Ranking Results Comparison Table

March 12, 2008 of the State Bank, regulations on classification of joint stock commercial banks. In decision 06/2008, article 11 ( appendix 10 ) stipulates standards for classifying banks into 4 types, from which the author proposes criteria for classifying joint stock commercial banks as shown in table 5.6.

Table 5.6: Classification criteria of commercial banks

Group

Classification standards

type of State Bank

Score 80 or more

and above

Reach from 60 to

79 points

Achieve from 50 to

59 points

Score below 50

point

Classification standards

type of thesis

p  0.2

0.2  p  0.4

0.4  p  0.5

p  0.5

Maybe you are interested!

Source: SBV and author

Based on the classification criteria in Table 5.6, the author conducts classification and compares the classification results of the study with the actual classification results of the State Bank, Appendix 11, the author compares the classification results in the two years 2011 and 2012.

In 2011, out of 4 banks that the State Bank classified as weak, the LA model correctly classified 3 banks. In 2012, out of 5 banks that the State Bank classified as weak, the LA model correctly classified 4 banks. Bank with code 3 in 2012 had a classification result according to the LA model of 3, while the State Bank's classification was 4.

From the results of the classification according to the Logit model built and the actual classification of the State Bank, the author summarizes the following results:

Table 5.7: Comparison table of classification results

Year

Overlapping rating ratio	1 level error rate	Group 4 classification ratio overlaps
2011	57.14%	17.14%	75%
2012	54.28%	14.28%	80%

Source: Author's calculation

Thus, the author's classification results in the thesis are quite similar to the classification results of the State Bank. For the years 2013 and 2014, based on the classification criteria in Table 5.6, the author has the classification results of the banks presented in Appendix 12 .

5.3. Some recommendations and policy implications

The growing economy facilitates banking operations, but banks also face more challenges. Determining the types of

The risks that the banking system faces are very important to the economy of a country, especially for a developing country like Vietnam. Identifying risks helps banks control the level of risk and limit the consequences caused by risks. One of the important things in risk control is estimating the risk of default of the bank. Therefore, successfully building an early warning model of default risk for Vietnamese commercial banks is meaningful in both theory and practice.

After building and testing a number of debt warning models for the Vietnamese commercial banking system, the author would like to propose a number of recommendations.

a) Recommendations for commercial banks:

+ Profit helps to offset lost loans and helps to make adequate provisions. Banks need to improve their profitability, the profitability of banks is shown in many indicators, however, in the forecasting model, only variable e11 contributes to forecasting the risk of default, the coefficient of variable e11 in the model is -2.014 less than 0, e11 has a negative impact on the risk of default. According to the model, if variable e11 increases by 1% while other variables remain unchanged, the Odds ratio decreases by 7.5 times. Therefore, to prevent and reduce the risk of bank default, solutions focus on improving the e11 indicator. Commercial banks should take a number of measures to increase the bank's marginal profit such as: increasing image promotion, expanding market share, attracting cheap deposits, and at the same time increasing credit expansion, searching for potential customers. Because the e11 index is calculated from the difference between interest income and interest payments, banks must also limit overdue debt. According to many authors, the NIM ratio is an important factor affecting the CAR ratio (Aktas et al.), high net interest income will help the bank's shareholders make a profit and thereby increase equity to prevent the risk of bankruptcy for the bank.

+ The overdue debt/total debt ratio in the model has a coefficient of 1.03 greater than zero, so d3 has a positive impact on the bank's risk of default. If the variable d3 increases by 1% while other variables remain unchanged, the Odds ratio increases by 2.81 times. First of all, the assessment and accurate classification of loans must be considered by the bank's specialized departments as a top priority task, only then can the scale and level of overdue debt be accurately reflected. Banks need to be aware that beautifying reports to conceal and deal with inspection and supervision agencies, shareholders, and investors only aggravates internal problems. After determining

The level of overdue debt banks need to focus on reducing overdue debt, especially bad debt as soon as possible: first of all, it is necessary to find financial sources to support adequate provisioning to compensate for possible losses. Then consider selling bad debts to businesses, organizations and individuals with sufficient capacity and power to handle the debt. On the other hand, banks need to take measures to limit new overdue debts arising right from the loan approval stage. Overdue debt includes group 2 debts, so banks need to closely monitor group 2 debts immediately, limiting the risk of transferring to bad debt groups. In the long term, banks need to focus on building and perfecting the early warning system for credit risks and bad debts. Banks need to periodically re-evaluate customers' credit ratings, analyze the ability to repay principal and interest according to the commitments of the credit contract, and pay attention to contract violations to assess changes in debt groups.

+ The variable net loans/customer deposits has a coefficient of 3.08 greater than 0, l3 has a positive impact on the risk of default. If the variable l3 increases by 1% while other variables remain unchanged, the odds ratio increases by 1.03 times. Banks should comprehensively consider the causes leading to the high ratio of net loans to customer deposits and then take measures to reduce this indicator. The average value of this indicator is currently 0.9, if the bank has a value of this indicator greater than 0.9, it is necessary to limit the credit growth rate while trying to maintain or increase the level of customer deposits.

+ The model results show that the coefficient of the RGDP variable in the model is -1.29, less than 0, RGDP has a negative impact on the risk of bank default. This suggests that management agencies as well as bank administrators have appropriate policies when macroeconomic conditions reflected in GDP growth rate change, specifically if GDP growth rate shows signs of decline, banks need to focus on ensuring the safety of banking operations because at this time the risk of default has increased due to the impact of macroeconomic factors, banks should improve capital safety ratio, reduce outstanding loans, focus on liquidity as a defense measure. The study also quantified the impact of changes in RGDP on the risk of default of commercial banks. Specifically, if GDP decreases by 1%, the ratio p/(1-p) increases by 3.65 times. Commercial banks with their own bank indicators can calculate in more detail the probability of default when the scenario of a 1% GDP decrease occurs, thereby having appropriate policies for their banks.

+ Also from the reality of data collection and model building, the author found that in order for the model estimation results to be highly reliable and meaningful, the input data must be collected accurately and completely. Commercial banks need to have measures to improve the internal information system, ensuring that information requirements are updated accurately and promptly. The information technology system of banks needs to be upgraded, which ensures that banks store data, as well as accurate data for the purpose of analysis and risk management.

+ Four banks with codes 22, 14, 19, 7 according to the author's calculation have large intercept coefficients:  22 10.265 ,  14 9.879 ,  19 8.206 ,  78.135 , which contains a higher risk of default, needs to be considered comprehensively to find specific, appropriate solutions to help banks reduce the risk of default. Banks with codes 10, 21, 6 have intercept coefficients  10 3.189 , 212.73 ,  6  The smallest 0.982 implies a lower risk of default under the same conditions of the variables.

+ The system of default warning indicators in the thesis is built on the basis of the indicators of the CAMELS model. These indicators help to comprehensively evaluate the operation and determine the level of default risk of the bank. According to the author, Vietnamese commercial banks should build and complete financial statements on the basis of the CAMELS model. Firstly, the CAMELS model will help banks manage risks well. Secondly, bank administrators can easily compare, evaluate and inspect commercial banks when banks agree on evaluation indicators according to international practices.

b) Recommendations to management agencies:

The State Bank is the State management agency of the banking industry, with the goal of supervising the operations of banks towards the stability and health of the system. From the results of building a model to warn of the risk of default of commercial banks, the thesis proposes a number of recommendations to management agencies and the State Bank:

+ According to the results of the Logit model of RGDP, the growth rate of total national income has an inverse effect on the risk of bank default. When economic growth is stable, it will create favorable conditions for banks to operate, increase income, and reduce the risk of default. Therefore, the Government should try to maintain the annual economic growth rate. When the economy declines, the growth rate needs to pay more attention to the safety of the banking system because banks will be at increased risk of default due to the economic downturn. With its supervisory role, the State Bank needs to develop scenarios on increasing

annual economic growth, thereby identifying banks that may default in scenarios to

early warning, monitoring

+ The State and Government of Vietnam, in addition to creating favorable conditions for the business environment for domestic commercial banks, need to provide legal support and administrative reforms. Monetary policies need to take into account the impacts on joint stock commercial banks, especially weak joint stock commercial banks. Currently, resolving bad debts is an urgent and important requirement to reduce the risk of default of joint stock commercial banks. In addition to the efforts of commercial banks themselves to speed up the recovery and settlement of bad debts, helping credit institutions reduce transaction costs and transaction times, the Government needs to soon complete the process of handling secured assets, shortening the time to resolve secured asset handling records. For example, as soon as the debt is transferred to group 3, the bank can implement some liquidation procedures for secured assets. For bad debts of enterprises that banks cannot transfer to debt trading companies and other organizations and individuals, the Government needs to have a mechanism for banks to proactively apply measures to participate in restructuring the finances and operations of enterprises, and can allow banks to participate in the process of restructuring enterprises, allowing debt to be converted into equity capital and participating in business operations. The Government should also consider introducing policies to mobilize more resources to participate in the process of handling bad debts to help speed up this process. For weak commercial banks, to avoid unnecessary negative impacts, the Government can declare banks bankrupt according to the law, and the procedures for bankruptcy of credit institutions also need to be reviewed and amended by the National Assembly accordingly.

+ To avoid the risk of a banking system collapse, the State Bank needs to encourage and eventually force banks to apply regulations according to international practices in their operations, improve their information systems, statistics and forecasting work. Strengthen inspection and supervision of debt classification and loan risk provisioning of commercial banks. Inspection and supervision by the State Bank need to be conducted regularly and with quality to avoid banks' financial statements being good while in fact the bank is on the brink of bankruptcy as in some cases in the past. There needs to be a mechanism that requires banks to report and provide information honestly. For banks with a high risk of default, the State Bank also needs to have requirements on strengthening internal governance, requirements on increasing equity as a defense measure, setting credit growth limits to limit lending, and increasing liquidity for banks.

+ The State Bank needs to develop a set of standards and indicators to assess the risk of default of commercial banks, establish an early warning model for bank defaults, and help banks forecast the risk of default and make appropriate adjustments.

c) Proposing a model and process for building a model to warn of the risk of debt default for commercial banks :

The author proposes to use the Logit model with fixed-effect panel data to warn of the risk of default for Vietnamese commercial banks, the proposal is based on the achieved empirical results. Specifically:

+ The Logit model has identified 4 indicators affecting the possibility of default of commercial banks. The indicators include: RGDP, Marginal interest; Overdue debt / Total liabilities; Net loans / customer deposits.

+ Logit model with array data increases the number of degrees of freedom and thus increases the reliability

accuracy of statistical inferences.

+ The panel data Logit model with fixed effects quantifies the specific characteristics of each bank that affect the probability of default.

If the NH grouping is not clear, it is possible to apply additional neural network and decision tree models to get more information to help determine the risk more accurately because the experimental results of the thesis show that the neural network model and decision tree model have higher classification efficiency than the Logit model.

Through the experimental construction of a debt warning model for commercial banks in the thesis, the author proposes a debt warning process.

The bank default warning process has the following steps:

Step 1: Clearly define the research objectives

- Identify the banks to be studied.

- Clarify and define groups of banks, such as banks in the status of "high risk of default" or "low risk of default". The initial classification or concept of "high risk of default" or "low risk of default" will have a great influence on the warning results.

- Identify specific banks that fall into the “high risk of default” or “low risk of default” group. The grouping of banks should be based on a combination of different analyses.

Step 2: Collect data

- Collect data, these data are mainly indicators in financial statements. Non-financial information is also analyzed to get more information about banks.

- Depending on the characteristics of the data to determine the sample size. When determining the sample size, it is also necessary to pay attention to be able to check the out-of-sample performance of the model.

Step 3: Calculate the indices and determine the indices using model estimation

- Calculate financial indicators from original data. Should calculate a fairly wide class of indicators.

- Check assumptions about input variables: correlation, distribution of variables, etc.

- Identify variables that are capable of distinguishing the two groups of banks.

Step 4: Estimation and model selection

- Estimate Logit function with array data, build neural network model, decision tree.

- Analyze model results. Calculate default probability, classify risks, compare grouping results of models at the same time, and then draw conclusions.

CONCLUSION AND FURTHER RESEARCH DIRECTIONS

Based on the need for forecasting the default of commercial banks and the existence of research gaps at home and abroad, this thesis applies the Logit regression model with array data and some non-parametric models (neural networks, decision trees) to build a model for warning the risk of default for Vietnamese commercial banks. To apply these models, the author chooses the bad debt index combined with the analysis of the performance of banks as criteria for determining the risk of default. The forecast variables in the thesis are mainly built from the indicators in the CAMELS model and are calculated from the financial statements of commercial banks in the period 2010-2015. The results achieved by the thesis are as follows:

+ Firstly: The thesis systematically reviews the methods and models of debt warning applied to companies, especially banks, from univariate analysis methods to methods using modern intelligent techniques that are currently widely used in debt warning analysis. Thereby, it points out the advantages and disadvantages of each method and model, and points out the research gap to consider choosing the Logit model with array data to build a debt warning model for Vietnamese commercial banks.

+ Second : The thesis builds a theoretical basis for the risk of default of Vietnamese commercial banks. The thesis analyzes and clarifies the current state of operations and the risk of default of Vietnamese commercial banks in the period 2010-2015. The author analyzes and proposes criteria to determine the risk of default for Vietnamese commercial banks based on the analysis of bad debts and the performance of banks.

+ Third: The thesis builds and selects a system of 39 micro indicators and 3 macro indicators used in warning of bank defaults. It identifies indicators that directly affect the risk of default of commercial banks. These indicators are: overdue debt / total liabilities; net interest margin, net loans / customer deposits. The study has demonstrated the influence and quantified the level of influence of the RGDP variable on the risk of default of commercial banks.

+ Fourth: The thesis proposes a debt warning model for Vietnamese commercial banks using the Logit regression model with array data. This model helps identify factors and indicators affecting the risk of debt default, and determines the probability of belonging to risk groups for banks in the sample. This model gives results that are economically appropriate, ensuring the standards of a good model. Experimental results of the thesis

Comment