Corporate Governance And Firm Performance In Australia: A Structural Equation Model Analysis

Assumptions Made

Corporate governance entails advocacy for law compliance as well as ethical conduct demonstration (Datta, 2018), which, is in order to ensure that the company is well aligned to both the society’s ethical concerns and existing regulations for both profitability and sustainability prospects. Over the recent years, business performance has been associated with a number of factors apart from the number of consumers for their products. Such factors include corporate governance. Beth (2003) in her article on corporate governance and firm performance notes that “the belief that governance best practices lead to superior firm performance is widespread” a notion which Beth refutes to not always be true. A number of studies to examine whether governance best practices are reflected in the performance of the company have been conducted, including Azim (2012) study on “Corporate governance mechanisms

and their impact on company performance”

1.2 Purpose of study

The purpose of our study is to replicate the studies data analysis and output done by Mohammed Azim on “Corporate governance mechanisms and their impact on company performance: A structural equation model analysis” Azim (2012) using Structural equation modelling and Ordinary Least Squares (OLS) Regression. Our focus will be to prove the theories presented by Azim (2012). Additionally, the study seeks to determine the nature of corporate monitoring mechanisms also addressed as structures in this study. The mechanisms include:

1. Shareholders
2. External auditors

• Executive board

1.3 Research questions

The research questions for our study are as those used in the original study by Azim.

1. Is there a substitution effect in the corporate monitoring structures? (Azim, 2012)
2. Is there a complementary effect among the featured monitoring mechanisms? (Azim, 2012)

1.4 Keywords

Structural equation modelling, Ordinary Least Squares regression, Corporate governance, Stakeholders

2.1 Data

The sample data used for this study was obtained from SIRCA and Morning Star DatAnalysis for 613 Australia’s big companies for the financial year of 2015. It contains details of shareholders’ activities as well as different monitoring bodies such as auditors. The data has 9 variables with various levels for the 613 companies, hence the total number of variables is 22.

2.2 Assumptions

During our study we made the following assumptions:

1. The data for the study is normally distributed
2. In order to ensure originality of data analysis we assumed that were no missing values in the data-set and therefore did not explore for missing data, impute or reconstruct

• We also assumed that the sample data was representative enough for estimation of the population distribution as well as sufficient for testing of hypotheses

1. The dependent and mediating variables for the SEM are continuous
2. The residuals from a SEM analysis are univariate normally distributed

From table 1, the top 20 shareholders hold approximately 59.2286% of the shares making them the majority shareholders while the top 1 have 22.81424 of the shares in the data-set of all the companies in our study. The average board size for our study was 10 where the minimum board size was 3 while the maximum board size was 33. The average number of board meeting were 10, however some company’s did not hold board meetings for the financial year 2015 as from the data provided, nevertheless, the maxim number of board meetings were 47. Approximately 22.7689% of the board of directors were independent in the companies. 24.59% of the board the board members in 2015 had some financial literacy. The average number of audit meetings convened were between 0-35. Additionally, the  

1- Descriptive statistics

Descriptive Statistics
	N	Minimum	Maximum	Mean	Std. Deviation	Variance	Kurtosis
	Statistic	Statistic	Statistic	Statistic	Statistic	Statistic	Statistic	Std. Error
BIG4	608	0	1	.55	.498	.248	-1.961	.198
PNAF	610	0	8	.40	.803	.645	43.732	.198
Top 20	613	.00	99.99	59.2286	22.81424	520.490	.154	.197
TOP1	546	.0367	.8848	.227689	.1664301	.028	2.149	.209
BSIZ	613	3	33	10.26	4.723	22.304	1.491	.197
PBFL	613	.000000000000	1.000000000000	.24590755629970	.177966572243531	.032	1.738	.197
PBIN	613	.000000000000	.800000000000	.26472272508730	.192304398302462	.037	-.798	.197
BM	613	0	47	10.04	5.044	25.438	8.333	.197
CHCE	613	0	1	.89	.318	.101	3.928	.197
ACM	612	0	35	2.83	2.648	7.010	35.702	.197
PAFL	613	.000000000000	1.000000000000	.28950904994951	.372833157671259	.139	-.628	.197
PAI	613	.000000000000	1.000000000000	.43898081255341	.440857511921781	.194	-1.707	.197
NCM	613	0	24	1.18	1.962	3.851	30.218	.197
PNI	613	0	1	.27	.409	.167	-.814	.197
RCM	613	0	24	1.68	2.209	4.880	17.660	.197
PRI	613	0	1	.36	.439	.192	-1.519	.197
Log( TA)	606	4	12	7.92	1.129	1.274	.786	.198
Valid N (listwise)	534

In testing for correlation between the various variables, using the Spearman correlation matrix, it is noted that there were a number of variables indicating high correlations. For instance from table 2 below, there is a high correlation between number of remuneration meetings and the proportion of independent members (0.708>0.6). Nevertheless, the use of Structural Equation Modelling will enable the management of multicollinearity problem.

Analysis Results

2-Spearman correlation test

Correlations
			BIG4	PNAF	Top 20	TOP1	BSIZ	PBFL	PBIN	BM	CHCE	ACM	PAFL	PAI	NCM	PNI	RCM	PRI
Spearman’s rho	BIG4	Correlation Coefficient	1.000	.190**	.115**	.047	.385**	-.072*	.186**	.157**	.064	.311**	.161**	.245**	.267**	.215**	.240**	.158**
Sig. (1-tailed)	.	.000	.002	.139	.000	.038	.000	.000	.058	.000	.000	.000	.000	.000	.000	.000
N	608	605	608	542	608	608	608	608	608	607	608	608	608	608	608	608
PNAF	Correlation Coefficient	.190**	1.000	.036	-.038	.298**	-.036	.092*	.149**	-.016	.183**	.120**	.125**	.190**	.141**	.212**	.168**
Sig. (1-tailed)	.000	.	.190	.188	.000	.187	.011	.000	.347	.000	.001	.001	.000	.000	.000	.000
N	605	610	610	544	610	610	610	610	610	609	610	610	610	610	610	610
Top 20	Correlation Coefficient	.115**	.036	1.000	.359**	.080*	-.022	-.077*	.023	-.049	.066	.038	.022	-.023	-.047	.053	.017
Sig. (1-tailed)	.002	.190	.	.000	.023	.290	.029	.288	.113	.051	.171	.292	.281	.124	.093	.334
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
TOP1	Correlation Coefficient	.047	-.038	.359**	1.000	-.071*	.006	-.195**	-.064	-.138**	-.091*	.016	-.091*	-.092*	-.086*	-.108**	-.118**
Sig. (1-tailed)	.139	.188	.000	.	.049	.441	.000	.067	.001	.016	.356	.017	.016	.022	.006	.003
N	542	544	546	546	546	546	546	546	546	545	546	546	546	546	546	546
BSIZ	Correlation Coefficient	.385**	.298**	.080*	-.071*	1.000	-.155**	.161**	.359**	.158**	.545**	.235**	.419**	.493**	.396**	.492**	.335**
Sig. (1-tailed)	.000	.000	.023	.049	.	.000	.000	.000	.000	.000	.000	.000	.000	.000	.000	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
PBFL	Correlation Coefficient	-.072*	-.036	-.022	.006	-.155**	1.000	.075*	-.069*	.009	-.046	.399**	-.051	-.068*	-.081*	-.056	-.023
Sig. (1-tailed)	.038	.187	.290	.441	.000	.	.033	.045	.412	.130	.000	.105	.045	.022	.085	.282
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
PBIN	Correlation Coefficient	.186**	.092*	-.077*	-.195**	.161**	.075*	1.000	.115**	.128**	.221**	.091*	.588**	.218**	.408**	.215**	.461**
Sig. (1-tailed)	.000	.011	.029	.000	.000	.033	.	.002	.001	.000	.012	.000	.000	.000	.000	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
BM	Correlation Coefficient	.157**	.149**	.023	-.064	.359**	-.069*	.115**	1.000	.152**	.358**	.138**	.217**	.243**	.194**	.342**	.266**
Sig. (1-tailed)	.000	.000	.288	.067	.000	.045	.002	.	.000	.000	.000	.000	.000	.000	.000	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
CHCE	Correlation Coefficient	.064	-.016	-.049	-.138**	.158**	.009	.128**	.152**	1.000	.199**	.151**	.161**	.155**	.078*	.108**	.089*
Sig. (1-tailed)	.058	.347	.113	.001	.000	.412	.001	.000	.	.000	.000	.000	.000	.027	.004	.014
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
ACM	Correlation Coefficient	.311**	.183**	.066	-.091*	.545**	-.046	.221**	.358**	.199**	1.000	.445**	.556**	.438**	.310**	.474**	.316**
Sig. (1-tailed)	.000	.000	.051	.016	.000	.130	.000	.000	.000	.	.000	.000	.000	.000	.000	.000
N	607	609	612	545	612	612	612	612	612	612	612	612	612	612	612	612
PAFL	Correlation Coefficient	.161**	.120**	.038	.016	.235**	.399**	.091*	.138**	.151**	.445**	1.000	.324**	.123**	.082*	.190**	.128**
Sig. (1-tailed)	.000	.001	.171	.356	.000	.000	.012	.000	.000	.000	.	.000	.001	.021	.000	.001
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
PAI	Correlation Coefficient	.245**	.125**	.022	-.091*	.419**	-.051	.588**	.217**	.161**	.556**	.324**	1.000	.281**	.364**	.307**	.434**
Sig. (1-tailed)	.000	.001	.292	.017	.000	.105	.000	.000	.000	.000	.000	.	.000	.000	.000	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
NCM	Correlation Coefficient	.267**	.190**	-.023	-.092*	.493**	-.068*	.218**	.243**	.155**	.438**	.123**	.281**	1.000	.708**	.571**	.374**
Sig. (1-tailed)	.000	.000	.281	.016	.000	.045	.000	.000	.000	.000	.001	.000	.	.000	.000	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
PNI	Correlation Coefficient	.215**	.141**	-.047	-.086*	.396**	-.081*	.408**	.194**	.078*	.310**	.082*	.364**	.708**	1.000	.389**	.536**
Sig. (1-tailed)	.000	.000	.124	.022	.000	.022	.000	.000	.027	.000	.021	.000	.000	.	.000	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
RCM	Correlation Coefficient	.240**	.212**	.053	-.108**	.492**	-.056	.215**	.342**	.108**	.474**	.190**	.307**	.571**	.389**	1.000	.664**
Sig. (1-tailed)	.000	.000	.093	.006	.000	.085	.000	.000	.004	.000	.000	.000	.000	.000	.	.000
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
PRI	Correlation Coefficient	.158**	.168**	.017	-.118**	.335**	-.023	.461**	.266**	.089*	.316**	.128**	.434**	.374**	.536**	.664**	1.000
Sig. (1-tailed)	.000	.000	.334	.003	.000	.282	.000	.000	.014	.000	.001	.000	.000	.000	.000	.
N	608	610	613	546	613	613	613	613	613	612	613	613	613	613	613	613
**. Correlation is significant at the 0.01 level (1-tailed).
*. Correlation is significant at the 0.05 level (1-tailed).

During analysis, data relating to 2015 financial year for 613 Australian companies was used for both Structural equation modelling and Ordinary Least Squares regression. The model was run to examine how monitoring variables affected the performance of the company. The data-set was tested using least squares and after w results compared against OLS.

Lewi (2017) argues that, “in order to find out whether and how the observed value of a given phenomena is significantly different from the expected value…” the chi-square ought to be used. The ch-square test involves comparison of how the computed p-value is different from the significance level provided which helps predict how well the distance between the fitted line and other data points is minimized. The model has high chi-square values that range from 31.2-2886.02 indicating that there is a relatively strong interrelationship between the monitoring variables. This is true given that the scaled Pearson chi-square is at 536.000 while the Pearson chi-square is at 588.007. The RMSE range from approximately 0.0028 to 0.0627 which according to Browne and Cudeck (1993) indicate a good fit since it is ?0.08.

3- Root mean square error of approximation

Variable	Obs	RMSE	F	P
pbv	536	0.0028	.3714427	0.8290
roe	536	0.0083	1.116927	0.3476
roa	536	0.0168	2.269835	0.0606
dy	536	0.0627	8.886511	0.0000

Goodness of Fitb
	Value	df	Value/df
Deviance	588.007	535	1.099
Scaled Deviance	536.000	535
Pearson Chi-Square	588.007	535	1.099
Scaled Pearson Chi-Square	536.000	535
Log Likelihooda	-785.369
Akaike’s Information Criterion (AIC)	1574.738
Finite Sample Corrected AIC (AICC)	1574.761
Bayesian Information Criterion (BIC)	1583.306
Consistent AIC (CAIC)	1585.306
Dependent Variable: Log( TA) Model: (Intercept)
a. The full log likelihood function is displayed and used in computing information criteria.
b. Information criteria are in small-is-better form.

In the study model, there are three groups of monitoring variables which comprise of:

1. Shareholders
2. External auditors

• Board of directors

The study question involves whether monitoring variables have a substitutional or complimentary effect. Jaffar and Zaleha (2016) in their research on the role of monitoring mechanisms on a firms performance argue that, generally, “The monitoring role of corporate governance as measured by the composition of independent board members have shown a positive significant effect on the company’s performance.” Practically, various monitoring mechanisms work together to streamline the common shareholder-executive interest clash. Ideally, the executive are mandated to check on the audits done by external auditors hence exacting an influence on the audit process. The process of auditing often results in reports which eventually reach the shareholder who acts as a watchdog for the auditors (both external and internal).

4- Relationship between Monitoring variables and Performance variables

Coefficientsa,b
Model	Unstandardized Coefficients	Standardized Coefficients	t	Sig.
B	Std. Error	Beta
1	(Constant)	-.610	.299		-2.041	.042
Top 20	.004	.004	.054	1.039	.299
TOP1	-.039	.486	-.004	-.080	.936
BSIZ	.028	.016	.085	1.720	.086
a. Dependent Variable: ROE
b. Weighted Least Squares Regression – Weighted by ACM

The table indicates a correlation between the performance variables and the monitoring variables. This is also true for the research done by Azim (2012) which indicated existence of correlation between the shareholders, auditors and the executive board as monitoring variables. The correlation coefficient between return on equity and shareholders is approximately 0.072424.Indicating that shareholders do affect the return on equity

In exploration of the relationship between the board, shareholders and the auditors in researching the existence of substitution and complementary relationship in the corporate monitoring tools, the study used Ordinary least squares regression for the performance variable Return on equity since it is the only performance variable that had enough observations for OLS regression. R² was 18.18% ;  adjusted R² = 15.65% while ANOVA: F = 7.19 with a p-value of 0.000. However, the SEM results and those of OLS regression are considerably different and hence inconsistent. Such may be due to factors such as:

1. Error mi-specification in the model which may include omission of an important explanatory variable or inclusion of a redundant exogenous variables
2. Incorrect specification of the model

1-Ordinary least Squares regression

When testing for robustness in the model, the structural equation modelling results show that there is a substitution and complementary relationship between the monitoring variables (shareholders, board of directors and auditors). The results were higher than those of Azim (2012).

5-Robust regression

	Number of obs	= 539
F( 9, 529)	= 8.34
Prob > F	= 0.0000
roe	Coef.	Std. Err.	t	P>t	[95% Conf.	Interval]
top20	.0015495	.0005109	3.03	0.003	.0005459	.0025532
top1	.1348009	.0662633	2.03	0.042	.0046293	.2649725
acm	.0163219	.0060309	2.71	0.007	.0044744	.0281694
pafl	-.0016783	.0298049	-0.06	0.955	-.0602288	.0568721
pai	.0462693	.0290546	1.59	0.112	-.0108072	.1033458
ncm	.0121012	.009268	1.31	0.192	-.0061054	.0303078
pni	-.0333545	.0335505	-0.99	0.321	-.0992631	.032554
rcm	.0128354	.0080525	1.59	0.112	-.0029833	.0286542
pri	.0238044	.0314163	0.76	0.449	-.0379116	.0855204
_cons	-.2215605	.034423	-6.44	0.000	-.2891831	-.1539379

When the financial crisis of 2007-2008 came, it seems evident that the monitoring structures of listed companies could not be able to prevent the decisions that were apparently risky to make. Azim (2012) argues that this the scenario had a thwarting effect on the investors’ trust on the executives ability to act as a monitoring variable. From his research, the effects of the GFC lowered the coefficients of the board. However the data from 2015 financial year for major listed companies in Australia indicate that the coefficients of the board is way higher than when Azim conducted the featured research at .0015495n which is higher than −0.073 (0.014) at a 0.05 level of significance.

Conclusion

Despite the scarcity of utilization of SEM, Azim (2012) through use of Ordinary Least Square regression and Structural Equation modelling manages to prove the existence of a cause-effect relationship between the monitoring structures of companies which include: shareholders, executive board, and the company auditors. As such our study has proven that there actually is a relationship between the monitoring structures and the performance of listed companies. However, the study shows that the monitoring variables have substitutional effects as they are complimentary. Depending on the context, this indicates that in presence of co-ordination between a company’s there is a likelihood to promote the company performance in case of GFC. Nevertheless, this study indicate a positive improvement between the trust of investors on the monitoring tools set up by a company, more-so the board following an improvement on the correlation coefficient between the monitoring structures (variables) of a company and the company performance, in line with Azim’s 2012 research which point out to “The practical reality of the diversity of the study…” (Azim, 2012). Therefore the research succeeded in replicating the results of Azim (2012). However there were a number of limitations for the study which included, the inability to transform the data entirely, for a better projection of the effect of monitoring factors on company performance, such limitations should be mitigated to enhance the whole process.

Azim, M. (2012). Corporate governance mechanisms and their impact on company performance: A structural equation model analysis. Australian Journal of Management. 37(3), 481 –505.

Grace, M.(2016). SPSS GLM: Choosing Fixed Factors and Covariates.[Online]. Available from: https://www.theanalysisfactor.com/spss-glm-choosing-fixed-factors-and-covariates/comment-page-1/. Accessed 27^th Jun 2018.

Agrawal, A., & Chadha, S (2005).Corporate Governance and Accounting Scandals.Journal of Law and Economics. 10(4), 56- 78

Appasamy, C., Lamport, M., Seetanah, B. &Sannasse, V.R. (2013).Corporate governance and firm performance: evidence from the insurance industry of Mauritius, Cambridge business and economics conference.