INTRODUCTION TO THE RESEARCH TOPIC
Research problem
In today's developing economy, the risk associated with banking is a growing concern The evolution of the banking system, particularly commercial banks, is closely tied to the economic progress of nations These banks play a crucial role in fostering development, enhancing economic models, and mobilizing domestic capital for production, business, and import-export activities Consequently, the adverse effects of commercial banks significantly impact the economic and social fabric of countries, especially in highly competitive markets where banking system risks are more likely to emerge.
Banking operations are susceptible to various risks, with credit risk being the most critical concern for both banks and the broader economy This type of risk not only impacts a bank's capital but also threatens its product offerings Commercial banks primarily generate income from lending activities, making the management of credit risk essential Additionally, changes in the macroeconomic environment—such as declining growth, rising unemployment, and fluctuating interest rates—can exacerbate credit risk, particularly in emerging economies Therefore, it is crucial to thoroughly assess credit risk before addressing overdue debts, exploring effective solutions to mitigate these risks and maintain them at manageable levels This article will analyze the implications of credit risk on the operations of commercial banks in Vietnam.
Research objectives
The general objective of this research is to research the factors affecting credit risk of commercial banks in Vietnam
• Modeling based on previous researches
• Checking the certificate activities of the factor risks of Vietnam Joint Stock Commercial
• Check the effect of instructions
• Proposing solutions and recommendations for joint stock commercial banks to improve the stability of credit risk of commercial banks in Vietnam, limiting unnecessary risks.
Research question
To achieve the research objectives, the thesis focuses on answering the main research questions: (i) What factors affect the credit risk of commercial banks in Viet Nam?
(ii)How have the above factors affected the Bank's credit risk?
The research model's findings suggest that the credit risk control policy will need to be tailored to align with the evolving activities of commercial banks in Vietnam This adaptation is essential for effectively managing financial risks and ensuring stability within the banking sector By implementing targeted strategies based on these results, banks can enhance their risk assessment processes and improve overall credit management practices, thereby fostering sustainable growth in the Vietnamese banking industry.
Object and scope of research
The object of this research is credit risk, the factors affecting credit risk of commercial banks in Vietnam
Research space: Data research was carried out on 31 commercial banks in Vietnam
Research time: In the research, data collected from 2010-2020 is used
Research Methods
In this research, used the following research methods:
This study utilizes secondary data sources, including the financial statements (balance sheets and income statements) of 31 banks from 2010 to 2020, annual reports from the State Bank during the same period, and relevant journal articles on credit risks, to facilitate comprehensive data collection and processing.
The study employed five estimation models—Ordinary Least Squares (OLS), Fixed Effects Model (FEM), Random Effects Model (REM), Generalized Least Squares (GLS), and Generalized Method of Moments (GMM)—to analyze the factors influencing credit risk and to identify the most appropriate model for the research.
Research significance
The findings of this thesis serve as a valuable resource for administrators, policymakers, and scholars, aiming to enhance the operational efficiency of banks and advance research and governance within the banking sector.
Structure of the topic
Chapter 1: INTRODUCTION TO THE RESEARCH TOPIC
This chapter outlines the research work, detailing the rationale behind the chosen topic, the core research problem, and the specific objectives guiding the study It presents the research questions, defines the research object and scope, and highlights the significance of the research Additionally, the structure of the research thesis is discussed to provide a clear framework for the study.
Chapter 2: THEORETICAL BASIS AND OVERVIEW OF PRIOR STUDIES
Chapter 2 presents the theoretical basis of the bank's credit risk, summarizes previous research models on the factors affecting the bank's credit risk to serve as a basis for building a new model study in the next chapter
Chapter 3 outlines the research model, variables, data, methods, and processes employed in the thesis, all grounded in the theoretical framework established in Chapter 2, to achieve results that align with the study's objectives.
Chapter 4: RESEARCH RESULTS AND DISCUSSION
This chapter conducts descriptive statistics on the model variables and evaluates the research model The findings are then analyzed to explore the correlations among the variables and identify the factors influencing the bank's credit risk.
Chapter 5 evaluation of the research results of the topic, limitations and directions for further development From there, make recommendations for Vietnamese commercial banks to avoid factors that affect the Bank's credit risk
Chapter 1 gave an overview of the research topic After analyzing the necessity of the research, the author has outlined the research objectives, clearly defined the subject and scope of research, research methods and finally the layout of thesis including 5 chapter.
THEORETICAL BASIS AND REVIEW OF PREVIOUS STUDIES
Credit of commercial banks
Credit refers to the relationship between a borrower and a lender, where the lender provides the borrower with the right to use borrowed money or goods for a specified period The borrower is then responsible for repaying the full amount, along with any applicable interest, by the agreed-upon due date.
Credit is an economic concept that describes a transactional relationship between two entities, where one party provides a specific amount of value to the other for a designated period The recipient is obligated to repay the borrowed amount within the agreed timeframe.
Credit is a contractual agreement where a borrower obtains money or valuable assets and repays the lender later, usually with interest It also encompasses the creditworthiness or credit history of individuals or companies In accounting terms, credit denotes a bookkeeping entry that either reduces assets or increases liabilities and equity on a company's balance sheet.
2.1.2 The role of credit for commercial banks
The role of the credit bank for import support operations can manifest through the following aspects:
+ Providing capital: like other supporting banks, banks are an important source of capital for import enterprises to purchase reserves, produce, consume goods, purchase machinery and equipment
+ Improve the production and business efficiency of enterprises in the market
Effective business practices are essential for economic accounting and are crucial for securing bank credit Consequently, access to bank credit encourages businesses to focus on enhancing their performance and boosting revenue.
+ Promote import activities to take place more smoothly and quickly
Credit in import business is inherently limited by risk factors, as both exporters and importers face uncertainties To mitigate these risks, banks play a crucial role by providing guarantees that protect both parties For exporters, bank guarantees ensure payment, while importers benefit from credit sources that facilitate transactions, even when their main accounts are unresponsive This banking support effectively bridges the gap between buyers and sellers, fostering a more secure trading environment.
Capital intermediaries, particularly banks, play a crucial role in facilitating foreign support for import activities Currently, the majority of assistance from international monetary and financial institutions is channeled through local commercial banks, making them essential for accessing these resources.
Credit risk
2.2.1 The concept of credit risk
Credit risk is the most critical risk encountered by banks, and their business success heavily relies on the precise assessment and effective management of this risk, more so than any other type of risk (Gieseche, 2004).
Chen and Pan (2012), credit risk is the degree of value fluctuations in debt instruments and derivatives due to changes in the underlying credit quality of borrowers and counterparties
According to Demirguc-Kunt and Huzinga (1999), credit risk management encompasses two key aspects: first, the acknowledgment that losses can become unmanageable after they occur; and second, the advancements in financing methods such as commercial paper and securitization, which have compelled banks to seek out viable loan borrowers amidst increasing non-bank competition.
Castro.V (2013), credit risk is defined as the risk of a loan not being repaid (in whole or in part) to the lender
According to Yurdakul (2014), the primary risk confronting banks is credit risk, which arises when borrowers are unable to repay loans on time or fail to meet their contractual obligations, potentially leading to significant financial losses for the bank.
Credit risk refers to the potential loss or damage to valuable assets due to counterparties' failure to fulfill their contractual obligations (Manab, Theng, and Md-Rus, 2015).
Overdue debt refers to unpaid obligations from customers, indicating a failure by the bank to collect these debts on time, which results in financial losses A higher ratio of overdue debt signifies increased credit risk for the bank.
Risky debt refers to the loans of customers showing signs of repayment difficulties or those who have missed payments, categorized from group 2 to group 5 This type of debt includes overdue payments, debts that have been rescheduled but remain within the new terms, and debts that are currently overdue A higher ratio of risky debt indicates an increased credit risk for the bank.
Bad debts, classified as group 3 to group 5, indicate a customer's diminished ability to repay, often resulting in multiple restructuring of repayment terms and prolonged overdue periods The risk and bad debt ratio serve as critical indicators of credit risk that banks must monitor closely A high level of risky and bad debts suggests significant challenges in loan recovery, prompting banks to implement effective strategies to manage their bad debt ratios.
Collateral serves as both an incentive for customers to repay their debts promptly, thereby preventing the liquidation of their assets, and as a means for banks to recover losses when customers fail to fulfill their obligations under credit agreements.
The provision for credit losses ratio indicates a bank's readiness to handle potential loan defaults by allocating a portion of its income annually for credit provisions This provision is determined by categorizing the bank's entire credit portfolio into various debt groups, with higher deduction rates reflecting increased risk levels A higher ratio signifies a greater credit risk associated with the bank's extensive credit portfolio.
Review of previous studies on factors affecting the bank's credit risk
Ninh and Ngoc (2012) conducted a study on credit risk in lending at six branches of the Banks for Investment and Development (BIDV) in the Mekong Delta prior to 2012, utilizing a Binary logistic model to identify key factors influencing credit risk Their findings revealed that micro-factors such as loan acquisition, loan capital utilization, credit officer experience, business activity diversification, primary income-generating activities, loan monitoring, loan history, and collateral play significant roles in determining credit risk The study emphasizes the importance for credit institutions to consider these factors when assessing the debt repayment capabilities of small and medium enterprises, aiming to reduce credit risk effectively.
In their 2014 study, Quy and Toan analyzed the factors influencing credit risk in Vietnam's commercial banking system from 2009 to 2012, utilizing data from 26 banks They employed both qualitative and quantitative models, specifically the OLS method, to assess the impact of various factors The findings revealed that three key variables—past bank credit risk, credit growth rate, and GDP growth rate—each with a one-year lag, significantly affect credit risk in Vietnamese commercial banks The study concluded that declines in GDP growth and credit growth, coupled with a history of low-quality loans, have heightened the credit risk faced by these banks.
Diep and Kieu (2015) conducted a study on the credit risk characteristics of Vietnamese commercial banks from 2010 to 2013, utilizing data from 32 banks Their analysis employed two regression models: fixed effects (FEM) and random effects.
A recent study utilized the Hausman test to identify the key factors influencing credit risk, ultimately selecting a model that includes three critical variables: credit growth, bank size, and the ratio of operating expenses to operating income The findings indicate that credit growth at commercial banks significantly affects credit risk, highlighting the importance of these variables in assessing financial stability.
Lanh and Ly (2016) conducted a study examining the effects of bad debt reduction on credit risk in 53 countries across Asia, Africa, Latin America, and the Middle East from 1993 to 2012 Utilizing the GMM estimation method, their findings indicate that reducing non-performing loans (NPLs) positively influences investment and facilitates higher economic growth rates in both developed and developing nations Furthermore, the research highlights that external debt related to credit risk adversely affects investment and growth, underscoring the importance of effective debt management strategies within each country.
Khoi and Thanh (2017) conducted a study on the micro factors influencing credit risk prior to 2014 at five Joint Stock Commercial Banks in Hau Giang province Utilizing the Multinomial Generalized Logit and binary logit models, they analyzed the relationship between various factors and credit risk Their findings revealed that key factors impacting credit risk include collateral, the borrower's financial capability, loan history, loan utilization, the primary income source for debt repayment, the experience of credit officers, and the processes of loan inspection and supervision Notably, the study concluded that diversification in business activities does not significantly affect credit risk at either level.
A study by Anh (2018) examined the relationship between bank capital, profitability, and credit risk in Vietnamese joint-stock commercial banks from 2009 to 2016, utilizing data from 15 banks The research employed the Hausman test model and multivariate regression analysis, revealing a negative correlation between equity and profitability, as well as credit risk The findings indicate that higher profitability is associated with increased credit risk, while lower profitability corresponds to decreased credit risk Additionally, the study highlights that bank size influences credit risk, showing both positive and negative correlations with GDP.
A study by Vinh and Sang (2018) examined the influence of macroeconomic and bank-specific factors on bad debt and credit risk within the commercial banking system of Southeast Asia, using data from 204 banks over the period of 2010 to 2015 and employing the differential GMM estimation method The findings indicated that both categories of factors significantly contribute to the rise in bad debt, thereby heightening credit risk for banks in the region Notably, current high levels of bad debt are linked to past bad debt, low profit margins, sluggish credit growth, substantial equity, and larger bank sizes Furthermore, macroeconomic factors were found to notably affect loan quality, with fiscal policy demonstrating a statistically significant negative impact on bad debt levels.
De Lis, Pagés & Saurina (2001) analyze the growth of bank credit and its prudential effects in Spain, highlighting its significance for banking supervisors due to the link between inadequate credit risk management and banking crises The study employs statistical reserves to assess the issue, revealing a trend of relaxed bank credit conditions driven by low bad debt levels However, this situation may lead to financial imbalances in the non-financial sector, with the low credit quality of loans becoming evident during economic recessions, typically surfacing about three years after the onset of a recession in Spain.
The study by Hazimi Bimaruci Hazrati Havidz and William Obeng-Amponsah (2020) examines the determinants of bank credit risk in Indonesia from both macroeconomic and bank-specific perspectives Utilizing panel data analysis through fixed effects, different GMM, and system GMM approaches, the research incorporates lagged determinant variables The findings indicate that Indonesian banks practice prudent credit risk management, resulting in bank-specific factors being more influential than macroeconomic variables in enhancing resilience to economic fluctuations Notably, half of the credit risk determinants, including GDP growth, lending interest rates, exchange rates, and various bank-specific factors, significantly affect credit risk across the analytical methods employed The study reveals that while bank-specific determinants show higher significance, macroeconomic determinants generally maintain a significance level of around ten percent Additionally, the persistence of credit risk is evident, as the lagged credit risk value is significantly correlated with current values, although this correlation is expected to diminish over time.
The study by YR Bhattarai (2016) investigates the impact of credit risk on the performance of Nepalese commercial banks using unbalanced panel data from fourteen banks over the period of 2010-2015 The findings reveal a significant relationship between bank performance and credit risk ratios, indicating that a higher non-performing loan ratio negatively affects bank performance, while a higher cost per loaned asset positively influences it This suggests that efficient loan distribution and earning interest at higher rates contribute to improved bank performance Additionally, the study highlights that a bank's size also impacts its performance It concludes that Nepalese commercial banks currently exhibit poor credit risk management, necessitating adherence to prudent practices to safeguard bank assets and protect stakeholder interests.
VV Shemetov's study (2020) focuses on modeling credit risk with inflation as a key factor, utilizing the Extended Merton Model (EMM) to assess corporate credit risk in inflationary contexts The findings indicate that inflation significantly influences firm survival, aligning with Keynesian theories on the nonlinear relationship between inflation and output growth The model reveals that this relationship stems from microeconomic characteristics, such as average annual returns, asset structure, and demand elasticity, which are often overlooked in macroeconomic analyses Notably, two firms facing identical macroeconomic conditions may react differently to inflation due to their distinct microeconomic environments Therefore, determining an optimal inflation rate requires consideration of microeconomic factors Ultimately, the study underscores that inflation is a critical aspect of the business landscape, affecting long-term corporate credit risk, where low inflation can enhance expected returns, bolster firm value, and reduce default probabilities.
A study by Dawood Ashraf, Yener Altunbas, and John Goddard (2007) investigates the factors influencing the use of credit derivatives by major U.S banks, revealing that these institutions incorporate credit derivatives into their risk management strategies The research indicates that the extent of equity ownership by managers does not impede the use of these derivatives It highlights that bank holding companies (BHCs) engage in credit derivatives trading primarily to mitigate potential losses from defaults, although trading credit derivatives is not seen as a substitute for traditional hedging methods Additionally, BHCs with significant maturities tend to prefer interest rate derivatives for hedging interest rate risks The study finds a positive correlation between the size of a BHC's commercial and industrial loan portfolio and the volume of credit derivatives traded.
Z Fungáčová and T Poghosyan (2011) investigate the factors influencing profitability in the Russian banking sector, highlighting the significance of bank ownership structure Analyzing single bank-level data from 1999 to 2007, the study reveals that ownership structure warrants a reassessment of previous empirical findings on interest margin determinants The research indicates that while operating size significantly impacts both domestic and foreign private banks, foreign banks tend to impose higher margins to mitigate risks, whereas domestic banks benefit from economies of scale, leading to lower profit margins on large-scale operations These findings align with theoretical expectations and are consistent with other studies on profitability determinants in emerging markets, underscoring the critical role of bank ownership in these contexts.
A study by AS Ahmed, C Takeda, and S Thomas (1999) reveals that loan loss provisions significantly impact non-performing loans positively This suggests that higher loan loss provisions correlate with increased credit risk and a decline in loan quality, ultimately adversely affecting bank performance.
Factors affecting the bank's credit risk
The Capital Adequacy Ratio (CAR) is a crucial financial metric that represents the minimum required capital a bank must hold, expressed as a percentage of its risk-weighted credit exposure This ratio measures a bank's available capital, ensuring it can absorb potential losses and maintain stability in the financial system.
The capital ratio reflects a bank's capitalization by comparing its equity to total assets, serving as an indicator of risk aversion A higher equity-to-assets ratio suggests that the bank is more risk-averse, as equity is generally more expensive than deposits Consequently, this increased proportion of equity is anticipated to result in higher profit margins for the bank.
Collateral refers to an asset that a lender accepts as security for a loan If the borrower defaults on their legal obligations, including the full repayment of principal and interest, the lender has the right to seize the collateral and sell it to recover potential losses.
Return on Assets (ROA) is a key financial metric that measures the efficiency of a company's use of its assets to generate profits It represents the ratio of net income to total assets, providing insights into how effectively a business is utilizing its resources in production and operational activities.
The size of a bank is determined by its total assets, which indicate whether the bank is experiencing expansion or contraction Fluctuations in total assets have a significant impact on banking operations, particularly in areas such as lending and deposit mobilization, ultimately influencing the bank's credit risk.
Credit growth refers to the percentage increase in money lent to individuals and organizations compared to the previous year, indicating the capital supplied to the economy Research on the relationship between credit growth and bad debt ratios yields mixed findings, with some studies suggesting that a rapid increase in credit correlates with higher rates of past due and bad debts.
Non-performing loan: Bad debt is a commonly used term in the world with words like
“Non-performing loans” (NPL), “bad debt”, “doubtful debt” for bad debts (Fofack, 2005) or loans problem loans (Berger & De Young, 1997) or defaulted loans that banks cannot profit from
Loans are classified as bad debt when principal and interest payments are overdue for 90 days or more, as noted by Ernst & Young (2004) and Rose (2004) However, there is currently no standardized rule or definition for bad debt across the industry.
Net interest margin (NIM) represents the net return generated from a bank's earning assets, such as loans, investment securities, and leases This financial metric is calculated by dividing the bank's interest income by its total assets, providing insight into its profitability and efficiency in managing interest-earning activities.
The Cost of Income Ratio (CIR) is a key metric that evaluates a bank's efficiency by measuring operating expenses as a percentage of operating income Typically, a lower CIR signifies greater operational efficiency; however, various factors such as the bank's business model and size can influence this ratio.
Return on equity (ROE) is a key financial metric that assesses the profitability of a company by measuring the rate of return on shareholders' equity It evaluates how effectively a firm generates profits from each unit of equity, indicating the efficiency of its investment strategies in driving earnings growth.
Asset quality, often referred to as FATA, is crucial for banks as it helps them evaluate the risk associated with their customer disclosures By assessing asset performance through this metric, banks strive to minimize non-performing loans, which can adversely affect profitability Additionally, asset quality reflects the proportion of fixed assets held by a bank relative to its total assets, providing insight into its financial health.
Liquidity (LIQ) measures the ratio of a bank's liquid assets to its demand liabilities, serving as an indicator of liquidity risk A higher proportion of liquid assets backing demand liabilities results in reduced liquidity risk and improved profit margins for the bank.
The personnel expenses to total assets ratio (PER) reflects how operational costs influence profit margins Banks with elevated operational costs often pass these expenses onto customers by raising interest margins, leading to a positive expected coefficient.
The logarithm of total assets (LA) serves as a proxy for operational size, with the theoretical model suggesting a positive correlation between operational size and profit margins This relationship indicates that larger operations, given the same levels of credit and market risk, are likely to face higher potential losses.
Nontraditional activities of banks (NTA), calculated by the ratio of noninterest income to total interest revenue, significantly impact the assessment of banks' efficiencies Excluding these activities leads to an understatement of individual banks' cost, technical, and allocative efficiencies, while also altering the rankings of ownership groups within the industry Notably, when non-traditional activities are included in the analysis, foreign banks demonstrate greater efficiency compared to public and private sector banks These findings underscore the importance of incorporating non-traditional activities in evaluating bank performance, as their omission can result in misestimations of output and distort efficiency assessments.
RESEARCH METHODS
Research model
The study utilizes a pooled data regression model to analyze 14 selected banks, effectively addressing the issue of heterogeneity among them This econometric approach ensures accurate estimation and insightful results.
In this study, we estimate the impact of credit risk on the performance of commercial banks using an econometric model represented by the equation Y = β0 + βXit + εit, where Y is the dependent variable, β0 is a constant, β represents the coefficients of the explanatory variables, Xit is the vector of these explanatory variables, and εit is the error term, which is assumed to have a zero mean and be independent across time periods.
Y: credit risk including CRI, NPL
X: CAP, COL, GROW, ROA, INEF, INF, GDP
Specific models are as follows:
CRI it = β 0 +β 1 CAP it +β 2 COL it + β 3 GROW it + β 4 ROA it +β 5 INEF it +β 6 INF t +β 7 GDP t + + i Model 2:
NPL it = β 0 +β 1 CAP it +β 2 COL it + β 3 GROW it + β 4 ROA it +β 5 INEF it +β 6 INF t +β 7 GDP t + + i
Credit risk, represented by the Credit Risk Indicator (CRI) and Non-Performing Loans (NPL), is the potential loss a bank faces when a customer fails to repay the principal and interest of a loan in full or makes late payments This risk is assessed for each bank in a given year and indicates the extent of risk for every hundred dong of loans issued Understanding these variables is crucial for evaluating a bank's financial health and managing credit risk effectively.
CRI it = The credit risk of bank i năm t
NPL it = Non-performing loan of i th bank in year t
CAP it = Ratio of capital of i th bank in year t
COL it = Collateral i th bank in year t
GROW it = Credit Growth of i th bank in year t
ROA it = Ratio of profitability of i th bank in year t
INEF it = Cost-effective operation of i th bank in year t
INF t = Inflation rate in year t
The equation for GDP growth in year t is represented as GDP t, where β 0 denotes the intercept or constant The coefficients β 1 through β 7 signify the slope, indicating how bank performance shifts in response to a one-unit change in the independent variable Additionally, eit represents the error component within the model.
Research data & variables
Collect panel data through an observational sample of 31 joint stock commercial banks in Viet Nam in the period from 2010 to 2020 This Bank data is collected from financial statements
Table 3 1-List of Commercial Banks
Number Full name Num Full name
Vietnam Joint Stock Commercial Bank for Industry and Trade (ViettinBank)
Joint Stock Commercial Bankfor Foreign Trade of Vietnam (Vietcombank)
Commercial Joint Stock Bank (SeABank)
3 Techcombank (Techcombank) 19 Orient Commercial Joint
Joint Stock Commercial Bank for Investment and Development of Vietnam
Commercial Bank (VPBank) 21 An Binh Commercial Joint
VietnamThuong Tin Commercial Joint Stock Bank (Vietbank)
Bank (SCB) 24 National Commercial Joint
9 Asia Commercial Joint Stock Bank
Stock Bank (Eximbank) 26 Viet A Commercial Joint
Stock Bank (SHB) 27 Kien Long Commercial
Bank (MSB) 28 Viet Capital Bank (Viet
Ho Chi Minh City Development Commercial Joint Stock Bank (HDBank)
Bao Viet Commercial Joint Stock Bank (BAOVIET Bank)
Saigon Industrial and Commercial Joint Stock Bank (SAIGONBANK)
15 Lien Viet Post Commercial Joint
PetrolimexPetroleum Commercial Joint Stock Bank (PG Bank)
16 Tien Phong Commercial Joint Stock
3.2.2.1 Credit risk of the bank
Credit risk represents the primary challenge for banks, significantly impacting their business success Effective measurement and management of this risk are crucial, more so than any other type of risk (Gieseche, 2004).
CRI =Value of annual provision for bank credit risk t
Total annual bank loan balance t
Non-performing loan: Bad debt is a commonly used term in the world with words like
Non-performing loans (NPLs), often referred to as bad debt or doubtful debt, occur when banks cannot profit from loans that are overdue for 90 days or more, as noted by various researchers (Fofack, 2005; Berger & De Young, 1997; Ernst & Young, 2004; Rose, 2004) Despite the significance of these loans in the financial sector, there is currently no uniform standard or rule for defining or discussing bad debt.
The capital ratio, representing a bank's capitalization, is determined by the relationship between equity and total assets This equity-to-assets ratio serves as an indicator of the bank's risk aversion, as equity typically incurs higher costs than deposits Consequently, a greater equity proportion in total assets suggests increased risk aversion, which is anticipated to result in higher profit margins.
Furlong and Keeley (1989), Van and Roy (2003), Berger et al (2013), and Jacob Oduor et al (2017) demonstrate a positive relationship with credit risk, leading the author to propose the following hypothesis.
H1: Ratio of capital has a possitive impact on credit risk
Collateral refers to an asset that a lender accepts as security against a loan If the borrower defaults on their legal obligations, such as failing to repay the principal and interest in full, the lender has the right to seize the collateral and sell it to recover their losses.
Berger and Udell (1990) with Gestel and Baesens (2009) showed that collateral has a negative impact on credit risk Based on that, the author develops the following hypothesis:
H2: Collateral has a negative impact on credit risk
Credit growth refers to the percentage increase in money lent to individuals and organizations over the past year, reflecting the capital available in the economy Research on the relationship between credit growth and bad debt ratios has yielded mixed findings, with some studies indicating that a rapid increase in credit correlates with higher rates of past due and bad debts.
A study by Robert T Clair (1992) indicates a negative correlation, which is further supported by Keeton's (1999) findings of a significant negative relationship between loan growth and credit risk Consequently, the author formulates the following hypothesis.
H3: Credit growth has a negative impact on credit risk
Return on Assets (ROA) is a key metric that measures the efficiency of a company's asset utilization by comparing its profits to the total assets employed in production and business operations This ratio serves as an important indicator of how effectively an enterprise generates returns from its assets.
A study by Zribi and Boujelbegrave (2011) reveals a positive and statistically significant relationship between return on assets (ROA) and credit risk, indicating that banks with higher profitability tend to take on more risk This finding supports the hypothesis that riskier banks achieve greater returns on their assets.
H4: Return on assets has a positive impact on credit risk
Berge and DeYoung (1997) explore the dual nature of the relationship between operating cost efficiency in banks and credit risk, highlighting both negative and positive correlations They present the "bad management" hypothesis, which suggests that poor management of operating expenses can lead to increased bad debt, resulting in lower operating efficiency Conversely, the "bad luck" hypothesis indicates that unforeseen bad debts can cause banks to incur high costs to address these issues, further diminishing efficiency In contrast, the "skimping" hypothesis posits that when banks cut costs in the short term, it can negatively impact loan quality over time, establishing a positive relationship between operating cost efficiency and bad debt.
Total operating income Based on that, the author develops the following hypothesis:
H5: Cost-effective operation has a positive impact on credit risk
The inflation rate (INF) is a crucial economic indicator that reflects the annual increase in prices When inflation rises, consumer spending tends to decline, resulting in decreased demand for goods This slowdown can create challenges for businesses, leading to lower profits and difficulties in loan repayment Consequently, the rise in bad debts among commercial banks increases credit risk, as highlighted in studies by Filip (2015), Do & Nguyen (2013), and KTNguyen & Dinh (2016).
INF = Inflation rate announced by the State agency Based on that, the author develops the following hypothesis:
H6: Inflation has a positive impact on credit risk
The economic growth rate, measured by Gross Domestic Product (GDP), represents the total dollar value of all goods and services produced within a country It encompasses all expenditures in the economy, including consumption, investment, government spending, and net exports The GDP rate indicates the percentage change in GDP over a specific period, typically one year.
According to researchers Tan and Floros (2012), economic growth may lead to a decrease in bank profitability, subsequently elevating credit risk This indicates a positive correlation between economic growth and credit risk, prompting the author to formulate the following hypothesis.
H7: Economic growth has a positive impact on credit risk
Value of annual provision for bank credit risk t Total annual bank loan balance t
Ong & Heng (2012); Daniel Foos&ctg(201 0); Hess&ctg (2009);
Total bad debt Total loan balancex100
Zribi, N., & Boujelbegrave , Y (2011) Fungáčová,Zu zana;
2 COL Collateral the loan amount on the total assets used as security for loans (-)
Pagés&Saurin a, 2001; Trương Đông Lộc&Nguyễn Thị Tuyết, 2011;
Phan Đình Khôi & Nguyễn Việt Thành, 2017
Luc Laeven& Giovanni Majnoni (2002); Clair,
4 ROA Return on assets Earnings after tax
Zribi, N., & Boujelbegrave , Y (2011) Kolapo, T et.al (2012)
Total operating costs Total operating income (+/-)
Inflation rate announced by the
Step1: Review of background theory and previous studies
Step 2: Build model and research method
Step 3: Analyze the impact of social responsibility on business performance
Step 4: Test the regression model
Step 5: Analyze the regression results and discuss the research results
Research method
The least squares estimation model, or OLS estimation model, is used to estimate the coefficients of explanatory variables on the dependent variable's mean by minimizing the sum of the squares of the residuals Residuals represent the difference between actual and predicted values of the dependent variable based on the explanatory factors.
When crossover units exhibit variability, the fixed effects model (FEM) is employed to analyze the influence of explanatory variables on the dependent variable while accounting for individual characteristics FEM posits that while the partial regression coefficients remain consistent across different cross units, the regression intercepts may vary.
The REM model computes distinct intercepts for each cross-unit while also assessing the overall impact of explanatory variables Each unit's cross-intercept is based on a consistent common intercept that remains constant across time and subjects, combined with a random variable that introduces a subject-specific, time-varying error component.
3.4.4 Feasible Generalized Least Square (FGLS)
To address issues of variable variance and autocorrelation in regression models, the Feasible Generalized Least Squares (FGLS) method can be employed This technique utilizes Ordinary Least Squares (OLS) to estimate the model parameters, even in the presence of these phenomena By calculating the variance-covariance matrix of the model errors, FGLS transforms the original variables, allowing for more accurate parameter estimation This approach effectively mitigates the complications arising from variable variance and autocorrelation, enhancing the reliability of regression analysis.
3.4.5 System Generalized Model of Moments (S-GMM)
Antoniou et al (2006) highlighted the effectiveness of the System GMM model for estimating dynamic models, emphasizing its ability to address endogeneity concerns By incorporating lagged variables, System GMM offers robust estimates even in the presence of variable variance or autocorrelation It is recommended to utilize the first degree of the dependent variable as the explanatory variable.
To ensure the appropriateness of System GMM estimates, the author emphasizes four critical conditions, including the Hansen test, which tests the hypothesis H0: the instrumental variable is suitable and free from endogeneity issues For the instrumental variable to be considered statistically significant, the P-value of this test must exceed 10%.
Hansen Sargan's test with the hypothesis H0: the instrumental variable is exogenous with the condition that the P-value is greater than 10%
The Arellano-Bond test (AR(2)) is utilized to assess autocorrelation across all levels, with the null hypothesis (H0) stating that there is no autocorrelation present A p-value exceeding 0.1 indicates that the null hypothesis cannot be rejected, suggesting that second-order autocorrelation is absent.
In addition, it is necessary to ensure that the number of instrument variables cannot exceed the number of units to be studied
3.4.6 Check for suitable model selection
H0: There is no difference between different subjects or time points
H1: There is a difference between objects or different times
If p-value ≤ with (𝛼 = 5%) then reject H0, FEM model is selected, otherwise if p-value ≥ 𝛼 the OLS model is selected
• Hausman test is performed to choose between FEM and REM models:
H0: Error of estimate does not include inter-subject deviations
H1: The error of the estimate includes deviations between objects
If p-value ≤ with (𝛼 = 5%) then reject H0, FEM model is selected, otherwise if p-value ≥ 𝛼 then the REM model is selected
• Breusch & Pagan test to choose OLS and REM:
H0: Error of estimate does not include inter-subject deviations
H1: The error of the estimate includes deviations between objects
If p-value ≤ với (𝛼 = 5%) then reject H0, REM model is selected, otherwise if p-value ≥ 𝛼 then the OLS model is selected
Once the suitable model is chosen, we utilize the Random Effects Model (REM) to analyze the outcomes, employing the Modified Ward test and assessing autocorrelation through the Wooldridge test.
• In the FEM model, the Modified Ward test is used to test the phenomenon of variance:
In statistical analysis, if the p-value is less than or equal to 0.05, we accept the alternative hypothesis (H1) and reject the null hypothesis (H0), indicating that the regression model exhibits variable variance Conversely, if the p-value is greater than or equal to 0.05, we accept the null hypothesis (H0), suggesting that the regression model does not display any change in variance.
• The test for autocorrelation in the FEM model is the Wooldridge test with the hypothesis:
H0: There is no autocorrelation in the model
H1: There is autocorrelation in the model
When the p-value is less than or equal to 0.05, we accept the alternative hypothesis (H1) and reject the null hypothesis (H0), indicating the presence of autocorrelation in the model In cases where autocorrelation and variance changes are detected, the Feasible Generalized Least Squares (FGLS) model is employed to effectively manage these issues.
SGMM addresses the endogeneity issue of certain explanatory variables by utilizing instrumental variables To evaluate the suitability of these instrumental variables, the Sargan test or Hansen test for over-identifying restrictions can be employed.
• Hansen test is used to test the over-identifying of instrumental variables (to determine whether there is a correlation between instrumental variables and residuals in the model) with the hypothesis:
H0: Instrumental variables are suitable (satisfactory over-identifying)
H1: Instrumental variables are not appropriate (not satisfactory over-identifying)
When the p-value exceeds 10%, we accept the null hypothesis (H0), indicating that the instrumental variables utilized in the model are appropriate Conversely, if the p-value is below 10%, we reject H0 in favor of the alternative hypothesis (H1), suggesting that the instrumental variables are unsuitable for the model.
• The second order autocorrelation test (AR2) to test the second order correlation of residuals in the model, with the hypothesis:
H0: There is no quadratic correlation of the residuals
H1: There is a quadratic correlation of the residuals
If the p-value is greater than 10%, we accept the null hypothesis (H0), indicating that the model's residuals do not exhibit second-order autocorrelation and that the model meets the necessary requirements Conversely, if the p-value is less than 10%, we reject the null hypothesis, suggesting that the residuals of the model do show second-order autocorrelation, which implies that the model is unsatisfactory.
In Chapter 3, the author outlines the research model and variables, detailing the data collection process utilized in the thesis The chapter also examines various research methods, including Ordinary Least Squares (OLS), Fixed Effects Model (FEM), Random Effects Model (REM), Feasible Generalized Least Squares (FGLS), and System Generalized Method of Moments (S-GMM), to ensure that the results align with the intended objectives and to identify the most effective model for analysis.
RESEARCH RESULTS AND DISCUSSION
Descriptive statistics
Table 4 1- Shows the summary of minimum, maximum, standard deviation and mean of the variable used
Variable Obs Mean Std Dev Min Max
Table 4.1 presents descriptive statistical results, including the number of observations, mean values, maximum and minimum values, and error levels for the variables in the study model In total, there are 249 observations for both the independent and dependent variable settings.
Between 2010 and 2020, the average credit risk of 31 Vietnamese commercial banks was 0.0118, with a minimum value of 0.002 recorded at Nam A Commercial Joint Stock Bank in 2019 and a maximum value of 0.0831 at Ban Viet Commercial Joint Stock Bank in 2011 Notably, significant disparities exist in credit risk levels among banks, particularly between larger and smaller institutions.
In 2010, the non-performing loan (NPL) ratio for Saigon Commercial Joint Stock Bank averaged 0.0211, with a minimum value of 0.0002 and a maximum of 0.1140, reflecting the bank's performance in managing loan defaults during that year Similarly, Tien Phong Commercial Joint Stock Company exhibited comparable NPL metrics in 2010.
In 2019, Saigon Commercial Joint Stock Bank reported a capital ratio (CAP) averaging 0.0927, with a minimum value of 0.029 and a maximum of 0.2554, while Kien Long Commercial Joint Stock Bank recorded its capital ratio in 2010.
Collateral with the representative variable COL has an average value of 0.5573, the minimum value and the maximum value are 0.2162 and 0.7881, respectively, of Southeast Asia
Commercial Joint Stock Bank in 2012 and Vietnam Commercial Joint Stock Bank in 2012
Vietnam Investment and Development Joint Stock Commercial Bank in 2020
Credit growth is expressed as GROW as a variable with a mean value of 0.2282, minimum and maximum values of 0.0001 and 0.9754 respectively for Asia Commercial Bank in
2012 and Tien Phong Commercial Joint Stock Bank 2013
Return on assets (ROA) has an average value of 0.0098 The minimum value of National
Commercial Joint Stock Bank in 2012, 2015 and Viet Capital Commercial Joint Stock Bank in
2016 is 0.0001 The largest value is 0.1283 of Technological and Commercial Joint Stock Bank in 2016
The average operating cost performance is 0.5260 The minimum and maximum values are 0.1415 and 0.9494 for Technological and Commercial Joint Stock Bank in 2017 and Vietnam
Thuong Tin Commercial Joint Stock Bank in 2012
The average inflation rate is 0.0554 The minimum and maximum values are 0.0063 and
0.1868 respectively for Baoviet Commercial Joint Stock Bank in 2015 and Vietnam Joint Stock
Commercial Bank for Industry and Trade in 2011
Economic growth has an average value of 0.603 The minimum and maximum values are
0.0291 and 0.0708 respectively for Vietnam Joint Stock Commercial Bank for Industry and
Trade in 2018 and Baoviet Commercial Joint Stock Bank in 2018.
Correlation analysis
Table 4 2-Correlation matrix of Model 1
CRI CAP COL GROW ROA INEF INF GDP
Table 4 3-Correlation matrix of Model 2
NPL CAP COL GROW ROA INEF INF GDP
The correlation coefficient is a statistical measure that quantifies the relationship between two variables, with values ranging from -1 to 1 A coefficient close to 0 indicates a lack of correlation between the variables, while values near -1 or 1 signify a strong negative or positive relationship, respectively.
The correlation coefficient indicates the strength and direction of a relationship between two variables A coefficient of -1 or 1 signifies an absolute relationship, while a negative value indicates that as one variable (x) increases, the other variable (y) decreases, and vice versa Conversely, a positive correlation coefficient (r > 0) suggests that when x increases, y also increases, and when x decreases, y decreases Among various correlation coefficients, the Pearson correlation coefficient is the most widely used.
The table indicates that all correlation coefficients are below 0.5, suggesting no correlation among the independent variables As a result, this dataset of independent variables is suitable for regression analysis to elucidate the model's dependent variable.
Multicollinearity test
Multicollinearity occurs when independent variables in a model exhibit linear correlation with one another This study examined the hypothesis of the absence of multicollinearity by applying the Variance Inflation Factor (VIF) criterion, with results detailed in the accompanying table.
The Variance Inflation Factor (VIF) for all independent variables is below 4, indicating that multicollinearity in the model is not a significant concern Therefore, the variables utilized in the analysis are deemed appropriate for evaluating their impact on the credit risk of commercial banks in Vietnam.
Result of the Ordinary Least Squares (OLS)
Table 4 5-Ordinary Least Squares Estimates Of Model 1
Source SS df MS Number of obs = 249
CRI Coef Std.Err t P>t [95%Conf Interval]
Table 4.5 shows the OLS results of the model 1, I found that ratio of capital (CAP), collateral (COL), cost-effective operation (INEF) have statistical significance at 1% Inflation
(INF) has statistical significance at 5% Credit growth (GROW) has statistical significance at
10% and no statistical significance was found in the remaining variables
Table 4 6- Ordinary Least Squares Estimates Of Model 2
Source SS df MS Number of obs = 249
NPL Coef Std.Err t P>t [95% Conf Interval]
In the OLS results for model 2, the inflation rate (INF) demonstrates statistical significance at the 1% level, while cost-effective operation (INEF) shows significance at the 5% level Additionally, the capital ratio exhibits statistical significance at the 10% level However, no statistical significance was identified in the other variables analyzed.
Result Of The Fixed Effect Model ( FEM)
Table 4 7-: Fixed Effect Estimates of Model 1
Group variable: bank Number of groups = 31
Obs per group: within = 0.1978 min = 1 between = 0.0538 avg = 8 overall = 0.1238 max = 11
CRI Coef Std.Err t P>t [95%Conf Interval]
The fixed-effects model regression results presented in Table 4.7 indicate that the ratio of capital (CAP), collateral (COL), credit growth (GROW), cost-effective operation (INEF), and inflation rate (INF) are statistically significant at the 5% level Furthermore, CAP, COL, GROW, INEF, and INF demonstrate even stronger statistical significance at the 1% level, while the other variables in the model do not show statistical significance.
Table 4 8-Fixed Effect Estimates of Model 2
Group variable: bank Number of groups = 31
Obs per group: within = 0.0909 min = 1 between = 0.1368 avg = 8 overall = 0.098 max = 11
NPL Coef Std.Err t P>t [95%Conf Interval]
Table 4.8 presents the FEM results for Model 2, revealing that the capital ratio (CAP), cost-effective operation (INEF), and inflation (INF) are statistically significant at the 10% level Notably, both the capital ratio (CAP) and inflation (INF) achieve statistical significance at the 5% level, while the other variables do not demonstrate statistical significance.
Result Of The Random Effect Model (REM)
Table 4 9-Random Effect Estimates of Model 1
Random-effects GLS regression Number of obs = 249
Group variable: bank Number of groups = 31
R-sq: Obs per group: within = 0.1930 min = 1 between = 0.0867 avg = 8 overall = 0.1387 max = 11
Wald chi2(7) = 50.8900 corr(u_i, X) = 0 Prob > chi2 = 0
CRI Coef Std.Err t P>t [95%Conf Interval]
The author conducted a regression analysis using the Random Effects Model (REM) to account for variations among banks The results, presented in Table 4.9, indicate that the ratio of capital (CAP), credit growth (GROW), and cost-effective operation (INEF) are statistically significant at the 1% level Additionally, collateral (COL) and inflation (INF) are significant at the 5% level, while other variables did not demonstrate statistical significance.
Table 4 10-Random Effect Estimates of Model 2
Random-effects GLS regression Number of obs = 249
Group variable: bank Number of groups = 31
R-sq: Obs per group: within = 0.0867 min = 1 between = 0.1897 avg = 8 overall = 0.1049 max = 11
Wald chi2(7) = 28.2400 corr(u_i, X) = 0 Prob > chi2 = 0.0002
NPL Coef Std.Err t P>t [95%Conf Interval]
In Model 2, the author employs a random effects model (REM) to account for variations among banks The results presented in Table 4.10 indicate that the ratio of capital (CAP), cost-effective operation (INEF), and inflation rate (INF) are statistically significant at the 10% level, with the inflation rate (INF) achieving significance at the 1% level, while the ratio of capital (CAP) and cost-effective operation (INEF) are significant at the 5% level Other variables in the model do not demonstrate statistical significance.
Estimating the regression model by Pooled OLS, FEM, REM methods
This study examines the correlation coefficient to understand the relationship between variables, followed by a regression analysis to assess the impact of independent variables on the dependent variable The analysis utilizes methods such as Pooled OLS, Fixed Effects Model (FEM), and Random Effects Model (REM), along with tests to determine the most suitable regression approach.
Table 4 11-Estimating the results by OLS, FEM, REM of Model 1
CRI Coef P-value Coef P-value Coef P-value
OLS & FEM FEM & REM OLS & REM
There is no difference between different subjects or time points
There is no correlation between the characteristic error between the subjects and the explanatory variables
The error of the estimate does not include the deviations between objects
P-value Prob > F = 0.000 Prob > Chi2 = 1.000 Prob > Chibar2 = 0.000
Table 4.10 presents the regression results, indicating that the findings from the random effects model (REM) align with those of the CRI model, which are utilized for the analysis.
Table 4 12-Estimating the results by OLS, FEM, REM of Model 2
NPL Coef P-value Coef P-value Coef P-value
OLS & FEM FEM & REM OLS & REM
There is no difference between different subjects or time points
There is no correlation between the characteristic error between the subjects and the explanatory variables
The error of the estimate does not include the deviations between objects
P-value Prob > F = 0.0003 Prob > Chi2 = 0.0057 Prob > Chibar2 = 1.000
Tables 4.11 and 4.12 present the regression results and tests, indicating that the findings from the random effects model (FEM) align with those of Model 2, which are utilized for the analysis.
From the regression table results of Pools OLS, FEM, REM with model 1 and 2, we compare and choose the model as follows:
The F-test is utilized to differentiate between Ordinary Least Squares (OLS) and Fixed Effects Model (FEM) models, testing the null hypothesis (Ho) that there is no significant difference among subjects or time points, which aids in selecting the appropriate pooled OLS model The analysis reveals that the results for both models, concerning the dependent variables Credit Risk Index (CRI) and Non-Performing Loans (NPL), yield a p-value of less than 5%, leading to the rejection of Ho and indicating the suitability of the FEM model.
The Hausman test is utilized to determine the appropriate model between Fixed Effects Model (FEM) and Random Effects Model (REM), with the null hypothesis (Ho) stating that there is no correlation between the characteristic error of the objects and the explanatory variables In cases where the dependent variable is Non-Performing Loans (NPL) and the p-value is less than 5%, the FEM model is deemed more suitable Conversely, when analyzing the dependent variable Capital Adequacy Ratio (CRI) with a p-value greater than 5%, the results support the selection of the REM model as appropriate.
The Breusch and Pagan test is utilized to determine the appropriate model between Pool OLS and REM In the case of model 1, the results indicate a p-value of less than 5%, suggesting that the REM model is more suitable Conversely, model 2 presents a p-value greater than 5%, providing justification for accepting the null hypothesis (Ho) and indicating that the OLS model is the appropriate choice.
The F-test, along with the Hausman and Breusch-Pagan tests, indicates that the random effects model aligns consistently with the fixed effects model in model 2 Therefore, for the remainder of model 2, the random effects model is deemed the most appropriate choice.
Test of variance and autocorrelation
4.8.1 Test for autocorrelation and variance of model 1
Table 4 13-Wooldridge test - Autocorrelation test of model 1
Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation F( 1, 26) = 24.469 Prob > F = 0.0000 ( Output of stata)
Table 4 14-Modified Wald test table - Check the variance of model 1
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model H0: sigma(i) 2 = sigma2 for all i chi2 (31) = 8.3e+33 Prob>chi2 = 0.0000 ( Output of stata)
In Model 1, the Modified Wald test indicates a significant variance phenomenon, as evidenced by the results showing Prob > chi2 = 0.0000, which is less than 5%, leading to the rejection of the null hypothesis (H0) in favor of the alternative hypothesis (H1) Additionally, the Wooldridge test results reveal a coefficient of Prob > F = 0.0000, also less than 0.05, resulting in the rejection of the null hypothesis regarding autocorrelation, confirming the presence of autocorrelation in the model.
4.8.2 Test for autocorrelation and variance of model 2
Table 4 15-Wooldridge test - Autocorrelation test of model 2
Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation F( 1, 26) = 36.554 Prob > F = 0.0000 ( Output of stata)
Table 4 16-Modified Wald test table - Check the variance of model 2
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model H0: sigma(i) 2 = sigma 2 for all i chi2 (31) = 1.4e+06 Prob>chi2 = 0.0000 ( Output of stata)
In Model 2, the Modified Wald test indicates that the null hypothesis (H0) of no variance phenomenon is rejected, with a Prob > chi2 value of 0.0000, which is less than 5% This leads to the acceptance of the alternative hypothesis (H1), confirming the presence of variable variance Additionally, the Wooldridge test results show a Prob > F value of 0.00000, also below 0.05, leading to the rejection of the null hypothesis (H0) concerning the absence of autocorrelation, thus confirming that the model exhibits autocorrelation phenomena.
Estimating the regression model by GLS
The fixed effects model exhibits issues such as variable variance and autocorrelation; therefore, this study employs the feasible generalized least squares (FGLS) method to address these challenges within the model.
Table 4 17-Estimating the FGLS of Model 1
Correlation: common AR(1) coefficient for all panels (0.4395)
Estimated covariances = 28 Number of obs = 246
Estimated autocorrelations = 1 Number of groups = 28
Estimated coefficients = 8 Obs per group: min = 2 avg = 8.7857 max = 11
CRI Coef Std.Err t P>t [95%Conf Interval]
Table 4 18- Estimating the FGLS of Model 2
Correlation: common AR(1) coefficient for all panels (0.5368)
Estimated covariances = 28 Number of obs = 246
Estimated coefficients = 8 Obs per group: min = 2 avg = 8.785714 max = 11
NPL Coef Std.Err t P>t [95%Conf Interval]
Note : Sig *** 0.01 ** 0.05 * 0.1 ( Output of stata)
Estimating the regression model by GMM
4.10.1 The GMM results of model 1
In Model 1, a dynamic panel data model is established by including the first-order lag of the dependent variable as an independent variable, alongside other independent variables This model is typically estimated using the System GMM method, which offers distinct advantages for handling endogeneity The results of this analysis are presented below.
Table 4 19-Estimating the GMM of Model 1
The dynamic panel data estimation analysis using the S-GMM method is presented in Table 4.19, following the xtabond2 approach by Roodman The model's validity is supported by having 24 instrumental variables, which is fewer than the 25 observation groups Additionally, the Sargan and Hansen tests confirm the model's validity with relatively high P-values The Arellano-Bond (AR(2)) test also yields a P-value of 0.273, indicating that the null hypothesis regarding the absence of second-order serial correlation is rejected.
The regression analysis reveals that cost-effective operation (INEF), ratio of capital (CAP), credit growth (GROW), return on assets (ROA), and economic growth (GDP) are statistically significant at the 1% level, while collateral (COL) is significant at the 5% level, with no variables showing significance at the 10% level Additionally, a negative correlation exists between INEF, COL, and ROA with credit risk index (CRI), whereas CAP, GROW, inflation (INF), and GDP positively influence CRI Furthermore, the first-order lag variable of CRI shows a P-value of 0.635, indicating that the previous year's CRI positively affects the current year's CRI Thus, the regression model 1 is established based on these findings.
CRI = 0.0219 + 0.0626 L1CRI + 0.0572CAP – 0.0151COL + 0.0264GROW- 0.0725ROA - 0.0254INEF + 0.0161INF + 0.0463GDP
4.10.2 The GMM results of model 2
Table 4 20-Estimating the GMM of model 2
The results of the dynamic panel data estimation analysis using the S-GMM method for Model 2 are presented in Table 4.20, which employs the xtabond2 command introduced by Roodman The validity of the model and the instrumental variables is demonstrated by the count of instrumental variables.
The analysis reveals that there are 24 observation groups of 25, with the Sargan and Hasen tests indicating model validity due to large P-values Additionally, the Arellano-Bond (AR (2)) test shows a P-value of 0.416, leading to the rejection of the null hypothesis concerning the absence of second-order serial correlation.
The regression analysis indicates that the model includes cost-effective operation (INEF), inflation (INF), and economic growth (GDP) as statistically significant variables at the 1% level, while credit growth (GROW) is significant at the 10% level Other variables did not show statistical significance Additionally, there is no negative correlation between the dependent variable and non-performing loans (NPL); instead, all dependent variables positively influence NPL The first-order lag variable of credit risk index (CRI) has a P-value of 0.489, with a positive regression coefficient, suggesting that the previous year's NPL positively impacts the following year's NPL Thus, the regression model 2 is formulated accordingly.
NPL = -0.0234 + 0.2558 L1NPL + 0.0149CAP + 0.0153COL + 0.0131GROW + 0.0724ROA + 0.0263INEF + 0.0889INF + 0.0956GDP
Research results and discussing research results
Table 4 21-Reseach results of Model 1
Dependent Variable CRI Independent variales Hypothesis Results Conclusion
Table 4 22-Reseach results of Model 2
Dependent Variable NPL Independent variales Hypothesis Results Conclusion
Regression analysis indicates that the equity ratio (CAP) positively influences the credit risk (CRi) of Vietnam Joint Stock Commercial Bank This finding aligns with the perspectives of Shrieves and Dahl (1992) and Jacques and Nigro (1997), who assert that a positive relationship exists between capital and credit risk due to effective market monitoring However, this conclusion contrasts with the views presented by Furlong and Keeley (1989) as well as Van and Roy.
(2003), Berger et al (2013), Jacob Oduor et al (2017) These researchers argue that equity helps to reduce risk and increase the financial stability of banks
Commercial banks typically hold only 5% of total capital, yet this capital is crucial for their stability and operations It serves as a foundational element for the bank's capacity to mobilize capital and engage in lending and guarantee activities The growth and size of a bank's owned capital directly influence its development potential, aiding in the mitigation of credit and financial risks Understanding the relationship between a bank's equity performance and its credit risk ratio is essential for assessing its overall stability and risk management.
Figure 4 1-Relationship between CRI & CAP
Between 2010 and 2020, the equity and credit risk ratios exhibited a similar trend, as illustrated in Figure 4.1 However, from 2009 to 2017, these ratios moved in opposite directions; the credit risk ratio rose steadily while equity capital saw a decline in 2011 Notably, in 2015, the equity ratio experienced a sharp decrease followed by a slight increase in 2016.
By 2020, evidence of an upward trend was still present, leading us to conclude that the equity ratio positively influences credit risk, based on the analysis of both variables and prior research findings.
Figure 4 2-Relationship between CRI & COL
Regression analysis indicates that collateral (COL) negatively impacts Credit Risk (CRi); specifically, as the amount of borrowed funds relative to total assets used as collateral increases, the likelihood of credit risk decreases This finding aligns with previous research by Berger and Udell (1990) and Gestel and Baesens (2009), which also highlighted the adverse effect of collateral on credit risk In Vietnam, commercial banks primarily utilize the judgment method for corporate lending, assessing credit risk through the "5Cs" framework Among these factors, collateral is deemed highly influential, with 92.6% of surveyed banks incorporating it into their analytical models Furthermore, 10.1% of respondents identified collateral as the most critical factor in lending decisions The effectiveness of collateral in mitigating credit risk stems from its straightforward assessment criteria, which include legality, liquidity, value, and management capability.
Relationship between CRI and COL
Figure 4 3-Relationship between CRI & GROW
Figure 4 4-Relationship between NPL & GROW
The graph illustrates the relationship between credit risk and its effects, indicating a significance level of 1% for Credit Risk Indicator (CRI) and 10% for Non-Performing Loans (NPL) Contrary to the initial hypothesis that credit risk negatively impacts itself, the results reveal a positive correlation This finding challenges earlier studies by Robert T Clair (1992) and Keeton (1999), while aligning with the conclusions drawn by Klein.
(2013), Do and Nguyen (2013) and V T H Nguyen (2015)
Relationship between CRI and GROW
Relationship between NPL and GROW
Figure 4 5-Relationship between CRI & ROA
Figure 4 6-Relationship between NPL & ROA
The analysis reveals varying impacts of return on assets (ROA) on credit risk across different significance levels In model 1, at a 1% significance level, ROA negatively influences credit risk, while at a 10% significance level, the effect is less pronounced Conversely, model 2 indicates a positive relationship between ROA and credit risk These findings suggest that ROA can exert both positive and negative effects on credit risk, corroborated by the research of Zribi and Boujelbegrave (2011) and Kolapo et al (2012).
Figure 4 7-Relationship between CRI & INEF
Relationship between CRI and INEF
Figure 4 8-Relationship between NPL & INEF
At a 1% significance level, the analysis reveals a complex relationship between operating cost efficiency in banks and credit risk In model 1, this relationship is negative, while model 2 indicates a positive correlation This suggests that operating cost efficiency can either positively or negatively influence credit risk, aligning with the findings of Berge and DeYoung (1997) They propose the "bad management" hypothesis, which posits that poor management of operating expenses can increase bad debt, and the "bad luck" hypothesis, which suggests that unforeseen circumstances may lead to bad debts Consequently, banks may incur additional costs to address these debts, resulting in decreased operating efficiency Overall, the interplay between operating cost efficiency and credit risk is multifaceted, highlighting the importance of effective management in mitigating financial risks.
The "skimping" hypothesis suggests that when banks prioritize short-term cost-cutting measures, it can result in a decline in loan quality over time, establishing a positive relationship between these two factors (Berge & DeYoung, 1997) Consequently, the interplay between operating cost efficiency and credit risk can yield either positive or negative outcomes.
Relationship between NPL and INEF
Figure 4 9-Relationship between NPL & INF
The regression analysis indicates a significant positive correlation between inflation and active credit risk, with a regression coefficient of 0.0886 for the INF variable at a 1% significance level Specifically, a 1% increase in the inflation rate leads to a 0.0886% rise in the non-performing loan (NPL) ratio This finding aligns with the research conducted by Ramadan, IZ, Kilani, QA, and Kaddumi, TA.
The inflation rate is positively correlated with the bad debt ratio, as rising inflation prompts the State Bank to adopt a tight monetary policy This leads to increased lending interest rates and higher input costs for businesses, ultimately diminishing their operational efficiency and ability to repay loans Furthermore, the banks' restrictive lending practices contribute to an illiquid economy and stagnant business activities, resulting in capital misappropriation among companies and heightened insolvency risks, particularly for small and medium-sized enterprises Consequently, this situation exacerbates the burden of bad debt on banks.
Relationship between NPL and INF
Figure 4 10-Relationship between CRI & GDP
Figure 4 11-Relationship between CRI & GDP
The findings indicate that GDP positively influences credit risk (CRi), with a coefficient of 0.0463, and a coefficient of 0.0925 for non-performing loans (NPL), both significant at the 1% level This suggests that higher GDP growth correlates with an increased likelihood of credit risk These results contradict earlier hypotheses that posit a negative relationship between GDP growth and credit risk, as suggested by researchers such as Kosmidou (2008) and Laeven & Majnoni (2002).
Relationship between CRI and GDP
Relationship between NPL and GDP
2013… ) but received the support and expectation of two researchers Tan and Floros (2012) They believe that economic growth will reduce bank profits, thereby increasing the risk of credit risk
Strong GDP growth is crucial for enhancing borrowing conditions for customers and boosting banks' capital repayment capabilities, ultimately leading to improved business performance and reduced credit risks However, economic growth also presents certain challenges, as it fosters a better business environment and lowers entry barriers for banks, intensifying market competition This heightened competition can diminish individual bank profitability, potentially undermining stability and increasing credit risk.
In Chapter 4, the author presents the analysis results of regression models used to test the research hypotheses This chapter also explores the factors influencing credit risk in joint-stock commercial banks in Vietnam from 2010 to 2020, highlighting the direction of impact for these credit risk factors as identified in the estimated regression model.
In the next chapter, the author will make some recommendations to reduce credit risk of joint stock commercial banks in Vietnam.