1 Introduction

The agency theory posits that good corporate governance (CG) is linked to better firm performance (Singh and Gaur 2009, 2013), and past scholarly works typically prescribed a one-size fits all approach when addressing CG issues (McCahery and Vermeulen 2014; Singh and Delios 2017). However, Gaur et al. (2015) argued that these past works on CG were based on several arguable assumptions such as the self-interested and opportunistic model of human behavior and the adoption of the rational systems approach which focuses on structural characteristics and limits the role of participant, social, cultural, and technological contexts. To overcome the limitation of these assumptions, Gaur et al. (2015) adopted a holistic approach by integrating agency theory, stewardship theory, resource dependence theory, and stakeholder theory to empirically examine the role board characteristics play in influencing firm performance. Although Gaur et al. (2015) proved that high-ownership concentration in a firm diminishes the positive effects of board size and board competence on firm performance, the generalizability of their study is limited only to CG systems within common-law countries.

Studies on CG are predominately done in the context of a conventional economy with little emphasis on an Islamic economy (Alnasser and Muhammed 2012; Judge et al. 2010), where firms are required to adhere to the good principles of a just, honest, fair, and balanced society (Abu-Tapanjeh 2009). Ethical and moral values serve as the foundation for Shariah, the Islamic religious legal system and hence, Islamic capital markets require their Shariah-compliant firms to adhere to principles of good as laid out by Islamic authorities and endorsed by market regulators. Shariah compliance limits a firm’s primary activities by excluding activities such as money lending, gambling, as well as the production and sale of goods and services prohibited in Islam including alcoholic drinks, non-halal meat, and immoral services such as prostitution, pubs, and discos (Securities Commission Malaysia 2017). Alnasser and Muhammed (2012) noted that most studies adopting the agency theory in examining CG in a conventional economy used a ‘typical way’ when describing the association between shareholders and the board of directors (BOD). They argued that what is a ‘typical way’ in a conventional economy may not hold in an Islamic economy because the former prioritizes maximization of shareholder return while the latter concentrates on maximizing return as well as benefitting society.

While there are some studies linking good CG to better firm performance (Singh and Delios 2017), others have reported otherwise. Some studies have found that the alignment of shareholders’ interest and management interest reduce agency costs and conflicts, thus enabling a firm to performance efficiently (Brown and Caylor 2004; Singh and Gaur 2009). However, the positive effect of CG on firm performance is not always consistent (Gaur and Delios 2015), such as the study by Heracleous (2001) who failed to find definite evidence that suggest CG enhances firm performance (Alnasser and Muhammed 2012). These mixed results trigger questions on the applicability of past studies in the unique context of Shariah-compliant businesses.

A “one-size fits all” CG approach based on agency theory is not reliable (Popli et al. 2017a, b; Singh et al. 2017) and there is a need to address CG in the unique context of Shariah-compliant economies. The current study contributes to the existing body of knowledge in three ways. Firstly, the study explores the role of CG in influencing firm performance among Shariah-compliant firms. Secondly, the study examines the validity of Gaur’s et al. (2015) multi-theoretic framework in a context that is influenced by a religious legal system, prioritizing resource dependency theory and stakeholder theory instead of agency theory. Thirdly, the study provides empirical evidence on the efficacy of the BOD among Shariah-compliant firms, a context that is much sought after due to the market growth of Islamic capital markets. The study’s hypotheses are tested using panel data comprising 200 top-performing Shariah-compliant firms listed in the Malaysia Stock Exchange from 2014 to 2017, with the aim of providing a benchmark reference to other Islamic capital markets.

2 Theories and hypotheses

As per steps suggested by Gaur and Kumar (2018), the following is a systematic review of literature on the relationship between CG and firm performance. In a firm, its ownership (principal) and management control (agent) are separate entities. This separation however contributes to agency problem, whereby a conflict of interest between the shareholders and the managers arises (Jensen and Meckling 1976). The agency theory suggests that managers (the board) are concerned about their own needs more than the interests of shareholders. Hence, there is a need to monitor and control the board to reduce the conflict of interest and ensure that managers do not pursue strategies that reduce the value of the firm (Gaur and Kumar 2009; Gaur et al. 2014; Kumar et al. 2012). To alleviate agency problem, ownership concentration is applied as a CG mechanism in firms. Gaur et al. (2015) noted that the role of the board in monitoring performance is critical especially when ownership is diluted with no sole owner garnering enough power to control the board. Although the principal-agent relationship persists in firms, the role of agency theory within a Shariah-compliant setting slightly differs from a conventional setting. In the former, the agency problem intensifies. For example, Shariah-compliant transactions limit managers’ options in identifying the best investment prospects which subsequently result in lower profit. More importantly, the existing of a Shariah-supervisory board (SSB) exerts a certain amount of control on the BOD in Shariah-compliant firms. SSB has significant role in influencing decision-making and hence, mitigates managers’ potential opportunistic behavior (Bindabel 2017).

Contrary to agency theory, stewardship theory posits that managers are not motivated by personal interest but instead prioritize the interest of stakeholders as a whole (Donaldson 1990; Singh and Delios 2017). Stewardship theory highlights the collective nature of managers (Davis et al. 1997) which enables them to effectively and subsequently improve the firm performance. In Shariah-compliant firms such as those that practice Islamic financing, collective decision-making practices are carried out through a consultation between the firms’ SSB and BOD (Alnasser and Muhammed 2012). Bhatti and Bhatti (2009) argued that in Shariah-compliant firms, managers also carry out their responsibilities according to Islamic values and principles. Inside directors and chief executive officers that also hold the position of board chairman (CEO duality) are important in stewardship theory as they are presumed to have a better understanding of the business and hence, make better business decisions and bringing forth better firm performance (Gaur et al. 2015).

Similar to stewardship theory, stakeholder theory postulates that in addition to profit maximization, managers also consider the social purpose of its stakeholders (Freeman 1984; Singh and Delios 2017). Stakeholders include individuals or groups who can affect or are affected by the firm. Stakeholders are also groups where the firm may have societal, environmental, and ethical implications influence over (Lee et al. 2017; Pattnaik et al. 2018; Freeman et al. 2004). The fact that Shariah-compliant businesses are primarily grounded by ethical and social responsibilities towards their stakeholders, the application of stakeholder theory in the context of Shariah-compliant firms is undisputable. Good stakeholder management, which includes board size and board competence, positively impacts firm performance (Gaur et al. 2011, 2015). In addition to stakeholder theory, previous scholarly works have also linked the role of board size and board competence to resource dependence theory (Singh et al. 2010). The resource dependence theory views board composition as value-added resources for the firm (Carpenter and Westphal 2001; Mukherjee et al. 2013). A larger board with professionally qualified members increase expertise and enhance the firm’s resources (Pfeffer 1972).

These theories give a different focus to the role of the board in influencing firm performance. Gaur et al. (2015) proposed an integration of these theories using ownership concentration (agency theory) as the vital mechanism to lessen agency problems. The current study opined that high ownership concentration limits a firm’s inclination to accommodate the preferences of all stakeholders and thus, limits the usefulness of stakeholder theory-related expectations. This study is set in the context of a Shariah economy which is unlike a conventional economy, whereby Shariah-compliant businesses focus on benefits to the society as stakeholders while maximizing return. Principles based on Islam emphasize the need for just and fairness of value and sensation of equality for stakeholders (Alnasser and Muhammed 2012). As such, besides the critical role of agency theory, the study also envisages the positive perspective of board members and adopt the resource dependency theory. Figure 1 illustrates the conceptual model for this study.

Fig. 1
figure 1

An integrative framework of Syariah firm governance and firm performance

Full size image

2.1 Ownership concentration

This study defines ownership concentration as the percentage of ownership held by top shareholders or a single owner either directly or indirectly through cross-shareholdings and/or pyramids (Gaur et al. 2014). While many studies have reported the positive influence of ownership concentration on firm performance, there are negative consequences that may arise when majority shareholders tunnel resources or transfer profits and assets systematically at the expense of minority shareholders. Hence, the premise of this study is built based on earlier findings of Malaysia (Amran and Ahmad 2013; Claessens et al. 2000) that have shown positive correlations between firm performance and ownership concentration.

H1

Ownership concentration has a positive influence on firm performance.

2.2 Board independence

There have been many conflicting arguments amongst scholars and policymakers on the influence of board independence on firm performance, either positive or negative, depending on which theory one holds—agency theory versus stewardship theory (Singh and Delios 2017). The proponents of stewardship theory (Boyd 1995; Charan 1998) argue that having insiders on the board decreases the chances of conflict within the board and makes the decision-making process more efficient as insiders have sound knowledge on the strength and weaknesses of the firm to provide strategic directions for the firm. On the other hand, the supporters of agency theory argued in favor of independence and an external Chair as a more effective form in monitoring firm performance and dissuaded managerial entrenchment since an independent board member and external Chair are free from management influence (Gaur et al. 2015; Bertoni et al. 2014). The Malaysian Code of Corporate Governance requires listed firms to have a majority or at least half of their board members to be independent directors. These criteria are also similar across Shariah-compliant firms in the country. As such, the following hypotheses are developed:

H2a

The presence of independent board members has a positive influence on firm performance.

H2b

The presence of an external Chair has a positive influence on firm performance.

H3a

The presence of inside board members has a negative influence on firm performance.

H3b

The presence of CEO duality has a negative influence on firm performance.

2.3 Board competence

Board size, defined as the total number of directors on the board (Gaur et al. 2015), indicates the monitoring and advisory roles of the board (Singh and Delios 2017; Piepenbrink and Gaur 2013). Thus far, there has been no consensus on the optimal number of board members, and the impact of board size on firm performance remains inconclusive. While some scholars and policymakers argue that a larger board size is preferable as it effectively monitors powerful managers, it is also subject to increasing costs and boardroom squabbles (Ujunwa 2012). Hence, the following hypothesis is proposed:

H4

Board size has a positive influence on firm performance.


It has been perceived that board members with higher levels of qualification provide strategic resources and act as a mix of competencies and capabilities to execute the governance function (Carpenter and Westphal 2001). Gaur et al. (2015) have used professional or specialized qualifications to proxy for high quality directors while Ujunwa (2012) used doctorate (Ph.D.) qualifications. Jermias and Gani (2014) used the criteria of university professors or government officers instead. The current study proposes that different qualification levels of board members impact firm performance. Considering the limited number of Ph.D. holders among board members of Shariah-compliant firms in the context of this study, the following hypotheses are proposed:

H5a

Board members with a bachelor’s degree have a positive influence on firm performance.

H5b

Board members with a master’s degree have a positive influence on firm performance.

H5c

Board members with other types of degrees have a positive influence on firm performance.

3 Methodology

3.1 Sample

The current study is conducted in one the first and fastest growing Islamic capital markets in the world, Malaysia. Shariah-compliant securities are managed under the Shariah Index and must comply to the Islamic-based requirements laid out by the Securities Commission of Malaysia. Pok (2012) cited Malaysia as the most liberal Islamic capital market in comparison with other prime Islamic capital markets such as Dow Jones Islamic Market (DJIM) and Financial Time Stock Exchange (FTSE) (Hooy and Ali 2017). The sample of 200 Shariah-compliant firms is drawn from 1362 listed firms on the Kuala Lumpur Stock Exchange (KLSE), and their detailed information including company financials, board of directors, corporate structures, ratings, stocks data, market research, as well as mergers and acquisitions are available from OSIRIS database. In addition, missing board characteristics data such as level of education and qualifications are supplemented using the firms’ annual reports. These data cover recent periods from 2014 to 2017, resulting in an initial sample of 569 firm-year observations. Data points that have large outliers are identified using Cook’s distance measure as the presence of outliers could distort the outcome and accuracy of the regression analysis. The Cook’s distance is calculated by measuring the observations’ leverage and residuals values—the higher the leverage and residuals, the higher is the Cook’s distance value. As general rule of thumb, if the Cook’s Distance is three times more than its mean, then the observation is considered as an outlier and will be removed. This deletion process reduced the sample size to 538 firm-year observations instead.

3.2 Variables

In this study, firm performance (dependent variable) is measured in terms of return on assets (ROA) and return of equity (ROE). The ROA is defined as the ratio of net income or operating benefit before depreciation and provisions to total assets, while the ROE is the ratio of net income or operating benefit before depreciation and provision to total equity capital. These two variables indicate management accomplishment given available assets (ROA) and shareholders’ equity (Contractor et al. 2016; Popli et al. 2017a, b). Instead of depending on one dependent variable, the inclusion of another dependent variable as an alternative performance indicator provides robustness to this empirical investigation (Gaur et al. 2015). However, it is arguable whether ROA is a better indicator of firm performance relative to ROE as there is greater variation in equity from year to year as compared to variation in assets (Gaur et al. 2015). It is to be noted that aside from ROA and ROE as measures of performance, other studies used a variety of performance measures including Tobin’s q (Dalton et al. 1998, 1999), return on revenues or sales (Daraghma and Alsinawi 2010), sales per employee (Gaur et al. 2007), firm survival (Gaur and Lu 2007), stock market returns (Gaur et al. 2013), managerial perception (Lee and Gaur 2013), and deal abandonment in the case of mergers and acquisitions (Popli et al. 2016).

The independent variables that represent BOD structure are classified into three main categories—ownership structure, board independence, and board competency. Ownership structure is represented by ownership concentration while board independence is represented by four variables: independence, CEO duality, insider representation, and external Chair. The ‘independence’ variable here is defined as the number of outside directors on the board while ‘insider representation’ is defined as the number of inside directors (i.e., managers or directors) on the board. As for ‘CEO duality’ and ‘external Chair’, both are binary variables representing the position of Chair of the board. External Chair takes the value of 1 if the Chair is an outsider and 0 otherwise, while CEO duality takes the value of 1 if the CEO is also a Chairman and 0 otherwise. Board competency is measured in terms of board size and board qualifications, which comprise bachelor’s degree, master’s degree, and other types of degrees.

Control variables refer to variables that have been identified to have a statistically significant impact on firm performance and in this study, firm age, firm size, long-term debt, and total assets are measured (Gaur et al. 2014). Firm age is defined as the number of years between the firm’s incorporation date and 2017. Berger and Udell (1998) argued that firms go through stages of financial cycle growth and capital structure transition as they age. New firms with less experience are struggling to position themselves in the market and may face diseconomies of scale as they grow while older firms can potentially reach the end of their product life cycle if they are not able to innovate themselves. These suggest the complexities of firm performance increase as firms age (Ujunwa 2012). For firm size, market capitalization is used as its proxy. Market capitalization, defined as the total market value of a firm’s outstanding shares, reflects the resource base of a firm (Gaur et al. 2015). In addition, long-term debt—loans and other financing obligations that mature after 1 year—is considered a control variable because debt holders influence the quality of governance when they impede managers from self-indulging behavior (Jensen and Meckling 1976). Long-term debt is of greater importance than short-term debt as the latter constantly changes. Finally, total assets refers to the resources owned by the firm.

The description of the variables are presented in Table 1 below. Natural logarithm is used on all the percentage variables in the study’s regression model in order to normalize the distribution and to interpret them in elasticity form. The count and binary variables remain as they are.

Table 1 Definition of variables
Full size table

3.3 Analytical procedure

The structure of this study’s data is a combination of time series data (year 2014–2017) and cross-sectional data (201 firms), forming a balanced panel data. The first step is to graphically examine the relationship between each independent variable and ROA/ROE to identify if functional transformation is needed to better capture the variation between them. In order to ease interpretation and normalize the distribution, the natural logarithm is used on non-binary, non-counting variables such as ROA, ROE, ownership concentration, long-term debt, firm size, and total assets (Gaur et al. 2007). The interpretation of log–log transformation variables will constitute elasticity. The remaining variables are left in their initial unit. Following this, the second step is to drop the outliers using Cook’s distance, which identifies high residuals by measuring how much an observation influences the overall model and how far an independent variable deviates from its mean (leverage). The high influence observations are those with distances, D > 4/Sample size. The elimination of outliers creates an unbalanced panel data.

Next, the general to specific (GTS) approach is adopted by specifying a full form ordinary least square (OLS) pooled model. In other words, a pooled OLS model brings together all time series data and cross-sectional data into one dataset and imposes a common set of parameters across them. Each variable that has highest insignificant p value is dropped one by one while ensuring that the adjusted R-square improves. From the final reduced form OLS pooled model, heterogeneity bias is tested. Although the pooled OLS model uses all information, it is the most restrictive in that the errors are assumed to be independent and identically distributed (i.i.d) and uncorrelated with the regressors. However, there is a possibility that each firm is different from one another (e.g., in terms of their managerial quality or board characteristics) and hence, the individual effect on firm performance may vary and create unobserved heterogeneity. To test for the existence of unobserved heterogeneity, the Breush-Pagan Lagrangian Multiplier test is used whereby the error term ε_t is decomposed into two independent components or composite error term: ε_it = λ_i + µ_it where λ_i is the individual-firm specific effect and µ_it is i.i.d with mean zero and variance 〖σ〗_t^2. The null hypothesis is Ho: 〖σ〗_t^2 = 0 and the alternative hypothesis is 〖σ〗_t^2 > 0. If the null hypothesis is rejected, this implies the existence of heterogeneity bias or individual firm specific effect. Next, the random effect and fixed effect panel data are tested using Hausman fixed test. The null hypothesis is Ho: Cov (λ_i,X_it) = 0 and the alternative hypothesis is Ho: Cov (λ_i,X_it) # 0. A rejection of Ho implies that the fixed effect panel data is preferable over the random effect panel data. It is to be noted that the fixed effect model will drop the dummy variables from its regression estimates.

Once the model is chosen, diagnostic tests are carried out to detect multicollinearity using variance-inflationary factor (VIF), autocorrelation using Wooldridge test for autocorrelation in panel data, and heteroscedasticity problem using modified Wald test for groupwise heteroskedasticity. To ensure the robustness of the model, ROE is replaced with ROA as the dependent variable.

The advantages of using panel data over pooled OLS model is mainly because it captures firm specific effects and hence, is able to control for individual heterogeneity. Furthermore, the dynamics of adjustment across time (within variation) and across observations (between variations) can be studied in panel data, which would otherwise not be identifiable with pure cross-sectional or pure time series data.

The full form pooled OLS model specificationFootnote 1:

$$\begin{aligned} & \Delta {\text{Log}}\left( {\text{ROE}} \right)_{\text{t}} =\upbeta_{0} +\upbeta_{1} {\text{Log}}\left( {\text{OC}} \right)_{\text{t}} +\upbeta_{2} {\text{Log(MarketCap)}}+\upbeta_{3} \log \left( {\frac{{\text{LTD}}}{{\text{TA}}}} \right)_{\text{t}} +\upbeta_{4} \log \left( {{\text{TA}}_{\text{t}} } \right. \\ & \quad +\upbeta_{5} \log \left( {\frac{{\text{IND}}}{{{\text{B.S}}}}} \right)_{\text{t}} +\upbeta_{6} \log \left( {\frac{{\text{IR}}}{{{{\text{B.S}}}}}} \right)_{\text{t}} +\upbeta_{7} \left( {\text{CEOD}} \right)_{\text{t}} +\upbeta_{8} \left( {\text{ExtChair}} \right)_{\text{t}} \\ & \quad +\upbeta_{9} {\text{Log}}\left( {\text{Age}} \right)_{\text{t}} +\upbeta_{10} {\text{Log}}\left( {\text{B.S.}} \right)_{\text{t}} +\upbeta_{11} \log \left( {\frac{{\text{B.Sc.}}}{{{\text{B.S.}}}}} \right)_{\text{t}} +\upbeta_{12} \log \left( {\frac{{\text{M.Sc.}}}{{{\text{B.S.}}}}} \right)_{\text{t}} +\upbeta_{13} \log \left( {\frac{{\text{Others}}}{{{\text{B.S.}}}}} \right)_{\text{t}} \\ & \quad +\upbeta_{14} {\text{Log}}\left( {\text{OC}} \right)_{\text{t }}\, * \,{\text{B.S.}} + \mathop \sum \limits_{{{\text{i}} = \left\{ {{\text{IND, IR, CEOD, SM, EXTCHAIR}}} \right\}}}\upbeta_{{15,{\text{i}}}} {\text{Log}}\left( {\text{OC}} \right)_{\text{t }} \,* \,{\text{i}} \\ & \quad + \mathop \sum \limits_{{{\text{s}} = \left\{ {{\text{B.Sc., M.Sc., Ph.D., Others}}} \right\}}}\upbeta_{{16,{\text{s}}}}\, {\text{Log}}\left( {\text{OC}} \right)_{\text{t }}\,*\,{\text{S }} +\upvarepsilon_{\text{t}} \end{aligned}$$
(1)

4 Results

4.1 Descriptive statistics

Table 2 shows the descriptive statistics of each variable used in this study. The average ROA and ROE for these 200 Shariah-compliant listed firms are 85% and 284% respectively. Interestingly, the maximum value ROE can reach is 716% while for ROA, it is 139%. The minimum value for both is less than 1%. On average, the number of board members is eight,Footnote 2 ranging from a minimum of one to a maximum of 15. Out of this, the average number of independent board members (47%) is twice the number of insiders (22%), scaled by average board size. However, the mean of independent board members does not reach the requirement of the Malaysian Code of Corporate Governance (MCCG 2017), which states that at least half of the board members should be independent. There are only 57 firms in the sample that satisfied this requirement. As for the leadership of the board, the results show that on average, 44% of these firms have an external Chair and 15% practice CEO duality. In the sample observations, the average age of firms is approximately 27 years, ranging from a minimum age of 1 year to a maximum age of 111 years.

Table 2 Descriptive statistics
Full size table

The average long-term debt is four times the total assets, and its maximum can be 1500 times more. Both total assets and market capitalization have averages close to RM1.2 million, while their maximum values being RM22 million and RM31 million respectively.

In terms of board competency, about 85% of the board members have a bachelor’s degree, 11% have additional qualifications including professional recognition, and only 3% have a master’s degree. The number of board members with a Ph.D. is very small indeed and hence, excluded from the study.

4.2 Correlation matrix

Table 3 presents 105 results of intercorrelation between the studied variables. The coefficients, whether positive or negative, are generally weak, indicating the possibility that the degree of a multicollinearity problem is too weak to influence the regression estimates. The highest correlation is 0.93 between ROA and ROE, implying a good substitution between these two variables for robustness testing. Similarly, the next highest correlation is 0.61 between total assets and market capitalization, implying a good substitution for these two variables as representatives of firm size. Another high correlation found is 0.46 between market capitalization and ownership concentration.

Table 3 Correlation matrix
Full size table

4.3 Regression results

Table 4 presents the reduced form models using pooled OLS (POLS), random effect (REM), and random effect generalized least square (GLS). The results present unstandardized beta coefficients and standard error in parenthesis. Based on the initial full form specification in Eq. (1), insignificant variables with high p value and high influence of multi-collinearity are removed and only relevant variables are left in Table 4. Since the Breush-Pagan Lagrange Multiplier null hypothesis is rejected, it is concluded that unobserved firm specific effects exist, and hence the random effect model is preferable over pooled OLS. The next step is to compare the random effect and fixed effect models using the Hausman test and since the study failed to reject the null hypothesis, this implies that the random effect model is preferable over the fixed effect model. The fixed effect model automatically excludes any binary variables or interaction terms that involve binary variables. Once model selection is specified, the diagnostic tests show that the model suffers from heteroscedasticity and serial correlation/autocorrelation problems. The problem of heteroscedasticity is corrected using robust standard errors or White-Huber standard errors. The problem of serial correlation is corrected by estimating the regression model using limited or no control for within cluster error correction, and obtaining the post-estimation again. The post-diagnostic tests estimations are shown in the fourth and fifth columns of Table 4.

Table 4 Regression results
Full size table

H1 predicts a positive relationship between ownership concentration and firm performance. All estimated regression models show a positive significant coefficient. However, the ownership concentration for ownership type ‘independent’ and ‘external Chair’ are statistically insignificant and hence, excluded from the model. Only the ownership concentration among insiders is statistically significant and found to be negatively affecting the ROE.

H2a suggests that the presence of independent board members will positively influence firm performance. All the four regression models (with exception to robust random effect model) statistically support this hypothesis. A 10% increase in the ratio of independent board members over total board members led to about 2% increase in the firm’s ROE. It is important to note that the independent board members need to have qualifications beyond a bachelor’s degree as having a bachelor’s degree only will negatively affect the ROE, as shown in the coefficient of interaction term ‘independent * bachelor’.

H2b suggests the presence of an external Chair will positively influence firm performance; this hypothesis is supported by the coefficient ‘external Chair’ in all estimated models. The results show that having an external Chair will improve the firms’ performance by 10% relative to not having an external Chair. However, similar to independent board members, the appointed external Chair needs to have a higher qualification because having a bachelor’s degree alone will negatively affect firm performance, as indicated in the coefficient of the interaction term, ‘external Chair * bachelor’.

H3a predicts a negative relationship between insider representation and firm performance. The positive significant coefficient of the insider representation variable in random effect GLS model did not support this hypothesis. The results instead show that having insiders on the board positively affect the ROE. It is also to be noted that when the ownership concentration is in the hand of the insiders, the effect of variable ‘ownership concentration * insider representation’ towards firm performance becomes negative. H3b posits the negative relationship of CEO duality and firm performance and this is supported by all estimated models.

For H4, which suggests that board size has a positive influence on firm performance, the results show that the impact is negative as the board expands. This implies that as the board size increases, the degree of its impact on ROE reduces. In terms of board competency, hypotheses H5a, H5b, and H5c suggest a positive relationship between level of education and ROE. The results show that the greater qualifications of board members can offer greater competencies, capabilities, and strategic resources in managing the firm, consistent with resource dependency theory.

For the control variables, the firm size as represented by market capitalization has a statistically positive effect on ROE but an expansionary of the firm’s total assets will not necessarily have a positive effect on ROE, as shown by the negative coefficient of the variable total assets. The negative impact of total assets suggests that the firm is failing to better utilize its assets and is most likely experiencing diseconomies of scale. Long-term debt is also shown not to be an efficient tool for ROE since the coefficient is negative. Firm age is negatively related to ROE, suggesting the possibility that greater experience in managerial knowledge will not necessarily improve firm performance.

4.4 Robustness test using ROA

Aside from diagnostic checking to ensure the consistency and efficiency in the estimated models, robustness testing is also conducted by substituting ROE with ROA. The results are presented in Table 5. One notable difference here is that the pooled OLS estimation does not suffer from heteroscedasticity and serial correlation problems. Hence, there is no need to re-estimate the model using robust and clustering standard errors. However, the random effect model does suffer from panel data groupwise heteroscedasticity and hence, is corrected with robust, clustering standard errors. Another difference is that ownership concentration and insider representation are not statistically significant in all the models despite having positive coefficients. The variables ‘independence’ is statistically significant on ROE but not on ROA in random effect robust estimation. Overall, in both cases, the key explanatory variables yield same signs on their coefficients although the magnitude differs.

Table 5 Robustness test using ROA as dependent variable
Full size table

5 Discussion and conclusion

This study started with three main objectives. Firstly, to explore the role of CG in influencing firm performance among Shariah-compliant firms. Secondly, to make a relative comparison to Gaur’s et al. (2015) multi-theoretic framework in the context that is influenced by a religious legal system, focusing on resource dependency theory and stakeholder theory instead of agency theory. Thirdly, to provide empirical evidence on the efficacy of the BOD among Shariah-compliant firms. The analysis covers 200 Shariah-compliant firms listed on the KLSE, Malaysia and the study began by specifying a full form model covering variables that represent ownership structure, board independence, board competency, debt structure, and firm characteristics. The statistical tests show that the preferred model was random effect with robust standard error that corrected for unobserved heterogeneity between firms and serial correlation in the model.

Consistent with Gaur et al. (2015), this study found that ownership concentration has a statistically positive influence on firm performance as measured by ROA and ROE. The study also found that ownership concentration in the hands of insider representation would affect the performance negatively. Between the external Chair and CEO duality, the findings show that CEO duality has a statistically negative effect on firm performance, supporting agency theory, while having an external Chair impact firm performance positively. However, the external Chair needs to be someone that has a qualification beyond a bachelor’s degree or otherwise the impact on firm performance will be negative instead. Both independent board members and inside board members will affect ROA and ROE positively in all estimation models except in random effect using robust standard error model. Similarly, independent board members must have a qualification beyond a bachelor’s degree or the effect will otherwise be statistically negative instead.

In terms of board competency, the statistically positive impact of board members with a bachelor’s degree, master’s degree or any other type of degree on ROA and ROE supports the resource dependence theory and implies the need for board members to have a high level of education. Achievement in higher education is associated with an individual’s ability for critical thinking, problem solving, innovation, and creativity. As such, Darmadi’s (2013) study in Indonesia’s emerging market, a neighboring country of Malaysia, found that postgraduate degrees held by BOD members significantly increase firm performance.

Contrary to Gaur et al. (2015), the current study’s findings indicate that the expansion of board size negatively affects firm performance. Ujunwa (2012) attributed the negative effect of an increased board size to rising cost and boardroom disputes. The current study’s findings show that an increase in board members with only a bachelor’s degree proves to have a negative correlation to firm performance. Instead, board members with additional qualifications such as a master’s degree or a professional certification will have a positive influence on firm performance. In addition, firm age has a negative correlation with profitability, implying that as firms get older, their earnings start to decline due to diseconomies of scale. This finding is against the findings of Berger and Udell (1998) and Gregory et al. (2005).

Overall, from a theoretical standpoint of agency theory, the current findings on ownership concentration do not differ much from findings done in a conventional economy. However, the “one-size fits all” CG approach based on agency theory is not entirely supported in the context of a Shariah-compliant economy. This study sees the need to adopt Gaur’s et al. (2015) multi-theoretic framework in examining CG in Shariah-compliant firms which are influenced by the Islamic legal system. Findings from this study recommends that in addition to agency theory, stakeholder theory or resource-based view is equally important in mitigating the relationship between stewardship theory and firm performance. With that argument in mind, the key findings support the requirement of for board members to have a qualification beyond a bachelor’s degree to ensure a positive impact on firm performance. This supports the stakeholder theory or resource-based theoretical argument. From a practical sense, board members with greater knowledge and skills will ensure compliance with Shariah law.

In order to draw a more distinctive conclusion between the performance of Shariah-compliant firms and their conventional counterparts, further research that includes board gender, board nationality, and board ethnicity are required. This study has drawn samples from the Shariah-compliant firms listed in the KLSE. To improve the generalizability of these findings, this study should be expanded to other countries that have Shariah-listed firms to provide better insights into their performance. Future research should also consider cross-country examination on the findings for benchmarking purposes.