STATE BANK OF VIETNAM MINISTRY OF EDUCATION AND TRAINING HOCHIMINH UNIVERSITY OF BANKING GRADUATION THESIS MAJOR FINANCE – BANKING TOPIC BANK SPECIFIC AND MACROECONOMIC DETERMINANTS OF CREDIT RISK IN.
INTRODUCTION
Reasons for choosing a research topic
Vietnam's market-based economy relies on monetary policy and a robust banking system as the vascular network of the national economy Banks act as intermediary financial institutions in credit relationships with businesses and individuals, serving as both borrowers and lenders to mobilize capital The banking sector is the lifeblood of the financial system, and as the economy integrates internationally, its development drives overall economic growth Bank credit profoundly shapes the social reproduction process by expanding production capacity, supporting industrial growth in a developing economy, and creating favorable conditions for industries Additionally, credit flows from banks contribute to the recovery and development of economies around the world through efficient financing and capital allocation.
The global financial crisis of 2008-2009 unleashed widespread economic turmoil, with the banking and finance sector bearing the brunt of the US-origin storm that swept through global markets As the turmoil spread, many large, once-strong banks in Western Europe and North America faced severe distress, shaking confidence in financial systems across advanced economies Yet outside the core crisis, many people perceived relatively little direct impact on our banking system, as we stood on the periphery of the storm and avoided its most destructive effects.
The emergence of a storm in the US banking sector and signs of its collapse have prompted Vietnamese banks to reevaluate their business models promptly Despite these introspections, Vietnam still grapples with macroeconomic fluctuations—rising inflation, higher unemployment, and slow GDP growth—that drive operating losses for many firms and push more businesses toward bankruptcy Consequently, bad debts to banks are climbing, placing a heavy burden on the Vietnamese banking system and underscoring the need for stronger risk management and resilience.
Vietnam's economy is currently affected by COVID-19, but a major concern today is the credit risk faced by banks The development of the commercial banking system has had a significant impact on the commodity economy, and as the economy advances toward a market-based model, banks become indispensable financial institutions that support growth Commercial banks play a crucial role in driving development by strengthening the banking sector, improving the business environment, mobilizing domestic capital to fund production and business activities, and financing imports and exports Therefore, commercial banks always face a range of risks in their operations.
In the process of formation and development, banks do not have the ability to avoid potential risks such as payment risks, foreign exchange risks, market risks, etc However, one of the most important risks is credit risk Credit risk is a concern of many commercial banks in particular and the national economy in general Credit risk causes a lot of damage such as financial losses, business operations are affected by credit levels, not only that credit risk if high can cause banks to go bankrupt Therefore, in order to improve the ability of commercial banks to manage credit risk, the analysis and study of macro and micro factors affecting the credit risk of banks is urgent in each period
Stemming from the above reasons, the author chose to carry out the research project "Macroeconomic and specific factors of the Bank determining credit risk at
Vietnamese banks" to study the factors that have affected the bank's financial capacity, Besides, proposing methods to improve the financial capacity of Vietnamese commercial banks This article focuses on the factors that affect the credit risk of Vietnamese commercial banks.
Research objectives
The overall objective of this study is to study macro and micro factors for the credit risks of Vietnamese commercial banks
Modeling is based on previous studies
• Check the risk confirmation activities of Vietnam Commercial Joint Stock Bank
• Check the effect of the instructions
• Propose solutions and recommendations for joint stock commercial banks to improve the credit risk stability of Vietnam's joint stock commercial banks, limiting unnecessary risks.
Research Questions
To achieve research objectives, the thesis focuses on answering the following key research questions:
(i) What micro and macro factors affect the credit risks of Vietnamese commercial banks?
(ii) What is the bank's model and method of measuring credit risk?
(iii) How have these factors affected the bank's credit risk?
(iv) How has credit risk affected Vietnam's commercial banks?
Objects and scope of study
The subject of this study is credit risk, factors affecting credit risk of commercial banks in Vietnam
Research space: Research data conducted on 31 Commercial Banks in Vietnam Study time: In the study using data collected from 2010-2020
Search for learning methods
To overcome the weaknesses of individual methods and enhance the reliability of the findings, this study adopts a mixed-methods approach by employing quantitative, qualitative, and other methods in parallel The quantitative component is used to detect the relationship and correlation between donations, while the qualitative component verifies and interprets the analysis results, providing triangulation and deeper understanding of the data.
This study uses a secondary data collection approach that involves building research models, designing research samples, and gathering data from published sources Data were extracted from commercial banking websites, including annual reports, cash flow statements, currency data, and reported business results, spanning the period 2010–2020.
This research uses a quantitative approach with Stata to perform panel data regression analysis, estimating pooled OLS, random effects (REM), and fixed effects (FEM) models, and selecting the best specification using tests such as the Preusch and Pagan test and the Hausman test To address issues of changing error variance and potential correlations in the data, the study applies Feasible Generalized Least Squares (FGLS), and endogeneity is tackled with the System Generalized Method of Moments (S-GMM) estimator.
Qualitative method: used to compare results from experimental analysis with results from previous studies to explain research objectives and research questions.
Scientific and practical implications
These research findings offer actionable insights for managers, policymakers, and academics seeking to improve the efficiency of bank operations and strengthen the banking sector, informing both banking research and administration They emphasize how evidence-based recommendations can uplift operational performance and policy effectiveness, while preserving references for future generations.
Structure of research paper
This chapter outlines the research scope and rationale, explaining why the topics were chosen, identifying the key research issues, and articulating the study’s objectives and research questions It defines the research objects and boundaries, discusses the implications of the work, and presents the overall structure of the research process.
THEORETICAL BASIS, RELEVANT EMPIRICAL RESEARCH
Theoretical basis
2.1.1 Overview of credit activities of commercial banks
Commercial banks are financial institutions that operate in the currency market, primarily collecting deposits and granting loans In the course of their activities, they face many uncertainties, with credit risk—the risk of loss from borrowers failing to repay—being a central concern of lending to customers Managing this risk involves rigorous credit evaluation, ongoing monitoring, and prudent capital management to sustain liquidity and profitability.
In Vietnam, loan capital accounts for more than 70% of banks’ asset value, making credit risk a highly sensitive issue that affects profitability, safety, and the stability of the banking system Banks and industry experts have conducted extensive studies on the factors driving credit risk and on options to minimize it In parallel, the State Bank of Vietnam has strengthened the legal framework to enhance banks’ risk resilience and boost their guarantee capacity.
Vietnam's commercial banks are contending with elevated bad debt ratios that undermine profitability and threaten financial stability In response to a Prime Minister directive, these banks are urged to implement robust and safe business policies, while systematically examining the factors driving credit risk, in order to design effective risk‑reduction strategies.
Credit is the relationship in which a lender provides funds or the value of an asset to a borrower, who agrees to repay the principal plus interest The repayment restores the original value and covers the interest charged for the use of funds In a loan or asset-transfer transaction, factors to consider include the loan duration, the level of trust between borrower and lender, and the borrower's ability to repay the principal along with the agreed interest rate.
Bank credit is the credit relationship between banks or credit institutions and borrowers, which can be businesses or individuals In this arrangement, the lender provides funds to the borrower for use over a defined period, and when the term ends the borrower must repay both the principal and the interest to the lending institution.
Bank credit includes the following characteristics:
Bank credit is built on trust Lenders extend loans only when borrowers intend to use the funds for the purposes stated in the contract and can demonstrate the ability to repay both principal and interest when the loan matures A borrower must have confidence that the loan will generate profits and that they can meet the scheduled principal and interest payments throughout the loan repayment term.
Secondly: credit is the transfer of an asset that has a loan term or is repayable
Loan repayment terms are defined by the contract signed between both parties The borrower must repay the loan on time in accordance with those terms If the repayment deadline is missed, the borrower may incur additional costs due to breach of contract.
Credit is built on the principle of repaying both the loan principal and the interest, reflecting the core function of banking where a loan creates an obligation for the borrower to repay the principal and to cover the interest The interest rate acts as compensation to the bank for providing funds and taking on risk, while lenders assess the borrower’s ability to repay to ensure the loan can be serviced.
Credit is a potentially high-risk activity for banks While credit operations can generate substantial profits, they also expose institutions to significant risk Assessing credit risk is difficult because some customers deliberately falsify records, making evaluation challenging and increasing the likelihood of loan losses In addition, macro factors such as epidemics, inflation, and environmental conditions can weaken borrowers’ repayment capacity, complicating loan recovery and elevating overall credit risk for banks.
At the core of lending is an unconditional repayment commitment The loan application process is conducted on strict legal grounds, including credit contracts, promissory notes, mortgage agreements, and guarantee contracts, in which the borrower—and any guarantor—pledge to unconditionally repay the loan to the bank when it is due.
Bank credit rests on two core principles: the loan must be used for its intended purpose and the borrower must have the ability to repay both the principal and the interest on time as stipulated in the contract.
2.1.1.3 The role of bank credit operations
Bank credit drives the efficient use of capital and reinforces the economic planning framework By enabling asset transfers between lenders and borrowers, bank loans give many customers the opportunity to profit through financing and leverage This mechanism increases available investment capital and improves the allocation of financial resources across the economy.
Bank credit plays a pivotal role in expanding and developing foreign economic relations by linking national economies with international credit networks It has become a key channel for credit flows among governments, between organizations and governments, and between individuals and businesses, supporting the growth of foreign trade and the broader international financial ecosystem As foreign trade activity expands and more players participate, robust financial services become an effective competitive tool alongside price, product quality, service, and trade policies By crossing national borders, bank credit accelerates internationalized production, helps form regional and global markets, and opens new avenues for cooperation and competition among countries.
Meeting customer needs by enabling access to capital for business financing and everyday consumption is central to our lending approach When borrowers have a clear purpose for the loan and meet all contract requirements, banks can offer loans more easily, helping customers grow their businesses and manage personal expenses with confidence.
One goal is to motivate individual and business customers to maximize the use of bank loans, encouraging them to deploy capital efficiently to generate higher profits and to ensure timely loan repayments and interest payments to the bank when due.
Overview of factors affecting previous empirical credit risk
Across analyses of credit risk determinants, many authors have used subprime factor groups as an indicator of risk levels, but most studies measure credit risk using metrics such as bad debt ratios or credit risk provisions.
De Lis, Pagés, and Saurina (2001) analyze the growth of bank credit in Spain and its implications for cautious credit risk management, a long-standing concern for banking supervisors Using loan loss reserves as a diagnostic tool, they show that easing bank credit conditions coincides with very low levels of bad debts, which can mask emerging financial imbalances in the non-financial sector The study warns that the true deterioration in credit quality tends to become evident only after defaults materialize, a pattern that is likely to surface with a delay of about three years in the event of a recession in Spain.
Hazimi Bimaruci Hazrati Havidz & William Obeng-Amponsah (2020) analyze the determinants of bank credit risk in Indonesia from both macroeconomic and banking-specific perspectives Using panel data analysis with fixed effects, various generalized method of moments (GMM) estimators, and system GMM to ensure robust results, the study employs a lagged dependent variable in the specification The findings indicate that banks maintain prudent credit risk management, which helps explain why bank-specific variables are more closely linked to credit risk than macroeconomic variables and can buffer banks against macroeconomic shocks Consequently, about half of the factors that determine credit risk show significant effects under fixed effects, difference GMM, and system GMM approaches The key determinants identified include GDP growth, lending rates, and exchange rates.
An analysis of 27 macroeconomic variables, along with loan provisions, net profit margins, productivity, and credit growth for specific variables, as identified by banking authorities, reveals that bank-specific determinants carry more explanatory power than macroeconomic determinants; while many macroeconomic determinants reach statistical significance at around the 10 percent level, most bank determinants have precise, bank-specific meanings that differentiate their impact The results also show that the bank's credit risk remains at its previous value when the lagged credit risk is statistically significant at about the 1 percent level of the current value, though the correlation between lagged and current credit risk values is likely to weaken over time.
YR Bhattarai (2016) examines the impact of credit risk on the operations of Nepalese commercial banks using an unbalanced panel dataset of 14 banks with 77 observations from 2010–2015 The results show a significant relationship between banking performance and credit risk ratios The study concludes that the bad debt ratio has a negative effect on bank operations, while cost per lending asset has a positive effect, suggesting that higher efficiency in distributing loans and collecting interest relative to operating expenses supports performance The cost per loan asset (often referred to as cost per lending asset) is identified as a key variable influencing improvements in bank operations In addition to credit risk measures, bank size also affects operations Overall, Nepalese commercial banks manage credit risk poorly, indicating a need for prudent credit risk management to protect assets and stakeholders’ interests.
VV Shemetov's 2020 Management study aims to model credit risk while accounting for inflation The author employs the Extended Merton (EMM) model to estimate a firm's credit risk under inflationary conditions and finds that inflation can either bolster or undermine business viability This nuance supports microeconomic reasoning and Keynes's newer insight into how nonlinear inflation can affect output growth The results indicate that inflation dynamics are a key determinant of credit risk, with nonlinear inflation potentially altering default probabilities in ways that depend on firm characteristics and the inflation regime.
The model shows that the New Keynesian non-linear link between output growth and inflation arises from microeconomic business characteristics—such as average value, average annual income, business guarantee payments, asset structure, demand elasticity for corporate goods, and fluctuations in corporate value—that are invisible at lower macroeconomic levels and depend on both exogenous conditions and the quality of the firm’s management team Under the same macro conditions, two companies with different microeconomic profiles can experience opposite inflation effects on their average trajectory and stability, indicating that the optimal inflation rate must be analyzed with micro-level conditions in mind At the microeconomic level, inflation becomes a central element of the business environment and should be considered when assessing long-term corporate credit risk, including scenarios of low inflation Finally, increasing the expected profit margin tends to bolster corporate value and support sustainable development, thereby reducing the probability of default.
An empirical study by Dawood Ashraf, Yener Altunbas, and John Goddard (2007) examines the determinants of credit-derivative usage by major U.S banks The authors find evidence that banks employ credit derivatives as part of their risk-management strategies, and that the level of managerial equity ownership does not hinder this use The study investigates factors likely to drive U.S banking holding companies (BHCs) to trade credit derivatives and to expand trading volumes Consistent with risk-management theory, the results show that BHCs can trade smaller credit derivatives to reduce losses from default, while there is no support for the idea that credit-derivative trading substitutes for hedging When BHCs have longer maturity profiles, they tend to use interest-rate derivatives rather than credit derivatives to hedge interest-rate risk Finally, there is a favorable relationship between the size of a BHC’s C&I loan portfolio and the volume of credit-derivative trading.
Z Fungáčová and T Poghosyan (2011) examine the determinants of profitability in the Russian banking sector with a focus on bank ownership structure, using bank-level data that covers the entire sector from 1999 to 2007 The analysis highlights ownership structure as a key determinant of profitability and calls for a reassessment of previous empirical findings on profit margins, especially those based on cross-country or data-driven studies The results show that the estimated effect of operating size is meaningful for domestic and foreign private banks, with foreign banks tending to report higher profit margins on riskier activities (adjusted for the higher risk), while domestic banks charge lower margins for large-scale operations due to economies of scale Overall, the findings are consistent with theoretical predictions and with results from other studies on profitability in emerging markets, underscoring that bank ownership plays an important role in profitability in emerging economies and should not be overlooked when analyzing profitability determinants.
A study by Ahmed, Takeda, and Thomas (1999) found that loan loss provisions have a significant positive relationship with bad debts Consequently, increases in loan loss provisions signal higher credit risk and deteriorating loan quality, which in turn negatively impacts bank operations and overall performance.
Alshatti (2015) analyzes the impact of credit risk management on the financial performance of Jordanian commercial banks during 2005–2013, using capital adequacy ratio, credit interest rate/credit ratio, hedging/net facility ratio, leverage ratio, and bad debt/total debt ratio as the independent variables Profitability is measured with ROA and ROE as the dependent variables The study concludes that all credit risk management indicators examined have a significant effect on the financial performance of Jordanian banks.
In Vietnam there are studies of Truong Dong Loc (2010); Truong Dong Loc & Nguyen Thi Tuyet (2011) on factors affecting the bank's credit risk This study used the probit model and the linear probability model (logit) The author uses financial/non-financial factors to include in the model from other records of customers applying for loans at the bank, in addition to observing the business activities of customers and the elements of the bank itself that the author observes in fact
Ninh and Ngoc (2012) studied credit risks in lending in the period before 2012 of 6 Investment and Development Joint Stock Banks (BIDV) branches of the Mekong Delta The two authors use the Binary logistic model to identify factors that influence credit risk and the results show that micro-factors that explain credit risk include: being borrowed; Use of loans; The experience of the credit officer; Diversifying business activities; The main business lines generate income to pay debts; Checking and monitoring loans; History of loans and collateral In addition, this result also implies that credit institutions need to pay attention to the above indicators to properly assess the ability of small and medium-sized enterprises to repay debts in order to minimize credit risks
Quy and Toan (2014) investigated the factors affecting credit risk in Vietnam’s commercial banking system from 2009 to 2012, using data from 26 banks They applied qualitative models based on bank data and quantitative models estimated with the ordinary least squares (OLS) method to measure the influence of each factor Their results show that three one-year-lag variables—past bank credit risk, the rate of credit growth, and GDP growth—significantly affect the credit risk of Vietnam’s commercial banks Subsequent research has indicated that GDP growth, reduced credit growth, and previously low-quality loans have together increased credit risk in Vietnam’s commercial banks.
Diep and Kieu (2015) analyze the credit risk characteristics of Vietnam's commercial banks over 2010–2013 using data from 32 banks They estimate both fixed-effects (FEM) and random-effects (REM) models to identify the determinants of credit risk and apply the Hausman test to select the most appropriate specification The results point to a model with three variables—credit growth, bank size, and the ratio of operating expenses to operating income—as determinants of credit risk The analysis indicates that credit growth at commercial banks affects credit risk.
RESEARCH METHODS
Research Model
This study employs a gross data regression model to conduct the analysis An aggregate data estimation technique is used to address inconsistencies among the 14 banks included in the sample The econometric model used in the study is provided below.
In this study, an econometric regression model is applied to estimate the impact of credit risk on the operation of commercial banks Let Y_t denote the dependent variable, β0 the intercept, β the vector of coefficients for the explanatory variables X_t, and ε_t the error term, which is assumed to have zero mean and to be independently distributed over time The regression equation is Y_t = β0 + β' X_t + ε_t, and the model is estimated to quantify how credit risk influences bank performance.
Y: credit risk including CRI, NPL
X: CAP, COL, GROW, ROA, SIZE, LIQ, CIR, INF, GDP
The 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 LIQ it + β 6 INF t + β 7
NPL it = β 0 + β 1 CAP it + β 2 COL it + β 3 GROW it + β 4 ROA it + β 5 LIQ it + β 6 INF t + β 7
For Year t, the credit risk (CRI) of First Bank captures the potential loss the bank could incur if borrowers fail to repay both principal and interest on their loans Non-performing loans (NPL) are a key indicator of this risk, representing loans where payments are overdue or borrowers are in default In essence, credit risk is the expected loss to the bank arising from customer nonpayment, and accurately measuring CRI and NPL helps with risk management, capital planning, and setting loan-loss reserves.
Credit risk is captured by a key indicator that tracks customers who delay paying both principal and interest after a loan is granted, with data considered both on- and off-balance sheet This indicator expresses the risk level per 100 dong of bank loans, helping lenders assess potential losses from borrower delinquency Credit risk comprises two main variables: the Credit Risk Indicator (CRI) and Non-Performing Loans (NPL).
CRI it = Bank's credit risk in 2015
NPL it = Bank bad debt of the year t
CAP it = The capital ratio of the bank in the year t
COL it = Bank collateral of the year t
GROW = Bank credit growth in 2015
ROA it = The profit margin of the 1st bank in the year t
LIQ it = Bank liquidity index of the year t
INF t = Inflation rate in year t
GDP t = GDP growth in 2015 β 0 = Block factor (constant) β 1 , β 2 , β 3 , β 4 , β 5 , β 6 , β 7 = Slope represents the degree of change in the bank's operations when the independent variable changes according to a unit variable eit = error component
The research model is formulated as follows:
CRI it = β 1 + β 2 CAP it + β 3 COL it + β 4 GROW it + β 5 ROA it + β 6 SIZE it + β 7
LIQ it + β 8 CIR it + β9 NPL it + β 10 INF t + β 11 GDP t + + i
CRIit represents the credit risk of bank j in year t Credit risk is the potential loss arising when a borrower fails to repay both the loan principal and interest, or pays late after credit is granted, including exposures on- and off-balance sheet This indicator expresses the risk level per hundred bank deposits, indicating how much risk the bank carries relative to its deposit base.
Data and research variables
Table 3 1- List of commercial banks Numbers Full name and name Numbers Full name and name
Vietnam Industrial and Commercial Joint Stock Bank (Viet t inBank)
17 International Commercial Joint Stock Bank (VIB)
Vietnam Foreign Trade Joint Stock Bank (Vietcombank)
18 Southeast Asia Joint Stock Bank (SeABank)
3 Techcombank 19 Eastern Joint Stock Bank
Vietnam Investment and Development Joint Stock Bank
20 North Asia Joint Stock Bank (Bac A Bank)
Stock Bank (VPBank) 21 An Binh Joint Stock Bank
Joint Stock Bank (MB) 22 DongABank (DongA
Saigon Commercial Tin Joint Stock Bank (Sacombank)
23 Vietnam Commercial Joint Stock Bank (Vietbank)
Stock Bank (SCB) 24 National Joint Stock Bank
Stock Bank (ACB) 25 Nam A Bank (Nam A
Stock Bank (Eximbank) 26 Viet A Joint Stock Bank
Stock Bank (SHB) 27 Kien Long Joint Stock
Bank (MSB) 28 Viet Capital Bank (Viet
Ho Chi Minh City Development Joint Stock Bank (HDBank)
Saigon Industrial and Commercial Joint Stock Company
Credit risk is the single most significant risk facing banks The success of a bank’s business hinges on precisely measuring and effectively managing this risk more than any other risk (Giesecke, 2004).
Bad debt is a term commonly used around the world with words such as "Bad debt" (NPL), "hard-to-claim debt", "unclaimed debt" for hard-to-claim debts (Fofack,
Definitions of bad debt vary across sources, ranging from problematic loans to unpaid debts that banks cannot profitably service Many authorities define bad debt as loans where principal and interest payments are overdue by 90 days or more, with references such as Berger & De Young (1997), Rose (2004), and Ernst & Young (2004) illustrating this threshold Currently, there is no uniform rule or standard governing what constitutes bad debt.
Capital ratio, defined as equity divided by total assets, measures a bank's capitalization and its ability to absorb losses The equity-to-assets ratio is a primary indicator of risk exposure, with a higher ratio signaling stronger capitalization and reduced risk Since equity is more expensive than deposits, banks that rely more on equity financing tend to be more risk-averse, a stance that is typically reflected in a higher profit margin.
Furlong and Keeley (1989), Van and Roy (2003), Berger et al (2013), Jacob Oduor et al (2017) show a positive impact on credit risk On that basis, the author formulates the following hypothesis:
H 1 : The capital ratio has a positive impact on credit risk
Collateral is the asset the lender accepts to secure a loan, giving the loan a valuable security interest If the borrower fails to meet the loan terms by paying the principal and interest in full and on time, the lender may seize the collateral and resell it to offset the loss.
Berger and Udell (1990) together with Gestel and Baesens (2009) have shown that collateral has a negative impact on credit risk On that basis, the author formulates the following hypothesis:
H 2 : The capital ratio has a negative impact on credit risk
Credit growth is the annual percentage change in lending to individuals and organizations, measured by comparing this year’s credit volume with last year’s It reflects the size of capital supplied to the economy and is a key focus of studies on financial conditions The evidence on its impact on delinquency and bad debt ratios is mixed: some research shows that rapid credit growth is associated with higher delinquencies and rising bad debts, while other studies find weaker or context-dependent effects.
Credit growth does not automatically elevate credit risk When loan growth results from higher demand or greater production output rather than an expansion of supply, it can actually lower risk across the total loan balance Key to this dynamic is increased credit demand, as borrowers seek to raise the share of bank financing in their overall business capital This shift occurs when funding from existing owners or from capital markets becomes more expensive than borrowing from banks.
In the face of rising credit demand, banks often raise lending rates or increase credit review standards On that basis, the author formulates the following hypothesis:
H 3 : Credit growth has a negative impact on credit risk
Return on assets (ROA) is a key profitability metric that expresses the profit-to-asset ratio, showing how efficiently a company uses its assets This indicator compares the profits generated from assets invested in production and business activities with the total assets employed, providing insight into asset utilization and overall enterprise performance By tracking ROA, businesses can evaluate how effectively their asset base turns into earnings and make informed decisions about asset deployment and investments.
Zribi, N., & Boujelbegrave, Y (2011) show that return on assets (ROA) is positively related to risk and statistically significant, indicating that banks at higher risk tend to be more profitable This finding reveals a favorable relationship between ROA and credit risk, where profitability increases with risk exposure among banks On that basis, the authors formulate the following hypothesis:
H 4 : The return on assets has a positive impact on credit risk
Variable banking scales are chosen by many academics to include in the research model, which is measured by taking the natural logarithm of total assets
Bank size is proxied by the natural logarithm of total assets An increase in total assets signals that the bank is in an expansion phase, yet larger asset bases tend to involve more risky assets, which can lower capital adequacy ratios Previous empirical studies show that bank size has a positive influence on financial performance, consistent with the findings of Gul and Zaman (2011) and San and Heng.
(2013), Duong & Nguyen (2021), Bao (2016) On that basis, the author hypothesizes the following:
H5: The size of the bank has a positive impact on financial performance
Liquidity is the ratio of total deposits to total assets, with deposits including both non-term deposits and term deposits This liquidity metric reflects a bank's ability to fulfill its obligations to depositors and directly influences its operating activity, shaping how well the bank can meet withdrawals and other depositor commitments.
Results from these experimental studies are contradictory For example, Owoputi et al (2014) found that liquidity management has a negative effect on CRI and NPL, a finding that leads to the formulation of the study’s fourth hypothesis.
H6: Liquidity has a negative impact on financial performance
3.2.2.9 Ratio of operating expenses to operating income ratios (CIR)
Quality management is calculated by Operating Expenses divided by Net Income plus Non-Income Interest = Earnings
According to research by Pain (2003), Salas & Saurina (2002), Hess & Associates (2010) Based on that, the authors formulated the hypothesis study:
H7: Management quality has a positive impact on financial performance
Inflation rate (INF) is the annual percentage change in the general price level, indicating how much prices rise from one year to the next When inflation increases, consumers’ purchasing power declines, leading to reduced spending and weaker demand for goods Consequently, businesses face higher costs, tighter margins, and slower sales, which can hamper profitability and investment across the economy.
Stagnant business operations have led to profits that are lower than expected These losses reduce the ability of firms to repay their debts, which in turn increases bad debts at commercial banks and elevates credit risk (Filip, 2015; Do & Nguyen, 2013; KTNguyen & Dinh, 2016).
On that basis, the author formulates the following hypothesis:
H 8 : Credit growth has a positive impact on credit risk
GDP growth rate measures how fast a country’s economy expands over a period, typically one year GDP is the total value in dollars of all goods and services produced in the economy, calculated as the sum of spending on consumption, investment, government purchases, and net exports The GDP growth rate is the percentage change in GDP from one period to the next, indicating whether economic activity is expanding or contracting.
Research by Gabriel Jimenez & Jesus Saurina (2006) in Spanish banks between
Evidence on how GDP growth affects bank credit risk is mixed: some analyses find opposite effects when comparing the current-year GDP growth to the GDP growth with a one-year lag, a pattern noted in studies from 1984 and 2002 Conversely, several studies do not detect a meaningful impact of GDP growth on bank credit risk For instance, Poudel (2013) investigates Nepal over 2001–2011 and reports no significant relationship, while Kalirai and Scheicher (2002) report similar results for Austria in 1990–2001 Given these inconclusive findings, the author formulates the following hypothesis:
H 9 : Economic growth has a negative impact on credit risk
Previ ous resea rch mark
The debt cannot be fulfilled
Zribi, N., & Boujelbegrav e, Y (2011) Fungáčová, Zuzana; Poghosyan, Tigran (2011)
Truong Dong Loc & Nguyen Tuyet, 2011; Phan Dinh Khoi & Nguyen Viet Thanh,
& Nguyen Thi Tuyet, 2011; Phan Dinh Khoi
Zribi, N., & Boujelbegrav e, Y (2011) Kolapo, T et.al (2012)
Log (Total assets of commercial banks) (+)
Gul and Zaman (2011); Bao (2016); San and Heng (2013); Duong & Nguyen
Getahun (2015); Desta (2016); Tomuleasa and Cocris (2014);
Inflation rate published by state agencies (+/-)
(Source: Compiled by the author)
Research process
Step 2: Build model and research method
Step 3: Analyze the impact of social responsibility on bussiness
Step 1: Review of background theory and previous studies
RESULTS OF RESEARCH AND DISCUSSION
Descriptive statistics
Table 4 1- Present a summary of the smallest, maximum, standard deviation, and average value of the variable used
Variable Obs Mean Std Dev Min Max cri 326 0.0058 0.0051 -0.0048 0.0363 npl 326 0.0219 0.0173 0.0002 0.214 grow 326 0.2649 0.2931 -0.2552 1.0966 cap 326 0.0928 0.0402 0.0293 0.2554 col 326 0.5468 0.125 0.1448 0.7881 roa 326 0.0077 0.0077 -0.0551 0.0621 size 326 8.0081 0.4901 6.9152 9.1809 liq 326 0.6412 0.1294 0.2508 0.8937 cir 326 0.484 0.1507 -0.7815 0.9337 inf 326 0.0581 0.0483 0.0063 0.1868 gdp 326 0.0601 0.0111 0.0291 0.0708
(Source: Calculation results from Stata software)
Table 4.1 presents the statistical results for the model study, detailing the number of observations, the mean value, the maximum and minimum values, and the degree of error for the variables The total observations cover both the independent and dependent variable settings, totaling 326 observations These metrics capture central tendency, data dispersion, and measurement precision, informing subsequent model evaluation and interpretation.
The credit risk of 31 Vietnamese commercial banks in the period of 2010-2020 has an average value of 0.0058166, the minimum and maximum value is -0.0048 and
In 2012, SHB Joint Stock Bank recorded a value of 0.0363, and in 2019 VPB Joint Stock Bank posted the same figure of 0.0363 Nevertheless, there are still significant disparities among banks, particularly between large and small institutions, due to differences in bank size.
In 2010, bad debts for NPL representative variables averaged 0.0219067 The smallest observed value was 0.0002 for TPB Joint Stock Bank, while SEA Joint Stock Bank recorded a value of 0.214 in the same year. -**Support Pollinations.AI:** -🌸 **Ad** 🌸Powered by Pollinations.AI free text APIs [Support our mission](https://pollinations.ai/redirect/kofi) to keep AI accessible for everyone.
Among 31 commercial banks in Vietnam, the average capital adequacy ratio (CAR) was 0.0928083 (about 9.28%) The CAR ranged from a minimum of 0.029 (2.9%) to a maximum of 0.2554 (25.54%), with the lowest value recorded for Saigon Commercial Joint Stock Bank in 2019 and the highest for Kien Long Joint Stock Bank in 2019.
Secured assets, evaluated with COL representative variables, have an average value of 0.5468604, with a minimum of 0.1448 and a maximum of 0.7881, as observed for TPB Joint Stock Bank in 2011 and for the Vietnam Investment and Development Joint Stock Bank.
Credit growth, denoted by GROW, has an average value of 0.2649466 The observed extremes are a minimum of -0.2552 for VIB Joint Stock Bank in 2012 and a maximum of 1.0966 for Tien Phong Joint Stock Bank in 2013.
The return on assets (ROA) has an average value of 0.0077681 Minimum value of TPB Joint Stock Bank in 2011 -0.0551 The largest value is 0.0621 belongs to
An Binh Joint Stock Bank in 2020
Bank size (SIZE) is measured through the logical nature of the entire asset, with an average value of 8.013 and a standard deviation of 0.494 The smallest observed value is 6.9152 for Ban Viet Joint Stock Bank (VCA/CAB) in 2010, while the largest value is 9.1808 for Vietnam Investment and Development Joint Stock Bank (BIDV) in 2020.
Liquidity (LIQ) has an average value of 0.6412301 with a Deviant Standard of 0.1294157 The minimum value was 0.2508 of Tien Phong Joint Stock Bank (TPBank) in 2012 and the maximum value was 0.8937 of Saigon Commercial Tin Joint Stock Bank (STB) in 2015
The management quality is represented by the CIR variable which has an average value of 0.4841405, which has a standard deviation of 0.1507299 The smallest value is -0.7815 of PVB Joint Stock Bank in 2020 and the largest value is 0.9337 of VBB Joint Stock Bank in 2015
Across the dataset, the average inflation rate is 0.0581693 The minimum value of 0.0063 was observed for BaoViet Joint Stock Bank in 2015, while the maximum value of 0.1868 was recorded for Vietnam Industrial and Commercial Joint Stock Bank in 2011 The standard deviation is 0.0483399, indicating moderate variability around the mean.
In 2018, economic growth values for the two banks averaged 0.0601822, with the Vietnam Industrial and Commercial Joint Stock Bank recording the minimum value of 0.0291 and BaoViet Joint Stock Bank the maximum value of 0.0708; the data exhibit a standard deviation of 0.0111212, indicating moderate variability around the mean.
Correlation Analysis
Table 4 2- Correlation matrix between CRI and independent variables cri grow cap col roa size liq cir inf gdp cri 1 grow -0.2441
(Source: Calculation results from Stata software)
Table 4 3- Correlation matrix of NPL and independent variables npl grow cap col roa size liq cir inf gdp npl 1 grow 0.1067
(Source: Calculation results from Stata software)
A correlation coefficient is a statistical measure that quantifies the relationship between two variables, ranging from -1 to 1; a value of 0 (or near zero) indicates no linear association, while -1 or 1 indicates a perfect, absolute relationship A negative correlation means that as one variable increases, the other tends to decrease, and vice versa, whereas a positive correlation means that as one variable rises, the other also tends to rise There are several correlation coefficients, with the Pearson correlation coefficient being the most commonly used.
All correlation coefficients among the independent variables are below 0.5, indicating low or negligible correlation and suggesting that multicollinearity is not a concern As a result, the dataset of these independent variables can be used for regression analysis to explain the model’s dependent variable.
Multi-line inspection
Multilinearity refers to a situation in which the independent variables in a statistical model are linearly correlated with one another To assess this, the study tests the null hypothesis of no multilinearity using the Variance Inflation Factor (VIF) criterion, and the results are presented in the table that follows.
Table 4.4: Multi-line Test Variable VIF 1/VIF size 2.72 0.367686 cap 2.47 0.404303 liq 2.41 0.41568 col 1.79 0.560037 inf 1.57 0.635094 roa 1.55 0.643957 cir 1.5 0.664553
(Source: Calculation results from Stata software)
All independent variables have a variance inflation factor (VIF) below 4, indicating that multicollinearity in the model is not a serious concern Consequently, the variables included in the analysis of the factors affecting the credit risk of Vietnamese commercial banks are well-suited and compatible with the study framework, enabling the generation of reasonable and credible results.
The result of the usual smallest square (OLS)
Table 4 5 Smallest binary squared estimate of model 1
Source SS df MS Number of obs = 324
Total 0.0087 323 00002706 Root MSE = 0.0043 cri Coef Std Err t P>t [95%
Conf Interval] grow -0.0038 0.0009 -4.14 0.000 -0.0056 -0.0020 cap 0.0376 0.0094 3.96 0.000 0.0189 0.0563 col 0.0099 0.0025 3.83 0.000 0.0048 0.0150 roa -0.0235 0.0387 -0.61 0.545 -0.0998 0.0528 size 0.0025 0.0008 3.07 0.002 0.0009 0.0041 liq -0.0048 0.0029 -1.66 0.097 -0.0106 0.0008 cir -0.0097 0.0019 -4.94 0.000 -0.0136 -0.005
(Source: Calculation results from Stata software)
Table 4.5 presents the ordinary least squares (OLS) results for Model 1, showing that credit growth (GROW), the capital ratio (CAP), collateral (COL), bank size (SIZE), management quality (CIR), and the inflation rate (INF) are statistically significant at the 1% level The liquidity ratio (LIQ) is statistically significant at the 10% level, while no statistical significance is found for the remaining variables.
Table 4 6 Smallest binary squared estimate of model 2
Source SS df MS Number of obs = 324
Total 0.0974 323 0.0003 Root MSE = 0.0168 npl Coef Std Err t P>t [95% Conf Interval] grow 0.0054 0.0035 1.54 0.124 -0.0015 0.012499 cap 0.0417 0.0366 1.14 0.256 -0.0303 0.113836 col -0.0050 0.0100 -0.51 0.612 -0.0248 0.014629 roa -0.1480 0.1497 -0.99 0.324 -0.4426 0.146599 size -0.0028 0.0031 -0.9 0.366 -0.0090 0.003357 liq 0.0140 0.0112 1.24 0.214 -0.0081 0.036209 cir 0.0046 0.0076 0.62 0.539 -0.0103 0.019714 inf 0.0726 0.0243 2.99 0.003 0.0248 0.120494 gdp -0.0682 0.0884 -0.77 0.441 -0.2423 0.105851
(Source: Calculation results from Stata software)
From the OLS results for Model 2, the inflation rate (INF) is statistically significant at the 1% level The credit growth rate (GROW) is statistically significant at the 10% level, while the remaining variables show no statistical significance.
The result of the fixed effect model (FEM)
Table 4.7: Model 1 Fixed Effect Estimate
Fixed-effects (within) regression Number of obs = 324
R-sq: Obs per group: within = 0.2836 min = 1 between = 0.2021 avg = 10.1 overall = 0.2643 Max = 11
F(9,283) = 12.45 corr(u_i, Xb) 0.0129 Prob> F = 0 cri Coef Std Err t P>t [95%
Conf Interval] grow -0.0027 0.0008 -3.09 0.002 -0.0044 -0.0009 cap 0.0341 0.0102 3.33 0.001 0.0139 0.0544 col 0.0134 0.0032 4.18 0.000 0.0071 0.0198 roa -0.1055 0.0359 -2.94 0.004 -0.1763 -0.0348 size 0.0016 0.0018 0.89 0.376 -0.0019 0.0052 liq -0.0004 0.0032 -0.15 0.880 -0.0069 0.0059 cir -0.0073 0.0017 -4.19 0 000 -0.0108 -0.0039
(fraction of variance due to u_i) sigma_e 0.0034 rho 0.4575
(Source: Calculation results from Stata software)
Table 4.7 presents the fixed-effects model (FEM) regression results for Model 1 across five variables—capital ratio (CAP), collateral (COL), credit growth (GROW), management quality (CIR), and return on assets (ROA)—significant at the 1% level, while the inflation rate (INF) is significant at the 10% level and the remaining variables are not statistically significant.
Table 4.8: Fixed Efficiency Estimate of Model 2
Group variable: BANK Number of obs 324
Obs per group: within = 0.0894 min = 1 between = 0.0571 avg = 10.1 overall = 0.0693 Max = 11
F(9,283) = 3.09 corr(u_i, Xb) = -0.2990 Prob > F = 0.0015 npl Coef Std Err t P>t [95%
68 col -0.0066 0.0153 -0.43 0.667 -0.0367 0.0235 roa 0.0099 0.1708 0.06 0.954 -0.3263 0.3462 size -0.0073 0.0087 -0.84 0.399 -0.0244 0.0097 liq 0.0304 0.0156 1.95 0.053 -0.0003 0.0612 cir 0.0027 0.0083 0.33 0.742 -0.0136 0.0191 inf 0.0737 0.0334 2.20 0.028 0.00787 0.1396 gdp -0.0703 0.0898 -0.78 0.434 -0.2470 0.1064
(fraction of variance due to u_i) sigma_e 0.0164 rho 0.1552
(Source: Calculation results from Stata software)
In model 2, Table 4.8 summarizes the FEM results for three variables: Management Quality (CIR) is statistically significant at the 10% level, Inflation (INF) is statistically significant at the 5% level, and the remaining variables are not statistically significant.
Results of random effect model (REM)
Table 4.9: Estimation of the proportional effect of Model 1
Random-effects GLS regression Number of obs 324
Group variable: BANK Number of groups = 32
R-sq: Obs per group: within = 0.2804 min = 1 between = 0.2547 avg = 10.1 overall = 0.2880 Max = 11
Forest chi2(9) = 122.45 corr(u_i, X) = 0 Prob > chi2 = 0
(assumed) cri Coef Std Err with P>z [95%
Conf Interval] grow -0.0029 0.0008 -3.61 0.000 -0.0044 -0.001 cap 0.0356 0.0092 3.86 0.000 0.0175 0.0538 col 0.0117 0.0028 4.16 0.000 0.0062 0.0173 roa -0.0903 0.0347 -2.60 0.009 -0.1585 -0.0221 size 0.00222 0.0011 1.95 0.051 -0.0007 0.0044 liq -0.0012 0.0029 -0.41 0.684 -0.0070 0.0046 cir -0.0079 0.0017 -4.57 0.000 -0.0113 -0.0045 inf -0.0123 0.0060 -2.04 0.041 -0.0241 -0.0005 gdp 0.0170 0.0187 0.91 0.364 -0.0197 0.0539
(fraction of variance due to u_i) sigma_e 0.0034 rho 0.3216
(Source: Calculation results from Stata software)
Using a random-effects model, the analysis indicates that observations differ across banks Table 4.9 reports the REM results for Model 1: credit growth (GROW), capital ratio (CAP), collateral (COL), and management quality (CIR) with return on assets (ROA) statistically significant at the 1% level Bank size (SIZE) and inflation rate (INF) are significant at the 5% level The remaining variables show no statistical significance.
Table 4.10: Estimation of The Proportional Effect of Model 2
Random-effects GLS regression Number of obs= 324
Group variable: BANK Number of groups= 32
70 within = 0.0818 min = 1 between = 0.0942 avg = 10.1 overall = 0.0820 Max = 11
Forest chi2(9) = 28.06 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0009 npl Coef Std Err with P>z [95% Conf Interval] grow 0.0056 0.0035 1.60 0.111 -0.0012 0.0126 cap 0.0457 0.0376 1.22 0.223 -0.0279 0.1195 col -0.006 0.0105 -0.57 0.570 -0.0266 0.0146 roa -0.1221 0.1512 -0.81 0.419 -0.4185 0.1742 size -0.0029 0.0033 -0.87 0.385 -0.0095 0.0037 liq 0.0168 0.0117 1.44 0.151 -0.0061 0.0398 cir 0.0045 0.0076 0.59 0.557 -0.0105 0.0195 inf 0.0736 0.0245 3.00 0.003 0.02551 0.1218 gdp -0.0678 0.0872 -0.78 0.437 -0.2388 0.1031
(fraction of variance due to u_i) sigma_e 0.0164 rho 0.0307
(Source: Calculation results from Stata software)
For model 2, the author uses a random-effects model regression since this approach captures cross-bank variation in the observations Table 4.10 reports the REM results for model 2, showing that the inflation rate (INF) is statistically significant at the 1% level and the credit growth rate is statistically significant at the 10% level, while the other variables are not statistically significant.
Estimation of regression models using OLS, FEM, REM synthesizes
After evaluating the correlation coefficient to establish the relationships among model variables, this study proceeds with regression analysis to quantify the direction and magnitude of the impact of the independent variables on the dependent variable The analysis uses regression specifications such as pooled OLS, fixed effects (FEM), and random effects (REM), together with specification tests to identify the most appropriate method for estimating the relationships.
Table 4 11- Evaluate results using OLS, FEM, REM of Model 1
CRI Coef P value Coef P value Coef P value grow -0.0038*** 0.000 -0.0029*** 0.002 -0.0027*** 0.000 cap 0.0376*** 0.000 0.0357*** 0.001 0.0342*** 0.000 col 0.0099*** 0.000 0.0118*** 0.000 0.0135*** 0.000 roa -0.0235 0.545 -0.0904*** 0.004 -0.106*** 0.009 size 0.0025*** 0.002 0.0022* 0.376 0.0016 0.051 full -0.0048* 0.097 -0.0012 0.880 -0.0004 0.684 cir -0.0097*** 0.000 -0.0079*** 0 000 -0.0073*** 0.000 inf -0.0104 0.002 -0.0123** 0.140 -0.0200*** 0.041 gdp 0.0213 0.354 0.0171 0.352 0.0176 0.364
OLS & FEM FEM & REM OLS & REM
There is no difference between different subjects or times
There is no correlation between the characteristic errors between the objects and the explanatory variables
Errors of estimates do not include deviations between objects
P value Prob> F = 0 000 Prob> Chi2 = 0 000 Prob> Chibar2 = 0 000
Conclusion section Reject H0 Accept H0 Reject H0
(Source: Calculation results from Stata software)
Table 4.11: reflects regression results and tests showing that the results of the randomized impact model (REM) are consistent with the CRI model and used to conduct the analysis
Table 4.12: Evaluation of results by OLS, FEM, REM of Model 2
OLS FEM REM npl Coef P value Coef P value Coef P value grow 0.0054 0.124 0.0052 0.002 0.0056 0.111 cap 0.0417 0.256 0.0604 0.001 0.0458 0.223 col -0.0050 0.612 -0.0066 0.000 -0.006 0.570 roa -0.148 0.324 0.0099 0.004 -0.122 0.419 size -0.0028 0.366 -0.0073 0.376 -0.0029 0.385 liq 0.0140 0.214 0.0304* 0.880 0.0168 0.151 cir 0.0047 0.539 0.0027 0 000 0.0045 0.557 inf 0.0727*** 0.003 0.0738** 0.140 0.0737*** 0.003 gdp -0.0683 0.441 -0.0703 0.352 -0.0678 0.437
OLS & FEM FEM & REM OLS & REM
There is no difference between different subjects or times
There is no correlation between the characteristic errors between the objects
Errors of estimates do not include deviations between objects
P value Prob> F = 0 0013 Prob> Chi2 = 0 1648 Prob> Chibar2 = 0 0307
Conclusion section Reject H0 Reject H0 Accept H0
(Source: Calculation results from Stata software)
Tables 4.11 and 4 12 reflects regression results and tests that show that the results of the random impact model (FEM) are consistent with model 2 and used to conduct the analysis
From the resulting regression table of OLS, FEM, REM pools with models 1 and 2, we compare and select the model as follows:
An F-test was used to compare a pooled OLS model with a fixed effects model (FEM) and to test the null hypothesis that there is no difference across objects or across time, i.e., that the pooled OLS specification is appropriate For the CRI specification with NPL as the dependent variable, the p-value is below 0.05, which leads to rejection of the null hypothesis and indicates that the FEM is the appropriate model for this data.
The Hausman test is used to select between the FEM and REM models with the
Using the Hausman specification test to compare fixed-effects (FEM) and random-effects (REM) models, the results indicate that model choice depends on the dependent variable For the NPL dependent variable, the null hypothesis of no correlation between entity-specific errors and the regressors is rejected (p < 0.05), suggesting FEM is the more reliable specification For the CRI dependent variable, the test yields p > 0.05, so we fail to reject the null and REM is appropriate In sum, FEM is preferred for NPL analyses, while REM is suitable for CRI analyses, consistent with the standard interpretation of the Hausman test.
Breusch-Pagan testing is used to choose between pooled OLS and random effects (REM) models In Model 1, the p-value is less than 0.05, suggesting that the REM model is more suitable In Model 2, the p-value exceeds 0.05, indicating that the pooled OLS model is appropriate.
Through F-testing, Hausman and Breusch and Pagan showed that the random impact model on FEM matched model 2 For the rest of model 2, choosing REM is the most suitable
4.8.1 Similar testing of the correlation and variance of model 1
Table 4 13-Check the cartridge - Test the correlation of the model 1
Wooldridge test for autocorrelation in panel data H0: no first-order autocorrelation
(Source: Calculation results from Stata software)
Table 4 14 -Modified Wald checklist - Check the variance of model 1
Modified Wald test for groupwise heteroskedasticity in fixed effect regression model
H0: sigma(i)^2 = sigma^2 for all i chi2 (32) = 10451.20
(Source: Calculation results from Stata software)
With Model 1, the corrected Wald test tests H0: there is no variance Table 4.13 yields prob> spend2 = 0.0000, which is below 5%, so we reject H0 in favor of H1, indicating that the variance changes (heteroskedasticity) in the model Similarly, the Wooldridge test results in Table 4.12 show Prob> F = 0.0000, less than 0.05, refuting the null hypothesis.
75 hypothesis: There is no similar phenomenon, i.e the model has a phenomenon similarly related
4.8.2 Analog testing and variance of model 2
Table 4.15 -Cartridge Test - Similar Test of Model 2
Wooldridge test for autocorrelation in panel data H0: no first-order autocorrelation
(Source: Calculation results from Stata software)
Table 4 16-Modified Wald Checklist - 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 (32) = 1469.59
(Source: Calculation results from Stata software)
For model 2, the Wald test evaluated H0: there is no variance Table 4.15 shows Prob>Chi2 = 0.0000, which is below 5%, so H0 is rejected in favor of H1, indicating that variance changes occur in the model (heteroskedasticity) At the same time, the Wooldridge test results (Table 4.14) yield Prob>F = 0.00000, also below 0.05, leading to rejection of the null hypothesis of no related phenomenon; thus the model exhibits the related variance phenomenon.
4 9 Estimate regression model by GLS
Thus, the fixed-effects model that exhibits variance changes and related phenomena should be analyzed using the feasible generalized least squares (FGLS) method to address these issues in the model By applying FGLS, researchers can account for heteroskedasticity and evolving error structures, resulting in more reliable estimates and robust inference.
Cross-sectional time-series FGLS regression
Correlation: common AR(1) coefficient for all panels (0.5215)
Estimated covariances = 31 Number of obs = 323 Estimated autocorrelations = 1 Number of groups = 31 Estimated coefficients = 10 Obs per group: min = 5 avg = 10.4193 max = 11
Prob > chi2 = 0.000 cri Coef Std Err with P>z 95% Conf Interval] grow -0.0019*** 0.0004 -3.84 0.000 -0.0028 -0.0009
Roa -0.0594*** 0.0227 -2.61 0.009 -0.1039 -0.0148 size 0.0012* 0.0006 1.91 0.056 -0.0000 0.0025 liq 0.0013 0.0020 0.65 0.516 -0.0026 0.0053 cir -0.0083*** 0.0013 -5.97 0.000 -0.0110 -0.0055 inf -0.0089** 0.0039 -2.26 0.024 -0.0166 -0.0011 gdp -0.0022 0 0121 -0.19 0.851 -0.0260 0.0215
(Source: Calculation results from Stata software)
Table 4.18 FGLS Estimate of Model 2
Cross-sectional time-series FGLS regression
Correlation: common AR(1) coefficient for all panels (0.4008)
Estimated covariances = 31 Number of obs = 323
Estimated autocorrelations = 1 Number of groups = 31
Obs per group: min = 5 avg = 10.41935
Forest chi2(9) = 54.55 Prob > chi2 = 0 npl Coef Std.Err with P>z [95% Conf Interval] grow -0.0006 0.0015 -0.4 0.693 -0.0036 0.0024 cap 0.0173 0.0214 0.81 0.42 -0.0247 0.0595 col -0.0066 0.0052 -1.26 0.207 -0.0170 0.0036 roa -0.111 0.0820 -1.36 0.175 -0.2723 0.0494 size -0.0049*** 0.0018 -2.74 0.006 -0.0084 -0.0013 liq 0.0193*** 0.0062 3.1 0.002 0.0070 0.0314 cir 0.0047 0.0041 1.13 0.26 -0.0035 0.0129 inf 0.0464*** 0.0113 4.08 0 0.0240 0.0686 gdp -0.0080 0.0341 -0.24 0.814 -0.0749 0.0588
(Source: Calculation results from Stata software)
GMM regression model estimate
System GMM (SGMM) is a dynamic panel estimator that addresses endogeneity by using lagged dependent variables as instruments and by combining equations to improve efficiency This article uses the SGMM estimator—the system generalized method of moments—to reach its final conclusions The SGMM framework emphasizes diagnostic tests such as the Sargan test, the Hansen test, and the Arellano-Bond AR(2) test, which assess instrument validity and the exogeneity of the included variables to ensure the reliability of the dynamic-panel results.
4.10.1 GMM results of model 1 and model 2
Model 1 treats the first-order latency of the dependent variable as a standalone predictor alongside other independent variables, forming a dynamic panel data model This dynamic model is typically estimated with the System GMM estimator, selected for its advantages in addressing endogeneity in dynamic panels The results are as follows:
Cri Coef Std Err t P>t [95% Conf Interval]
L1 -0.0366 0.2452 -0.15 0.882 -0.5374 0.4642 grow 0.0031 0.0021 1.5 0.145 -0.0011 0.0074 cap 0.0802 ** 0.0338 2.37 0.025 0.0110 0.1493 col 0.0230 * 0.0116 1.98 0.057 -0.0007 0.0467 roa -0.4739 ** 0.2734 -1.73 0.093 -1.0322 0.0844 size 0.0065*** 0.0026 2.45 0.02 0.0010 0.0120 liq -0.0303* 0.0150 -2.02 0.053 -0.0611 0.0003 cir -0.0103* 0.0053 -1.95 0.061 -0.0211 0.0004 inf -0.0357** 0.0162 -2.2 0.035 -0.0688 -0.0026 gdp -0.0153 0.0235 -0.65 0.519 -0.0633 0.0327
Check out Hansen Prob > chi2 = 0.520
(Source: Calculation results from Stata software)
Table 4.19 shows the dynamic panel data estimates obtained with the S-GMM method using the xtabond2 command introduced by Roodmand The instrument set comprises 13 instruments, which is 18 fewer than the 31 observation groups Sargan and Hansen tests support the model’s validity with relatively large P-values The Arellano-Bond AR(2) test yields P = 0.383, indicating that we fail to reject the null hypothesis of no second-order serial correlation.
Based on the regression results, the model identifies capital ratios (CAP), return on assets (ROA), and the level of play (INF) as statistically significant at the 5% level; in addition, mortgage variables (COL), liquidity (LIQ), and management quality (CIR) are statistically significant at the 10% level, with bank size (SIZE) demonstrating significance at the 1% level, while the remaining variables are not statistically significant Regarding correlations, ROA, LIQ, CIR, GDP growth, and INF are inversely related to CRI, whereas CAP, credit growth (GROW), and COL positively influence CRI; CRI’s first-order lag (the previous year’s CRI) has a P value of 0.519 and a positive coefficient, indicating persistence of CRI from year to year Overall, the regression model confirms these relationships and the persistence of CRI over time.
1 of the study is drawn as follows:
CRI = -0.03549 - 0.0366 L1CRI + 0.003158 GROW + 0.080204 CAP + 0.023016 COL-0.47392 ROA + 0.006561 SIZE - 0.03037 LIQ - 0.01033 CIR - 0.03572 INF - 0.01533 GDP
Table 4.20: MODEL 2 GMM Review npl Coef Std Err t P>t [95%
L1 0.0495 0.058 0.84 0.407 -0.0708 0.1700 grow 0.0083 ** 0.0034 2.4 0.023 0.0012 0.0155 cap 0.0861 *** 0.0187 4.61 0.000 0.0479 0.1243 col 0.0023 0.0066 0.36 0.721 -0.0111 0.0159 roa -0.330 ** 0.1440 -2.29 0.029 -0.6246 -0.0362 size 0.0023 * 0.0013 1.78 0.085 -0.0003 0.0050 liq 0.0087 0.0070 1.25 0.222 -0.0055 0.0230 cir 0.0045 0.0027 1.64 0.112 -0.0011 0.0102 inf 0.2037 *** 0.0165 12.33 0.000 0.1700 0.2375 gdp 0.0533 ** 0.0234 2.27 0.030 0.0053 0.1012
Check out Hansen Prob > chi2 = 0 533
(Source: Calculation results from Stata software)
Model 2’s analysis of the dynamic-panel data using the S-GMM method, implemented with xtabond2 as introduced by Roodmand, is presented in Table 4.20 The validity of the model and its instruments is reflected by 29 instrument variables, which correspond to 31 observation groups In addition, the Sargan and Hansen tests indicate model validity with relatively large P-values The Arellano–Bond test for AR(2) also yields P-values.
= 0.951, meaning that the initial hypothesis of the absence of a 2-tier sequence correlation is refuted
Based on the regression results, the model including credit growth (GROW), return on assets (ROA), and economic growth (GDP) is statistically significant at the 5% level, while the capital ratio (CAP) and inflation (INF) are significant at the 1% level, and bank size (SIZE) is significant at the 10% level; the remaining variables are not statistically significant In terms of correlations, ROA is the only variable inversely correlated with non-performing loans (NPL); the other predictors show no inverse relationship with NPL, while past bad debts have a positive association with current NPL Additionally, the lagged bad-debt variable has a P-value of 0.721 and a positive regression coefficient, indicating that higher bad debts in the previous year are linked to higher bad debts in the following year Therefore, the study presents Regression Model 2 as follows:
NPL= - 0.0267 + 0.0495 L1NPL + 0.008 GROW + 0.0861 CAP + 0.0023 COL
- 0.3304 ROA + 0.0023 SIZE + 0.0087 LIQ + 0.0045 CIR + 0.2037 INF + 0.0533 GDP