Changes in corporate governance after changes in accounting
quality
Name: Erik van de Burgt Student number: 11369620
Thesis supervisor: Dr. Réka Felleg Date: 25 June 2017
Word count: 14088
MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Erik van de Burgt 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.
Acknowledgements
I would like to thank my supervisor Réka Felleg for guiding me in the right direction with helpful feedback during this thesis process.
Abstract
This study examines whether corporate governance changes as a result of changes in accounting quality. When accounting quality changes, the demand of investors and other users of the financial statements for useful information and monitoring activities change. Accounting quality is determined by calculating the annual change in discretionary accruals. Using board size, board independence, CEO duality, the presence of an audit committee and the presence of a corporate governance committee as proxies for corporate governance, I find that firms are more like to install a corporate governance committee as a result of increased discretionary accruals. Furthermore, board size and CEO duality decrease as a result of decreased discretionary accruals. However, firms are also more likely to install a corporate governance committee as a result of decreased discretionary accruals. These results imply that changes discretionary accruals and thus in accounting quality are not causing major changes in corporate governance.
Keywords: Corporate governance, Accounting quality, Agency theory, Information asymmetry, Discretionary accruals
Contents 1 Introduction ... 7 2 Literature review ... 10 2.1 Agency Theory ... 10 2.2 Accounting quality ... 10 2.3 Corporate governance ... 11 3 Hypothesis development ... 14 4 Research methodology ... 15 4.1 Sample ... 15 4.2 Empirical model ... 15
4.2.1 Dependent variable: Board size ... 16
4.2.2 Dependent variable: Board independence ... 16
4.2.3 Dependent variable: CEO duality ... 17
4.2.4 Dependent variable: Audit committee ... 17
4.2.5 Dependent variable: Corporate governance committee... 18
4.2.6 Independent variable: Change in accounting quality ... 18
4.2.7 Independent variable: ∆𝐷𝐴𝐶𝐶 dummy ... 20
4.2.8 Control variables ... 20 4.2.9 Test specifications ... 21 5 Empirical findings ... 22 5.1 Descriptive statistics ... 22 5.2 Regression results ... 26 5.2.1 Hypothesis 1 ... 26 5.2.2 Hypothesis 2 ... 28 5.3 Robustness test... 30
References ... 37 Appendix ... 41
1 Introduction
According to the agency theory, information asymmetry between the principal and the agent arises because the agent (management) has more information about the firm than the principal (Jensen & Meckling, 1979). Solving this problem of asymmetric information between firms and investors and other users of the financial statements can be accomplished by providing high quality accounting information (Healy & Palepu, 2001). To make good investments decisions investors and other users of the financial statements demand high accounting quality (Biddle et al., 2009). In other words, investors require high accounting quality which will decrease the information asymmetry to make good investment decisions. High accounting quality is accomplished through providing financial statement information that give a good overall representation of the economic conditions and is useful for investors to make good investment decisions (Chen et al., 2010). Corporate governance practices are proven to have a positive effect on accounting quality. The existence of corporate governance is essential for improving the quality of financial reporting (Beasley, 1996; Dechow et al., 1996; Carcello & Neal, 2000). Further, Johnson et al. (2002) uncover that governance practices, which mitigate agency problems by aligning incentives and reducing information asymmetry, are expected to improve the quality of financial reporting. A set of well working corporate governance mechanisms could be implemented to ensure that the quality of the financial information and the financial reporting itself will improve. Moreover, corporate governance mechanisms support the audit procedures and the audit operations within a firm (Larcker & Richardson, 2004; Beasley et al., 2009; Pergola & Joseph, 2011; Zaman et al., 2011). Furthermore, information asymmetry is reduced between the firm, investors and other users of the financial statements through the increased ability of investors to monitor managerial incentives and decisions (Biddle et al., 2009). Therefore, corporate governance is an effective way to mitigate for information asymmetry (Armstrong et al., 2010). In sum, corporate governance has a mitigating effect on information asymmetry and is able to align incentives and therefore results in an improved accounting quality.
While in prior literature the effect of corporate governance on accounting quality is clearly examined, in this study I argue for reverse causality between corporate governance and accounting quality. I examine whether the extent of corporate governance changes as a result of accounting quality changes. More specifically, I analyze how several corporate governance mechanisms change as a result of changes in accounting quality. I expect that when accounting quality decreases, the information asymmetry increases which result in an increased demand from investors and other users of the financial statements to increase the monitoring activities and therefore corporate governance will increase. Furthermore, I expect that when accounting
quality increases, information asymmetry decreases and as a result there will be less demand from investors and other users of the financial statements for extra monitoring activities in terms of corporate governance.
I test my hypotheses, using a sample of 13,272 firm-year observations from the U.S. for the years 1996-2006. The change of discretionary accruals as proxy for changes in accounting quality is utilized in this study. Furthermore, five different proxies for corporate governance: board size, board independence, CEO duality, the presence of an audit committee and the presence of a corporate governance committee are used. The findings provide no evidence that board size, board independence, CEO duality and the presence of an audit committee increase when discretionary accruals increase and thus accounting quality decreases. This means that increased discretionary accruals have no effect on these corporate governance mechanisms. This finding can be explained that investors and other users of the financial statements do not react to changes in information asymmetry as strongly as I was expecting, or not react at all. The presence of an audit committee is explained by the fact that it is obligatory by U.S. law since the introduction of SOX in 2002 (Defond & Francis, 2005). However, I do find a positive significant association between an increase in discretionary accruals, thus a decrease in accounting quality and the presence of a corporate governance committee. Indicating that a an increase in discretionary accruals leads to an increase in corporate governance. This can be explained that investors demand more monitoring activities after a decrease in accounting quality to make good investment decisions (Biddle et al., 2009). Second, I find a positive significant association that board size and CEO duality decrease after a decrease in discretionary accruals and thus an increase in accounting quality. Furthermore, I find no evidence that board independence and the presence of an audit committee decrease when discretionary accruals decrease and thus accounting quality increases. This can be explained by the asymmetric effect of cost stickiness of corporate governance mechanisms. An interesting finding is that the presence of a corporate governance committee increases when discretionary accruals decrease and thus accounting quality increases. This implies that corporate governance increases as a result of increased accounting quality. This could be explained because there is not much variation in the data according the presence of a corporate governance committee and it is a relative new mechanism. Third, a robustness test is constructed to test whether there were changes in the coefficients of accounting quality before and after the introduction of the Sarbanes-Oxley act in 2002. I find that within this study the introduction of SOX does not have an effect on the measured changes in accounting quality using discretionary accruals. However, SOX could explain that managers
introduction of SOX or that the use of real earnings management cannot be detected and therefore no effect is shown.
This study contributes to prior literature by examining the reverse causality between corporate governance and accounting quality which gives more insight in demand of investors and other users of the financial statements as a reaction on accounting quality changes. Previous literature has only examined the effect of corporate governance on accounting quality (Bushman & Smith, 2001; Byard et al., 2006; Armstrong et al., 2010). Furthermore, this study adds to existing literature new information about the cost stickiness of corporate governance mechanisms. Corporate governance mechanisms are installed within firms for the long term. Discretionary accruals are short-term volatile measures which are not affecting changes in corporate governance that strong, even though investors and other users of the financial statements demand it. Furthermore, this study contributes to existing literature by adding a new insight in the reaction of investors on changes in accounting quality. This study implies that investors and other users of the financial statements do not demand extra monitoring after a decrease in accounting quality. Also, I contribute to existing literature by creating a new empirical model to examine changes in corporate governance. This model is determining the changes in corporate governance. The model shows that some corporate governance mechanisms are affected by changes discretionary accruals and thus accounting quality.
The remainder of the paper proceeds as follows. In section 2, I provide an overview of the related literature and theory. Section 3 contains the development of the hypotheses. In section 4, I introduce the data sample, the selection procedures, and the empirical model. Section 5 contains the empirical findings of this research. Section 6 concludes with contributions in relation to previous research.
2 Literature review
2.1 Agency TheoryThe underlying theory that is used in this research is the agency theory. Within this study the agency theory is applied for the relationship between the users of financial statements of a firm (principals) and management of the firm (agents). The agency theory has an important assumption that everyone wants to maximize their own interest (Jensen & Meckling, 1979). However, agents should not be driven by self-interest but by the interest of the principal. Information asymmetry between the principal and the agent arises because the agent (management) has more information about the firm than the principal (Jensen & Meckling, 1979). Consequently, information asymmetry causes two problems: moral hazard and adverse selection (Pauly, 1974). Moral hazard possibly arises between management and current investors. Hereby one party is willing to take more risks because this party knows that he is not accountable for the costs of those risks. Therefore, investors and other capital providers want to monitor management to act in their best interest (Pauly, 1974). On the other hand there is adverse selection. Adverse selection occurs between potential investors and a company when potential investors have to decide whether to invest in the company. Within this relationship one of the two parties have more information than the other regarding the product that is sold (Pauly, 1974). The information asymmetry causes incentive problems for management to not provide the financial information that investors need or demand. High accounting quality can mitigated these information asymmetry problems (Healy & Palepu, 2001).
By disclosing useful information to investors and other users of the financial statements the problems that occur with information asymmetry can be mitigated (Healy & Palepu, 2001). Investors demand high accounting quality due to disclosure of useful information. Regulated and monitored financial statement information provide relevant and useful information to investors (Healy & Palepu, 2001).
2.2 Accounting quality
There is not an unanimous definition or description of what accounting quality means even though it has been extensively examined in prior literature (Cohen et al., 2004). According to the International Accounting Standards Board (IASB) accounting information should have at least the following characteristics to be useful for investors. First, the information should be relevant
also be understandable for the users as well as comparable and consistent. Third, it should be free from material misstatement and it should be presented in a faithful manner. Fourth, the information should be material and useful (Iatridis, 2011). Moreover, Chen et al. (2010) define accounting quality as in what way the information from the financial statements give a good overall representation of the economic conditions. In other words, to what extent the accounting information is useful for users to make decisions. In some cases it is difficult to observe the economic conditions and therefore accounting quality can be determined in terms of timely loss recognition, earnings management and value relevance (Barth et al., 2008; Paananen et al., 2008; Chen et al., 2010). Firms disclose financial and non-financial information to investors and other users of the financial statements. These disclosures are a signal of high accounting quality and contain confirmable information about the financial conditions of the firm such as, losses and investment projects which are in most cases in poor projects (Iatridis, 2011). Consequently, less verifiable information is a more easy catch for manipulation and therefore unfavorable and less valuable for the investors who use this information (Iatridis, 2011).
Information asymmetry influences the overall representation of the underlying economic condition due to problems as adverse selection and moral hazard (Healy & Palepu, 2001). In other words, information asymmetry affects the usefulness of information disclosed to investors and other users of the financial statements and therefore also the accounting quality (Healy & Palepu, 2001). The investor is not provided with the needed and demanded information. Useful accounting practices and higher quality accounting information will reduce the information asymmetry between the principal and the agent (Healy & Palepu, 2001; Biddle & Hilary, 2006). Next to good accounting practices, corporate governance systems can also mitigate the information asymmetry problems and therefore also influence the good overall representation of the financial statement information and as a result increase accounting quality (Armstrong et al., 2014).
In sum, high accounting quality mitigate information asymmetry problems (Healy & Palepu, 2001). High accounting quality is demanded by investors and other users of the financial statements (Biddle et al., 2009). Accounting quality can be increased through effective monitoring in terms of corporate governance mechanisms (Armstrong et al., 2014).
2.3 Corporate governance
According to Lin and Hwang (2010) corporate governance is a broadly defined concept. Lin and Hwang (2010) state that corporate governance mechanisms are all principles by which the firm is monitored and managed. Other prior literature define corporate governance in numerous ways.
The reason why corporate governance is defined in multiple ways is because this concept has two different functions, an internal and an external function. The internal function, is seen as a corporate governance system that is mainly focused on the internal mechanisms within the company such as the control procedures and the internal system of law that influences the operations within the company (Gillan & Starks, 1998). Specifically, the internal function is focused on how successful the company is controlling and managing their internal operations. This set of internal corporate governance mechanisms provide certain boundaries within the firm about who are involved in certain operations in the company (Gillan & Starks, 2003). For example, internal control mechanisms are the board size within an organization, the structure of the board of directors or internal auditors (Beasley et al., 2000). The external function is seen as corporate governance mechanisms that ensure that the company is acting and performing in order to make sure that the investors are getting their returns on investment (Shleifer & Vishny, 1997). In other words, the external corporate governance is focused more on the interests of the external users, in many cases the investors. External corporate governance mechanisms do provide mitigating effects for problems that arise within financial reporting. For example, external corporate governance mechanisms can be government regulations or the market for corporate control which is used when internal corporate governance mechanisms have failed (Daily et al., 2003).
Well working corporate governance systems can mitigate agency problems, especially problems regarding information asymmetry (Lin & Hwang, 2010; Ashbaugh et al., 2003; Jensen & Meckling, 1976). Specifically, it focuses on how corporate governance can ensure managers to act in the principals’ interest and prevent them to engage in earnings management. Corporate governance looks for instance at structures of the board or at the presence of an audit committee to make sure that the incentives between the principal and the agent are aligned and that earnings management cannot be accomplished (Klein, 2002; Xie et al., 2002; Bedard et al., 2004). Consequently, Johnson et al. (2002) find that good governance practices are expected to improve the quality of financial reporting due to the fact that good corporate governance practices are mitigating the agency problems and information asymmetry. Moreover, a set of well working corporate governance mechanisms make sure that the quality of financial information and financial reporting itself will improve by supporting the audit procedures and the audit operations within a firm. So the use of effective corporate governance can increase the quality of internal controls and audits which will result in an increased accounting quality. (Larcker & Richardson, 2004; Beasley et al., 2009; Pergola & Joseph, 2011; Zaman et al., 2011).
To make investment decisions, investors require and demand useful and high quality accounting information (Biddle et al., 2009). Because of high accounting quality information asymmetry reduces between the company and the external investors. High accounting quality could be achieved by an increased ability of investors to monitor managerial incentives and decisions (Biddle et al., 2009). Therefore, corporate governance is an effective way to mitigate for information asymmetry and managerial decisions (Armstrong et al., 2010).
In sum, corporate governance is an effective way of mitigating agency problems and information asymmetry problems, which can result in a higher accounting quality (Armstrong et al., 2010).
3 Hypothesis development
Discretion enables managers to act in their own interest which leads to information asymmetry and earnings management (Carlin & Finch, 2010). The increase in information asymmetry leads to incentive problems for managers to not provide the information that investors and users demand which results in information that lowers accounting quality (Healy & Palepu, 2001). Investors and other users of the financial statements demand useful and high quality accounting information to make investment decisions (Biddle et al., 2009). By monitoring the incentives and decisions of management it is possible to increase accounting quality and as a result reduce information asymmetry (Biddle et al., 2009). Corporate governance is an effective way to ensure high accounting quality and to monitor information asymmetry and managerial decisions (Armstrong et al., 2010). When accounting quality is low information asymmetry is high, which results in more demand for monitoring activities in terms of corporate governance by investors and users. This is in line with the findings of Bushman et al. (2004) that limited transparency and an increased information gap between managers and shareholders increases the demand for corporate governance systems. Therefore, the first hypothesis is:
H1: The level of corporate governance increases after accounting quality decreases.
High accounting quality causes a decrease in information asymmetry between firms and investors and other users of the financial statements (Healy & Palepu, 2001; Armstrong et al., 2010). A decreased information asymmetry mitigates incentive problems so that managers have to provide the information that investors and other users of the financial statements demand (Healy & Palepu, 2001). When accounting quality increases and as a result information asymmetry decreases, investors and other users of the financial statements are provided with more useful information and therefore are able to make better investment decisions (Biddle et al., 2009). When investors and other users of the financial statements are provided with useful and high quality information to make decisions, there is no demand for extra monitoring management in terms of corporate governance. Therefore, the second hypothesis is:
4 Research methodology
4.1 SampleI use the Institutional Shareholder Services (ISS) database to obtain corporate governance data of U.S. firms, issued for the years 1996 to 2006. Year 1996 is taken as starting point because that is the first year available in the Directors legacy database. Year 2006 is the last year of observation because this is the last year available in the Directors legacy database. To obtain data for the calculation of the discretionary accruals I use the Compustat database. Data for all firms is collected for the years 1995-2006. Furthermore, I collect data for the year 1995 because for the calculation of the change in discretionary accruals one prior year has to be included. This results in an initial sample of 102,160 firm-year observations. I use annual data because financial information is reported at the end of each fiscal year and that is the most important date of reporting (Oyer, 1998).
The sample size is reduced due to missing values for corporate governance. This reduces the sample size by 87,701 firm-year observations. Furthermore, there were missing values for total assets and the calculation of the discretionary accruals. This reduces the sample size by 4,188 firm-year observations. Consequently, the final sample contains 13,272 firm-year observations. Table 1 summarizes the sample selection.
[Insert table 1 about here]
4.2 Empirical model
My dependent variable is corporate governance. Within this study I aim to examine whether the level of corporate governance changes as a result of changes in accounting quality. I expect that after a decrease in accounting quality, corporate governance increases and once accounting quality increases I expect a decrease in accounting quality. Within prior literature corporate governance is used as the variable of interest and not as a dependent variable. Therefore, prior literature does not provide methodology where corporate governance is the dependent variable. As a result, I rely on the model of Arena and Braga-Alves (2013). Arena and Braga-Alves (2013) use a model to determine corporate governance variables, for example board size or the proportion of independent directors. This regression uses firm-level characteristics, firm fixed effects, year fixed effects and fixed effects for the CEO to determine the dependent variable which is one of the corporate governance variables. For this study, I modified the model, by adding an accounting quality variable, to test my hypotheses whether the level of corporate governance changes as a result of changes in accounting quality as follows:
𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑖𝑡 = 𝛽0+ 𝛽1∆𝐷𝐴𝐶𝐶𝑖𝑡+ 𝛽2 ∆𝐷𝐴𝐶𝐶𝑑𝑢𝑚𝑚𝑦𝑖𝑡+ 𝛽3∆𝐷𝐴𝐶𝐶𝑖𝑡∗ ∆𝐷𝐴𝐶𝐶𝑑𝑢𝑚𝑚𝑦𝑖𝑡 + 𝛽4𝐶𝑉𝑠𝑖𝑧𝑒𝑖𝑡+ 𝛽5𝐶𝑉𝐿𝑂𝑆𝑆𝑖𝑡+ 𝛽6𝐶𝑉𝐿𝐸𝑉𝑖𝑡+ 𝑌𝑒𝑎𝑟𝑑𝑢𝑚𝑚𝑦
+ 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑑𝑢𝑚𝑚𝑦 + 𝐶𝑜𝑢𝑛𝑡𝑟𝑦𝑑𝑢𝑚𝑚𝑦 + 𝜀𝑖𝑡 (𝟏)
Where 𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑖𝑡 represents one of the five different proxies for corporate governance:
Boardize, board independence, CEO duality, the presence of an audit committee or the presence of a corporate governance committee. The independent variable that is used to test the hypotheses is the change in discretionary accruals referred to as ∆𝐷𝐴𝐶𝐶𝑖𝑡. The variables CVSize, CVLoss, CVLeverage, Yeardummy and the Industrydummy are the control variables that I expect to affect the extent of corporate governance and are included in the model.
4.2.1 Dependent variable: Board size
Within this study the board size is the number of members that is in the board of directors. Prior literature is mixed about whether small or large board sizes is leads to more benefits for the company and its stakeholders. Most prior studies find that small board have a more positive effect on firm performance compared to large boards (Yermack, 1996 and Eisenberg et al., 1998). Problems that arise due to larger boards are free riding and social loafing. Free riding increases as the board becomes larger. As a result, large board sizes reduces effectiveness and efficiency of the board (Lipton & Lorsch, 1992). Moreover, smaller boards are more efficient in decision making compared to larger boards. This is because smaller boards have better coordination and communication amongst the board (Jensen, 1993). Based on these findings, I also assume that smaller boards provide a higher accounting quality than larger boards. In this study, I assume that small boards are boards with less than 12 board members.
𝐵𝑂𝐴𝑅𝐷𝑆𝐼𝑍𝐸𝑖𝑡 1 if the number of board members in the board of directors is
below 12 in year t for each firm i in year t, 0 otherwise.
4.2.2 Dependent variable: Board independence
In this study, board independence is the number of outside directors within the board of directors. This is in line with prior literature (Beasley, 1996). According to Klein (2002), the board becomes more independent when the number of outside directors relatively to other directors increases. Furthermore, Klein (2002) finds that firms which decreased the percentage of outside board members experienced an increase in discretionary accruals. Osma (2008) find that
boards may lack incentives to monitors managerial decisions which could result in decreased accounting quality. Since investors and other users of the financial statements require a high accounting quality to make decisions (Biddle et al., 2009), I expect that investors and other users of the financial statements demand the board of directors to be independent.
𝐵𝑂𝐴𝑅𝐷𝐼𝑁𝐷𝑖𝑡 The percentage of independent board members for each firm i in year t.
4.2.3 Dependent variable: CEO duality
The effectiveness of the board of directors relies for a significant part on the position of the CEO (John & Senbet, 1998). CEO duality is an important corporate governance mechanism. With CEO duality, the CEO of a firm is also the chairman of the board of directors (Rechner & Dalton, 1991). The CEO has in that situation combined power, which can lead to behavior of self-interest (Jensen & Meckling, 1976). According to Rechner and Dalton (1991), CEO duality has a negative influence on firm and organizational performance and it can result in a conflict of interest. More specifically, the main issue is the classic tradeoff between the focus and control by the CEO and in this cases also the chairman of the board and the responsibility that the board of directors have to provide independent corporate governance (Rechner & Dalton, 1991). Zhang (2008) finds that in case of CEO duality, the CEO builds up a strong leadership within the firm but that decreases the monitoring role of the board of directors. Therefore, a conflict of interest can arise and as a result an increase in information asymmetry. In addition, an outside chairman will protect his own reputation as manager more and therefore it is more likely that the outside chairman will discharge managers when they perform badly and gives the chairman more incentives to monitor strictly (Fama & Jensen, 1983; Johnson et al., 1993; Warner, Watts, & Wruck, 1988). This increase in information asymmetry results in a decrease of accounting quality (Armstrong et al., 2014). To provide independent monitoring and to maintain accounting quality it is important to have an outside chairman (Fama & Jensen, 1983; Johnson et al., 1993).
𝐶𝐸𝑂𝐷𝑈𝐴𝐿𝐼𝑇𝑌𝑖𝑡 1 if the chairman of the board of directors of each firm is
someone else than the CEO in year t for each firm i in year t, 0 otherwise.
4.2.4 Dependent variable: Audit committee
Next to the board of directors and supervisory board there are other committees that are responsible for monitoring important decisions and processes that are made by the board of directors (Klein, 2002). An audit committee is one of those committees and is responsible for
supervision and monitoring of a company’s financial reporting process. It meets on a regular basis with the external auditors of the firm and also with the internal financial auditors and managers. The audit committee reviews next to the financial statements, the whole audit process and the internal auditing controls (Klein, 2002). In addition, Kanagaretnam et al. (2007) find that the presence of an audit committee will result in a decrease of earnings management and an increase corporate disclosure quality as a result of the decrease in information asymmetry when firms have an effective audit committee. Consequently, a good working and effective audit committee can result in a decrease of information asymmetry, a decrease in agency costs and in higher accounting quality (Li et al., 2008).
𝐴𝑈𝐷𝐼𝑇𝐶𝑂𝑀𝑖𝑡 1 if the firm has an audit committee in year t for each firm i in
year t, 0 otherwise.
4.2.5 Dependent variable: Corporate governance committee
Corporate governance committee’s role is to enable boards of directors to perform their control role effectively. Furthermore, the presence of the a corporate governance committee is negatively associated with the number of inside directors within the board. This implies that corporate governance committees have an positive effect on the board independence (Ruigrok et al., 2006). As a result of increased board independence, accounting quality will increase (Osma, 2008). 𝐶𝐺𝐶𝑂𝑀𝑖𝑡 1 if the firm has a corporate governance committee in year t for
each firm i in year t, 0 otherwise.
4.2.6 Independent variable: Change in accounting quality
In this study, I use discretionary accruals as a proxy for accounting quality. As described above, accounting quality can be measured in terms of earnings management. Accruals is one of several ways to manage earnings (Barth et al., 2008; Paananen et al., 2008; Chen et al., 2010). Accruals can be used for the recognition of cash flows over time in a way that accruals can adjust or shift this recognition. Consequently, the adjusted financial numbers, and in a lot of cases earnings, measure the firm performance better (Bissessur, 2008). Managers have to estimate and make some assumptions to predict and to measure these cash flows of the future period in the current period. These estimations about the future cash flows are very uncertain and will cause errors, therefore it will lead to noise. As a result, when the errors will become of greater impact, the lower the quality of the financial information in terms of accruals will be. Therefore will be a poor reflection of firm performance (Dechow & Dichev, 2002). For determining the accruals,
discretionary accruals are the types where the managers don’t have any form of discretion. On the other hand, there are the discretionary accruals which are possibly used by managers when they are engaged in earnings management.
In prior studies, the most common model that is used for determining discretionary accruals is the cross-sectional modified Jones model. This model is determining the discretionary accruals as a result of total accruals. In this model, the discretionary accruals together with the non-discretionary accruals form the total accruals. It is important to separate the discretionary accruals form the non-discretionary accruals because normal business conditions can’t be influenced by managers. By using this model, it can be detected if firms are engaged in manipulating earnings by using discretionary accruals and therefore affect the accounting quality.
The model that I used for this study is the modified Jones model as described by Kothari et al. (2005). This model calculates the amount of discretionary accruals as follows:
𝑇𝐴𝑖𝑡 = 𝛼0+ 𝛼1( 1
𝐴𝑆𝑆𝐸𝑇𝑆𝑖𝑡−1) + 𝛼2(∆𝑅𝐸𝑉 − ∆𝑅𝐸𝐶)𝑖𝑡+ α3PPEit+ εit (2)
Where:
𝑇𝐴𝑖𝑡 Total accruals change in non-cash current assets minus the change in current liabilities excluding the current portion of long-term debt, minus depreciation and amortization, scaled by lagged total assets.
𝐴𝑆𝑆𝐸𝑇𝑆𝑖𝑡−1 Total assets for year t-1 for each firm i in year t.
∆𝑅𝐸𝑉𝑖𝑡 Change in revenue from prior year for each firm i in year t.
∆𝑅𝐸𝐶𝑖𝑡 Change in accounts receivables from prior year for each firm i in year t.
PPEit Gross property, plant and equipment for each firm i in year t. εit Error term, the residual of the estimated model and is interpreted
as discretionary accruals for each firm i in year t (𝐷𝐴𝐶𝐶𝑖𝑡 ). Total accruals (𝑇𝐴𝑖𝑡) are determined against the non-discretionary variables. This modified Jones
model determines the expected level of accruals, based on both firm and industry specific factors and is estimated for every industry group with at least 10firms in a given year. Within this study I use 10 firms per industry group which are based on 2-digit Standard Industrial Classification codes (SIC). I eliminated industry groups with include less than 10 firms to estimate discretionary accruals more adequately. The residual is the absolute value of discretionary accruals referred to
as 𝐷𝐴𝐶𝐶𝑖𝑡 . To test my hypotheses, that relies on the change of accounting quality, I’ve taken the change of 𝐷𝐴𝐶𝐶𝑖𝑡 , referred to as ∆𝐷𝐴𝐶𝐶𝑖𝑡. The ∆𝐷𝐴𝐶𝐶𝑖𝑡 reflects a change in the absolute
value of discretionary accruals in year t compared to year t-1. In this way the changes in accounting quality can be determined. If ∆𝐷𝐴𝐶𝐶𝑖𝑡 increases in year t compared to year t-1 it
indicates a decrease in accounting quality because this increases the information asymmetry between investors and the firm. On the other hand, if the ∆𝐷𝐴𝐶𝐶𝑖𝑡 decreases in year t compare
to year t-1 it indicates a increase in accounting quality because it reduces information asymmetry.
4.2.7 Independent variable: ∆𝐷𝐴𝐶𝐶 dummy
In this study, I include a dummy variable of the change in discretionary accruals referred in the model as ∆𝐷𝐴𝐶𝐶𝑑𝑢𝑚𝑚𝑦. This dummy is set up specifically to address the hypotheses. I expect different effects for positive and negative change in discretionary accruals. A positive change in discretionary accruals is a decrease in accounting quality and a negative change in discretionary accruals is an increase in accounting quality. This dummy captures the potential asymmetric effects of increased or decreased discretionary accruals.
∆𝐷𝐴𝐶𝐶𝑑𝑢𝑚𝑚𝑦𝑖𝑡 1 if the change in discretionary is positive in year t for each firm i in year t, 0 otherwise.
4.2.8 Control variables
Previous research has shown that different variables influence both the dependent variable and the main variables of interest in this research: corporate governance and accounting quality. These variables will be included in the empirical model as control variables, in order to ensure that these do not bias the results of this research.
First, I included size as control variable because the size of the firm could affect discretionary accruals as it affects earnings management (Frankel et al., 2002; Ashbaugh et al., 2003; Cahan et al., 2008; Kim et al., 2012). According to Becker et al. (1998) size also affects corporate governance. In addition, Myers et al. (2003) find that bigger firms could have more constant accruals. Within this research size is measured the natural logarithm of total assets. Second, previous literature expects that firm which reporting negative net income have increased incentives to engage in earnings management by using discretionary accruals (Frankel et al., 2002; Cahan et al., 2008), therefore loss is included as control variable. LOSS equals 1 if net income in year t for firm i is negative and 0 otherwise. Third, leverage is taken as control variable because there are mixed results about leverage has a positive or negative association with accrual-based
measured by long-term debt scaled by total assets. Time dummies are created to control for year effects. In addition, industry dummies are created to control for industry fixed effects.
4.2.9 Test specifications
For this study I use an OLS regression to test my hypotheses. To confirm my first hypothesis, I expect a positive significance on 𝛽1, the coefficient of the change of absolute values of the
discretionary accruals, in regression (1). This will be consistent with the expectation of H1 that the level of corporate governance increases after accounting quality decreases. To confirm my second hypothesis, I expect a positive significance on 𝛽3, the coefficient of the interaction term,
in regression (1). This will be consistent with the expectation of H2 that the level of corporate governance decreases after accounting quality increases.
The five regressions used in this study are ran robust to control for heteroscedasticity. Furthermore, the regressions are ran clustered for company code, Cusip.
5 Empirical findings
5.1 Descriptive statisticsThe descriptive statistics for all the dependent, independent and control variables used in this study for the main tests are provided in table 2.
The discretionary accruals (DACC) have a mean of 0.025. This shows that the average of the absolute value of discretionary accruals is positive. This is in line with Geiger and North (2006) who find a mean for discretionary accruals of 0.0133. The mean level for Delta_DACC is 0.004. This implies the change of discretionary accruals on average is positive and therefore accounting quality decreases. The mean for the board size dummy is 0.820. In this study, this means that in 82 percent of the total firm-year observations there is a board size with less than 12 board members. The mean of the real number of board size members is 9.197. This is in line with Sun et al. (2014) who also found 9 board members as their mean. The mean of board independence is 0.658. This indicates that around 65.8 percent of the board members are outside directors. This is in line with Xie et al. (2003). They find 67 percent of the board members are outside directors. The mean of the real number of independent board members is 5.925, which implies that within this sample there are on average around 6 independent board members in the board of directors. Within this study, the minimum amount of independent board members is 0 and the maximum amount of independent board members is 22. CEO duality has a mean of 0.857 which means that within this study in around 85.7 percent of the firm-year observations the CEO of the firm is not the chairman of the board of directors. This is not in line with the findings of Xie et al. (2003) who found that in 85 percent of their observations, there is CEO duality. The mean for audit committee is 0.820 which means that in 82 percent of the firm-year observations there is an audit committee appointed. This could be explained by the fact that after introduction of SOX in 2002, an audit committee is mandatory which means that after 2002 every firm within this sample should have an audit committee (Defond & Francis, 2005). The mean for corporate governance committee is 0.432 which means that in 43.2 percent of the total firm-year observations there is a corporate governance committee. This is much lower than the mean of the audit committee, which could be caused by the fact that a corporate governance committee is not compulsory (Defond & Francis, 2005).
[Insert table 2 about here]
same underlying construct, the correlation coefficient could be higher. None of the variables of interest and none of the control variables are correlated with Delta_DACC. This implies for both the corporate governance proxies and the control variables that there is no relationship with Delta_DACC. This can be interpreted that changes in discretionary accruals and thus accounting quality do not have an effect on each of the corporate governance individually.
Board size is not correlated with DACC and with Delta_DACC. This can be interpreted that when the board of the firm decreases it does not have an effect on DACC and on Delta_DACC. This is a surprising result because Yermack (1996) and Rahman & Ali (2006) found a significant relationship, i.e. when the board becomes bigger, the monitoring function becomes more ineffective and therefore discretionary accruals increase.
Board independence is positively correlated with DACC (0.03, p<0.01), but does not have a relationship with Delta_DACC. This can be interpreted that when the board has more independent board members discretionary accruals will increase. This is a surprising result as I expect that independent board members monitor more effectively and therefore DACC would decrease. Xie et al. (2003) find that the amount of outside directors within a board increases, discretionary accruals decrease. Board independence does not have an effect on board size. This implies that when the amount of independent board members increases, it has no effect on the size of the board. This could be explained that the new independent board member replaces the existing not dependent board member.
CEO duality does not have a relationship with DACC and with Delta_DACC. This implies that when the CEO and the chairman of the board of directors become separated it has no effect on discretionary accruals and the change of discretionary accruals. CEO duality is negatively correlated with board independence (-0.04, p<0.01). This can be interpreted that when the CEO and the chairman become separated, the amount of independent board members in the board decreases. This is a surprising result because independent board members are outside directors so when a CEO leaves the board, the relative amount of independent board members should increase.
Audit committee is positively correlated with DACC (0.057, p<0.01). This implies that firms that install an audit committee have higher discretionary accruals. This is surprising because when there is an audit committee there should be more monitoring and therefore less discretionary accruals. This is also not in line with Klein (2002) who finds that an audit committee decreases the discretionary accruals. Further, audit committee does not have a relationship with the change in discretionary accruals. There is a positive correlation between audit committee and board size (0.07, p<0.01) and between audit committee and board
independence (0.1537, p<0.01). This could be interpreted as when a firm installs an audit committee the size of the board increases and the board becomes more independent which is not a logical result. An audit committee is a separate committee to monitor the board of directors, therefore the board of directors cannot increase when an audit committee is installed (Defond & Francis, 2005). Furthermore, audit committee is negatively correlated with CEO duality (-0.09, p<0.01).
Corporate governance committee is positively correlated with DACC (0.057, p<0.01). This implies that firms that install a corporate governance committee have higher discretionary accruals. This is surprising because when there is a corporate governance committee there should be more monitoring and therefore less discretionary accruals. Corporate governance committee does not have a relationship with the change in discretionary accruals. Corporate governance committee is negatively correlated with board size (-0.05, p<0.01). This indicates that when a firm installs a corporate governance committee the size of the board is lower. Corporate governance committee is positively correlated with board independence (0.434, p<0.01). This can be interpreted that when a firm installs a corporate governance committee the board becomes more independent. Which is a logical result since a corporate governance committee monitors the board which increases the independency of the board. Corporate governance committee is negatively correlated with CEO duality (-0.04, p<0.01). Corporate governance committee is positively correlated with audit committee (0.4, p<0.01). This can be interpreted that when a firm installs a corporate governance committee the firm is also more likely to install an audit committee. This can be explained due to that firms are more likely to install both committees to ensure a certain level of monitoring.
CVSize is positively correlated with DACC. This implies that bigger firms have more discretionary accruals. This is in line with the findings of Frankel et al., (2002) and Cahan et al., (2008). CVSize does not have a relationship with Delta_DACC. This implies that the size of a firm does not have an effect on the change in discretionary accruals. CVSize is negatively correlated with board size (-0.413, p<0.01). This can be interpreted that bigger firms have smaller boards. This not in line with the findings of Eisenberg et al. (1998). They find that bigger firms have bigger board of directors. An explanation could be that smaller boards are more effective than larger board (Yermack, 1996). So bigger firms tend to decrease their board size to become more effective in terms of monitoring. In addition, when firms become bigger they have no incentive to increase their board size. Furthermore, CVSize is positively correlated with board independence (0.180, p<0.01). This can be interpreted that bigger firms have more independent
more demand for independence. CVSize is positively correlated with corporate governance committee (0.258, p<0.01). This implies that bigger firms have corporate governance committees installed. This can also be explained that bigger firms serve stakeholders and society, so there is more demand for independence.
CVLoss is negatively correlated with DACC (-0.199, p<0.01) and CVLoss is negatively correlated with Delta_DACC (-0.078, p<0.01). This indicates that firms which have losses are have less discretionary accruals and there is also a smaller change in discretionary accruals. This is a surprising result because firms which make losses are more likely to have higher discretionary accruals, because then firms can let investors believer that they are performing properly (Frankel et al., 2002; Cahan et al., 2008). Furthermore, CVLoss is negatively correlated with CVSize (-0.164, p<0.01). This implies that firms with losses are smaller firms. This could be explained by that smaller firms could be in the starting phase and less stable and have to make more investments which results in a negative net income and thus a loss.
CVLeverage is negatively correlated with DACC (-0.02, p<0.01). This implies that firms which have more debt have less discretionary accruals. CVLeverage has no relationship with Delta_DACC, which implies that firms which have more debt do not affect the change in discretionary accruals. Prior research also shows that leverage could have either a positive or a negative effect on earnings management and discretionary accruals (Frankel et al., 2002; Ashbaugh et al., 2003). Furthermore, CVLeverage is positively correlated with CVSize (0.199, p<0.01) and CVLeverage is positively correlated with CVLoss (0.140, p<0.01). This can be interpreted that firms which have a higher debt to equity ratio are bigger firms and are firms which have more losses.
The correlation between my dependent corporate governance proxies and the specifications of my independent variable are all very low. Additionally, only for two corporate governance proxies the coefficient is in the right direction. However, since the correlation analyses only provide univariate evidence, I rely on the multivariate regressions in the next section of this study.
[Insert table 3 about here]
In addition to the correlation analysis, I tested the variance inflation factor (VIF). This test contains a multivariate analysis to check whether high correlation coefficients will lead to multicollinearity in the regression used. In other words, the VIF tests whether a variable of interest can be a linear combination of other variables of interest. A VIF higher than 10 implies
such a linear combination. Within this study no variable has a VIF higher than 3.13. As a result, multicollinearity is not an issue within this study.
5.2 Regression results
To test the hypotheses the regression model is tested with five corporate governance proxies: board size, board independence, CEO duality, audit committee and corporate governance committee as dependent variable. The results of the regressions are shown in table 4. The first analysis has an adjusted R-squared of 0.215, the second 0.257, the third 0.016, the fourth 0.978 and the fifth 0.505. These numbers imply that respectively 21.5%, 25.7%, 1.6%, 97.8% and 50.5 % of the variation in corporate governance variables used in this study is explained by these particular models. For board size, board independence and the presence of a corporate governance committee the adjusted R-squared represents a good model fit. CEO duality has low adjusted R-squared which implies that a lot of other not defined factors are explaining the model. The adjusted R-squared for the presence of an audit committee is surprisingly high which implies that the model explains almost 98%.
As described earlier, the model that I use is not used in this way in prior literature. In prior literature corporate governance is used as a variable of interest and not as the dependent variable. Therefore, comparing results with prior literature is difficult. I first discuss the regression results separately for the different corporate governance proxies followed by the overall results.
5.2.1 Hypothesis 1
To test the first hypothesis a regression analysis based on model 1 is made for each of the five corporate governance proxies as dependent variables. The results are shown in table 4.
Column 1 shows no significant relation between Delta_DACC and board size, which indicates that an increase in discretionary accruals and thus a decrease in accounting quality does not affect the size of the board. This is not in line with hypothesis 1. This can be explained that an increase in discretionary accruals is not enough to trigger a change in the size of the board. The costs to install the existing board can be an incentive to not change it as a result of an increase in discretionary accruals.
Column 2 shows that there is no significant relationship between Delta_DACC and board independence. This implies that an increase in discretionary accruals and thus a decrease in accounting quality does not affect the independence of the board. This is not in line with
accruals and board independence. However, this contains discretionary accruals and this study is focusing on the change of discretionary accruals it can be argued that cost stickiness can explain why board independence is not affected. According to Balakrishnan and Cruca (2008), cost stickiness is an asymmetric effect. The costs are increasing when activity increases, however the costs do not decrease with the same amount when activity decreases. It could be argued that the costs of attracting independent board members are sticky and therefore firms are not easily willing to change the board structure.
Column 3 shows that no significant relationship is found between Delta_DACC and CEO duality. This can be interpreted that when discretionary accruals increase and thus accounting quality decreases CEO duality is not affected. This is not in line with hypothesis 1. This is line with the finding that board independence is not affected by an increase in discretionary accruals. When the functions of chairman and CEO become separated, the independence of the board increases (Rechner & Dalton, 1991). For this finding it could also be argued that the cost of separating the two functions as a result of increases discretionary accruals is too costly for the firm because the change of discretionary accruals is to volatile. It could be possibly not realistic to change the structure of the board every year as a result of changes in discretionary accruals.
Column 4 shows that I find no significant relationship between Delta_DACC and audit committee. This implies that increased discretionary accruals and thus decreased accounting quality does not affect the presence of an audit committee. This is not in line with hypothesis 1. This can be explained in a way that an increase in discretionary accruals is not enough to trigger the installation of an audit committee. In addition, the level of monitoring by the existing audit committee could possibly be affected when discretionary accruals increase and thus accounting quality decreases. Furthermore, within the U.S. an audit committee is obligatory (Defond & Francis, 2005), therefore it could be argued that a change in the presence of an audit committee can not be affect by a change in discretionary accruals because it is obligatory by U.S. law.
Column 5 shows that there is a positive significant relationship between Delta_DACC and corporate governance committee (𝛽1=0.053, p=0.011). This implies that when discretionary
accruals increase and thus accounting quality decreases that firms are more likely to install a corporate governance committee. This is in line with hypothesis 1. The expected increase is based on the demand of investors and other users of the financial statements for high accounting quality to make good investment decisions (Biddle et al., 2009 and Armstrong et al., 2010).
Only the presence of a corporate governance committee has a positive significant association with Delta_DACC. None of the other corporate governance proxies have an
association with Delta_DACC. This can possibly explained by the fact that an audit committee is obligatory by U.S. law. It can be argued that investors and other users of the financial statements do not react that strongly on changes in discretionary accruals by demanding more corporate governance because a part of the corporate governance is already captured by law. In sum, only the presence of a corporate governance committee support my first hypothesis.
5.2.2 Hypothesis 2
To test the second hypothesis the same regression based on model 1 is used for each of the five different corporate governance proxies.
Column 1 shows that there a positive significant relationship between the Delta_DACC and Delta_DACC_dummy interaction term and board size (𝛽3=0.048, p=0.059). This implies that when discretionary accruals decrease and thus accounting quality increases the size of the board decreases. This is in line with my second hypothesis that after accounting quality increases corporate governance decreases. This decrease in board size can be explained by the findings of Eisenberg et al. (1998) and Yermack (1996) that smaller boards are more effective and report less discretionary accruals.
Column 2 shows that I find no significant relationship between the interaction term and board independence. This can be interpreted that when discretionary accruals decrease and thus accounting quality increases it has no effect on the amount of independent board members. This is not in line with my second hypothesis that after accounting quality increases corporate governance decreases. This result can be explained by cost stickiness (Balakrishnan & Cruca, 2008). The costs of firing independent board members and therefore decrease the independence of the board are costs that firms possibly are not willing to take.
Column 3 shows that in line with the second hypothesis there is a positive significant relationship between the interaction term and CEO duality (𝛽3=0.033, p=0.100). This implies
that when discretionary accruals decrease and thus accounting quality increases CEO duality decreases. This is in line with my second hypothesis that after accounting quality increases corporate governance decreases. CEo duality leads to a situation of combined power, which can lead to behavior of self-interest (Jensen & Meckling, 1976). When this is experienced by investors and other users of the financial statement, they possible demand a separation between the position of CEO and chairman.
Column 4 shows that there is no significant relation between the interaction term and the presence of an audit committee. This implies that when discretionary accruals decrease and thus
line with my second hypothesis that after accounting quality increases corporate governance decreases. This can be explained that in the U.S. an audit committee is obligatory (Defond & Francis, 2005), therefore it could be argued that a change in the presence of an audit committee can not be affect by a change in discretionary accruals because it is obligatory by U.S. law.
Column 5 shows that I find a negative significant relation between the interaction term and the presence of a corporate governance committee (𝛽3=-0.098, p=0.011). This indicates that when discretionary accruals decrease and thus accounting quality increases firms are more likely to install a corporate governance committee. This is not line with my second hypothesis that after accounting quality increases corporate governance decreases. This is a surprising result because it implies that when accounting quality increases firms get an incentive to install a corporate governance committee to increase monitoring.
For board size and board independence, a positive significant relationship with the interaction term is found. These corporate governance proxies support my second hypothesis that corporate governance decreases as a result of decreased discretionary accruals and thus increased accounting quality. However, for CEO duality, the presence of an audit committee and the presence of a corporate governance committee no positive significant relationship with the interaction term is found. These corporate governance proxies do not support my second hypothesis.
Furthermore, CVSize is positive significant associated with board independence (𝛽4=0.018,
p=0.000), the presence of an audit committee (𝛽4=0.001, p=0.081). and the presence of a
corporate governance committee (𝛽4=0.063, p=0.000). This implies that bigger firms have more independent board members and that bigger firms are more likely to have audit committees and corporate governance committees. CVSize is negative significant associated with board size (𝛽4=-0.104, p=0.000) and with CEO duality (𝛽4=-0.016, p=0.000). This can be interpreted that
bigger firms have smaller boards and that bigger firms have less CEO duality. Furthermore, none of the corporate governance proxies are significantly associated with CVLoss, which implies that losses do not affect the corporate governance proxies. Board size has a positive significant association with CVLeverage. This implies that firms with a higher leverage have larger boards. It could be argued because larger boards are less effective in their monitoring function that it is easier to create more debt. Therefore it could be that firms with higher leverage have larger boards.
In sum, the main effect can be explained by an overall line of arguments. First, it is possible to argue that the asymmetric effect of cost stickiness has an important role in changes in corporate governance mechanisms. The costs of installing corporate governance mechanisms are significant costs and are mostly focused for long-term perspectives. Discretionary accruals are volatile short-term measures for accounting quality. Because corporate governance mechanisms are set up for the long-term and require big investment it is arguable that these mechanism do not change that quickly over time because of changes in discretionary accruals. Consequently, changes in discretionary accruals could possibly not trigger firms enough to make such big and costly changes in terms of corporate governance mechanisms every year. Second, the demand of investors and other users of the financial statements for monitoring activities can be limited by the fact that since the introduction of SOX audit committees are obligatory. It could be that investors and other users of the financial statements react less intense because part of the corporate governance as possible solution for the information asymmetry is provided by U.S. law. Third, the results can possibly be explained by the introduction of Sarbanes-Oxley act. In 2002, SOX was introduced in the U.S. which had a huge impact on accrual-based earnings management, i.e. discretionary accruals. Moreover, Cohen et al. (2008) find that in the post SOX period managers are more intensively using real earnings management instead of using accrual-based earnings management. This could explain that after 2002 the effects of discretionary accruals do not have that much impact anymore. Since the time span used in this study is from 1996 until 2006 it is arguable that is could have consequence. An effect that could be thought of is the trade-off effect. It possible that within the years of before the introduction of SOX and within the years after the introduction of SOX accounting quality is affected differently because of changes from accrual-based earnings management to real earnings management. It could be that the two effects balance each other out and as a result give an overall effect which is not as strong as they are separately. It is also possible that in one of the two periods there is not an effect at all. Subsequently, when the two periods are merged together it could result in an different overall effect. Based on the overall effect, is arguable that it is not possible to draw conclusions on the both parts.
[Insert table 4 about here]
5.3 Robustness test
I perform a robustness test to determine whether the introduction of the Sarbanes-Oxley act in 2002 has an effect on measuring accounting quality by using discretionary accruals. As described,
conducting a survey that managers prefer real earnings management above accrual-based earnings management. The shift from accrual-based earnings management to real earnings management took place because real earnings management is more difficult to detect and is less likely to be inspected by regulators and accountants (Graham et al., 2005). Especially since the introduction of SOX more firms and managers switched from accruals-based earnings management to real earnings management (Cohen and Zarowin, 2010). Within this study I use discretionary accruals to measure accounting quality over the years 1996 to 2006. Because the change from accrual-based earnings management to real earnings management after the introduction of SOX in 2002 it could have different effects in terms of accounting quality. Because the introduction of SOX is within the time span that I examine, my results can be affect in multiple ways. This could imply that before and after 2002 different effects for accounting quality can possibly be observed. As described earlier there are effects that can be thought of. An effect that could be thought of is the trade-off effect. It possible that within the years of before the introduction of SOX and within the years after the introduction of SOX accounting quality is affected differently because of changes from accrual-based earnings management to real earnings management. It could be that the two effects balance each other out and as a result give an overall effect which is not as strong as they are separately. It is also possible that in one of the two periods there is not an effect at all. Subsequently, when the two periods are merged together it could result in an different overall effect. Based on the overall effect, is arguable that it is not possible to draw conclusions on the both parts.
For this robustness test I use a chi-square test to test whether there is a difference in coefficients between the period before the introduction of SOX and the period after the introduction of SOX. To test the robustness, I used model 1 for the two different time periods with board size as dependent variable. 1996-2001 is taken as period before the introduction of SOX and 2002-2006 as period after the introduction of SOX.
I expect less changes in accounting quality after the introduction of SOX because I measure accounting quality through discretionary accruals and managers are shifting away from accruals-based earnings management to real earnings management, therefore I expect a significant Chi2
value.
The results are shown in table 5. Column 3 shows that for both periods there is no difference in the coefficients (Chi2 1&2 = 0.010, p = 0.927) and (Chi2 1&2 = 0.050, p = 0.822).
This implies that statistically both periods are not different. This can be interpreted that within this study the introduction of SOX does not have an effect on the accounting quality measured using discretionary accruals. These result can be explained in a way that managers are still using
discretionary accruals after the introduction of SOX as much as they did before. Additionally, it is arguable that the use of real earnings management increased while the use of discretionary accruals stayed the same. Another explanation could be that because real earnings management is more difficult to detect and that there is not that much scrutiny to real earnings management, that it does not affect accounting quality to the extent where investors and other users of the financial statements are aware of.
