Pearson Correlation – Univariate Correlation Between Variables

After performing descriptive statistics of the data series, the author will continue to take the logarithm of the data series of GDP, stock market index, and total bank asset size. Taking the logarithm of the data series is to smooth (reduce volatility) the data series in the model. In addition, the data series taken from the logarithm will have the unit of percentage when expressing the meaning, so this will also be a way to unify the units of the variables in the regression model.

4.2. Univariate correlation between variables

The author uses the univariate correlation coefficient to examine the relationship between the independent variables and the dependent variables in the model. From the analysis results, the author will see the univariate correlation between the dependent variables and the independent variables in the model.

Table 4.2. Pearson correlation – univariate correlation between variables

Correlation GCR LGDP INF UNEMP LCK ROA LSIZE CAR1 LIQ

Probability

GCR

1,000 -----
GDP	-0.340***	1,000
INF	(0.000) -0.139*	----- -0.615	1,000
	(0.050)	(0.000)	-----
UNEMP	0.346***	-0.902	0.432	1,000
	(0.000)	(0.000)	(0.000)	-----
LCK	0.486***	-0.009	-0.437	0.024					1,000
	(0.000)	(0.893)	(0.000)	(0.737)	-----
ROA	0.199**	-0.541	0.313	0.470	-0.010	1,000
	(0.005)	(0.000)	(0.000)	(0.000)	(0.890)	-----
LSIZE	-0.159**	0.470	-0.272	-0.442	-0.081	-0.172	1,000
	(0.024)	(0.000)	(0.000)	(0.000)	(0.253)	(0.015)	-----
CAR1	-0.117	-0.269	0.245	0.249	-0.076	0.312	-0.669	1,000
	(0.100)	(0.000)	(0.000)	(0.000)	(0.285)	(0.000)	(0.000)	-----
LIQ	-0.009	-0.338	0.246	0.298	-0.101	0.292	-0.152	0.230	1,000

Maybe you are interested!

(0.899) (0.000) (0.000) (0.000) (0.156) (0.000) (0.032) (0.001) -----

Source: Author synthesized from analysis on Eviews Software.

Note: The variables in the results table correspond to the following: GCR : Dependent variable showing the credit growth rate of banks; GDP: Gross domestic product;

real domestic output – independent variable ; INF: Inflation rate – independent variable; UNEMP: Unemployment rate of the economy – independent variable; CK: Stock market index – independent variable; ROA: Return on total assets – independent variable; LSIZE: Natural logarithm of bank size – independent variable; CAR1: Tier 1 capital adequacy ratio – independent variable; LIQ: Bank liquidity – independent variable . In parentheses () is the result of statistical value p – value. The symbols *, ** and *** show that the variables are statistically significant at the 10%, 5% and 1% levels, respectively.

The results of the correlation matrix between the variables are presented in Table 4.2. With a statistical significance level of 10%, in the univariate relationship of the independent variables with the dependent variable GCR, it can be seen that:

- Credit growth has a statistically significant negative correlation with the independent variables logarithm of GDP (LGDP), inflation rate (INF), logarithm of total assets of banks (LSIZE). This shows that when GDP increases, inflation rate increases or total assets of banks increase, it will reduce the credit growth rate of banks.

- Credit growth has a statistically significant positive correlation with the independent variables unemployment rate (UNEMP), logarithm of the stock market index (LCK), and return on bank assets (ROA). This shows that when the unemployment rate increases, the stock market index increases or the return on total bank assets increases, it will lead to an increase in the credit growth rate of banks.

- Credit growth has no correlation with the independent variables of capital adequacy ratio tier 1 (CAR1) and liquidity (LIQ). This shows that increasing the capital adequacy ratio tier 1 or bank liquidity will not affect the growth rate of bank credit.

4.3. Analysis of multivariate relationships between variables and credit growth rate

After analyzing the univariate relationship between independent variables and dependent variables, the author continues to analyze the multivariate relationship between macroeconomic factors and internal factors of enterprises with the growth rate of bank credit. First, the author will perform theoretical model regressions in turn according to 03 analysis methods: Pooled OLS regression model, fixed effects regression model (FEM) and random effects regression model (REM). Next, the author will perform tests in turn to see which analysis method is most suitable in the conditions of the author's data sample, from which the author will analyze the results according to the most suitable regression analysis method. In this section, the author will present the results of the most suitable regression analysis method and the tests show that this regression method is the best in the conditions of the data sample collected by the author. The results of the remaining two models will be presented in the appendix of the thesis.

4.3.1. Multivariate regression results of factors affecting bank credit growth

Table 4.3: Regression results of factors affecting bank credit growth

Independent variable

Dependent variable: GCR
	Impact Factor Z-Statistic Value	p-value
Constant	13,527 2.40	0.017
GDP	-0.924*** -2.60	0.009
INF	-0.008*** -3.14	0.002
UNEMP	0.001 0.02	0.981
LCK	0.533*** 4.73	0.000

ROA

3,288*	1.77	0.078
CAR1	-1,115***	-3.22	0.001
LIQ	-0.499	-0.97	0.330
LSIZE	-0.057*	-1.66	0.097
Number of observations	200

R 2 0.4357

Source: Author's synthesis from analysis on Stata Software.

Note: The variables in the results table correspond to the following: GCR : Dependent variable showing the credit growth rate of banks; GDP: Gross domestic product - independent variable ; INF: Inflation rate - independent variable; UNEMP: Unemployment rate of the economy - independent variable; CK: Stock market index - independent variable; ROA: Return on total assets - independent variable; LSIZE: Natural logarithm of bank size - independent variable; CAR1: Tier 1 capital adequacy ratio - independent variable; LIQ: Bank liquidity - independent variable . In parentheses () are the results of the statistical value p - value. The symbols *, ** and *** indicate that the variables have statistical significance levels of 10%, 5% and 1%, respectively.

Table 4.3 shows the impact of macroeconomic variables and internal bank variables on credit growth rate. At 10% significance level, the results show that:

For macro-economic impacts:

- Regarding the impact of GDP: The results show that the coefficient of the variable LGDP is -0.924 and is highly statistically significant (p – value = 0.009). This shows that a 1% increase in GDP will slow down the credit growth rate at commercial banks by about 0.00924% (0.924/100) compared to the previous year.

- For the impact of inflation rate (INF): The results show that the coefficient of the variable INF

= -0.008 and is highly statistically significant (p – value = 0.002). This shows that when the domestic inflation rate increases by 1%, commercial banks will reduce their credit growth rate by about 0.008% in their operations.

- Regarding the impact of the stock market : The results show that the stock market index has a positive impact on credit growth of banks with a high statistical significance (p – value = 0.000). The impact coefficient of the stock market index on credit growth at commercial banks is 0.533, showing that when the stock market index increases by 1%, the credit growth rate at commercial banks will increase by 0.00533%.

- Regarding the impact of unemployment rate (UNEMP) : The impact coefficient of the UNEMP variable on credit growth of banks is not statistically significant (p – value = 0.981), showing that it is not possible to draw a conclusion about the relationship between unemployment rate and credit growth of commercial banks.

For the micro impacts of commercial banks:

- Regarding the impact of ROA: The results show that the coefficient of the ROA variable is 3.288 and is statistically significant (p – value = 0.078). This shows that when the return on total assets of commercial banks increases by 1%, the credit growth rate of commercial banks will increase by 3.288%.

- Regarding the impact of capital adequacy ratio tier 1 (CAR1): The results show that the coefficient of variable CAR1 = -1.115 and is highly statistically significant (p – value = 0.002). This shows that when the capital adequacy ratio tier 1 of banks increases by 1%, commercial banks will reduce their credit growth rate by about 1.115% in their operations.

- Regarding the impact of bank size (SIZE) : Bank asset size has a statistically significant negative impact on credit growth. The impact coefficient = -0.057 shows that when total asset size increases by 1%, commercial banks will reduce credit growth by about 0.00057% (0.057/100).

- Regarding the impact of liquidity (LIQ) : The impact coefficient of liquidity variable does not affect the credit growth of banks (p – value = 0.330). This shows that it is not possible to draw a conclusion about the relationship between liquidity and credit growth of commercial banks.

From the above regression results, it can be seen that the ROA factor has the most significant impact on the credit growth rate of commercial banks, followed by the impact coefficient of CAR1. This shows that the credit growth rate of Vietnamese commercial banks is significantly affected by internal factors of the bank rather than by macroeconomic factors of the economy.

4.3.2. Results of testing to select the most suitable model 1

4.3.2.1. Testing the choice between the Pooled OLS model and the fixed effects model (FEM)

Table 4.4 Results of the test of choice between the Pooled OLS model and the fixed effects model (FEM)

Target

Value
F-statistics	2.44
P-value	0.0156

Source: Author's synthesis from analysis on Stata Software.

Hypothesis H 0 : There is no difference between banks in the survey data sample (There is no difference between Pooled OLS model and FEM model)

At the 10% significance level, the F-statistic test results in choosing between the Pooled OLS model and the fixed effects model (FEM) show that the null hypothesis H 0 is rejected , i.e. the FEM model is better than the Pooled OLS model. This shows that when adding

1 The remaining basic tests of the regression model are presented in the results appendix.

The bank dummy variable will better reflect the differences between banks in the sample.

4.3.2.2. Testing the choice between the Pooled OLS model and the random effects model (REM)

Table 4.5 Results of the test of choice between the Pooled OLS model and the random effects model (REM)

Target

Value
Chi-square statistics	2.38
P-value	0.0371

Source: Author's synthesis from analysis on Stata Software.

Hypothesis H 0 : There is no potential error in the survey data sample (There is no difference in error between the Pooled OLS model and the REM model)

At the 10% significance level, the Chi-square test results for the choice between the Pooled OLS model and the random effects model (REM) show that the null hypothesis H 0 is rejected , i.e. the REM model is better than the Pooled OLS model. This shows that there are some potential error components in the regression model that the Pooled OLS model does not reflect.

4.3.2.3. Testing the choice between the random effects model (REM) and the fixed effects model (FEM)

Table 4.6 Results of the test of choice between the random effects model (REM) and the fixed effects model (FEM)

Target

Value
Chi-square statistics	6.20
P-value	0.6245

Hypothesis H 0 : There is no correlation between the intercept and the independent variables in the regression model (There is no significant difference between the FEM model and the REM model)

At the 10% significance level, the results of the Chi-square test (Hausman test) in choosing between the fixed effects model (FEM) and the random effects model (REM) show that the null hypothesis H 0 is accepted , that is, the REM model is better than the FEM model. This shows that in the regression model, there is no correlation between the intercept coefficient and the independent variables in the model.

4.4. Discussion of research results

From the results of the above model, the author will give some discussions on the research results as follows:

- Regarding macroeconomic factors of the economy: There are 3 macroeconomic factors of the economy that have an impact on credit growth of commercial banks, in which GDP and inflation rate have a negative impact, stock market index has a positive impact. In the above results, the negative inflation rate occurs in accordance with the theoretical expectation of the author and the studies of Pouw and Kakes (2013), H. Vu and D. Nahm (2013), A. Singhn and A. Sharma (2016) presented above. Meanwhile, the results on the negative impact of economic output on credit growth seem to be slightly different from theoretical expectations but quite similar to the research results of H. Vu and D. Nahm (2013). This result is also consistent with the characteristics of Vietnam, a country with a Central Bank under the Government. Vietnam's monetary and fiscal policies are mainly implemented by the Government. During the period from 2008 to 2015, the Government's top priority was to stabilize prices and curb inflation, while economic growth and employment rates were the next priorities. Therefore, the Government and the State Bank have implemented many administrative measures to influence the credit growth limits of joint stock commercial banks in order to curb inflation. Meanwhile, economic growth comes from a number of other driving forces, especially the activities of foreign corporations and FDI enterprises investing in Vietnam, so economic growth and

Comment