewczyk, & Varma, 2004). To confirm this conclusion we have compare fraud firms with control firms.
4.1.1 Comparing means
TABLE 3-COMPARISONS OF BOARD OF DIRECTORS CHARACTERISTICS
Variable Difference initial and final year
Firm Initial year Final Year Mean T-statistic Significance
outsideDir% Fraud Firm 0.699 0.777 -0.078 -3.212 0.003
Control Firm 0.723 0.775 -0.052 -2.828 0.007 T-statistic -0.691 0.650 Significance 0.492 0.949 #Matched pairs 41 41
#OutsideDir Fraud Firm 6.610 7.340 -0.732 -2.357 0.023
Control Firm 6.440 7.070 -0.634 -3.329 0.002 T-statistic 0.323 0.518 Significance 0.747 0.606 #Matched pairs 41 41
#Directors Fraud Firm 9.490 9.460 0.024 0.084 0.934
Control Firm 8.710 8.930 -0.220 -1.177 0.246 T-statistic 1.512 1.093 Significance 0.134 0.278 #Matched pairs 41 41
Table 3 (vertically) shows that there is no significant difference in the percentage of outside directors between fraud firms and control firms for the initial nor final year. The number of outside directors in the initial year is higher for fraud firms than control firms however, this is not a significant difference. The table does show that the mean of fraud firms for outside director percentage in the initial year is lower than the mean of control firms.
Table 3 (horizontally) shows also the difference in the percentage of outside directors between the initial year and the final year for the fraud firms and control firms separately. The table shows that the change in means of the percentage of outside directors between the initial year and the final year for fraud firms is 0.078, which is greater than the changes in means of the control firms, which is 0.052. Also, the last column also shows a greater significance for fraud firms than control firms. This result supports the expectation of hypothesis 1, since the expectation is that fraud firms will experience a greater increase in outside director percentage comparing the initial year and the final year with each other.
TABLE 4-COMPARISONS OF AUDIT COMMITTEE CHARACTERISTICS
Variable Difference in Years
Firm Initial year Final Year Mean T-statistic Significance
#AuditComMeet Fraud Firm 10.730 8.390 2.341 2.967 0.005
Control Firm 7.560 7.880 -0.317 -0.534 0.597 T-statistic 3.951 0.642 Significance 0.000 0.523 #Matched pairs 33 41
#AuditComMbrs Fraud Firm 3.850 3.880 -0.024 -0.154 0.878
Control Firm 3.490 3.710 -0.220 -2.038 0.048 T-statistic 1.754 0.969 Significance 0.083 0.335 #Matched pairs 41 41
#AuditComOut Fraud Firm 3.830 3.850 -0.024 -0.144 0.886
Control Firm 3.410 3.680 -0.268 -2.314 0.026 T-statistic 1.880 0.959 Significance 0.064 0.341 #Matched pairs 41 41
#FinExpert Fraud Firm 1.780 1.950 -0.171 -1.155 0.255
Control Firm 1.560 2.020 -0.463 -2.960 0.005 T-statistic 0.975 -0.261 Significance 0.332 0.795 #Matched pairs 41 41
Table 4, shows the comparisons of audit committee characteristics between fraud firms and control firms (vertically) as well the comparisons for each sort of firm in years (horizontally). The mean of the variable AuditComMeet is in the initial year higher for fraud firms compared to control firms, it is even a statistically significant difference. The final year of the variable AuditComMeet shows also a greater mean for fraud firms compared to control firms, but for the final year this difference is not statistically significant. For the variable AuditComMeet, fraud firms have a higher number of meetings prior to the detection (10.730) compared to three years after the detection (8.390). At the same time control firms have the opposite, a small increase comparing the initial year with the final year. This result does not support the expectation of hypothesis 3, since the expectation is that fraud firms experience a greater increase in audit committee meetings.
With the variable ‘AuditComMembers’ fraud firms have a greater mean than the control firms for both years. Also, the difference is statistically significant for the initial year between the firm types, but not for the final year. Although the mean of the number audit committee members is higher for fraud firms than for control firms, it does not have a greater increase in the number of audit committee members. The average AuditComMembers is between eight and eleven members. With the variable ‘AuditComOutsideDir’ fraud firms have a greater mean than the control firms for both years. Also, the difference is significant for the initial year between the firm types, but not for the final year. Although the mean of the number outside audit committee members is higher for fraud firms than for control firms, it does not have a greater increase in the number of outside audit committee members. Therefore, the result does not support the expectation of hypothesis 2, since the expectation is that fraud firms would have experienced a greater increase in AuditComOutsideDir than control firms. The changes for the control firms between the initial year and the final year are significant. The average AudtComOutsideDir is between three and four members.
For the variable ‘FinExpert’ fraud firms have a greater mean than the control firms for the initial year, although the difference is not enough to be statistically significant. For the final year control firms have a greater mean than fraud firms. This result does not support the expectation of hypothesis 4, since the expectation is that fraud firms would have experience a greater increase in financial experts than control firms. The changes for the control firms between the initial year and the final year are statistically significant. The average finanial experts in an audit committee are between one and two members.
TABLE 5-COMPARISONS OF OTHER GOVERNANCE CHARACTERISTICS
Variable Difference in Years
Firm Initial year Final Year Mean T-statistic Significance Big4% Fraud Firm
0.878 0.829 0.049 1.000 0.323 Control Firm 0.756 0.780 -0.024 -1.000 0.323 T-statistic 1.429 0.552 Significance 0.157 0.583 #Matched pairs 38 41
CEODuality Fraud Firm
0.415 0.390 0.024 0.298 0.767 Control Firm 0.561 0.610 -0.049 -0.703 0.486 T-statistic -1.324 -2.012 Significance 0.189 0.048 #Matched pairs 41 41
Block% Fraud Firm
0.277 0.304 -0.027 -0.738 0.465 Control Firm 0.303 0.315 -0.013 -0.509 0.614 T-statistic -0.494 -0.241 Significance 0.622 0.810 #Matched pairs 41 41
InstOwn% Fraud Firm
0.008 0.018 -0.010 -1.348 0.185 Control Firm 0.003 0.013 -0.010 -1.138 0.262 T-statistic 0.767 0.332 Significance 0.446 0.741 #Matched pairs 41 41
InsideOwn% Fraud Firm
0.161 0.160 0.002 0.089 0.930 Control Firm 0.176 0.196 -0.020 -0.517 0.610 T-statistic -0.944 -0.508 Significance 0.348 0.614 #Matched pairs 41 41
Table 5, shows the means of the remaining variables of this study. Of all the variables in this table there is only one difference that is significant, which is for the variable ‘CEODuality’. The significant change of this variable is only for the final year, where there is a difference between fraud and control firms that is significant. The rest of the variables do not show a change that is significant.
Fraud firms have a greater mean for the variable Big4% than control firms for both years. On the other hand, fraud firms have a small decrease in means comparing the initial year with the final year, while control firms have a small increase in means comparing the initial and final year with each other.
The variable ‘CEODuality’ shows in this table that control firms have more CEOs that also represent the function Chairman of the Board than fraud firms have, as well in the initial year as the final year. For the final year the differences between the fraud and control firm is significant. Comparing the initial year and the final year with each other, control firms have an increase in the combined function. Fraud firms on the other hand, have a decrease in the combined function comparing the initial year and final year with each other. This result supports the expectation of hypothesis 5, since the expectation is that a fraud firms experiences a greater decrease in CEODuality than control firms. The increase of the combined function, CEO representing also as the Chairman of the Board, at control firms can perhaps be explained that in a lot of proxy statements control firms were implying that they believe in a combined function.
Another variable in this table is blockholders, this table shows that fraud firms compared to control firms have a lower mean in both initial and final year. Even though fraud firms tend to have a greater increase in blockholders compared to control firms from the initial to the final year. Further the variable institutional ownership shows in this table that the increase of both fraud and control firms are the same amount. The table also shows that the average of fraud firms is greater than the average of control firms. Lastly, the variable inside ownership shows in this table that fraud firms have a small decrease on inside ownership between the initial and final year compared to control firms. On the other hand, control firms tend to have an increase on inside ownership between the initial and final year.
4.2 Correlations
A correlation analysis is performed to see if there is a relationship between the variables. Table 6 shows a correlation analysis between the independent variables and table 7 shows a correlation analysis between the control variables.
TABLE 6-INDEPENDENT VARIABLES
Correlations
82 observations
OutsidDir% #Directors ComMeet #Audit ComMbrs #Audit #AuditCom OutsideDir Experts #Fin
outsideDir% 1 #Directors 0.153 1 #AuditComMeet 0.03 0.149 1 #AuditComMbrs .358** .415** 0.131 1 #AuditComOut .408** .409** 0.156 .965** 1 #FinExpert 0.194 0.204 0.153 0.166 0.165 1
Table 6 shows that there is a significant strong positive relationship between the number of outside audit committee members and the total number of audit committee members. This can be explained that in the database of this study, where it rarely happened that audit committee members were not outside audit committee members.
Therefore it is logical that the two variables have a strong relationship with each other. Another observation that can be made is that there are some variables that do have a significant medium positive relationship with each other. It shows that if the number of audit committee member increases, the percentages of outside board members increases as well. The same observation can be made for the number of outside audit committee members. The table also shows that when the number of audit committee members increases, the total number of director increases. This seems understandable since the total number of directors also includes audit committee members. On the other hand, also observe that board of director members are joining less board committees, since the number of directors increases if the number of audit committee members increases.
TABLE 7-CONTROL VARIABLES
Correlations
82 observations
Big4% CEODUALITY Block% InstOwn% InsideOwn% Market Value
Big4% 1 CEODUALITY 0.02 1 Block% -0.177 -0.079 1 InstOwn% 0.09 0.087 0.058 1 InsideOwn% -.338** 0.03 .418** 0.054 1 Market Value 0.131 0.139 -0.248* -0.047 -0.164 1
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the level 0.05 level (2-tailed).
Table 7, shows that there are three significant correlations. Two of them have to do with inside ownership. It shows that the percentage of inside ownership is negatively correlated with the percentages of having a Big-4 as an independent auditor. This means that when the percentage of inside ownership increases, the percentages of having a Big-4 as an independent auditor decreases. Another significant correlation is between the percentages of ownership and the percentages of blockholders. The table shows that there is a positive relationship between the two variables, which means that if the percentage of ownership increases, the percentages of blockholders increases as well. Another significant correlation is between market value and the percentage of blockholders. Those two variables are negatively correlated with each other, which means that when the variable market value increases, the percentages of blockholders decreases.
4.3 Regressions
As explained in the previous chapter, we have used a logistic regression because the dependent variable is dichotomous. We use a binary logistic regression in order to investigate the relationship between a response variable and one or more explanatory variables. A binary logistic regression finds the probability that a value of the independent variable belongs to the category given by the dependent variable.
The exponential beta (Exp B) in the logistic regression represents the factor by which the odds of fraud occurring change when the independent variable increases by one unit. If the value of the exponential beta is below one, it indicates a negative correlation. If the value of the exponential beta is above one, it indicates a positive correlation. The p-value is based on the Wald test. The Wald test shows if there is a significant association between the variables that have been measured. If the variable is significant, it indicates that it improves the model. If the variable is not significant, it indicates that the variable does not improve the model.
Due to multicollinearity between the number of audit committee members and the number of outside audit committee members, we have performed two regression models. 4.3.1 Regression model with audit committee members
The first model of interest can be expressed by the following regression specification:
FraudFirmi = β0 + β1OutsideDiri + β2Boardsizei + β3AuditComMeeti + β4AuditComMbrsi + β5FinExpi +
β6Marketvaluei + β7Big4i + β8Blocki + β9InstOwni + β10CEODualityi + β11InsideOwni +εi
Table 8 represents the results of several regression models, where each regression model represents a different test. Model (1) represents the results of the independent variables only, where one variable, ‘AuditComMeet’ gives a positive significant change. This implies that for every additional meeting of the audit committee, the probability of being a fraud firm becomes more likely.
The other variables in model (1) do not have a significant impact on the likelihood of a firm being fraudulent. The coefficient of the variable ‘Boardsize’ shows that every additional member in the board of directors results in an increase in the probability that a firm is a fraud firm, but this coefficient is not significant. The coefficient of the variable ‘FinExpert’ shows that for every additional financial expert in the audit committee there is an increase in the probability that the firm is a fraud firm. For the variable ‘OutsideDir%’ the coefficient is negatively correlated but does not have a significant impact. The coefficient shows that for each additional
outside board member in percentages there is a decrease in the probability of being fraudulent. The last variable of this model is the variable ‘AuditComMbrs’, the coefficient of this variable is positively correlated, but does not show a significant impact. The coefficient shows that for every additional member in the audit committee it gives an increase in the probability that the firm is a fraud firm. The Pseudo R2 of model (1) shows that the model accounts for 30.9% of the
variability in being a fraud firm.
TABLE 8-INCLUDING AUDIT COMMITTEE MEMBERS -EXCLUDING OUTSIDE AUDIT COMMITTEE MEMBERS
Model (2) consists of all the control variables and also the independent variable, ‘AuditComMbrs’. In this model none of the coefficients of the variables have a significant effect on the dependent variable. The coefficients of the variables ‘AuditComMbrs’, ‘Big4’, and ‘InstOwn’ have a positive correlation with the dependent variable. The coefficient of the variable ‘InstOwn’ shows a large number. This may be explained by the fact that a lot of companies do not have institutional ownership, so if the variable increases the impact is large, nevertheless the coefficient does not show a significant effect.
The variable ‘Marketvalue’ has no change in odds. This indicates that the variable ‘Marketvalue’ has zero correlation with the dependent variable. The coefficients of the variables
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
INDEPENDENT EXP B P-value EXP B P-value EXP B P-value EXP B P-value EXP B P-value EXP B P-value EXP B P-value
OutsideDir% 0.240 0.560 0.018* 0.075 0.039 0.141 0.022* 0.091 0.018* 0.075 0.041 0.148 Boardsize 1.074 0.544 1.087 0.515 1.080 0.541 1.131 0.366 1.087 0.516 1.111 0.425 AuditComMeet 1.349*** 0.001 1.389*** 0.001 0.793 0.572 1.413*** 0.001 1.389*** 0.001 0.860 0.722 AuditComMeet2 1.032 0.179 1.028 0.245 AuditComMbrs 1.646 0.119 1.523 0.115 1.567 0.176 1.597 0.166 0.110 0.374 1.565 0.188 0.194 0.525 AuditComMbrs2 1.363 0.290 1.279 0.416 Fin Expert 1.112 0.687 1.100 0.733 1.142 0.648 1.056 0.850 1.086 0.931 1.047 0.964 Fin Expert2 1.003 0.989 1.012 0.956 CONTROL Marketvalue 1.000 0.447 1.000 0.195 1.000 0.169 1.000 0.167 1.000 0.197 1.000 0.157 Big4 1.876 0.349 1.130 0.878 1.383 0.697 1.585 0.600 1.131 0.878 1.767 0.530 Block% 0.709 0.766 0.550 0.657 0.656 0.756 0.887 0.933 0.549 0.657 0.540 0.303 InstOwn% 3617. 0.338 138.722 0.699 235.927 0.667 19.334 0.814 139.928 0.699 49.052 0.758 CEOduality 0.514 0.161 0.564 0.285 0.538 0.254 0.570 0.297 0.562 0.325 0.930 0.959 InsideOwn% 0.613 0.774 0.385 0.681 0.641 0.848 0.274 0.583 0.383 0.684 0.464 0.751 OTHER Constant 0.072* 0.087 0.209 0.188 0.126 0.312 0.496 0.752 9.926 0.617 0.129 0.399 14.36 0.588 Pseudo R2 0.309 0.131 0.368 0.391 0.382 0.368 0.399 Observations 82 82 82 82 82 82 82
