NHTM, this goal is achieved through models (3.1), (3.13), (3.14) and (3.15) the thesis uses the generalized panel data estimation method based on the system GMM moment proposed by Arellano and Bond (1991) and Blundell and Bond (1998). The reason the thesis chooses the GMM method is because: there are two important potential problems related to the error component in a panel data model. That is the correlation between the explanatory variables with the individual effects and the correlation between the explanatory variables with the noise error component. The existence of 1 of 2 (or both) of these problems makes the OLS estimation results biased or inefficient.
With the above equations, when lagged variables are included, the estimation by Fixed Effects FE will be biased when the model has a small panel data time series T (Judson and Owen, 1999). Nickell (1981) and Kiviet (1995) explain that the regression coefficients will be unbiased when T approaches infinity. That is, FE only gives good results when the panel data time series is large. Therefore, there are some problems that arise when estimating the above equations: (i) The variables can be considered endogenous. Because the causal relationship can occur in two directions: from the explanatory variables to the explained variables and vice versa. Regressing these variables can lead to correlation with the error, that is, endogeneity occurs and biases the results; (ii) Fixed effects containing errors in the above equation include the specificity of unobserved variables and the specific observed errors; (iii) The presence of lagged variables in the equation will lead to autocorrelation and (iv) Panel data in studies often have short time periods (short T) and large spatial arrays (large N).
The GMM method is commonly used in estimating linear dynamic panel data or panel data that violates the properties of heteroscedasticity and autocorrelation. At that time, the classical linear estimates of panel data models such as FE or RE will no longer give reliable and efficient estimation results. The original GMM method (Hansen 1982) is based on Fisher's maximum likelihood estimation (MLE). In addition, GMM allows
to solve problems where classical MLE fails, and there are many such problems in economics. Hansen has constructed estimators with good statistical properties such as consistency, asymptotic normality, and efficiency.
However, according to Blundell and Bond (1998), the above estimates will suffer from the weak representativeness problem when the coefficients approach 1. When the coefficient = 1, the moment conditions are completely unrelated to the real parameters and the behavioral nature of the estimates depends on time T. To solve this problem from the original GMM method, scholars have improved many versions of GMM that are more suitable for empirical studies. Most notably, the two difference GMM methods built by Arellano and Bond (1995) based on previous studies by Anderson and Cheng (1982), Holtz-Eakin et al. (1988) and the system GMM method built by Blundell and Bond (1998) based on the ideas of Arellano and Bover (1995) by adding some constraints to the difference GMM.
In addition, the GMM method also exploits the pooled data of the panel and does not constrain the length of the time series of the panel units in the data panel. This allows the use of an appropriate lag structure to exploit the dynamic characteristics of the data. However, the limitation of the model is that with short-term samples and high persistence, the accuracy of the estimation is low (Blundell and Bond, 1998). To overcome this problem, the system GMM method is used because it performs better than the difference GMM with small samples and high persistence. It uses the lag differences of the predictor variables as instrumental variables and the differences of strictly exogenous variables (Blundell and Bond, 1998; Roodman, 2009). At the same time, the study applies the two-step GMM because it will give better results than one step if there is serial correlation or heteroscedasticity in the series components. According to Windmeijer (2005), two-step GMM will use the variance-covariance matrix adjustment method.
To conduct the GMM test, the instrumented variables and the instrumental variables are distinguished. If the variables are expected to be endogenous (equivalent to non-strictly exogenous), they are classified as instrumented variables according to the GMM approach and then only the lagged values of these variables are appropriate instruments (Judson and Owen, 1996). If the explanatory variables are determined to be strictly exogenous as well as the added instrumental variables, they are classified as instrumental variables. Variables that are considered strictly exogenous have both their current values and their lagged values as appropriate instruments. The thesis chooses the lagged values of the explanatory variables as instrumental variables in the research model.
In summary, GMM estimates are appropriate in the following cases: (i) Panel data with small T, large N (many observations with few time points); (ii) There is a linear relationship between the dependent variable and the explanatory variables; (iii) Dynamic models with one or both sides of the equation containing lagged variables; (iv) Independent variables are not strictly exogenous, meaning they may be correlated with the residuals (current or previous) or there is an endogenous variable in the model; (v) There is heteroscedasticity or autocorrelation in measurement errors; (vi) There are separate fixed effects; (vii) There is heteroscedasticity and autocorrelation within each subject (but not between subjects).
The Sargan or Hansen test determines the appropriateness of the instrumental variables in the GMM estimation. It is a test of the endogeneity of the model. The Sargan test assumes that the instrumental variables are exogenous, that is, they are not correlated with the model errors. To test the correlation under the hypothesis Ho: no autocorrelation, the Arellano-Bond test is applied to the differenced residuals.
3.3. Research data
The data set of the topic is collected from the annual reports and financial statements of commercial banks in the time series from 2005 to 2015. The number of banks in the research sample includes 34 banks, including 5 state-owned commercial banks and 29 joint-stock commercial banks. The number of samples is 34 out of the total 35 joint-stock commercial banks today, so the sample is representative of the group of joint-stock commercial banks in Vietnam. Due to the characteristics of disclosing business information and a number of newly established banks as well as mergers and consolidations during the research period, the number of banks in the research sample is not balanced. Due to the fluctuations in the commercial banking system during the research period, the sample of 34 banks and the corresponding data collection stages of each commercial bank in the panel data are presented specifically in Appendix 1.
3.4. Data collection sources
To seek quantitative evidence to answer the research questions, the thesis exploited and used the data set of 34 Vietnamese commercial banks in the period 2005-2015. Specific data related to bank operations were collected secondary from annual reports, financial statements, and reports of the board of directors of 34 Vietnamese commercial banks. Macro data such as GDP, inflation rate, average lending interest rate, and exchange rate were collected from the IFS Database of the International Monetary Fund, in which 2015 data were collected from the IMF's Regional Economic Outlook. The housing price index was collected from the General Statistics Office of Vietnam. The data sources for each variable are presented in Appendix 2.
Chapter 3 Conclusion
In this chapter, the study presented the main research hypotheses of the thesis. At the same time, it proposed a specific research model, how to choose variables and explained the reasons for choosing variables in the model. At the same time, the author also clarified the data analysis method and the method of testing the research results, which is the GMM dynamic panel data estimation method to overcome the endogeneity problem of the model, the parameter envelopment method to determine cost efficiency. This chapter also clarified the specific data sources as well as macro data and gave an overview of the bank sample in the study. Finally, Chapter 3 presented an overview of the data and the GMM estimation method for panel data to achieve the research objectives and serve as a basis for the next empirical steps.
Chapter 4 will present the research results on the impact of factors on bad debt as well as the impact of bad debt on banking operations along with a discussion of the significance of the results achieved.
CHAPTER 4
DISCUSSION OF RESEARCH RESULTS
Introduce
To achieve the research objectives, the thesis conducts experiments to analyze the factors affecting bad debt as well as the impact of bad debt on banking operations. First, the thesis focuses on assessing the impact of factors on bad debt at Vietnamese commercial banks. Then, the author examines the impact of bad debt on factors such as: efficiency, capital safety and credit growth. To achieve the goal of evaluating the regression coefficients of variables in the model, the author uses a two-step system GMM dynamic panel data model. Before presenting the results, the study also summarizes the current status of bad debt at Vietnamese commercial banks and analyzes the correlation of variables through graphs as well as performs the necessary tests of the model.
Previous empirical studies on the impact of bad debt on banking operations consider macroeconomic factors as control variables. In addition, cost efficiency has not been given much attention in these studies. Therefore, the thesis will clarify the research results on cost efficiency through the non-parametric data envelopment method DEA in this chapter. Finally, the thesis presents the results of the analysis of the impact of macroeconomic and specific factors on bad debt of Vietnamese commercial banks as well as testing the impact of bad debt on banking operations through the GMM method. This is also the basis for the thesis to propose policy suggestions in the next chapter.
4.1. Overview of the operations of Vietnamese commercial banks
4.1.1. Development of Vietnam's commercial banking system
The Vietnamese commercial banking system has grown extremely strongly in terms of quantity, scale and diversity in ownership structure and types of operations since the single-tier banking system of Vietnam was separated into a central bank.
represented by the State Bank and state-owned commercial banks in 1988. The research period is from 2005 to 2015, in which 2005 is associated with the conversion of rural joint stock commercial banks into urban joint stock commercial banks operating nationwide. Decision 1557/QD-NHNN in 2006 of the State Bank on the Project on restructuring rural joint stock commercial banks with the goal of consolidating and rearranging rural joint stock commercial banks to increase competitiveness in new conditions, avoiding risks in the integrated economy. Since then, 12 rural joint stock commercial banks have been converted into urban joint stock commercial banks. Table 4.1 shows that the number of commercial banks in Vietnam has continuously increased during the research period.
Table 4.1. Number of Vietnamese commercial banks in the period 2005 - 2015
2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 | 2015 | |
State Commercial Bank | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 5 | 4 | 7 |
Joint Stock Commercial Bank | 35 | 35 | 34 | 40 | 39 | 38 | 35 | 32 | 33 | 31 | 28 |
NHLD&NN | 38 | 38 | 46 | 49 | 50 | 58 | 59 | 60 | 60 | 55 | 58 |
Total | 78 | 78 | 85 | 94 | 94 | 101 | 99 | 97 | 98 | 90 | 93 |
Maybe you are interested!
-
Overview of Business Activities of Lao Agricultural Bank -
Overview of Credit Activities and Credit Risks at Vietnam Bank for Agriculture and Rural Development -
Some solutions to limit risks in international goods trading activities of Vietnamese traders - 10 -
![Experience in Bank Marketing Activities in Some Countries [4] & [24]](https://tailieuthamkhao.com/en/uploads/2024/08/22/experience-in-bank-marketing-activities-in-some-countries-4-24-120x90.jpg)
Experience in Bank Marketing Activities in Some Countries [4] & [24] -
Analysis and proposal of solutions to expand retail activities at Military Commercial Joint Stock Bank - 1
Source: SBV, Statistics from SBV website, section System of credit institutions
In addition to the increase in quantity, the legal capital of Vietnamese commercial banks on the books has also increased strongly. On November 22, 2006, the Government issued Decree 141/2006/ND-CP stipulating that commercial banks must increase their legal capital to a minimum of VND 1,000 billion by the end of 2008 and VND 3,000 billion by the end of 2010. Equity is the part of a bank's debt assets that does not need to be repaid and thus acts as a source of protection and a buffer in case the value of the bank's assets declines and business makes a loss. According to international practice, ensuring adequate equity is regulated by the minimum capital adequacy ratio (CAR), which requires commercial banks to have sufficient equity corresponding to the scale of asset value after adjusting for the risks of each asset group. Table 4.2 shows the sharp increase in charter capital of the Vietnamese banking system, from VND 28,928 trillion in 2005 to VND 460,279 trillion in 2015.
Looking back at the process of increasing charter capital, although commercial banks faced many difficulties, they still found ways to increase capital on the books (Nguyen Xuan Thanh, 2016). However,
The growth of charter capital has some negative aspects: (i) Due to the pressure to increase capital, commercial banks have borrowed money from one commercial bank to contribute to another commercial bank, creating a cross-ownership structure. Cross-ownership causes credit institutions and major shareholders to not comply with credit safety regulations, which will lead to risks for the commercial banking system.; (ii) Large economic groups holding bank shares during the capital increase process will become major shareholders or owners of the bank. A private bank with major shareholders from corporations will be under pressure to give preferential credit to these shareholders. This leads to consequences of violating the bank's risk management principles.
4.1.2. Operational situation of Vietnam's commercial banking system
From 2005 to 2015, Vietnamese commercial banks experienced many periods of fluctuations. Total outstanding credit increased at an average rate of 27% per year, equivalent to more than 96% of GDP by the end of 2010. Vietnamese commercial banks also mobilized a significant amount of capital, by the end of 2015 the entire system had mobilized more than 5 million billion VND (Table 4.2). The credit boom in the period of 2005 to 2010 could be the result of loose monetary policy. The SBV's policy interest rates were kept unchanged throughout 2007 (refinancing rate at 6.5%, discount rate at 4.5% and base rate at 8.25% in Figure 4.1).