Impacts of corporate governance on risks and financial performance of commercial banks in Vietnam

Economic growth Bank risk Z-Score NPL ROA ROE NIM Financial 11 Figure 3.1. Research process Research issues Impact of Corporate Governance on risks and financial efficiency Objectives of the study - Testing the impact of CG on risks of commercial banks in Vietnam. - Testing the impact of CG on financial efficiency of commercial banks in Vietnam. - Proposing a number of policy implications to improve the CG management capacity, limit risks and improve financial efficiency of commercial banks in Vietnam. - Theoretical foundations of CG, risks and financial efficiency - Research overview of the impact of CG on risks and financial efficiency of banks Conclusions and recommendations Results and discussions Using the SGMM method for goal 1 Using GLS method for goal 2 Research Methods Collecting and processing data, analyzing and estimating models Study gaps and analytical framework Proposing an experimental research model 12 3.2. Research data The study uses secondary data with a data sample of 29 commercial banks in Vietnam from 2011-2017. As of December 31, 2017, according to the State Bank’s statistics, the number of commercial banks is 35 banks (including 7 State-owned commercial banks and 28 joint-stock commercial banks). The total assets of 35 commercial banks as of December 31, 2017 are VND 8,598,594 billion, while the total assets of the 29 commercial banks used by the author as of December 31, 2017 are VND 7,761,728 billion, accounting for 90.3% of total assets of commercial banks. Thus, 29 commercial banks are selected by the author which ensure to represent commercial banks in Vietnam (Appendix 1). In which, the State-owned commercial banks include 7 banks: Vietnam Bank for Agriculture and Rural Development, Vietnam Joint Stock Commercial Bank for Industry and Trade (CTG), Joint Stock Commercial Bank for Foreign Trade of Vietnam (VCB), Joint Stock Commercial Bank for Investment and Development of Vietnam (BID), Vietnam Construction Commercial Joint Stock Bank, Global Petro Sole Member Limited Commercial Bank, Ocean Limited Company. In the sample, due to limited information, the state-owned commercial banks are included 3 banks by the author: CTG, VCB and BID. The remaining 26 banks of the sample belong to joint stock commercial banks. According to updated data as of December 31, 2017, the number of banks listed on HOSE and HXN exchanges was 10 banks, including: CTG, VCB, BID, ACB, EIB, MBB, NCB, SHB, STB and VPB. The remaining samples are unlisted banks. Data for calculating internal variables inside the bank are collected from annual reports, audited consolidated financial statements, corporate governance reports, annual shareholder meeting documents of commercial banks. Data for calculating external factors in the macro environment are collected from official sources such as the World Economic Outlook (WEO) data set of the International Monetary Fund (IMF), General Statistics Office of Vietnam. The data is collected and selected after eliminating banks that does not disclose information or disclose incomplete information, resulting in a sample of balance sheet data study consisting of 29 banks with 203 observations to be used for research purposes. Therefore, the data set will be in a balanced form and is presented in Appendix 2. 3.3. Measuring the impact of corporate governance on the risks of commercial banks in Vietnam 3.3.1. Research models + Based on the research models of Pathan and Faff (2013), Dong et al (2017), this study applies the following regression model 1 as follows: 𝐑𝐢𝐬𝐤𝑖𝑡 = 𝛼0 + 𝛼 ∗ 𝐑𝐢𝐬𝐤𝑖𝑡−1 + 𝛾 ∗ 𝐶𝐺𝑖𝑡 + 𝛿 ∗ 𝑋𝑖𝑡 + 𝜀𝑖𝑡 (1) Of which: 13 𝛼0: Coefficient of original coordinates; i: Cross data of banks; t: Current year (t = 1,.,k); Risk𝑖𝑡: Risk of bank i (Z-score, NPL) at time t; CGit: The variables representing the Corporate Governance of bank i at time t, including: size of the Board of Directors (Bsize), Number of Independent Members of Board of Directors (Bindep), Number of female members of Board of Directors (Femdir), The percentage of foreign members of Board of Directors (Fordir), the percentage of members of Board of Directors taking part in management (Execdir), and Education level of Board of Board of Directors (Edu). The variable adjusted to suit the Vietnamese context in the research model is the variable of the percentage of members of Board of Directors taking part in management (Execdir), as defined in Paragraph 1, Article 34 of the 2010 Law on Credit Institutions of Vietnam: “Chairman of the Board of Directors, Chairman of the Board of Members of a credit institution must not concurrently be the executive of that credit institution and of other credit institutions”. Therefore, in this thesis, the author uses the variable of the members of Board of Directors to participate in the management as compared to the previous studies, the majority used the duality variable (Chairman of the Board of Directors cum CEO). Xit: The control variables include bank characteristics and macro variables: bank size (SIZE), size of lending activities (LAR), equity size (CAP), Loan to Deposit ratio (LDR), Bank Liquidation (LIQ), Management Effectiveness (CTI), Listed Bank (List) and Economic Growth (Ecogrow). 𝛼 ,γ,δ: Are the estimated coefficient vectors. ε𝑖𝑡:: Is the standard error. 3.3.2. Measurement of variables in the research model 3.3.2.1. Risk dependent variable The thesis measures the risks of Vietnamese commercial banks by the Z-score bankruptcy risk index inherited from the research of Boyd and Graham (1986), Goyeau and Tarazi (1992); Barry et al (2011) and Lepetit and Strobel (2013) and ratio of bad debts (NPL). - Z-score is calculated based on the following formula: 𝑍𝑖𝑡 = 𝑅𝑂𝐴𝑖𝑡 + 𝐸𝑇𝐴𝑖𝑡𝜎(𝑅𝑂𝐴) In which: + ROA: Return on assets of bank i at time t. + ETA: Equity to asset ratio of bank i at time t. 14 + σ(ROA): Standard deviation of returns on the assets of the entire sample. A higher Z-score indicates that the bank is more stable and less risky. Because the Z-score has a high deviation, according to Laeven and Levine (2009), to reduce the bias, the natural logarithm of Z-score should be used. Z-score is often used in studies to measure bank risk (for example: Angkinand and Wihlborg, 2010; Barry et al, 2011; Demirgüç-Kunt and Huizinga, 2013; Laeven and Levine, 2009) . - The traditional risk of banks is often related to lending and is measured by the ratio of bad debt to total outstanding debt (NPL), which reflects the quality of a bank’s assets (Demirgüç –Kunt et al, 2006; Shehzad et al., 2010; and Delis and Kouretas, 2011). Because bad debts cause losses for banks, high NPL ratios lead to high credit risks (Delis and Kouretas, 2011). 3.2.2.2. Independent variables for corporate governance in the model a) Size of Board of Directors Hypothesis 1a (H1a): Large size of Board of Directors is positively correlated with the bank’s Z-Score. Hypothesis 1b (H1b): Large size of Board of Directors is negatively correlated with the bank’s NPL. b) Independent members of Board of Directors Hypothesis 2a (H2a): The percentage of independent members of the Board of Directors is positively correlated with the bank’s Z-Score. Hypothesis 2b (H2b): The percentage of independent members in the Board of Directors is negatively correlated with the bank’s NPL. c) Female members of Board of Directors Hypothesis 3a (H3a): The high percentage of female board members is positively correlated with the bank’s Z-Score. Hypothesis 3b (H3b): The high percentage of female members in the Board of Directors is negatively correlated with the bank’s NPL. d) Foreign members of Board of Directors Hypothesis 4a (H4a): The high percentage of foreign members of Board of Directors is positively correlated with the bank’s Z-Score. Hypothesis 4b (H4b): The high percentage of foreigner members of Board of Directors is positively correlated with the bank’s NPL. e) Percentage of members of Board of Directors taking part in management Hypothesis 5a (H5a): The high percentage of board members taking part in management is negatively correlated with the bank’s Z-Score. 15 Hypothesis 5b (H5b): The high percentage of members of Board of Directors taking part in management is negatively correlated with the bank’s NPL. f) Education level of Board of Directors Hypothesis 6a (H6a): The high percentage of members of Board of Directors with postgraduate degrees is significantly correlated with the bank’s Z-Score. Hypothesis 6b (H6b): The high percentage of board members with postgraduate degrees is significantly correlated with the bank’s Z-Score. 3.3.2.3. Control variables in the model Table 3.1. Description of the variables used in the regression model 1 Variable Measurement method Scientific basis Expected mark Z-Score NPL Dependent variable (Risk) NPL NPL to total outstanding loan PL NPL to total outstanding loan Dong et al (2014), Berger et al (2014), Berger et al (2016), Calomiris and Carlson (2016), Dong et al (2017), Skała and Weill (2018) Z-Score = ln (𝑅𝑂𝐴𝑖𝑡 + 𝐸𝑇𝐴𝑖𝑡 𝜎(𝑅𝑂𝐴) ) Pathan (2009); Anginer et al (2014); Dong et al (2014); Chan et al (2016); Berger et al (2016); Mollah et al (2017); Ben Zeineb and Mensi (2018); Skała and Weill (2018); Setiyono and Tarazi (2018) Independent variable (Corporate Governance - representative variable of Corporate governance) Bsize Natural logarithm of BOD Berger et al (2014), Dong et al (2017), Ben Zeineb and Mensi (2018) + - Bindep Percentage of independent members of Board of Directors/Total members Chan et al (2016), Dong et al (2017) + - 16 Variable Measurement method Scientific basis Expected mark Z-Score NPL of Board of Directors Femdir Percentage of female members of Board of Directors/Total members of Board of Directors Dong et al (2014), Dong et al (2017) + - Fordir Percentage of foreign members of Board of Directors/Total members of Board of Directors Dong et al (2017) + - Execdir Percentage of members of the Board of Directors taking part in management/Total number of members of Board of Directors The author proposes to comply with Paragraph 1, Article 34 of the Law on Credit Institutions in 2010 of Vietnam. - - Edu Percentage of members of Board of Directors with postgraduate level/Total number of members of Board of Directors Berger et al (2014), Chan et al (2016), Setiyono and Tarazi (2018) + - Control Variables Size Natural logarithm of total assets Pathan (2009), Berger et al (2014), Dong et al (2014), Chan et al (2016), Ben Zeineb and Mensi (2018), Setiyono and Tarazi (2018) + - LAR Outstanding loans divided by total assets Berger et al (2014), Dong et al (2017) + - CAP Equity divided by total assets Chan et al (2016), Mollah et al (2017) + - LDR Outstanding loans divided by customer deposits Dong et al (2017), Ben Zeineb and Mensi (2018) - - LIQ Liquidated assets divided by total assets Dong et al (2017) + - 17 Variable Measurement method Scientific basis Expected mark Z-Score NPL CTI Total operating cost divided by total income Dong et al (2014) - + List By 1, the listed bank, by 0 otherwise Dong et al (2014) + - Ecogrow Annual growth ratio of GDP Dong et al (2017), Ben Zeineb and Mensi (2018) + - 3.4. Measuring the impact of corporate governance on the financial performance of commercial banks in Vietnam 3.4.1. Research models + Based on the research models of Pathan and Faff (2013), Dong et al (2017), this study proposes the Regression Equation 2 as follows: FPit = α0 + γ * CGit + δ * Xit + εit (2) In which: α0: Coefficient of original coordinate; i: Cross data of banks; t: Current year (t = 1, ...., k); FPit: Financial performance of bank i (ROA, ROE, NIM) at time t; CGit: Are variables representing the Corporate Governance of bank i at time t, including: size of the Board of Directors (Bsize), the percentage of Independent Members of Board of Directors (Bindep), Number of female members of Board of Directors (Femdir), the percentage of foreign members of Board of Directors (Fordir), the percentage of members of Board of Directors taking part in management (Execdir), and Education level of Board of Directors (Edu). The variable adjusted to suit the Vietnamese context in the research model is the variable of the percentage of members of Board of Directors taking part in management (Execdir), as defined in Paragraph 1, Article 34 of the 2010 Law on Credit Institutions of Vietnam: “Chairman of the Board of Directors, Chairman of the Board of Members of a credit institution must not concurrently be the executive of that credit institution and of other credit institutions”. Therefore, in this thesis, the author uses the variable of members of Board of Directors taking part in management as compared to the previous studies, the majority used the part-time variable (Chairman of Board of Directors cum CEO). 18 Xit: The control variables include bank characteristics and macro variables: bank size (SIZE), size of lending activities (LAR), equity size (CAP), Loan to Deposit ratio (LDR), Bank Liquidation (LIQ), Management Effectiveness (CTI), Listed Bank (List) and Economic Growth (Ecogrow). 𝛼 ,γ,δ: Are the estimated coefficient vectors. ε𝑖𝑡:: Is the standard error. 3.4.2. Measure of variables in the research model 3.4.2.1. The variable depends on financial performance - The thesis measures the financial performance of Vietnamese commercial banks by the ratios: Return on Assets (ROA) is a key parameter of management performance. It shows the ability of the Board of Directors in the process of converting a bank’s assets into net income and inherited from research by De Andres and Vallelado (2008); Lin and Zhang (2009); Grove et al (2011); Adams and Mehran (2012); Liang et al (2013); García-Meca et al (2015). ROA is calculated based on the following formula: 𝑅𝑂𝐴 = 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑎𝑓𝑡𝑒𝑟 𝑡𝑎𝑥 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 - Return on equity (ROE) is an indicator measuring the percentage of income for the bank’s shareholders. It represents the income that shareholders receive from investing in a bank and inherited from research by Staikouras et al (2007); Lin and Zhang (2009); Rowe et al (2011); Westman (2011); Fahlenbrach and Stulz (2011); Aebi et al (2012); Liang et al (2013); Elyasiani and Zhang (2015). ROE is calculated based on the following formula: 𝑅𝑂𝐸 = 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑏𝑒𝑓𝑜𝑟𝑒 𝑡𝑎𝑥 𝐸𝑞𝑢𝑖𝑡𝑦 - Net Interest Margin (NIM) is one of the most important measures for measuring financial performance in a deposit taking institution (Golin, 2001). Because it usually accounts for 70-85% of a bank’s total income, the higher the percentage, the higher its profit. In particular, in Vietnam, credit activities make up the main profit in banking activities. Net Interest Margin is calculated by the following formula: NIM = Interest returns – Interest paid Total assets 3.4.2.2. Independent variables for corporate governance in the model a) Size of Board of Directors Hypothesis 1c (H1c): The size of the Board of Directors has positive impacts on the financial performance of the bank. 19 b) Independent members of Board of Directors Hypothesis 2c (H2c): The percentage of independent members in the Board of Directors has positive impacts on the financial performance of the bank. c) Female members of Board of Directors Hypothesis 3c (H3c): The high percentage of female members in the Board of Directors has positive impacts on the financial performance of the bank. d) Foreign members of Board of Directors Hypothesis 4c (H4c): The high percentage of foreign members in the Board of Directors has positive impacts on the financial performance of the bank. e) Ratio of members of Board of Directors taking part in management Hypothesis 5c (H5c): The high percentage of members of the Board of Directors taking part in management has negative impacts on the financial performance of the bank. f) Education level of Board of Directors Hypothesis 6c (H6c): The percentage of members of the Board of Directors with postgraduate degrees has positive impacts on the financial performance of the bank. 3.4.2.3. Control variables in the model 3.5. Measuring the relationship between risks and financial performance of commercial banks in Vietnam + Regression equation 3 is as follows: 𝐑𝐢𝐬𝐤𝑖𝑡 = 𝛼0 + 𝛼 𝐅𝐏𝑖𝑡 + 𝛾 ∗ 𝐶𝐺𝑖𝑡 + 𝛿 ∗ 𝑋𝑖𝑡 + 𝜀𝑖𝑡+ (3) + Regression equation 4 is as follows: 𝐅𝐏𝑖𝑡 = 𝛼0 + 𝛼 ∗ 𝐑𝐢𝐬𝐤𝑖𝑡 + 𝛾 ∗ 𝐶𝐺𝑖𝑡 + 𝛿 ∗ 𝑋𝑖𝑡 + 𝜀𝑖𝑡 (4) In which: α0: Coefficient of original coordinates; i: Cross data of banks; t: Current year (t = 1, ...., k); Riskit: Risks of bank i (Z-score, NPL) at time t; FPit: Financial performance of bank i (ROA, ROE, NIM) at time t; CGit: Are variables representing the Corporate Governance of bank i at time t, including: size of the Board of Directors (Bsize), the percentage of independent members of Board of Directors (Bindep), Number of female members of Board of Directors (Femdir), the percentage of foreign members of Board of Directors (Fordir), the percentage of members 20 of Board of Directors taking part in management (Execdir), and Education level of Board of Directors (Edu). Xit: The control variables include bank characteristics and macro variables: bank size (SIZE), size of lending activities (LAR), equity size (CAP), Loan to Deposit ratio (LDR), Bank Liquidation (LIQ), Management Effectiveness (CTI), Listed Bank (List) and Economic Growth (Ecogrow). 𝛼 ,γ,δ: Are the estimated coefficient vectors. ε𝑖𝑡:: Is the standard error. 3.6. Methods of data analysis and processing 3.6.1. Methods of estimation 3.6.3. Testing the errors of the model 3.6.4. Handling endogenous phenomena of the model However, the weakness of the above regression models is unable to handle the potential endogenous phenomena in the model. To solve this problem, previous studies have used instrumental variables (IV) estimation. However, the problem arising when using the instrumental variable estimation is that it is often difficult to find the appropriate instrumental variables because if you choose weak instrumental variables, the IV estimation may be skewed (Mileva, 2007). In other words, using the IV estimation without choosing the appropriate instrumental variables, the problems of OLS estimation will not be improved. Since then, the GMM dynamic panel data model is proposed to be used according to the research of Arellano and Bond (1991). One of the advantages of the GMM model over the instrumental variable estimation model is that it is easier for the GMM model to select instrumental variables because using exogenous variables at other time or taking the latency of the variables can be used as instrumental variables for endogenous variables at the present time. Therefore, GMM has introduced many instrumental variables to easily achieve the condition of a standard instrumental variable (Overidentification of Estimators). Moreover, the estimation of Arellano and Bond are consistent with short panel data with small T time series (7 years) and large N (29 banks). Therefore, the GMM method introduced by Arellano and Bond (1991) will be used in this study. Specifically, the data of the project is as follows: The first, after checking the data, it shows that the change of variable phenomenon has occurred for the research model. To eliminate this phenomenon, the regression model is run with robust command in Stata software if it detects that this phenomenon occurs in the model. The second, the author examines the multi-collinearity and finds that this was not a problem for the analysis of the topic through the results of the correlation coefficients between the variables and presented in the data content description. The last, the relationship between Corporate Governance and bank risk 21 can occur as endogenous phenomena because of the possible causal relationship between Corporate Governance and bank risk (Lehn et al. 2009; Wintoki et al, 2012). Therefore, it commonly selects FEM regression models to reduce endogenous problems in case the research does not find appropriate instrumental variables to handle (Cheung et al., 2010). Studies of Pathan (2009), Dong et al (2014), Chan et al (2016), Dong et al (2017, Moilah et al (2017) use the 2-step GMM method to measure the impacts of Corporate Governance on bank risk. Therefore, in this study, the author uses the 2-step GMM analysis method to handle potential endogenous problems in the model of measuring the impacts of Corporate Governance on bank risk. Studies of Mollah et al (2017), Kusi et al (2018) use the GLS method to measure the impacts of Corporate Governance on financial performance of banks. Therefore, in the model to measure the impacts of Corporate Governance on financial performance of banks, the author uses the GLS analysis method for analysis. Conclusion of chapter 3 CHAPTER 4: STUDY RESULTS AND DISCUSSION 4.1. Actual situation of commercial banks in Vietnam in the period of 2011-2017 4.2. Descriptive statistics of study variables Table 4.2. Table of descriptive statistics of study variables Variable Observation number Average value Standard deviation Minimum value Maximum value Z-Score 203 29.9281 0.7711 0.5081 126.7510 NPL 203 0.0236 0.0142 0.0034 0.088 ROA 203 0.0063 0.0066 -0.0551 0.0253 ROE 203 0.0692 0.0847 -0.8200 0.2682 NIM 203 0.0256 0.0120 -0.0064 0.0742 Bsize 203 6.9891 0.2473 5 15 Bindep 203 0.1439 0.0710 0 0.4 Femdir 203 0.1794 0.1619 0 0.625 Fordir 203 0.0930 0.1251 0 0.4286 Execdir 203 0.1528 0.1271 0 0.4444 Edu 203 0.5391 0.2559 0 1 SIZE 203 89.349 1.0938 13.224 1.202.283 LAR 203 0.5284 0.1271 0.1473 0.7313 22 Variable Observation number Average value Standard deviation Minimum value Maximum value CAP 203 0.0970 0.04192 0.035 0.2384 LDR 203 0.8391 0.2015 0.3719 1.805 LIQ 203 0.1878 0.0959 0.0452 0.611 CTI 203 0.9880 6.0192 0.2875 86.3019 List 203 0.3448 0.4764 0 1 GDP 203 0.0608 0.0054 0.0525 0.0681 4.3. Analyzing the correlation between variables 4.4. Measuring the impacts of corporate governance on the risks of commercial banks in Vietnam in the period of 2011 - 2017 Table 4.4. Regression analysis results by the 2-step SGMM method Variable Z-Score NPL Z-Scoret-1 0.8924 *** (0.000) NPLt-1 0.3082 *** (0.000) Bsize -0.0613 (0.264) -0.0014 (0.601) Bindep -0.4723 ** (0.035) -0.0197 (0.158) Femdir 0.3121 *** (0.000) -0.0043 ** (0.028) Fordir 0.2316 ** (0.015) 0.0002 (0.954) Execdir -0.2592 ** (0.031) -0.0017 (0.646) Edu 0.0275 (0.746) -0.0018 (0.302) SIZE 0.1313 *** (0.000) -0.0004 (0.693) LAR 0.3475** (0.034) -0.0080 (0.360) CAP 5.9788 *** (0.000) 0.0764*** (0.007) LDR -0.3608 ** (0.000) 0.0066 (0.142) LIQ 0.0982 (0.625) -0.0145 (0.106) CTI -0.0144 0.0102** 23 Variable Z-Score NPL (0.856) (0.021) List 0.0334 (0.271) 0.0034 ** (0.011) GDP 6.2399 *** (0.001) -0.5134 *** (0.000) Constant -2.8070 (0.000) 0.0501 (0.056) AR(1) 0.050 0.003 AR(2) 0.661 0.223 Hansen test 0.203 0.245 F-test 0.000 0.000 Note: *, ** and *** are statistical significance at 10%, 5% and 1% respectively The suitability of regression by the SGMM method was assessed through F test, Hansen test and Arellano-Bond test (AR). The F test examines the statistical significance of the estimated coefficients. Hansen test examines excessive constraints, the rationality of representative variables. The AR test determines whether there is a residual correlation of the model. In both models, the Hansen test with p-value of 0.203 and 0.245 respectively is greater than 0.1, so we accept the Hypothesis H0: the model is correctly defined, the representative variables are reasonable. The F test in both models has p-value of 0.000 which is less than 0.01, so we reject the Hypothesis H0: all the estimated coefficients in the equation are equal to 0, or estimated coefficients of explanatory variables with statistical significance. So both models are appropriate. The AR test (1) of both models with p-value of 0.050 and 0.003 respectively, is less than 0.1, so we reject the Hypot