MINISTRY OF EDUCATION AND TRAINING THE STATE BANK OF VIETNAM HO CHI MINH UNIVERSITY OF BANKING LAM THI DO HUYEN FACTORS AFFECTING LOAN LOSS PROVISION OF JOINT STOCK COMMERCIAL BANKS IN VIETNAM GRADUAT.
INTRODUCTION
Reasons for choosing the topic
Banks have long functioned as essential intermediaries that allocate critical financial resources within an economy, making stable banking systems a priority for nations worldwide, including Vietnam The experience of past crises—such as the Great Depression of 1929 and the global financial crisis of 2008—shows that a failure in banks often stems from inadequate credit risk management As Delis and Pagés have noted, poor handling of credit risk was a contributing factor in these crises Therefore, maintaining sound credit risk governance and robust risk management practices is crucial to preserving financial stability and ensuring banks can reliably support economic activity in Vietnam and beyond.
Vietnam is gradually entering the current stage of industrialization and modernization, with the banking sector expanding into more diverse and sophisticated activities However, credit still accounts for a large proportion of bank operations, making lending a core focus of Vietnamese banks (Saurina, 2000).
In commercial banks, lending drives the majority of income yet carries the highest level of risk Non-performing loans and credit risk arise when borrowers default or become insolvent, exerting substantial pressure on banks' operations and compromising financial stability Effective credit risk management is therefore essential to protect earnings, sustain liquidity, and support long-term resilience in the banking sector.
To limit the impact of credit risk on business operations, banks commonly deploy loan loss provisions Properly sized provisions help protect the safety and continuity of bank operations while supporting profitability Beatty and Liao (2009) note that loan loss provisioning policies play a significant role in stabilizing the financial system.
Empirical studies on credit risk provisions and loan loss provisioning have consistently identified key determinants such as bank size, non-performing loans, pre-tax income and provisioning, and macroeconomic indicators like GDP growth, as shown by Abu-Serdaneh Jamal Abdel-Rahman, Hussainey Khaled, and Abdul Rasit Zarinah (2018); Bryce Cormac, Dadoukis Aristeidis, Hall Maximilian, Nguyen Linh, Simper Richard (2015); Soedarmono Wahyoe, Pramono Sigid Eko, Tarazi Amine (2016) In the Vietnamese banking sector, research similarly investigates factors affecting credit risk provisions, including Pham Dinh Tuan and Nguyen Thi Thu Hien’s 2014 study of Vietnam’s commercial banking system and Nguyen Thi Minh Hieu’s 2015 analysis of provision determinants at Vietnamese joint-stock banks.
Loan loss provision policies play a key role in bank operations by buffering credit risk and maintaining financial stability Exploring the determinants of provisioning for bank risks is highly relevant, given their impact on profitability, capital adequacy, and risk management This research, titled "Factors affecting loan loss provision of joint-stock commercial banks in Vietnam," aims to identify and analyze the drivers of loan loss provisions in Vietnamese joint-stock commercial banks Through empirical analysis of bank data and macroeconomic indicators, the study seeks to provide actionable insights for policymakers and bank managers on how provisioning behavior responds to credit risk, regulatory requirements, and bank-specific factors.
Research objective
The general objective of this thesis is to study factors affecting loan loss provision of Vietnamese joint-stock commercial banks
This thesis investigates the factors affecting loan loss provision (LLP) in Vietnam's joint-stock commercial banks and evaluates how these determinants influence LLP levels across the banking sector By analyzing macroeconomic conditions, bank-specific risk profiles, credit quality, capital adequacy, and regulatory factors, the study measures the relative impact of each factor on loan loss provisioning decisions in Vietnamese banks The findings demonstrate the varying influence of these variables on LLP rates and offer evidence-based guidance for risk management practices within joint-stock banks Additionally, the article provides practical recommendations for bank managers to optimize loan loss provisions and for policymakers to fine-tune regulatory and supervisory frameworks, with the aim of improving LLP accuracy and preserving financial stability in Vietnam's banking system.
Research questions
To achieve the research objectives, the thesis focuses on answering the main research questions:
What factors affect to loan loss provision of joint-stock commercial banks in Vietnam?
To what extent do the factors affect loan loss provision of joint-stock commercial banks in Vietnam?
What solutions help improve the loan loss provision ratio of joint-stock commercial banks in Vietnam?
Research scope
About research space: The thesis is conducted to obtain analytical data from financial statements such as balance sheets, income statements, etc of 25 Vietnamese joint-stock commercial banks
Research time: Research data over a 10-year period from 2011 to 2020.
Research methods
The research method used in this thesis is quantitative research methods The thesis uses previous studies as the basis for determining factors affecting loan loss provision
Using Stata 14, this study analyzes panel data through multiple regression specifications—pooled OLS, random effects, and fixed effects—and applies feasible generalized least squares (FGLS) to refine the model It conducts diagnostic tests for multicollinearity, autocorrelation, heteroskedasticity (variable variance), and endogeneity to ensure robust estimates, then reports the estimation results and draws conclusions about the most appropriate specification Finally, the findings are compared with real-world conditions to assess practical relevance and to propose actionable solutions to the research problem.
Using analytical methods such as statistics, descriptive graphs, comparisons, evaluations, referencing, and both inductive and deductive reasoning, the paper assesses the current status of loan loss provisions in Vietnamese joint-stock commercial banks and argues, explains, and evaluates the issues presented in the thesis.
The study was conducted according to the following steps:
Picture 1.1: The process of researching
(Source: compiled by the author)
At the initial stage, the thesis identifies and clarifies the research problem and objectives, systematizes the relevant theories and prior studies, and uses these insights to build the research model and plan data collection The variables are clearly defined, measured, and the expected impact of the independent variables on the dependent variable is specified From the collected raw data, values for the independent and dependent variables are computed, after which regression analysis is carried out in Stata to estimate the model and test specifications to identify the most suitable model Finally, the regression results are analyzed and discussed to draw conclusions about the research findings.
Structure of the research
The thesis consists of 5 chapters:
Chapter 1 provides an overview of the research topic, detailing the rationale for topic selection, the research objectives and questions, the research object and scope, the selected research methods, and the study’s scientific and practical significance It articulates why the topic matters, presents precise objectives and research questions, defines the study’s boundaries, and describes the methodological approach used to explore the topic The chapter concludes with an outline of the expected contributions and a concise preview of the thesis structure.
This chapter presents the theoretical framework of loan loss provisions, detailing their concept, measurement methods, and the financial meaning of provisioning It analyzes the factors affecting loan loss provisions, drawing on global empirical studies as well as country-specific evidence from Vietnam The discussion surveys how macroeconomic conditions, bank risk exposure, asset quality, and regulatory environments influence provisioning decisions It also identifies gaps in previous research and articulates the new contributions this thesis offers to the literature, including theoretical refinements and empirical insights into credit risk management and financial reporting in Vietnam.
Chapter 3: Research methods and data
Chapter 3 will focus on presenting the research method of the topic, research hypotheses research data, research model, and research process
Chapter 4: Research results and discussion
This chapter presents the research findings by analyzing and processing the collected data with Stata software, including descriptive statistics, regression analysis, and model selection tests It clarifies how the independent variables influence the dependent variable, outlining the direction, magnitude, and statistical significance of their effects based on the empirical results.
This final chapter presents an assessment of the research results, identifies the influencing factors, and explains their impact on loan loss provisions, while outlining the study’s limitations and directions for future research It consolidates how each factor affects provisioning levels and discusses implications for measurement, modeling, and risk management within banks The chapter also offers actionable recommendations for bank administrators, showing how to use these factors to manage loan loss provisions more effectively and to bolster the bank’s operational efficiency By clarifying the relationships between factors and provisioning outcomes, the findings inform policy design, parameter selection, and ongoing monitoring Additionally, it highlights avenues for future study to close gaps and build on the robustness of provisioning analyses.
Chapter 1 clarifies the reasons for choosing the topic "Factors affecting loan loss provision of joint-stock commercial banks in Vietnam" The project was studied to analyze the factors affecting the loan loss provision of Vietnamese joint-stock commercial banks by panel data regression method Specifically, the thesis uses secondary data from 25 commercial banks between 2011 and 2020 for quantitative research using Stata 14 software.
LITERATURE REVIEW
Credit risk
Banking activities encompass the regular provision of services such as accepting deposits, granting credit, and offering payment services, with lending being the most important and profitable line of business When a bank extends credit, it evaluates the borrower's ability to meet debt obligations to ensure repayment of principal and interest Credit risk arises when a borrower defaults on part or all of the loan, potentially impairing the bank's cash flow and threatening its solvency.
Credit risk is defined as the degree of value fluctuations in debt instruments and derivatives due to changes in the underlying credit quality of borrowers and counterparties (Jose & Marc, 2000)
Credit risk is the possibility that a debtor or issuer of a financial instrument—whether an individual, a company, or a country—will fail to repay the principal and other cash flows as outlined in a credit agreement In banking, this risk can lead to delayed or non-payment, causing cash flow problems and potentially undermining a bank’s liquidity.
Credit risk is the possibility that a counterparty will fail to meet its obligations under the terms of an agreement Also called default risk, performance risk, or counterparty risk, it encompasses the potential impact of credit events on a firm’s transactions, reflecting the likelihood that contractual obligations will not be fulfilled and the ensuing financial consequences.
According to Circular 11/2021/TT-NHNN, credit risk in banking activities is the potential loss of debts held by credit institutions and foreign bank branches that arises when borrowers are unable to repay part or all of their obligations under a contract or agreement with the lending institution or its foreign branch.
Credit risk includes transaction risk and portfolio risk
Transaction risks stem from limitations in the transaction process, loan approval, customer assessment, and ongoing loan monitoring These risks are organized into three core components: selection risk, which concerns choosing inappropriate borrowers or deals; assurance risk, relating to gaps in verification and confidence in data; and operational risk, arising from weaknesses in processes and execution By understanding these elements, lenders can pinpoint vulnerabilities, strengthen controls, and enhance risk management throughout the lending lifecycle.
Selection risk is a risk related to the credit assessment and analysis process when banks choose effective loan options to make lending decisions
Assurance risk arises from security standards that govern lending, including the terms of loan contracts, the types of collateral assets, the collateral subject itself, the method of guarantee, and the relationship between the loan amount and the collateral value.
Operational risk is the risk associated with loan management and lending practices, including the use of risk rating systems and problem loan handling techniques
Portfolio risk arising from limitations in the management of a bank's loan portfolio is divided into two categories: intrinsic risk and concentration risk
Intrinsic risk stems from the unique internal factors and characteristics of each borrower, as well as from the distinctive traits of different economic sectors or fields It arises from the operating characteristics and capital-use patterns of borrowers, shaping the risk profile based on inherent differences rather than external conditions.
Concentration risk arises when a bank concentrates too much lending capital on a few borrowers, a limited set of enterprises in the same industry or economic sector, loans tied to a particular geographical area, or exposure to a single type of high‑risk loan This risk amplifies potential losses if those borrowers, sectors, or regions underperform, making diversification across customers, industries, geographies, and loan types essential to protect financial stability Effective risk management for concentration risk focuses on monitoring exposure concentrations, setting concentration limits, and diversifying portfolios to reduce vulnerability to sectoral or regional downturns.
The reason is from the bank that the bank staff does not strictly comply with the credit regime and lending conditions
Key issues include a lending policy and process that lack rigor, an absent or weak risk management framework, insufficient customer analysis, and no formal credit risk classification to assess loan conditions and repayment ability For both small business and individual loans, decisions rely mainly on experience rather than applying a credit scoring tool or other standardized underwriting metrics.
Because banks sometimes focus too much on profits, placing loans with higher returns than healthy loans
Credit risk management remains weak in forecasting, analyzing, and underwriting credit, and in detecting and handling problem loans—especially in industries that demand high professional knowledge—leading to mistakes in loan processing and lending decisions Even when a loan decision is sound, a lack of post-disbursement monitoring and controls can allow borrowers to misuse funds, with banks often unable to intervene in time Strengthening credit analysis through industry-specific risk assessment, robust post-loan surveillance, and proactive post-disbursement controls is essential to reduce credit risk, improve loan portfolio quality, and safeguard financial institutions.
Credit risk within banks stems from several intertwined factors, including the limited capacity and questionable ethical standards of some credit officers It is further aggravated by inadequate management practices and insufficient remuneration for bank staff, which collectively undermine risk controls and increase the likelihood of bad loans.
On the borrower’s side, internal factors drive repayment risk, including poor financial autonomy, weak management capacity, and an inefficient business administration system These issues can lead to improper use of loan funds, higher business losses, and a reduced ability to repay loans Poor customer management further compounds the problem by weakening operational performance and debt-servicing capacity In some cases, a lack of goodwill in repaying bank loans underpins the repayment failure, undermining both the borrower's reliability and the lender’s ability to recover funds.
Due to the customer's intentional fraud, and incompetence as a legal entity, to appropriate the bank's capital, these customers use a type of collateral to borrow from many places
Due to the lax legal environment, lack of synchronization, there are many loopholes leading to failure to control fraudulent phenomena in borrowing and using the capital of customers
Due to fluctuations in the economy such as economic recession, exchange rate fluctuations, increased inflation, it affects the production and business situation of customers as well as banks
Due to natural disasters, epidemics, fires, etc., both banks and customers could not respond in time
Failure to recover debts—principal, interest, and fees—erodes commercial banks' capital and reduces profitability, since banks still incur interest costs on mobilized funds even when borrowers default If profits do not cover these losses, banks may need to absorb the shortfall with their own capital, potentially reducing capital adequacy and constraining growth This dynamic can limit lending activity, hinder the bank's operational scale, and negatively affect overall financial performance.
High overdue debt undermines banks' reputations and erodes confidence among investors and customers, narrowing their ability to mobilize capital With constrained funding, banks edge toward insolvency, posing a threat to the stability of the entire banking system.
Customers who cannot repay the principal and interest owed to banks lose access to bank capital and, more broadly, to financial resources in the economy because their credibility is damaged This credibility loss reduces their credit access and makes it harder to secure future loans, thereby limiting their participation in economic activity.
Other borrowers' access to bank capital is also more limited as credit risks force commercial banks to tighten lending and even downsize their operations
Subjects depositing money in banks are at risk of not recovering their deposits and interest if the banks fall into bankruptcy
Loan loss provision
A loan loss provision is an expense charged to banks’ income statements, enabling banks and certain financial institutions to deduct it from current earnings while building reserves on their balance sheets to absorb potential loan losses (Azira, Herman, Noor Shahieda, Ariza, & Mohd, 2015).
Provisioning for loan losses is the process of identifying potential loan losses that may arise when borrowers fail to repay principal and interest When a bank determines that a loan cannot be recovered, it sets aside reserves to cover expected credit losses On the balance sheet, the allowance for loan losses is recorded against loans, reducing their net value, while on the income statement the provision is recognized as an expense that reduces the bank’s net income and equity These provisions reflect the cost of credit risk and help banks absorb future losses.
Under Circular 11/2021/TT-NHNN on asset classification and the use of provisions to manage risks in credit activities, provisions for risks are amounts set aside and charged to operating expenses to cover potential losses on the debts of credit institutions and branches of foreign banks These provisions consist of two types: specific provisions and general provisions A specific provision is the amount set aside to cover risks for each individual debt, while a general provision covers potential risks that have not yet been identified when making specific provisions.
Provisioning for credit losses is the process of identifying loan losses and assessing the risk of borrower defaults on principal and interest When banks determine that a loan cannot be fully recovered, they establish reserves to cover expected credit losses On the balance sheet, the allowance for credit losses is recorded as a contra-asset that reduces total assets and reflects asset impairment before actual losses occur In the income statement, the provision for credit losses is a non-cash expense that reduces net income and can impact a bank's equity.
According to Article 2 of Circular 11/2021/TT-NHNN on asset classification and use of provisions to handle risks in credit activities
Credit institutions, including commercial banks and non-banking credit institutions, except for those under special control, must comply with the provisions of the Law on Special Control of Credit Institutions, including branches of foreign banks.
Branches of foreign banks may apply the risk provisions used by foreign banks for debt classification, off-balance-sheet commitments, and provisioning, but such application must be approved by the State Bank under specified terms and conditions The risk provision standards employed by foreign banks are more advanced and superior to those specified in Article 10 of this Circular.
Foreign bank branches that have been approved by the State Bank to apply the risk provisions policy before the effective date of Circular 11/2021/TT-NHNN shall comply with the foreign bank's regulations in the country that approves such application under Clause 2 of this Article During inspection and supervision, if the State Bank assesses that a foreign bank's risk provisions policy does not fully reflect the credit risk level in Vietnam's actual banking operations, it may require the foreign bank branches to classify debts, off-balance sheet commitments, and to establish and use risk provisions in accordance with this Circular.
According to Circular 11/2021/TT-NHNN, the specific provision amount to be deducted for each customer is calculated according to the following formula:
R: Total specific provision to be deducted from each customer
Ri: is the specific provision amount to be deducted from the customer for the principal balance of the ith debt Ri is determined by the formula:
Ri = (Ai - Ci) x r Ai: The i principal outstanding balance
Ci represents the deductible value of security assets—including financial lease assets, negotiable instruments, and other valuable papers—used in the discounting and resale of government bonds (hereinafter referred to as government securities) associated with debt i, while r denotes the group-specific provisioning rate.
In case of Ci > Ai, then Ri is calculated as 0 (zero) Specific provisioning rates for each debt group are as follows: group 1 (0%); group 2 (5%); group 3 (20%); group 4 (50%) and group 5 (100%)
The collateral for deduction when calculating the specific provision amount (R) must meet the following conditions: a) Credit institutions, and branches of foreign bank have the right to handle security assets under the security contract and according to the provisions of law when the customer fails to perform his/her obligations as agreed; b) The expected time for handling the collateral is not more than 01 (one) year for the security property which is not real estate and not more than
02 (two) years for the collateral which is real estate, since credit institutions, and branches of foreign bank have the right to handle collaterals; c) Collaterals must comply with the provisions of the law on secured transactions and other relevant laws; d) Where the collateral does not satisfy the conditions specified at Points a, b, and c of this Clause, the deductible value of that security must be considered as zero (zero)
Credit institutions and branches of foreign banks shall determine, at their own discretion, the deduction rate for each type of security asset based on an assessment of recoverability when handling such collateral, but the rate must not exceed the maximum deduction rate for each specified type of collateral.
Value of collaterals for deduction when making provision for risks is determined as follows: Gold bars are valued at the purchase price at the head office of the enterprise or credit institution that owns the gold bar on the day of the transaction immediately preceding the provisioning date; for listed securities including listed stocks, fund certificates, derivative securities, and covered warrants, the deduction uses the closing price on the most recent trading day before provisioning, with the value to be determined as prescribed at Point e if there has been no trading within 30 days before provisioning or if the securities are delisted or suspended from trading; for shares registered for trading on the Upcom system, the reference price on the nearest trading day before provisioning announced by the Stock Exchange is used, with the same 30‑day no‑trading or delisted/suspended rules applying as prescribed at Point e; for government bonds listed on the Stock Exchange, the deduction uses the average price of transaction prices in the offering session in accordance with government regulations on issuance, custody, listing, and trading, and if there is no offering transaction price, the price for deduction is the average of secondary market transaction prices over the last 10 working days up to the provisioning date.
If there is no transaction within the nearest 10 working days up to the time of making provision for risk, the credit institution or foreign bank branch shall determine the value of the security asset at its par value For local government bonds, government-guaranteed bonds, and corporate bonds (including those of credit institutions) that are listed and registered for trading, the value is based on the average price of transactions in the secondary market within the last period.
Ten working days before the date of making provisions for risks announced by the Stock Exchange, the value of the applicable assets is determined If there is no transaction within 10 days up to the provision date, the credit institution or foreign bank branch shall determine the security asset’s value at par value Securities not listed on the Stock Exchange, promissory notes, bills, and certificates of deposit issued by enterprises (including credit institutions and foreign bank branches) are calculated at par value For financial leasing assets, the value is either the value of the financial leasing assets valued according to the provisions of point h of this Clause or the residual value over the lease period, calculated by the formula: value of financial leasing assets divided by the lease term under the contract multiplied by the remaining lease term under the contract.
Credit institutions and foreign bank branches determine the collateral deduction rate for each type of collateral based on price volatility, with higher volatility yielding a lower deduction rate The maximum deduction rates for collateral are as follows: a) customers’ balances of deposits and certificates of deposit in Vietnamese dong at the credit institution or foreign bank branch: 100%; b) government bonds and gold bars in accordance with the law on gold trading, and the balance of deposits and certificates of deposit of customers in foreign currencies at credit institutions or foreign bank branches: 95%; c) local government bonds, bonds guaranteed by the Government; negotiable instruments, promissory notes, bills, and bonds issued by credit institutions themselves; and the balance of deposits, certificates of deposit, promissory notes, and bills of exchange issued by other credit institutions or foreign bank branches:
- Having a remaining term of less than 1 year: 95%;
- Remaining term from 1 year to 5 years: 85%;
Factors affecting loan loss provision
Gross Domestic Product (GDP) is a key indicator of an economy's health It represents the monetary value of all final goods and services produced within a country's borders over a defined period, typically one year.
Gross domestic product (GDP) is a key indicator of the health of a national economy When GDP grows, it signals stronger business activity, higher production and consumption of goods and services, and improved efficiency across firms This environment supports rising profits, greater investment, and an enhanced capacity to repay debts Firms with on-time debt payments and low non-performing loans typically carry lower loan-loss provisions Conversely, a recession—defined by a falling GDP—reflects weaker demand and reduced production and sales, making it harder to repay existing bank loans and leading to higher non-performing loans and larger provisions for credit risk.
Numerous studies have explored how GDP growth relates to credit provision and loan loss provisions, with mixed results Hasni Abdullah, Ismail Ahmad, and Imbarine Bujang (2014) report a negative relationship between credit provision and GDP growth, and Soedarmono Wahyoe, Pramono Sigid Eko, and Tarazi Amine (2016) find GDP growth also negatively affects loan loss provisions On the other hand, Olszak Małgorzata, Roszkowska Sylwia, and Kowalska Iwona (2018) demonstrate a positive correlation between GDP and loan loss provisions Taken together, these studies indicate that GDP's impact on lending risk indicators can vary by context and methodology, yielding both negative and positive associations.
Most studies indicate that bank size, as reflected by total assets, influences loan loss provisioning When banks grow their total assets, their lending activity tends to increase, and faster credit growth is associated with more non-performing loans, which in turn elevates loan loss provisions Consequently, an expansion in total assets typically leads to greater lending and a corresponding rise in provisions for loan losses.
Many studies include the bank size factor in their models, with evidence showing that asset size, typically measured by total assets, affects loan loss provisioning (LLP) According to Mahmoud O Ashour (2011), expanding assets drives credit activity, increases the potential for non-performing loans (NPLs), and consequently raises LLP This suggests a positive relationship between bank asset size and the cost of provisioning for credit risk Given the wide variation in bank sizes, researchers often log-transform assets using the natural logarithm (ln) to stabilize regression results and limit data fluctuations.
Net interest margin (NIM) is the ratio that captures the growth rate of a bank's interest income relative to the growth rate of its expenses It represents the spread between interest income and interest expenses that a bank can achieve through tight management of profitable assets and access to low-cost funding A higher NIM signals that the bank has maximized interest income while minimizing interest expenses, reflecting efficient asset selection and funding strategies that support stronger profitability.
An uptick in the net interest margin (NIM) indicates banks are prioritizing profit through interest income and expanding their lending activities, which increases credit risk and, in turn, loan loss provisions This relationship is supported by multiple studies, including Banks’ Net Interest Margin in the 2000s: A Macro-Accounting international perspective (2011) by Germán López-Espinosa, Antonio Moreno, and Fernando Pérez de Gracia; The effect of commercial banks' internal control weaknesses on loan loss reserves and provisions (2016) by Cho Myojung and Chung Kwang-Hyun; and additional evidence from Nguyen Van Thuan and Duong Hong Ngoc (2015) and Jesús Gustavo Garza-García (2010).
Net return on assets (ROA) is a profitability metric that measures how much profit a company earns for each dollar of assets It shows the profit earned per dollar of assets and reflects how efficiently a business uses its resources to generate earnings A higher ROA indicates more efficient asset utilization and stronger overall profitability, making ROA a useful indicator for investors evaluating a company's ability to manage assets and sustain earnings.
An increasing ROA indicates that a bank is utilizing its assets efficiently to generate profits, reflecting strong lending practices and better asset quality As ROA improves, the bank’s lending decisions tend to be more effective, resulting in lower credit risk and smaller loan loss provisions This positive link between ROA and risk management is documented in several studies, including Cho Myojung and Chung Kwang-Hyun (2016); Hasni Abdullah, Ismail Ahmad, and Imbarine Bujang (2014); and Nguyen Van Thuan, Duong Hong Ngoc (2015).
Loan growth represents the year-over-year increase in a bank’s loan portfolio When loan growth rises, it signals expanded credit activity, which can raise credit risk and, in turn, inflate the loan loss provision This relationship is supported by research from Hasni Abdullah, Ismail Ahmad, and Imbarine Bujang (2014), as well as Nguyen Thi Hieu (2017).
Non-performing loans (NPLs) are debts classified in groups 3–5, including sub-standard and doubtful debts as well as debts with the potential for capital loss NPLs are a major driver of higher loan loss provisioning costs because, under the provisioning framework, loans in groups 3–5 carry specific provisioning rates of 20%, 50%, and 100%, respectively As a result, there is a clear correlation between the NPL ratio and loan loss provision expenses This relationship has been confirmed by numerous studies, including Soedarmono Wahyoe, Pramono Sigid Eko, and Tarazi Amine (2016) and Nguyen Thi Diem Kieu (2013).
The ratio of outstanding loans to total assets is often used as a warning indicator of default risk and greatly influences loan loss provisions A high ratio indicates that shareholders are pursuing a debt-intensive policy and increases bank risk, which is not favorable for safety or for attracting capital There are two opposing findings in the literature on the effect of the credit risk coefficient on loan loss provisions: Soedarmono Wahyoe, Pramono Sigid Eko, Tarazi Amine (2016) and Phan Van Tan (2015) report a positive relationship between the credit risk coefficient and loan loss provisions, while Ngo Hoang Trong (2020) and Nguyen Duy Anh Thu (2018) report a negative relationship.
Debt-to-equity ratio, calculated as total debt divided by common equity, is a key financial indicator of a bank’s operating capacity and management quality A higher debt-to-equity ratio signals greater perceived risk, which can prompt bank managers to reduce loan loss provisions and bolster retained earnings to increase common equity, thereby lowering the ratio and mitigating risk perceptions This relationship is supported by the study by Hasni Abdullah, Ismail Ahmad, and Imbarine Bujang (2014).
Previous studies relating to loan loss provision
Across 35 bank-year observations, Ahmed, Takeda, and Thomas (1999) study loan loss provisions as a lens on capital management, earnings management, and signaling in both Islamic and conventional banks They find that managers cut loan and investment provisions when the gap between the required legal reserve and the current reserve is large, freeing earnings to fund reserves and shrink the gap They also find that as the loans-to-deposits ratio rises, managers reduce these provisions to lower perceived risk and attract external funds The results do not support the income-smoothing view, i.e., provisions are not lowered whenever pre-tax-and-provision ROA falls below the prior ROA; nor do they support the leverage-risk view that higher debt-to-equity leads banks to lower provisions to mitigate risk.
Wall, Larry D., and Iftekhar Hasan (2004) in “Determinants of the Loan Loss Allowance: Some Cross-Country Comparisons” use a quantitative approach to test whether bank managers adjust loan loss provisions to influence reported net income and equity, acknowledging that hedging factors, while present, can still be estimated The study examines factors affecting credit risk provisions among banks in the United States, Canada, and Japan, finding that the loan-to-total assets ratio and the fee reduction ratio are statistically significant for US banks but not for non-US banks; profit before tax and provision is significant overall except in Canada, and non-performing loans affect provisioning with a coefficient that varies across samples, likely reflecting country-level differences in the expected loss on non-performing loans.
Entitled 'Loss of reserves of weakly provisioned banks: evidence from major Tunisian banks,' the 2010 paper by Neila Boulila Taktak, Abdelkader Boudriga and Dhouha Nefla Ajmi analyzes 66 Islamic banks across 19 countries from 2001 to 2006 Using a pooled panel-data approach that combines time-series and cross-sectional data, the authors estimate a basic model with three methods—OLS with robust standard errors, feasible generalized least squares (FGLS), and PCSE (panel-corrected standard errors)—to examine the determinants of credit provisioning in Tunisian banks Their results for the full sample indicate a negative relationship between loan loss provisions and real GDP growth, a positive link between provisioning and both the non-performing loan ratio and the cost of provisioning, and a significant effect of bank type on loan loss provisioning.
Frank Packer and Haibin Zhu (2012) examine loan loss provisioning practices of Asian banks, drawing on data from 2000–2009 for 240 banks across 12 Asian economies to study credit risk provisions The study's main objective is to identify the determinants of loan loss provisions, with explanatory variables including factors that affect credit quality—the non-performing loan (NPL) ratio, the loan-to-asset ratio, and bank loan growth—together with the capital adequacy ratio, GDP growth, and pre-tax earnings and provisions as a percentage of total assets In addition, country and year fixed effects are included to control for cross-country and time variation in provisioning behavior.
Empirical analysis shows that the coefficients for both the non-performing loan (NPL) ratio and the loan asset ratio have the expected positive signs, with only the NPL ratio statistically significant at the 99% level; the factor of safety is inherently negative and statistically significant This finding suggests that Asian banks make higher loan loss provisions when credit is riskier, consistent with standard accounting principles and results from prior studies The loan growth coefficient is negative and statistically significant, indicating that provisioning tends to be lower even when rapid loan growth signals increased credit risk Regarding macro factors, GDP is not significant in the panel OLS regression, but the GMM regression results show a positive and statistically significant effect.
In 2015, Azira Abdul Adzis, Herman Shah Anuar, and Noor Shahieda Mohd Hishamuddin published a study titled “Malaysian Commercial Banks: Do Income Smoothing, Capital Management, Signaling, and Pro-Cyclicality Exist Through Loan Loss Provisions?” The research aims to determine whether Malaysian commercial banks exhibit income smoothing, capital management, signaling, and pro-cyclical behavior through their loan loss provisions, analyzing evidence from the period 2002–2012.
In addition, this study also tests whether the global financial crisis in the years 2007-
The 2009 period had a significant impact on loan loss provisions among Malaysian commercial banks, with findings indicating that these provisions are used to smooth reported income The study found no concrete evidence that capital management, signaling, or pro-cyclicality drive loan loss provisions in Malaysian banks After controlling for the 2007–2009 global financial crisis, the results show that loan loss provisions were indeed influenced by the crisis itself Malaysia was not insulated from the 2007–2009 global recession and experienced a decline in real GDP growth in the first quarter of 2009 This context helps explain why Malaysian banks might have set higher loan loss provisions during the crisis—to absorb expected loan losses and dampen volatility in net income.
Curcio Domenico and Hasan Iftekhar (2015) re‑examine earnings and capital management and the signaling explanations for banks’ use of loan-loss provisions (LLPs) in a study of 218 banks from 11 euro‑area countries over 1996–2006, using data from Bureau van Dijk’s Thompson Bank Scope database, and they also develop a comparison between these euro‑area banks and another group of banks.
This study analyzes 273 credit institutions from non-euro-area countries to compare the provisioning policies of euro-area banks and non-euro-area banks, and to assess how cross-country determinants such as bank regulation and creditor protection influence income smoothing It examines how banks account for impaired loans, a practice that affects reported earnings and capital and has been widely studied, though empirical results remain inconsistent and largely focused on United States banks The authors investigate whether commercial banks located in euro-adopting countries exhibit different provisioning behavior compared with banks in non-euro-area countries, with the aim of deepening international understanding of provisioning practices and making reporting standards more reliable from an international standpoint, thereby supporting the supervisory objective of leveling the playing field from regulators’ perspective.
The study finds that loan loss provisions (LLPs) reflect changes in the expected quality of banks’ loan portfolios, as measured by non-performing loans (NPLs); earnings management is a significant factor affecting provisioning decisions for euro area banks but not for non-euro area banks; the desire to signal private information to outsiders helps explain provisioning policies for non-euro area banks but not for euro area intermediaries; and creditor protection systems have a different impact on income smoothing practices across euro area and non-euro area banks, reducing incentives to smooth earnings in the euro area banking systems.
According to the study titled “Determinants of loan loss provisions of commercial banks in Malaysia” by Mohd Isa Mohd Yaziz, Voon Choong Yap, Yong Gun Fie David, Abdul Rashid Md Zabid Hj, Zain Mustaffa Mohamed, and Johari RazanaJuhaida (2018), loan loss provisioning practices in Malaysian banks are shaped by determinants such as non-performing loans, interest income, loans and advances, net profit, and GDP, with the moderating effect of credit risk management and the intervening effects of relevance and faithful representation This framework helps banks capture expected losses and continuously reassess loss expectations as borrower conditions change.
Results indicate that Credit Risk Management positively moderates the relationship between the independent variables and loan loss provisions, strengthening their impact on the dependent variable This moderating effect is more pronounced in banks with a higher frequency of credit risk management meetings Moreover, higher levels of relevance and faithful representation of non-performing loans, interest income, net profit, loans and advances, and Gross Domestic Product (GDP) are associated with lower loan loss provisions, whereas lower levels of these indicators are associated with higher loan loss provisions.
In Vietnam, the study titled "Factors affecting provision for credit risk" by Nguyen Thi Thu Hien and Pham Dinh Tuan (2014) uses regression analysis to model the determinants of loan loss provisioning, conducting a panel data regression with 212 observations from 27 Vietnamese commercial banks during 2008–2012 to identify factors affecting LLP The results show that LLP is positively correlated with bank size (SIZE) and non-performing loans (NPL), and negatively correlated with the financial risk coefficient The study also notes that the variables included in the model cannot fully explain LLP, as many factors are not included and the analysis relies only on financial variables, without macro-control variables or performance-management-related variables.
A 2015 study titled "Factors affecting provision for credit risks at Vietnamese joint-stock commercial banks" by Nguyen Thi Minh Hieu analyzes the factors, trends, and levels of impact on loan loss provisions for Vietnamese commercial banks during 2006–2014 The findings identify five influential factors: interest rate and credit growth have negative effects, while non-performing loans, bank size, and income before tax and provision have positive effects on loan loss provisions These insights help bank managers formulate policies aligned with Basel principles and State Bank regulations to support the stability and development of Vietnamese banks.
Nguyen Thi Hieu's 2017 study analyzes the provisions for loan losses among Vietnamese joint-stock commercial banks over 2007–2015 Building on prior research by both international and domestic authors, it identifies key determinants—provisions for loan losses in the previous year, non-performing loans, pre-tax income and provisions, credit growth, the credit risk coefficient, bank size, the equity ratio, and GDP growth—and employs quantitative tools to test and measure their impact on banks’ credit risk forecasts The results show that the prior year's risk provision, the ratio of non-performing loans to total assets, bank size, and credit growth are statistically significant factors in the model.
RESEARCH METHODS AND DATA
Research method
Using Stata 14, the study computes and synthesizes annual financial indicators and applies regression methods to analyze how bank size, net interest margin, return on assets, loan growth, non-performing loans, credit risk coefficient, debt-to-equity, and GDP growth influence loan loss provisions of commercial banks in Vietnam, with the goal of selecting the most meaningful variables for the model The analysis runs three regression specifications—Pooled OLS, Fixed Effects (FEM), and Random Effects (REM)—and employs descriptive statistics to characterize the data F-tests determine whether the Pooled OLS or Fixed Effects specification is more appropriate, while the Hausman test assesses whether FEM or REM is preferable in this study.
Research model
Drawing on theory and prior empirical studies on the factors affecting loan loss provisions in commercial banks, this study adopts and integrates the research models of Hasni Abdullah et al (2014) and Nguyen Thi Hieu to examine the determinants of provisioning behavior in the banking sector.
(2017) as the main model for this research
Hasni Abdullah and colleagues (2014) examine macroeconomic and other determinants of loan loss provisions in Malaysian commercial banks over 2004–2012, a market with notable parallels to Vietnam Nguyen Thi Hieu (2017) analyzes the determinants of loan loss provisions for 18 Vietnamese joint-stock banks during 2007–2015 Both studies rely on established variables that prior research has shown to influence loan loss provisions, testing models that integrate macroeconomic indicators with bank-specific factors Collectively, the research highlights how broader economic conditions and internal bank characteristics shape provisioning behavior in Southeast Asian banking systems.
In addition, the author also synthesizes independent variables taken from other studies such as the net-interest margin variable taken from the study of Germán et al
(2011) to further complete the research model
LLP it = β 0 + β 1 SIZE it + β 2 NIM it + β 3 ROA it + β 4 LGit + β 5 NPLit + β 6 CEit + β 7 DEit + β 8 GDPit +ε i
LLP it : Provision ratio for loan losses of bank i, year t
SIZE it : Bank size of bank i, year t
NIM it : Net interest margin of bank i, year t
ROA it : Return on asset of bank i, year t
LG it : Loan growth ratio of bank i, year t
NPL it : Non-performing loan ratio of bank i, year t
CE it : Credit risk coefficient of bank i, year t
DE it : Debt-to-equity ratio of bank i, year t
GDP t : Gross domestic product in year t.
Research variables
The research of Abu-Serdaneh, J., Jamal, A.-R., Hussainey, K., & Abdul Rasit,
Income smoothing is primarily used to dampen earnings volatility over time In good years, bank managers may deliberately reduce current profits by applying discretionary loan loss provisions (LLP), building up funds that can lift reported earnings when credit conditions worsen A substantial body of research argues that banks use LLP to smooth income, including work by Anandarajan, Hasan, and McCarthy (2007); Balboa Marina, López-Espinosa Germán, and Rubia Antonio (2013); and Curcio Domenico, Hasan Iftekhar (2015) At the same time, Ahmed, Takeda, and Thomas (1999) show that movements in LLP reflect meaningful changes in the expected quality of loan portfolios, making LLP a management tool to mitigate future credit risk; consequently, the optimal LLP level should reflect credit-risk factors that managers cannot arbitrarily decide The LLP ratio, typically measured as the loan loss provision divided by total assets, is defined in empirical studies using formulas advanced by Packer and Zhu (2012); Ahmed, Raza ul Mustafa, Riaz Hussain Ansar, Muhammad Umair Younis (2012); Hasni, Abdullah, Ismail, Imbarine Bujang (2014); and Azira Abdul Adzis, Herman Shah Anuar, Noor Shahieda Mohd Hishamuddin (2015).
LLP = Loan loss provision / total assets
It noted that the provision for loan losses and the total assets are taken in the financial statements
Independent variables are potential factors that can affect the provision ratio of commercial banks The study divided independent variables into two categories: bank- specific variables and macroeconomic variables
Bank size (SIZE) is measured by the natural logarithm of total assets According to previous studies, there are some impacts of bank size on the loan loss provision Abu-Serdaneh Jamal Abdel-Rahman, Hussainey Khaled, Abdul Rasit Zarinah (2018) suggest that Bank size is embraced in much-related literature as an important control or bank-specific factor that may affect LLP This relationship is justified for different reasons, for example, large banks with a high volume of business transactions require high LLP, and another reason is related to political effects and the intervention of regulatory authorities in high fluctuation performances, the study also found that the control-variables size displays insignificant relationships with the provision The research by Cho Myojung, Chung Kwang-Hyun (2016) study of the effect of commercial banks' internal control weaknesses on loan loss reserves and provisions found that the size of the bank had a positive effect on the provision for loan losses The study of Nguyen Van Thuan, Duong Hong Ngoc (2015) found that bank size has a positive effect on the loan loss provision ratio In theory, larger banks can often be more effective at managing credit risk by diversifying their loan portfolios At the same time, large-scale banks are also focused on investment by large financial institutions, therefore, large banks often have stable long-term investment strategies This is also a factor that makes large banks willing to accept high risks because the expectation of government protection when these banks fail causes the credit risk ratio to increase to ensure safety for payment to customers, the credit risk provision must be increased The study expects there will be a positive correlation between size variables and the loan loss provision ratio of Vietnamese banks
Hypothesis 1: The larger the size of the bank, the higher the loan loss provision of Vietnam's commercial bank
Net interest margin (NIM) is measured by dividing net interest income by total earning assets According to previous studies, there is a contradiction in the results about the impact of net interest margin on loan loss provision Research by Hess Kurt, Grimes Arthur, Holmes Mark J (2008) found that net interest margin has a negative relationship with the level of credit risk while other studies such as the study by Cho Myojung, Chung Kwang-Hyun (2016) also found that credit provision has a positive impact on net interest margin Germán López-Espinosa, Antonio Moreno, Fernando Pérez de Gracia (2011) found that credit provision has a positive relationship with net interest margin, the authors said that loan loss provisions are a measure of credit risk, and it is thus sensible to find the tradeoff between risk and asset returns and loan loss provisions are robust positive predictors of NIM, thus implying that banks increase their NIM internationally in response to higher asset risks Research by Garza-García, J
Research from 2010 indicates that banks with higher lending volumes face greater risk, requiring larger loan loss provisions This obligation to provision for expected losses can squeeze profits, as banks must offset these provisions in their earnings Accordingly, this study expects a positive relationship between net interest margin (NIM) and the loan loss provisions of Vietnamese banks.
Hypothesis 2: The higher net interest margin, the higher the loan loss provision of Vietnamese commercial banks
Return on asset (ROA) is determined by dividing profit after tax by total assets
ROA measures the profit earned per dollar of assets and serves as an indicator of asset-management effectiveness Studies on the relationship between ROA and loan loss provision (LLP) among Vietnamese banks yield mixed results: Misman and Ahmad (2011) report a negative relationship between LLP and ROA, indicating that higher credit provisions reduce bank income; in contrast, Karimiyan, Nasserinia, and Shafiee (2013) find a positive relationship between LLP and future income and profit; Nguyen Van Thuan and Duong Hong Ngoc (2015) show that bank income on total assets has a negative effect on the LLP ratio, suggesting that during difficult economic conditions the risk of unrecoverable loans rises, prompting higher provisions; therefore, the expected impact of ROA on LLP for Vietnamese banks in this study is negative.
Hypothesis 3: The higher the return on asset, the lower the loan loss provision of Vietnamese commercial banks
Loan growth (LG) is the percentage rise in banks’ loans in the current period relative to the same period last year, typically measured as the change in gross loan balances divided by this year’s total assets A growing body of evidence shows that loan growth is positively related to loan loss provisions, with Hasni Abdullah, Ismail Ahmad, and Imbarine Bujang (2014) reporting a positive and significant link between the loan growth ratio and loan loss provisions In the process of economic development, banks compete aggressively for lending market share, which drives high credit growth and can raise non-performing loans, prompting banks to set aside more for credit risk Daniel Foos, Lars Norden, and Martin Weber (2010) argue that loan loss provisions cover the unsecured fraction of a loan unlikely to be repaid, after accounting for collateral pledged by the borrower; the unsecured portion is typically determined by applying a haircut to collateral values to reflect fluctuations and liquidity risks If banks lower collateral requirements to spur lending, the unsecured fraction increases, and loan loss provisions and write-offs are likely to rise as well Across banks, these dynamics are associated with a significant positive relationship between past credit growth and the loan loss provision ratio, supporting the hypothesis that higher loan growth tends to increase loan loss provisions.
Hypothesis 4: The higher the ratio of loan growth of banks, the higher the loan loss provision of Vietnamese commercial banks
Non-performing loan (NPL) is measured by the ratio of total non-performing loan groups 3, 4, and 5 to total assets A substantial body of prior studies demonstrates that non-performing loans have a positive effect on loan loss provision (LLP) According to the IMF (2006), the ratio of non-performing loans to total outstanding loans is a core indicator for assessing the financial health of commercial banks; if this ratio is high, it can push the bank toward bankruptcy Faridah Najuna Misman and Wahida Ahmad (2011) showed a positive relationship between LLP and the NPL ratio, meaning that as the non-performing loan ratio increases, the risk provision ratio also increases to compensate for potential risks The studies by Soedarmono Wahyoe, Pramono Sigid Eko, Tarazi Amine (2016) and Hasni Abdullah, Ismail Ahmad, Imbarine Bujang (2014) also found that non-performing loans are positively and significantly correlated with loan loss provision Based on these findings, the impact of non-performing loan on the loan loss provision of Vietnamese banks in this study is expected to be positive.
Hypothesis 5: The higher the non-performing loan ratio, the more loan loss provision ratio Vietnamese commercial banks are
Credit risk coefficient (CE) is measured as loan outstanding balance divided by total assets, and the literature shows mixed results on its relationship with loan loss provisions (LLP) Balboa Marina, López-Espinosa Germán, and Rubia Antonio (2013) argue that loans-to-assets is a bank-specific control proxy that captures credit quality conditions and the non-discretionary component of LLP, with provisions acting as a buffer against credit losses and expected to rise when credit quality deteriorates or derivative exposure is high Soedarmono Wahyoe, Pramono Sigid Eko, and Tarazi Amine (2016) similarly find that LLP remains influenced by the loan-to-assets ratio Conversely, Nguyen Thi Thu Hien and Pham Dinh Tuan (2014) report a negative correlation between the credit risk coefficient and LLP Thus, the hypothesis is expected that the credit risk coefficient has a positive relationship to loan loss provision.
Hypothesis 6: Credit risk coefficient is positively related to the provision for loan losses of Vietnamese commercial banks
Debt to equity (DE) is calculated as total debt divided by total equity There is a contradiction in the observed relationship between the DE variable and loan loss provision (LLP) Hasni Abdullah, Ismail Ahmad, and Imbarine Bujang (2014) found that LLP is positively and significantly correlated with DE, implying that higher non-performing loans, greater total loan exposure, and faster loan growth contribute to a higher LLP In contrast, the excerpt references Mahmoud O Ashour's research, but its results are not detailed here.
The debt-to-equity (D/E) variable is used to test the D/E hypothesis and is predicted to have a negative relationship with loan loss provisions (LLP) A higher D/E ratio signals greater perceived bank risk, which would motivate managers to decrease LLP to boost common equity, thereby lowering the ratio and the perceived risk However, the results do not support the notion that banks manipulate LLP to affect their D/E ratio; there is no evidence that banks with high D/E ratios cut LLP to increase net income and retained earnings, pushing D/E down and reducing perceived risk The hypothesis that a higher D/E ratio corresponds to higher bank risk and that banks set lower loan and investment provisions to decrease risk is also rejected Consequently, the study expects the debt-to-equity ratio to have a negative impact on the loan loss provisions of Vietnamese banks.
Hypothesis 7: The higher debt to equity, the lower the loan loss provision of Vietnamese commercial banks
GDP is the year-over-year change in economic output, and the annual GDP rate is used to examine how macro conditions relate to a bank’s loan loss provisions A weak economy can deteriorate loan portfolio quality, reduce bank profits, and prompt higher loan loss provisions, while favorable conditions can improve borrower solvency and increase credit demand from households and firms The relationship between GDP and loan loss provision is mixed: Bryce Cormac, Dadoukis Aristeidis, Hall Maximilian, Nguyen Linh, and Simper Richard (2015) report an insignificant link between GDP growth and provisioning, indicating no countercyclical or procyclical provisioning behavior; in contrast, Soedarmono Wahyoe, Pramono Sigid Eko, Tarazi Amine (2016) and Olszak Małgorzata, Roszkowska Sylwia, Kowalska Iwona (2018) find that provisioning can deepen recessions, with loan loss provisions rising in response to falling growth Based on these prior results, this study expects the impact of GDP on the loan loss provision of Vietnamese banks to be negative.
Hypothesis 8: The higher GDP, the lower the loan loss provision of Vietnamese commercial banks
Picture 3.1: Vietnam’s GDP growth from 2011 to 2020
(Source: compiled by the author from the World Bank)
Table 3.1: Signs of expectation of variables in the model
Variable Description Measure Signs of expectation
LLP Loan loss provision Loan loss provision/ Total assets
SIZE Bank size Logarithm of total assets +
NIM Net interest margin Net interest income/ Total earning assets + ROA Return on asset Profit after tax / Total assets -
LG Loan growth (Gross loan balancet -Gross loan balance t-1) /
NPL Non-performing loan Non-performing loan/ Total assets +
CE Credit risk coefficient Outstanding loan balance / Total assets +
DE Debt-to-equity Total debt / Total equity -
(Source: Summary by the author)
Research data
Data for this study was collected from audited financial statements of 25 Vietnamese joint-stock commercial banks in the period from 2011 to 2020
Table 3.2: List of joint-stock commercial banks in the study
Num Stock code Full name
1 ABB An Binh Commercial Joint Stock Bank
2 ACB Asia Commercial Joint Stock Bank
3 BID Joint stock Commercial Bank for Investment and Development of
4 BVB Viet capital Commercial Joint Stock Bank
5 CTG Vietnam Joint Stock Commercial Bank for Industry and Trade
6 EIB Vietnam Export Import Commercial Joint Stock Bank
7 HDB Ho Chi Minh City Development Joint Stock Commercial Bank
8 KLB Kien Long Commercial Joint - Stock Bank
9 LPB LienViet Post Joint Stock Commercial Bank
10 MBB Military Commercial Joint Stock Bank
11 MSB Vietnam Maritime Commercial Join Stock Bank
12 NAB Nam A comercial Join Stock Bank
13 NVB National Citizen Commercial Joint Stock Bank
14 OCB Orient Commercial Joint Stock Bank
15 PGB Petrolimex Group Commercial Joint Stock Bank
16 SGB Saigon Bank For Industry And Trade
17 SHB Saigon – Hanoi Commercial Joint Stock Bank
18 SSB Southeast Asia Commercial Joint Stock Bank
19 STB Saigon Thuong Tin Commercial Joint Stock Bank
20 TCB Vietnam Technological and Commercial Joint Stock Bank
21 TPB Tien Phong Commercial Joint Stock Bank
22 VAB Vietnam Asia Commercial Joint Stock Bank
23 VCB Joint Stock Commercial Bank for Foreign Trade of Vietnam
24 VIB Vietnam International Commercial Joint Stock Bank
25 VPB Vietnam Prosperity Joint Stock Commercial Bank
(Source: Summary by the author)
Most empirical studies in economics aim to explain how a dependent variable Y relates to one or more explanatory variables X1, X2, , Xk In short, they seek to determine the effect that each Xi has on Y, including both the direction (positive or negative) and the magnitude (the size) of that effect By estimating these relationships, researchers quantify how changes in the explanatory variables are associated with changes in Y, providing evidence that can inform policy decisions, business strategy, or theoretical development The core task is to specify a model that links Y to X1 through Xk and to interpret the estimated coefficients to understand the strength and direction of the impact.
Panel data is data that scales in both time and space The tabular data structure is composed of two components: a cross-section component and a time series component
Combining two types of data offers several key advantages for analysis, especially when you want to observe how research object groups change after events or over time and to analyze differences between study subject groups This integrated approach enables robust longitudinal and cross-sectional insights, capturing trajectories, post-event effects, and inter-group variations By merging data sources, researchers can improve measurement accuracy, increase statistical power, and generate evidence that informs targeted interventions and decision-making.
Panel data track the same units over time, so heterogeneity across these units is inherent Estimation techniques for panel data address this heterogeneity by incorporating unit-specific variables, enabling more accurate modeling of differences between units.
By combining time series of cross-observations, panel data gives researchers
“more useful information, more variability, less multicollinearity between variables, more degrees of freedom and more efficiency.”
By studying repeated observations of cross units, panel data are more suitable for studying the time-varying dynamics of these cross units
Panel data enhances the ability to detect and measure effects that pure time-series or cross-sectional data may miss, because it follows multiple units over time By observing changes within the same entities across periods, panel data enables researchers to uncover more complex behavioral patterns and dynamic relationships that would be invisible in either time-series or cross-sectional analyses alone This structure improves estimation accuracy, controls for unobserved heterogeneity, and supports rigorous analysis of both short-term and long-term effects, making panel data a powerful tool for empirical research.
Panel data, which brings together observations for thousands of units across multiple time periods, can minimize the bias that arises when researchers include individuals or firms tied to high-risk variables By leveraging both cross-sectional and time-series variation and tracking the same entities over time, panel data delivers more reliable estimates and reduces errors due to unobserved heterogeneity, enhancing the robustness of empirical analyses.
Panel data combines the advantages of both time series and cross-sectional data, enabling richer insights and more precise estimates Yet it brings estimation and inference challenges such as heteroskedasticity (variable variance), autocorrelation, and cross-sectional dependence across units observed at the same time points The two most prominent methods to address these problems are the fixed effects model (FEM) and the random effects model (REM), which provide different ways to account for unobserved heterogeneity and correlation structures in panel data.
A fixed effects model (FEM) accounts for unit-specific, time-invariant characteristics that may influence the explanatory variables It does this by analyzing the correlation between each unit’s residuals and the explanatory variables, which controls for these individual effects and isolates them from the explanatory variables, enabling an accurate estimation of the net impact of the explanatory variables on the dependent variable.
Consider an economic relationship, with the dependent variable, Y, and two observable explanatory variables, X1 and X2, and one or more unobserved variables
Panel data for Y, X1, and X2 consists of N subjects observed across T time periods, producing N×T observations The classical linear regression model without the cutoff coefficient is specified by the standard regression equation that links Y to the regressors X1 and X2 across individuals and time, typically written as Y_it = β0 + β1 X1_it + β2 X2_it + ε_it, with i = 1, ,N and t = 1, ,T, and where ε_it represents the error term capturing unobserved factors.
Y it = β 1 X1 it + β 2 X2 it + μ it with i= 1, 2, , N and t= 1, 2, ,T Where:
Y_it denotes the price value of Y for object i at time t, X1_it denotes the value of X1 for object i at time t, X2_it denotes the value of X2 for object i at time t, and μ_it is the error term for object i at time t The fixed-effects regression model, an extension of the classical linear regression model, accounts for unobserved, time-invariant differences across objects by including an object-specific intercept α_i The standard specification is Y_it = α_i + β1 X1_it + β2 X2_it + μ_it, where β1 and β2 capture the effects of X1 and X2 on Y within each object over time after controlling for α_i This approach yields consistent estimates of the relationships between Y and the predictors when omitted variables are constant over time but vary across objects.
Y_it = β1 X1_it + β2 X2_it + ν_i + ε_it is a panel data regression where the error is split into two components, μ_it = ν_i + ε_it The ν_i term represents unobservable, time-invariant factors that differ across subjects, capturing persistent individual effects The ε_it term represents unobservable factors that vary across subjects and over time, capturing idiosyncratic shocks that change with each period This decomposition clarifies how observed regressors X1_it and X2_it influence Y_it while accounting for both stable heterogeneity and time-varying disturbances.
Consider an economic relationship in which the dependent variable is Y and the observable explanatory variables are X1 and X2 We have panel data for Y, X1, and X2 across N cross-sectional units observed over T time periods, yielding a total of N×T observations The random-effects model is written in the standard panel-data form to analyze how X1 and X2 influence Y while accounting for unobserved heterogeneity across objects.
Y it = β 1 X1 it + β 2 X2 it + ν i +ε it with i= 1, 2, , N and t= 1, 2, , T
The classical error term is decomposed into two components: a subject-specific effect v_i, which captures unobservable factors that vary between subjects but remain constant over time, and an idiosyncratic error ε_it, which captures unobservable factors that vary between objects and over time We assume v_i is given by:
In this model, v_i is decomposed into two components: an uncertain component a0 and a random component ω_i For each subject, ω_i is drawn from an independent probability distribution with zero mean and constant variance, i.e., E[ω_i] = 0 and Var(ω_i) = σ^2.
N random variable ω i is called random effects The random-effects model can be rewritten:
Y it =α 0 X1it+ β 2 X2 it +μit Where μ it = ω i + ε it
An important assumption in the random-effects model is that the error component μit is not correlated with any of the explanatory variables in the model
Figures will be computed and presented as table data that encompass both dependent and independent variables Each table will list variable names and the corresponding observation counts, along with descriptive statistics such as mean values, mode (most frequent values), standard deviations, maximum and minimum values, and measures of distribution shape including skewness and kurtosis.
Check the correlation of the variables in the model
To explore how variables relate to one another and identify the predictors that most influence the model, the analysis begins with constructing a correlation matrix This matrix highlights strong associations and helps select influential variables for the model while discarding weak or redundant ones To increase confidence in these relationships, multicollinearity among the predictors is assessed using the Variance Inflation Factor (VIF) By detecting and mitigating high VIF values, the approach ensures more reliable coefficient estimates and more accurate interpretations of the correlations among study variables.