FEM (Fixed Effect Model) and REM (Random Effect Model) were used to perform regression. To select the optimal model, the author used the F-test to choose between Pooled OLS and FEM, and the Hausman test to choose between the fixed effect model (FEM) and the random effect model (REM). The regression results are presented below:
Table 4.4 Regression results by L1
transform
Pool OLS | Fix Effect | Random Effect | |
CAP | -0.1574045** | -0.1133719 | -0.122249** |
NPL | -1.246979** | -1.707346*** | -1.639494*** |
ROE | 0.0932373** | 0.1585879** | 0.149992** |
CAR | -0.0135327** | -0.0147613* | -0.0147028** |
GDP | -2.415594** | -2.567123** | -2.544181** |
MIR | 1.360746*** | 1.323747*** | 1.328771*** |
_cons | 0.4328807*** | 0.4584307*** | 0.4562048*** |
Adj R-squared | 0.1826 | 0.1984 | 0.1997 |
F Statistics | 9.6800 | 19.0100 | |
Prob (F-statistic) | 0.0000 | 0.0000 | |
F test | F(25, 202) = 7.98 Prob > F = 0.0000 | ||
Hausman test | Prob>chi2 = 0.8873 | ||
Maybe you are interested!
-
Testing the credit risk tolerance of Vietnamese commercial banks - Case study of Vietnam Joint Stock Commercial Bank for Industry and Trade - 21 -
Evaluating customer satisfaction with savings deposit services at National Citizen Commercial Joint Stock Bank NCB Tan Huong Transaction Office - 3 -
Organizational Chart of Lien Viet Post and Commercial Joint Stock Bank -
Investment Situation in Non-Credit Services at Saigon Commercial Joint Stock Bank in the Period of 2015-2019 -
Credit risk management at Vietnam International Commercial Joint Stock Bank - Hanoi Branch - 12
*statistically significant at 10% Source: Author processed with Stata 13
**statistically significant at 5% level
***statistically significant at 1% level
Table 4.5 Regression results by L2
transform
Pool OLS | Fix Effect | Random Effect | |
CAP | 0.3426365** | 0.4777865*** | 0.4499393*** |
NPL | -2.094013*** | -2.418151*** | -2.365211*** |
ROE | 0.1383944** | 0.2408883** | 0.2255552** |
CAR | -0.0094827** | -0.0127725 | -0.0115033** |
GDP | -1.62143** | -1.498051 | -1.551057** |
MIR | 2.262981*** | 2.13785*** | 2.168702*** |
_cons | 0.2845068** | 0.3080785* | 0.2976684** |
Adj R-squared | 0.2853 | 0.2980 | 0.3000 |
F Statistics | 16.5000 | 27.4700 | |
Prob (F-statistic) | - | - | |
F test | F(25, 202) = 6.36 Prob > F = 0.0000 | ||
Hausman test | Prob>chi2 = 0.8176 | ||
*statistically significant at 10% Source: Author processed with Stata 13
**statistically significant at 5% level
***statistically significant at 1% level
Table 4.6 Regression results by L3
transform
Pool OLS | Fix Effect | Random Effect | |
CAP | 0.3750097** | 0.1782048 | 0.199314** |
NPL | 1.183496** | 0.4143958 | 0.459672** |
ROE | 0.2281253** | 0.0156434 | 0.0409049 |
CAR | 0.0142889** | -0.0000632 | 0.003482 |
GDP | -0.0627535 | -0.2865151 | -0.3501687 |
MIR | -1.33461*** | -1.281094*** | -1.264346*** |
_cons | 0.3817849** | 0.6063261*** | 0.5638247*** |
Adj R-squared | 0.0779 | 0.0648 | 0.3000 |
F Statistics | 4.2800 | 5.7700 | |
Prob (F-statistic) | 0.0004 | - | |
F test | F(25, 202) = 12.83 Prob > F = 0.0000 | ||
Hausman test | Prob>chi2 = 0.3233 | ||
*statistically significant at 10% Source: Author processed with Stata 13
**statistically significant at 5% level
***statistically significant at 1% level
Table 4.7 Regression results according to L4
transform
Pool OLS | Fix Effect | Random Effect | |
CAP | 1.165061*** | 0.9533545*** | 0.9769671*** |
NPL | 1.239809** | 0.6512734 | 0.6580931** |
ROE | 0.3604252** | 0.1061424 | 0.1384669** |
CAR | 0.0251285** | 0.0008097 | 0.0068517 |
GDP | 0.1036653 | 0.292373 | 0.1274603 |
MIR | -0.7097228** | -0.7501403** | -0.7083918** |
_cons | 0.2296651** | 0.5497763*** | 0.4833583*** |
Adj R-squared | 0.1140 | 0.0747 | 0.3000 |
F Statistics | 4.8700 | 5.4900 | |
Prob (F-statistic) | 0.0001 | - | |
F test | F(25, 202) = 12.47 Prob > F = 0.0000 | ||
Hausman test | Prob>chi2 = 0.7171 | ||
*statistically significant at 10% Source: Author processed with Stata 13
**statistically significant at 5% level
***statistically significant at 1% level
4.5.4 Testing for violations of regression assumptions
4.5.4.1 Multicollinearity test
Table 4.8, results of multicollinearity test by VIF variance inflation factor method, using auxiliary regression models of explanatory variables of 3 models L1, L2, L3, L4. The results are as follows:
Table 4.8 Variance inflation factor VIF
Secondary regression
VIF value | R-Squared | Result | |
CAP | 1.89 | 0.4706 | No multicollinearity |
NPL | 1.13 | 0.1140 | No multicollinearity |
ROE | 1.18 | 0.1490 | No multicollinearity |
CAR | 2.03 | 0.5068 | No multicollinearity |
GDP | 1.17 | 0.1460 | No multicollinearity |
MIR | 1.23 | 0.1838 | No multicollinearity |
Source: Author processed with Stata 13
Looking at table 4.8, the VIF variance inflation factor shows that all VIF values are less than 10, indicating that models L1, L2, L3, L4 do not have multicollinearity (Hoang Trong and Chu Nguyen Mong Ngoc, 2008).
4.5.4.2 Autocorrelation test
To examine the autocorrelation phenomenon in the model, the author uses the Wooldridge method as the result of testing the residuals of the 3 models L1, L2, L3, L4 (random effects model REM). With the hypothesis H 0 : no autocorrelation phenomenon, hypothesis H 1 : autocorrelation phenomenon. The test results are as follows:
Table 4.9 Autocorrelation test results
Model
Wooldridge test | |||
F Statistics | Prob > F | Test results | |
L1 | 18,914 | 0.00 | There is autocorrelation phenomenon. |
L2 | 12,457 | 0.00 | There is autocorrelation phenomenon. |
L3 | 62,055 | 0.00 | There is autocorrelation phenomenon. |
L4 | 16,194 | 0.00 | There is autocorrelation phenomenon. |
Source: Author processed with Stata 13
The Wooldridge values of the three models L1, L2, L3, L4 of the Wooldridge test are all between 1 and 3, so there is no basis to reject the hypothesis H0 , concluding that there is no autocorrelation phenomenon in the model.
4.5.4.3 Heteroscedasticity test
Table 4.10 Results of heteroscedasticity test
Random Effect Model
Breusch and Pagan Lagrangian | |||
Chibar2(01) | Prob > F | Test results | |
L1 | 169.58 | - | There is a phenomenon of heteroscedasticity. |
L2 | 121.67 | - | There is a phenomenon of heteroscedasticity. |
L3 | 273.4 | - | There is a phenomenon of heteroscedasticity. |
L4 | 270.29 | - | There is a phenomenon of heteroscedasticity. |
4.5.4.4 Generalized method of squares (GLS)
Table 4.11 GLS method regression results
transform
L1 | L2 | L3 | L4 | |
CAP | -0.1336 | 0.2195 | 0.4458*** | 0.9822*** |
NPL | -1.0484** | -1.6321*** | 0.888* | 1.0297 |
ROE | 0.0855 | 0.1645 | 0.1591** | 0.2161*** |
CAR | -0.0095*** | -0.0131* | 0.0146* | 0.0226*** |
GDP | -3.4285*** | -3.7023 | -0.1597 | 0.0168 |
MIR | 1,513*** | 2,632*** | -1.1483*** | -0.5775* |
_cons | 0.4278*** | 0.4146* | 0.3725*** | 0.2906** |
*statistically significant at 10% Source: author processed with Stata 13
**statistically significant at 5% level
***statistically significant at 1% level
4.6 Discussion of research results
The model is rewritten as follows:
Model 1:
L1 = 0.4278 – 1.0484*NPL - 0.0095*TOA – 3.4285*GDP + 1.513 *MIR
Model 2:
L2 = 0.4146 – 1.6321*NPL – 0.0131*TOA + 2.632 *MIR
Model 3:
L3 = 0.3725 + 0.4458*CAP + 0.888*NPL + 0.1591*ROE + 0.0146*TOA – 1.1483*MIR
Model 4:
L4 = 0.2906 + 0.9822*CAP + 0.2161*ROE + 0.0226*TOA - 0.5775 *MIR
The L1 model shows that liquid assets divided by total assets (L1) are explained by non-performing loans (NPL), bank size (TOA), economic growth (GDP) and monetary policy instruments (MIR). The results show that non-performing loans (NPL), bank size (TOA), economic growth (GDP) have negative impacts on liquid assets divided by total assets (L1). When non-performing loans (NPL), bank size (TOA), economic growth increase/decrease by 1%, liquid assets divided by total assets (L1) decrease/increase by 1.0484%, 0.0095%, 3.4285% at the significance level of 5%, 1% and 1%, respectively. And when monetary policy instruments increase/decrease by 1%, liquid assets divided by total assets (L1) increase/decrease by 1.513% at the significance level of 1%.
The L2 model shows that the liquid assets divided by customer deposits and short-term loans (L2) are explained by the non-performing loan ratio (NPL), bank size (TOA) and monetary policy instruments (MIR). The results show that the non-performing loan ratio (NPL), bank size (TOA) and economic growth have negative impacts on the liquid assets divided by customer deposits and short-term loans (L2). On the contrary, monetary policy instruments have positive impacts on the liquid assets divided by customer deposits and short-term loans (L2). When the non-performing loan ratio (NPL), bank size (TOA) increases/decreases by 1%, the liquid assets divided by customer deposits and short-term loans (L2) decrease/increase by 1.6321%, 0.0131% respectively with the significance level of 1% and 5%. And when the monetary policy instrument increases/decreases by 1%, the liquid assets divided by customer deposits and short-term loans increase/decrease by 2.632% at the 1% significance level.
The L3 model shows that debt divided by total assets (L3) is explained by the ratio of equity to total assets (CAP), the ratio of non-performing loans (NPL), the rate of return (ROE), the size of the bank (TOA) and the monetary policy instrument (MIR). The results show that the monetary policy instrument (MIR) has an inverse effect on debt divided by total assets (L3). And the ratio of equity to total assets (CAP), the ratio of non-performing loans (NPL), the rate of return (ROE), the size of the bank (TOA) have a positive effect on debt divided by total assets (L3). When the monetary policy instrument (MIR) increases/decreases by 1%, the debt divided by total assets (L3) decreases/increases by 1.1483% at the 1% significance level. And vice versa, when the ratio of equity to total assets (CAP), bad debt ratio (NPL), profitability ratio (ROE), bank size (TOA) increases or decreases by 1%, the debt divided by total assets (L3) increases/decreases by 0.4458%, 0.888%, 0.1591%, 0.0146% respectively at the significance levels of 1%, 10%, 5% and 10%.
Model L4 shows that the outstanding balance of customer deposits and short-term loans (L4) is explained by the ratio of equity to total assets (CAP), rate of return (ROE), bank size (TOA) and monetary policy instruments (MIR). The results show that the monetary policy instrument (MIR) has an inverse effect on the outstanding balance of customer deposits and short-term loans (L4). And conversely, the ratio of equity to total assets (CAP), rate of return (ROE), bank size (TOA) have a positive effect on the outstanding balance of customer deposits and short-term loans (L4). When the monetary policy instrument (MIR) increases/decreases by 1%, the outstanding balance of customer deposits and short-term loans (L4) decreases/increases by 0.5775% at the significance level of 10%. When the ratio of equity to total assets (CAP), rate of return (ROE), bank size (TOA) increases/decreases by 1%, the outstanding balance divided by customer deposits and short-term loans (L4) increases/decreases by 0.9822%, 0.2161%, 0.0226% at the significance level of 1%, 1% and 1%, respectively.
Equity to Total Assets Ratio (CAP)
The research results show that the ratio of equity to total assets (CAP) has a positive impact on the balance of debt divided by total assets (L3) and the balance of debt divided by customer deposits and short-term loans (L4). In addition, the ratio of equity to total assets (CAP) has a negative impact on the liquid assets divided by total assets (L1) and a positive impact on the liquid assets divided by customer deposits and short-term loans (L2), but it is not statistically significant. Thus, the equity ratio has a negative impact on bank liquidity. And this result is consistent with the research results of Lucchetta, M. (2007); Nguyen Thi My Linh (2016); Deléchat, C. et al (2012); Cucineli, D. (2013); Moussa,
MA . (2015); Vodová, P., (2011b); Truong Quang Thong and Pham Minh Tien (2014); Tran Hoang Ngan and Pham Quoc Viet (2016). Meanwhile, Teixeira, D. (2013) found an unclear impact of the equity ratio to total assets (CAP) on liquidity. The equity ratio to total assets (CAP) has a negative impact on bank liquidity, which can be explained as follows: according to Nguyen Thi My Linh (2016), banks with low capital ratios are banks with higher liquidity ratios, which can be understood as the fact that Vietnamese commercial banks have low equity, under the pressure of Basel III, they must maintain a high liquidity ratio to ensure safety in payment. When capital increases, it invisibly increases the CAR ratio for banks, helping commercial banks to be more bold in supplying capital to the market. In fact, in
In the period of 2008 - 2016, many banks lent more than they mobilized, such as: CTG, EI, OCB, SCB, VCAP, especially in the period of 2008 - 2010, when banks lent heavily, leading to a high increase in bad debt, and thus reducing the efficiency of bank operations. At that time, banks were forced to reduce cash reserves and reduce liquid assets or borrow additional capital in the money market to compensate for liquidity. Therefore, when equity capital increases, credit growth increases, while reducing bank liquidity.
Non-performing loan ratio (NPL)
The research results show that the non-performing loan (NPL) ratio has an inverse impact on liquid assets divided by total assets (L1) and liquid assets divided by customer deposits and short-term loans (L2) and an inverse impact on outstanding loans divided by total assets (L3). In addition, the NPL ratio has a positive impact on outstanding loans divided by customer deposits and short-term loans (L4) but is not statistically significant. Thus, this result is similar to the hypothesis and similar to recent studies Vodová, P., 2011a; Deléchat, C. et al, 2012; Vodová, P., 2012; Truong Quang Thong and Pham Minh Tien, 2014. The research results are consistent with the recent reality at Vietnamese commercial banks, high NPL ratios cause the reputation, trust and financial potential of banks to decline, leading to a reduction in the bank's ability to mobilize capital. Banks with high rates of overdue debt and bad debt are often banks that have difficulty in payment. This research result is contrary to the research of Malik, MF et al, 2013; Tran Hoang Ngan and Pham Quoc Viet, 2016.
Return on equity (ROE)
The research results show that return on equity (ROE) has a negative impact on debt divided by total assets (L3), debt divided by customer deposits and short-term loans (L4). The regression models L3 and L4 give results contrary to the expected hypothesis and have the same results as the studies of Aspachs, O., et al, 2005; Nguyen Thi My Linh, 2016; Vodová, P., 2012; Vodová, P., 2011b; Tran Hoang Ngan and Pham Quoc Viet, 2016. According to Nguyen Thi My Linh (2016), banks with high profitability often face high risks, including liquidity risk, which means low liquidity ratios due to banks accepting risky investments.
Risky loans, or loans with high risk but high profitability, lead to a decrease in liquid assets.
Bank Size (TOA)
The research results show that bank size (TOA) has an inverse effect on liquid assets divided by total assets (L1), liquid assets divided by customer deposits and short-term loans (L2). And a positive effect on outstanding loans divided by total assets (L3), outstanding loans divided by customer deposits and short-term loans. Thus, bank size (TOA) has an inverse effect on bank liquidity, this result is contrary to the hypothesis and contrary to the research of Malik, MF et al, 2013; Deléchat, C. et al, 2012; Cucineli, D., 2013; Truong Quang Thong and Pham Minh Tien, 2014 but similar to the research of Lucchetta, M., 2007; Vodová, P., 2013; Nguyen Thi My Linh, 2016; Diana Teixeira 2013; Vodová, P., 2012; Vodová, P., 2011b. The larger the bank, the lower the liquidity ratio. In Vietnam, small banks are often under greater pressure on liquidity than large banks, so small banks proactively maintain a high liquidity ratio to meet payment requirements as well as cope with market fluctuations. According to Vodová, P., 2013, it is considered "too big to fail", large banks rely on brand advantages, lower capital mobilization and more investment in risky assets, so liquidity is low. According to Nguyen Thi My Linh, 2016, there are also arguments that small banks will have more difficulty accessing capital from the market. On the contrary, large banks will have an easier time mobilizing capital thanks to their reputation and extensive branch network, so they only need to maintain a low liquidity ratio .
Economic growth (GDP)
The research results show that GDP growth has a negative impact on liquid assets divided by total assets (L1). Thus, economic growth (GDP) has a negative impact on bank liquidity, this result is consistent with the hypothesis and consistent with recent research results (Aspachs, O., et al, 2005; Deléchat, C. et al, 2012; Cucineli, D., 2013; Vodová, P., 2011b; Truong Quang Thong and Pham Minh Tien, 2014). During the economic development period, enterprises expand their scale, increase