Involvement of Board Chairmen in Audit Committees and Earnings Management: Evidence from Malaysia

Abstract

This paper investigates the effect of the involvement of the board chairman in the audit committee (AC) on earnings management (EM). It examines Bursa Malaysia-listed companies with the lowest positive earnings for the years 2013 to 2015. The Modified Jones Model by Kasznik (1999) was used to determine discretionary accruals. An AC that includes its board chairman as an ordinary member is associated with greater discretionary accruals. However, a board chairman who is also the chairman of the AC does not seem to influence discretionary accruals. This paper supports the agency theory and policy-makers' efforts to prevent board chairmen from sitting on ACs. It is the first study that uses the agency theory to describe the association between the board chairman's involvement in the both AC and EM. This study alerts policy-makers, stakeholders and researchers to the influence of a board chairman serving on the AC in curbing EM. Furthermore, it provides empirical evidence that the majority of Malaysian companies whose board chairmen are involved in the AC appoint the chairman as an ordinary member of the AC. This indicates that executive directors may affect such actions. Hence, more policies are needed to improve AC independence.

1. Introduction

The high incidence of accounting scandals worldwide has pointed out the problem of earnings management (EM). It is argued that such scandals are typically followed by confirmation of EM activities (Teh, San Ong, & Ying, 2017). EM takes place when managers use their judgment in preparing financial reports or in structuring the firm’s transactions in order to alter earnings reports, either to mislead stakeholders about the actual performance of business or to influence the contractual outcomes that in most companies depend on accounting figures (Healy & Wahlen, 1999). It is stated that the management of troubled companies typically starts with EM, but eventually leads to misrepresentation or fraud when the situation is uncontrollable (Jones, 2011). In other words, early EM may be an indication of frauds to come.

The EM problem is not new, but it has increased in many countries (Leuz, Nanda, & Wysocki, 2003), and Malaysia is no an exception, where the extent of EM was greater in Malaysia (and other countries in the region such as Hong Kong and Singapore) (Leuz et al., 2003) than in other regions (see more recently Enomoto, Kimura, and Yamaguchi (2015)). EM has therefore been much studied (Nguyen & Duong, 2020) and continued to be a popular subject in accountancy research (Soliman & Ragab, 2014). It needs to be explored further, especially in Malaysia (Teh et al., 2017).

The reasons for practicing EM is the conflict of interests between agent and owner (Jensen & Meckling, 1976). To overcome this problem, various reforms and standards for corporate governance (CG) have been developed worldwide and in Malaysia. However, the CG mechanisms among Malaysian companies currently appear to be inadequate in preventing EM and there is therefore a need for further reform (Mohammad, Wasiuzzaman, & Salleh, 2016). Essentially, more responsibility needs to be given to the audit committee (AC) to constrain EM. Even in companies that implement the Malaysian Code of Corporate Governance (MCCG), ACs are not effective in constraining EM (Chandrasegaram, Rahimansa, Rahman, Abdullah, & Mat, 2013) and not yet effectively performing its oversight role (Abdul-Rahman & Ali, 2006). Hence, there is a need for more policies to reinforce the AC independence (Al-Absy, Ku Ismail, & Chandren, 2018a); as previous research has failed to posit a definitive association between AC and EM, more studies are required (Inaam & Khamoussi, 2016).

One characteristic of an AC could influence its effectiveness: its independence. The MCCG version of 2007, for example, requires that the majority of AC members must be independent, and its recent amendment (MCCG, 2017) recommends that an AC should comprise only independent directors. The Malaysian Securities Commission is now even more serious about the need to ensure AC independence as the recent MCCG (2017) also requires that the AC chairman should not be the same person as the board chairman. This is a timely and laudable move, and is in line with the requirements of other countries such as the UK, Australia, and Russia. It indicates that the appointment of the board chairman to an AC is widespread not only in Malaysia but also in other countries.

Even though the separation of the roles of board chairman and AC chairman by Malaysian regulators is a desirable initiative, in the researcher’s opinion the new regulation will not completely solve the problem of AC independence because the board chairman can still become involved in an AC by being an ordinary member, meaning that the AC cannot be seen as truly independent. A board chairman is very powerful and tends to dominate the board, especially if he is an influential figure or the founder of the company (Satkunasingam, Yong, & Cherk, 2012).

Therefore, the current study expects that the involvement of the board chairman in the AC, either acting as chairman or just an ordinary member, could affect the AC’s effectiveness and independence in mitigating EM. Secondly, the study expects that the influence of a board chairman who chairs the AC should not equal the influence of a board chairman who is just an ordinary member of the AC. Hence, this paper investigates the effect on EM practices of the board chairman’s involvement in the AC, and whether a board chairman who chairs the AC and one who is an ordinary member have the same influence on EM.

The current study argues that the involvement of the board chairman either as an AC chairman or a member is not appropriate in a country where there is deference to a wellconnected or dominant person; this could result in board committees being subservient to the chairman’s directives, even if the latter does not practice good CG (Satkunasingam et al., 2012). Al-Absy, Ku Ismail, and Chandren (2018b) found that EM is higher in companies whose board chairman dominates their nomination committee, either was the committee chairman or an ordinary member in the committee.

This study will contribute to the body of knowledge in different ways. It is the first study to use the agency theory to explain the relationship between the board chairman’s involvement in the AC and EM. Second, it introduces a new proxy for AC independence, namely, the board chairman’s involvement in the AC. Companies whose ACs include board chairmen are labeled 1, otherwise 0. Two further measures were extracted from this proxy: a company in which the board chairman acts as the AC chairman is given the value 1, and 0 otherwise; and where the board chairman is an ordinary member of the AC, it is given the value 1, and 0 otherwise. Lastly, to the best of the researcher’s knowledge, no studies have empirically investigated the effect on EM of a board chairman who either acts as chairman of the AC or is an ordinary member. The reason for selecting Malaysia is because of the low percentage of independent directors on boards. Thus, controlling shareholders of some companies in Malaysia still dominate both the board nomination committee and the selection of directors (The World Bank Report, 2012, p. 29). Therefore, it is common for family groups and controlling shareholders to elect a board chairman who belongs to the family or group and also to appoint him as an AC chairman or member.

2. Literature Review

The figure for earnings is one of the most important elements of an income statement for the majority of readers of financial statements, because it is a measurement of a company’s performance. It is prepared under the accruals basis of accounting, based on accounting principles. However, in some cases, the deployment of accruals may bring some issues such as EM (Dechow, 1994). The management can normally exercise some discretion concerning income recognition (Dechow, 1994), because accounting measurements often require some judgments to be made by managers (Xie, Davidson, & DaDalt, 2003). Earnings manipulation via accounting principles and methods is called accruals earnings management (AEM) and may be the preferred instrument for managing opportunistic profits because it has no general direct cash flow outcomes and is somewhat difficult to unravel (Peasnell, Pope, & Young, 2005).

Numerous researchers have studied the extent to which EM occurs in companies, but the results are mixed (Xie et al., 2003). Importantly, incentives for EM are many (Healy & Wahlen, 1999), including (i) capital market expectations or valuation; (ii) contracts based on accounting figures; and (iii) antitrust or other government regulations. Regardless of the different definitions and incentives associated with EM, manipulation of earnings in the financial reporting process provides deceptive reports to users. The existence of the agency problem has led to the need to look for effective mechanisms to enhance the alignment between the respective interests of owners and agents (Fama & Jensen, 1983). International efforts have been made to hamper managers’ opportunistic behaviours by developing and implementing appropriate CG mechanisms, such as the various committees of the board of directors (Fama & Jensen, 1983), the AC (Klein, 2002), the nominations committee and the remuneration committee.

An AC, as an essential mechanism of CG, plays a vital role in ensuring the financial integrity and the quality of a company’s financial reporting. A well-functioning, active, and structured AC may play a significant role in mitigating EM (Xie et al., 2003). The majority of regulators have recommended that publicly-held companies should establish ACs consisting of outside directors. According to Abdullah and Nasir (2004), the issue of the AC in Malaysia emerged in the mid-1980s around the time of the financial failure of Bumiputra Malaysian Finance Ltd, following which in 1986, the Malaysian Central Bank (Bank Negara Malaysia) made it mandatory for all financial institutions under its control to establish an AC. The requirement was further extended in 1990 to include all insurance companies operating in the country. Subsequently, in 1993 the Kuala Lumpur Stock Exchange made the AC compulsory for all its listed companies (Abdullah & Nasir, 2004).

3. Audit Committee Independence and Hypotheses

In Malaysia, although the level of independent AC directors has increased, the scope and effectiveness of the committee are still questionable (Mohammad et al., 2016). Following the establishment of an AC in listed companies in Malaysia, ACs have yet to have any particular success in terms of their monitoring role (Abdul-Rahman & Ali, 2006). This is especially the case for family-owned companies in which an increase in Type II agency problems (Interest disputes between controlling shareholders and minority shareholders) and a decline in the effectiveness of CG mechanisms have been observed (Cheung & Chan, 2004). This study believes that the AC cannot be effective when the chairman of the board is also involved in the AC, either as chairman or an ordinary member.

The Finance Committee of Malaysia referred to the role of powerful board chairmen who seek to control the board decision due to their social or political status, or especially if they were the founders of the corporation (Satkunasingam et al., 2012). The authors argued that involvement of the board chairman in the AC, either as a chairman or a member, is inappropriate, especially in a country where there is deference to a well-connected or dominant board chairman. Thus, to improve the effectiveness of the board’s evaluation of AC activities, the chairman of the board should not be involved in the AC, either as chairman (Council, 2007; KPMG, 2012; MCCG Draft, 2016; Smith, 2003) or member (Smith, 2003).

With respect to the functions of an AC, it is impractical for the board’s chair to sit on it. This is because the AC reports to the board on how it has fulfilled its obligations and makes some recommendations. In addition, the board should review the AC’s effectiveness annually (Financial Reporting Council, 2012). It is thus highly likely that AC independence will be compromised when the board chairman sits on the AC. Where this occurs, an AC would not be able to fulfil its role effectively, which in turn may lead to managers practising unethical behaviour such as EM in companies whose board is not really independent. This argument has been empirically supported by Al-Absy et al. (2018b), who found that companies whose board chairman is part of the nomination committee i.e., chairman of the committee or just a member, are significantly related to a high level of EM. Based on agency theory and the discussion above, the following hypothesis is suggested:

H₁ . The board chairman’s involvement in the audit committee is positively related to earnings management.

This hypothesis does not consider the different personal characteristics and impact of the board chairman either as AC chairman or an ordinary member. The literature argues that AC effectiveness often depends on the role of its chairman, who has a significant effect on the committee. Therefore, there is a need for a separate investigation to see whether or not the board chairman who chairs the AC and the board chairman who is just a member has the same influence on the EM. Regulators have focused more on the AC chairman’s characteristics, most requiring the AC to be chaired by an independent director, and others also requiring the chairman to possess accounting expertise. Indeed, the AC chairman has more responsibilities than any ordinary member. Therefore, an AC chairman may not give his best to the committee if he is also the board chairman (MCCG Draft, 2016). This is because it is extremely difficult for one person to do two tasks.

In addition, it is impossible for the board to evaluate the AC function when the AC is chaired by the board chairman (MCCG, 2017). Also, the chairman of the AC must be answerable to the chairman of the board regarding the activities and responsibilities of the AC (Financial Reporting Council, 2012; Smith, 2003). Thus, when the chairman of the AC and the board is the same person, there are no checks and balances, as the chairman of the AC is answerable to himself. This chairman duality may affect the performance of the committee, especially when the board has a powerful chairman (Satkunasingam et al., 2012), which is often the case in family-owned companies. However, when the chair of the AC is not the same person as the board chairman, he or she can offer full commitment to their role and to exercise independent judgement (MCCG Draft, 2016).

Recognising the problem of chairman duality, various countries have introduced regulations that prevent the chairman of a board from becoming an AC chairman. The Australian Securities Exchange, for example, requires that the AC should, among other things, be chaired by an independent person, who is not sitting as the chair of the board (Council, 2007). In Russia, the Audit Committee Institute also recommends that the AC chairman should be independent and must not also be the chairman of the board (KPMG, 2012). Recently, in the Malaysian context, the MCCG recommended that AC chairman shall be an independent director and not be a chairman of the board (MCCG, 2017; MCCG Draft, 2016). Consequently, the Bursa Malaysia revised its Corporate Governance Guidelines requiring all listed companies to avoid the duality problem. Based on agency theory and the discussion above, the following hypothesis is suggested:

H_1a. The dual role of board chairman and audit committee chairman is positively related to earnings management.

In contrast to the above, the study expects that companies would be more likely to appoint their board chairman as an ordinary member of the AC rather than as the committee’s chairman. Appointing as an ordinary member a board chairman who is not really independent or may not be well qualified can be seen as an alternative choice for executive directors, who in most cases are family members, to continue to dominate the AC’s decisions. This situation could continue while the MCCG (2017) does not prevent the board chairman from becoming an ordinary member of the AC. Furthermore, companies are likely to appoint the board chairman as an ordinary member in AC instead of chairman to escape from the requirements imposed by regulators on appointing of AC chairman.

Most regulators require the AC chairman to be independent, but few require the AC itself to be fully independent. Thus, companies can readily appoint the nonindependent board chairman as an ordinary member of the AC if there is a restriction on a non-independent board chairman becoming AC chairman. Therefore, to enhance the effectiveness and independence of the AC, several CG codes, such as Bangladesh, Ireland, Nigeria, South Africa, Thailand, the United Arab Emirates, do not allow the board chairman to become even an ordinary member of the AC. Based on agency theory and the discussion above, the following hypothesis is suggested:

H_1b. The dual role of board chairman and audit committee member is positively related to earnings management.

4. Research Methodology

4.1. Sample Selection

The occurrence of EM is high in companies whose earnings are close to zero because managers are usually inspired to avoid reporting negative earnings (Huynh & Nguyen, 2019) by converting the small losses into a small profit (Roychowdhury, 2006; Yuliana, Anshori, & Alim, 2015). Hence, previous studies focused on companies whose return on assets (ROA) is near to zero (0 to 0.005) (Roychowdhury, 2006; Yuliana et al., 2015). Accordingly, this study covers only companies with low positive earnings, in average for the years 2013 to 2015 (Al-Absy et al., 2018a, 2018b; AlAbsy, Ku Ismail, & Chandren, 2019a, 2019b, 2019c).

To obtain the sample, first, 54 financial companies, closeend funds, special-purpose acquisition companies, and realestate investment trusts are excluded, as are the 79 companies whose ROA data was not available, and the 200 companies with negative earnings in one or more years. Next, the average ROA for each of the remaining listed companies (475) is calculated. Finally, 300 companies with the lowest average ROA for the three years were selected. In addition, during data collection, a further 12 companies were excluded because they belong to industries with fewer than six observations (Subramanyam, 1996) or have missing data. Therefore, the final sample consists of 864 company-observations. The selection of 2013 as the base year was because it was the year following the introduction of MCCG 2012.

4.2. Measurement of Earnings Management

Researchers are increasingly using an accruals-based measurement as a key proxy of EM (Enomoto et al., 2015). Hence, the Modified Jones Model (MJM) by Kasznik (1999) is applied to determine the discretionary accruals (DA). Kasznik (1999) includes the change in cash flows from operations (CFO) as an explanatory variable that increases the power of the model (Shuto, 2007). Kasznik (1999) indicates that CFO is negatively correlated with total accruals (see. Dechow, 1994). In calculating DA and by applying Kaszik’s MJM, total accruals are first calculated as earnings before extraordinary items minus CFO. Furthermore, an ordinary least squares (OLS) test was run for a specific year and industry to estimate the coefficients values (α₁, α₂, α₃ and α₄) from the following model:

\(\begin{aligned} \frac{\mathrm{TAC}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}} &=\alpha_{0}+\alpha_{1}\left(\frac{1}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\alpha_{2}\left[\frac{\left(\Delta \mathrm{REV}_{\mathrm{it}}-\Delta \mathrm{REC}_{\mathrm{it}}\right)}{\mathrm{TA}_{\mathrm{it}-1}}\right] \\ &+\alpha_{3}\left(\frac{\mathrm{PPE}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\alpha_{4}\left(\frac{\Delta \mathrm{CFO}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\varepsilon_{\mathrm{it}} \end{aligned}\)       (1)

where: TA_it-1 is total assets in the past year, ΔREV is the revenues in year_t minus revenues in year_t-1, ΔREC is net receivables in year_t minus net receivables in year_t-1, PPE is gross property, plant and equipment in year_t, ΔCFOit is cash flows from operations in year_t less year_t-1, ε_it is regression error terms. Subsequently, the coefficients α₁, α₂, α₃ and α₄ and derived from Equation 1 are used in the subsequent equation in order to estimate the non-discretionary accruals value (NDA_it):

\(\begin{aligned} \mathrm{NDA}_{\mathrm{it}} &=\alpha_{1}\left(\frac{1}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\alpha_{2}\left[\frac{\left(\Delta \mathrm{REV}_{\mathrm{it}}-\Delta \mathrm{REC}_{\mathrm{it}}\right)}{\mathrm{TA}_{\mathrm{it}-1}}\right] \\ &+\alpha_{3}\left(\frac{\mathrm{PPE}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\alpha_{4}\left(\frac{\Delta \mathrm{CFO}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\varepsilon_{\mathrm{it}} \end{aligned}\)       (2)

Lastly, the DA is calculated by extracting NDA_it from the total accruals, for which the next equation is used:

\(\mathrm{DA}_{\mathrm{it}}=\frac{\mathrm{TAC}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}-\mathrm{NDA}_{\mathrm{it}}\)       (3)

Importantly, following previous studies (e.g., Abdullah & Ku Ismail, 2016; Mohammad et al., 2016), the absolute value of DA is used to reflect the extent of EM. In addition, to test the robustness of the results, this study uses two different proxies of DA, namely the MJM by Dechow, Sloan, and Sweeney (1995) and Jones (1991) using the following cross-sectional models, respectively:

\(\begin{aligned} \frac{\mathrm{TAC}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}} &=\alpha_{1}\left(\frac{1}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\alpha_{2}\left[\frac{\left(\Delta \mathrm{REV}_{\mathrm{it}}-\Delta \mathrm{REC}_{\mathrm{it}}\right)}{\mathrm{TA}_{\mathrm{it}-1}}\right] \\ &+\alpha_{3}\left(\frac{\mathrm{PPE}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\varepsilon_{\mathrm{it}} \end{aligned}\)       (4)

\(\begin{aligned} \frac{\mathrm{TAC}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}} &=\alpha_{1}\left(\frac{1}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\alpha_{2}\left[\frac{\Delta \mathrm{REV}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right] \\ &+\alpha_{3}\left(\frac{\mathrm{PPE}_{\mathrm{it}}}{\mathrm{TA}_{\mathrm{it}-1}}\right)+\varepsilon_{\mathrm{it}} \end{aligned}\)       (5)

In addition, a robustness test is conducted using real activities manipulation. Roychowdhury (2006) introduced three variables that capture the influence of real activities manipulation: the abnormal levels of cash flow from operations (ABCFO), the abnormal levels of production costs (ABPROD) and the abnormal levels of discretionary expenses (ABDISX). The normal levels of each variable are estimated by running the following cross-sectional regression to estimate coefficients for every industry and year for each variable respectively:

\(\frac{\mathrm{CFO}_{\mathrm{t}}}{\mathrm{TA}_{\mathrm{it}-1}}=\alpha_{0}+\alpha_{1}\left(\frac{1}{\mathrm{~A}_{\mathrm{t}-1}}\right)+\beta_{1}\left(\frac{\mathrm{S}_{\mathrm{t}}}{\mathrm{A}_{\mathrm{t}-1}}\right)+\beta_{2}\left(\frac{\Delta \mathrm{S}_{\mathrm{t}}}{\mathrm{A}_{\mathrm{t}-1}}\right)+\varepsilon_{\mathrm{t}}\)       (6)

\(\begin{aligned} \frac{\mathrm{PROD}_{\mathrm{t}}}{\mathrm{TA}_{\mathrm{it}-1}} &=\alpha_{0}+\alpha_{1}\left(\frac{1}{\mathrm{~A}_{\mathrm{t}-1}}\right)+\beta_{1}\left(\frac{\mathrm{S}_{\mathrm{t}}}{\mathrm{A}_{\mathrm{t}-1}}\right)+\beta_{2}\left(\frac{\Delta \mathrm{S}_{\mathrm{t}}}{\mathrm{A}_{\mathrm{t}-1}}\right) \\ &+\beta_{3}\left(\frac{\Delta \mathrm{S}_{\mathrm{t}-1}}{\mathrm{~A}}\right) \varepsilon_{\mathrm{t}} \end{aligned}\)       (7)

\(\frac{D I S E X P_{t}}{T A_{i t-1}}=\alpha_{0}+\alpha_{1}\left(\frac{1}{A_{t-1}}\right)+\beta\left(\frac{S_{t-1}}{A_{t-1}}\right)+\varepsilon_{t}\)       (8)

where: S_t is sales during period_t, S_t–1 is sales at the end of period_t-1, ΔS_t is sales during period_t minus sales for the previous period_t, ΔS_t–1 is one year lag of ΔS_t, PROD_t is cost of goods sold and change in inventory during period_t, DISEXP_t is the sum of expenses for research and development, advertising, sales, general and administration in period_t.

Subsequently, the abnormal level of each variable is its actual level minus its normal level. Essentially, this study standardised the value of each variable of ABCFO, ABPROD and ABDISEXP (Chen, Cheng, & Wang, 2010; Cohen & Zarowin, 2010; Haji-Abdullah & WanHussin, 2015). Previous studies provide evidence that companies who engage in increase-earnings manipulation are most definitely to own low CFO and low discretionary expenditures while they are most likely to show high production costs. Accordingly, to attain consistency across variables, ABDISEXP and ABCFO are multiplied by (–1) where the higher values of ABCFO and ABDISEXP reflect the higher real activities manipulations (Cohen & Zarowin, 2010; Haji-Abdullah & Wan-Hussin, 2015). Therefore, this study combined the standardised values of ABCFO, ABPROD and ABDISX to represent the aggregate value of abnormal levels of real earnings management (REM) according to the equation below:

\(\mathrm{REM}=\mathrm{ABCFO}^{*-1}+\mathrm{ABPROD}+\mathrm{ABDISX}^{*-1}\)       (9)

Moreover, to be consistent with DA’s proxies, this study uses the absolute value of REM to reflect the level of real activities manipulation as recommended by Chen et al. (2010) and following Kwon, Na, and Park (2017).

4.3. Empirical Model

The OLS regression is run to test the influence of the board chairman’s involvement in the AC (BCHINV) on the practice of EM (Hypothesis 1). To control the relationship, this study includes some variables related to CG characteristics – board independence (BIND), board size (BSIZE), frequency of board meetings (BMEET), AC independence (ACIND), AC size (ACSIZE), frequency of AC meetings (ACMEET), ownership concentration (OWNCON), and audit quality (BIG4). It also includes some variables related to the company-specific characteristics – company size (SIZE), leverage (LEV), and return on assets (ROA). Hence, the following regression is applied:

\(\begin{aligned} \mathrm{DA}=& \beta_{0}+\beta_{1} \mathrm{BCHINV}+\beta_{2} \mathrm{BIND}+\beta_{3} \mathrm{BSIZE}+\beta_{4} \mathrm{BMEET} \\ &+\beta_{5} \mathrm{ACIND}+\beta_{6} \mathrm{ACSIZE}+\beta_{7} \mathrm{ACMEET} \\ &+\beta_{8} \mathrm{OWNCON}+\beta_{9} \mathrm{BIG} 4+\beta_{10} \mathrm{SIZE}+\beta_{11} \text { LEV } \\ &+\beta_{12} \mathrm{ROA}+\varepsilon \end{aligned}\)(Model 1)

where DA is the absolute value of DA using MJM by Kasznik (1999). The measurement details for all variables are displayed in Table 1.

Table 1: Summary of the Operationalisation of the Variables

Secondly, the study expects that the influence of a board chairman who chairs the AC will not be the same as that of a board chairman who is an ordinary member of the AC. Subsequently, this study categorises board chairman involvement in an AC into two styles: board chairman duality (BCHDUAL) and board chairman as AC member (BCHMEM). Thus, to examine the influence on EM of board chairman duality and board chairman membership (H_1a and H_1b), the following regression is run:

\(\begin{array}{r} \mathrm{DA}=\beta_{0}+\beta_{1} \mathrm{BCHDUAL}+\beta_{2} \mathrm{BCHMEM}+\beta_{3} \mathrm{BIND}+ \\ \quad \beta_{4} \mathrm{BSIZE}+\beta_{5} \mathrm{BMEET}+\beta_{6} \mathrm{ACIND}+\beta_{7} \mathrm{ACSIZE}+ \\ \quad \beta_{8} \mathrm{ACMEET}+\beta_{9} \mathrm{OWNCON}+\beta_{10} \mathrm{BIG} 4+\beta_{11} \mathrm{SIZE} \\ \quad+\beta_{12} \mathrm{LEV}+\beta_{13} \mathrm{ROA}+\varepsilon \end{array}\)(Model 2)

Robustness tests are also conducted using the previous two models, but using a different proxy for DA, the Jones Model (1991) and the modified of this Model by Dechow et al. (1995) as well as the sum value of the three proxies of REM presented by Roychowdhury (2006).

5. Results and Discussion

Table 2 displays the average (mean) value of the absolute DA using the MJM by Kasznik (1999), which is 0.036. Regarding the involvement of board chairman in AC, Table 2 shows that almost one-third (31.25%) of company observations appointing the board chairman as either AC chairman or AC member. This is very high considering that AC independence is a major concern among policymakers worldwide. Specifically, 19.44% of the companyobservations indicate appointing the board chairman as a member of the AC, and 11.81% as AC chairman. This shows that the percentage of board chairmen who serve on an AC is reasonably high, especially as an ordinary member. Upon further scrutinising the data, it is found that the phenomenon is common among family companies.

Table 2: Descriptive Statistics (N=864)

To eliminate the outliers, this study winsorizes extreme observations of ACMEET and ACSIZE in the bottom and top 1%; and extreme observations of BMEET using 2%. Moreover, skewness and kurtosis are used to check the normality assumption. Table 2 shows no significant violation of the normality in the dataset of individual variables because kurtosis and skewness of each variable are lower than ±10 and ±3 respectively. In addition, Pearson’s correlation is used to check the multicollinearity issue among the variables. Accordingly, Table 3 shows that none of the correlation matrices between variables exceeds 0.80.

Table 3: Pearson Correlation Analysis (N=864)​​​​​​​

The Breusch-Pagan/Cook-Weisberg and DurbinWatson tests are applied to check the heteroscedasticity and autocorrelation problem, respectively. All our models suffer from heteroscedasticity problems. Consequently, we applied the OLS regression with a robust function. Column A in Table 4 shows no evidence that board chairman involvement in AC, either a chairman or ordinary member is significantly associated with DA. Hence, this does not support H₁, which states that a board chairman’s involvement in the AC is positively related to EM. Therefore, the study expects that the influence of and characteristics of a board chairman who chairs the AC could not be the same as those of a board chairman who is just an ordinary member. This may be an explanation of the insignificant result. Therefore, investigating separately the influence of each of them on EM could provide beneficial results.

Table 4: Regression Result of the Main Model (Kasznik’s Model)​​​​​​​

Column B of Table 4 provides no evidence that the dual role of board chairman and AC chairman is significantly associated with DA. Hence, this does not support H_1a, which states that the dual role is positively related to EM. This result is in line with Al-Arussi and Shamkhi (2016) who found no significant relationship between the board chairman also being chairman of the AC and the low level of Internet financial disclosures. However, it is inconsistent with Al-Absy et al. (2018b), who found that the dual role of board chairman and nomination committee chairman is significantly positively related to high EM.

The explanation of this result may be that board chairman duality imposes a double responsibility for any potential financial risk. Thus, a board chairman who is also the AC chairman may pay greater attention to controlling EM and is less likely to engage in it; although the result is not significant the direction is negative. In particular, we note that 39% of those board chairmen who are also AC chairmen possess accounting knowledge. Moreover, the board chairmen who are also AC chairmen are independent directors and are not related to any of the directors. This is because the MCCG requires companies to appoint an AC chairman who is an independent director.

However, in terms of the board chairman being an ordinary member of the AC, Column B of Table 4 shows that this is associated with higher DA. This result supports H_1b, which states that the board chairman also being an AC member is positively related to EM. This finding indicates that the presence of the board chairman on the AC as a member may jeopardise the AC’s independence, and thus weaken the effectiveness of the AC in monitoring the company’s activities, including EM practices. This result is consistent with Al-Absy et al. (2018b), who found that board chairman also being an ordinary member of the nomination committee is significantly positively related to high EM. Appointing a board chairman as an ordinary member of an AC can be seen as an alternative to the board chairman also chairing the AC, given the restriction on a non-independent board chairman becoming an AC chairman. That is, when companies are not allowed to appoint a non-independent board chairman as the AC chairman, they may instead choose to appoint the former as an ordinary member, and by doing so the influence of the board chairman persists.

It is observed that 25% of companies appoint a nonindependent board chairman as an ordinary member of an AC (42 from 168 company-observations). In addition, 23% of companies appoint a board chairman who is a family member of directors or shareholders in AC as an ordinary member (38 from 168 company-observations). It is more likely that the purpose of appointing the board chairman as an ordinary member of the AC is to dampen the effectiveness and independence of the AC rather than to monitor the financial reporting process, as it is observed that only 11% of board chairmen who are ordinary members of the AC possess accounting knowledge (18 from 168 companyobservations). That is, it appears that the appointment is not mainly to monitor the financial reports, but rather to dominate the AC through their social or political status, or especially if they are the founders of the corporation.

As for the control variables, the frequency of board meetings (BMEET) is associated with high DA, which is consistent with Mansor, Che-Ahmad, Ahmad-Zaluki, and Osman (2013). However, the number of board directors (BSIZE) is associated with low DA, which is consistent with Xie et al. (2003). Bigger boards may involve more independent directors; this may reduce the extent of EM practices. The results of this study show that a high proportion of board independent directors (BIND) is associated with low DA. This result is consistent with Prencipe and BarYosef (2011) and Xie et al. (2003).

Table 4 also shows that high frequency of AC meetings (ACMEET) is associated with low DA, which is in line with Xie et al. (2003). However, high proportion of independent directors on the AC (ACIND) is associated with high DA. The result indicates that ACIND fails to question or act against management activities (Mohammad et al., 2016), which may due to the information asymmetry problem. This study found no significant relationship between ACSIZE and EM, consistent with most studies in Malaysia such as Abdullah and Ku Ismail (2016) and Saleh, Iskandar, and Rahmat (2007). In addition, there is no significant relationship between ownership concentration (OWNCON) and EM, which in line with Abdul-Rahman and Ali (2006) and Mohammad et al. (2016). Similarly, this study finds no significant relationship between audit quality (BIG4) and EM, again supporting the majority of studies such as Abdullah and Ku Ismail (2016) and Mohammad et al. (2016).

Table 4 shows that leverage (LEV) is associated with high DA. The positive result for LEV on DA is in line with prior studies in Malaysia such Saleh et al. (2007). This is expected because companies with financial difficulties are most likely to manage earnings. Further, it is found that company size (SIZE) is negatively associated with DA. Lastly, this study finds a significant positive relationship between ROA and DA, which is consistent with other studies (Abdullah & Ku Ismail, 2016; Saleh et al., 2007).

6. Robustness Tests

6.1. Using Different Measurements of Earnings Management

For the robustness test, as explained earlier, this study uses a different proxy for EM, which provides reliable results and avoids bias towards one measurement (Al-Jaifi, 2017). Table 5 shows the regression results of Model 1 and Model 2 using different proxies for EM: the absolute value of DA calculated by the Jones Model (1991) as shown in Column A, the MJM by Dechow et al. (1995) as shown in column B, and the absolute value of REM as shown in Column C. The results for Model 1 in all Columns of Table 5 show no evidence that board chairman involvement in an AC (either as a chair or a member) is associated with the absolute DA and REM, which is consistent with the result of Model 1 (using Kasznik’s model) in Column A of Table 4.

Table 5: Regression Result Using Different Measurements of Earnings Management​​​​​​​

Regarding Model 2, the results are consistent with those (using Kasznik’s model) shown in Column B of Table 4. In more detail, Model 2 in Columns A and B of Table 5 show no evidence that board chairman duality is associated with a high level of absolute DA. However, Model 2 in Column C of Table 5 shows evidence that board chairman duality is associated with a low level of absolute REM. In terms of a board chairman who is only an ordinary member, Model 2 in all Columns of Table 5 show evidence of association with a high level of absolute DA and REM. This result matches those for Model 2 (using Kasznik’s model) in Column B of Table 4. As for the control variables, the results of Column A using the Jones Model (1991) and Column B using the MJM by Dechow et al. (1995) are similar to the earlier ones recorded in Table 4 (using Kasznik’s model), except for BIND, BSIZE and ACMEET. Regarding Column C using REM, the result of control variables in the majority of cases appears to be different from the proxies of DA.

6.2. Using Lagged Independent Variables of the Main Model

Despite the wide range of control variables used in the current study’s models to reduce the possibility of endogeneity and mis-specification (Prencipe & Bar-Yosef, 2011), there is a need for testing the potential reverse causality between CG mechanisms and the level of earnings (Xie et al., 2003). Therefore, to control the potential reverse causality problem, Models 1 and 2 (using Kasznik’s model) were re-estimated using the lagged independent variables values (Al-Jaifi, 2017). Column A of Table 6 shows the results, which are consistent with the previous results of Models 1 and 2 in Table 4, regarding BCHINV, BCHDUAL and BCHMEM. This indicates that reverse causality is impossible to occur.

Table 6: Regression Result of the Main Model (Kasznik’s Model) Using Lagged Independent Variables and FGLS Estimator​​​​​​​

6.3. Using Different Regression Estimator for the Main Model

It is argued that random-effects feasible generalized least squares (FGLS) is asymptotically more efficient and its estimator of the asymptotic variation takes the normal form (Wooldridge, 2010). Thus, the study re-runs Models 1 and 2 (Kasznik’s model) using FGLS regression with a panel-specific heteroscedastic error structure (Yoshikawa & Rasheed, 2010) instead of OLS regression to test the robustness of the results. Column B of Table 6 shows the results of the re-run for Models 1 and 2 using FGLS regression, which are consistent with the results of Models 1 and 2 in Table 4 using OLS regression.

7. Conclusion

One important feature of an AC is its independence, and attempts have been made to associate AC independence with earnings quality. Another aspect of AC independence that most researchers have ignored is the involvement of board chairmen on the committee. It is only recently that various jurisdictions around the world have instituted the requirement that a company’s board chairman should not chair the AC, or even sit on the AC. So, this study investigates whether board chairman involvement in an AC was associated with EM practices in Malaysia before the new regulation was introduced.

Consistent with the agency theory, it is found that an AC that includes its board chairman as an ordinary member of the AC is associated with greater absolute DA. However, a board chairman who also acts as the AC chairman does not influence absolute DA. A robustness test is conducted by using two different proxies of DA, the Jones Model and MJM by Dechow et al. (1995). The results confirm that BCHMEM is associated with greater DA while BCHDUAL is not associated with DA. Furthermore, a robustness test is conducted using a proxy of REM, the sum value of the three proxies announced by Roychowdhury (2006). The results show that the presence of BCHMEM is more likely to increase the level of REM. However, the presence of BCHDUAL is more likely to reduce the level of REM.

Although the Securities Commission and Bursa Malaysia have taken a laudable step in enhancing AC independence in the MCCG 2017 and the proposed Corporate Governance Guidelines, respectively, our results suggest that the regulators should prevent the board chairman not only from being an AC chairman, but also from sitting on the AC. Our study aids policy-makers, not only in Malaysia but in other parts of the world, in enhancing their regulations with respect to AC independence. In addition, the findings provide valuable information for the business community, academics and researchers. Although this study finds that chairman duality does not increase the prevalence of EM, it is believed that other factors such as family ownership and the chairman’s accounting expertise may act as a moderating influence. We therefore suggest that more research is needed to investigate ways of reducing the extent of EM in order to better serve stakeholders’ interests.