Factors affecting bank profitability, The case of commercial banks, in Vietnam 2022
INTRODUCTION
INTRODUCTION
The banking system plays a vital role in economic development, serving as a key channel for prosperity by ensuring efficient capital circulation, mobilizing financial resources for the state, creating jobs, and linking economic actors As domestic banks evolve and expand, assessing profitability and its drivers becomes essential to gauge overall performance and inform policy measures that strengthen the economy Commercial banks operate within an economic accounting framework with the profit motive, yet many institutions pursue broader goals such as productivity gains, product quality, reputation, and market expansion, meaning profit targets may be less prominent than these objectives Consequently, short-term performance metrics may diverge from long-term success, even as both strive to sustain enduring profitability and financial stability A comprehensive assessment should reconcile short- and long-term indicators to support strategic decisions and policy design.
However, not all commercial banks invest deeply; instead, most focus on the immediate goals of survival and profit, yet these goals are not easily achieved as banks face a number of barriers Consequently, to address the practical need to understand how financial factors influence the profitability of Vietnamese commercial banks, the author chooses this topic.
This study explores the factors affecting bank profitability in Vietnam, focusing on Vietnamese commercial banks The research results clarify how key determinants influence the financial performance and overall financial condition of Vietnam’s banking sector, offering evidence-based insights into profitability drivers Furthermore, the thesis presents governance implications for bank administrators, outlining practical recommendations to strengthen governance, risk management, and strategic decision-making in Vietnam’s banks Together, these findings support policymakers and bank leaders in enhancing profitability, financial stability, and long-term resilience in the Vietnamese banking industry.
RESEARCH OBJECTIVE
This study investigates how potential factors affect bank profitability, with a particular emphasis on liquidity creation as a key determinant Liquidity creation is a relatively new factor whose impact has been underexplored in previous studies To achieve this overall goal, the research sets forth several specific objectives that structure the analysis and clarify how liquidity creation and other factors influence bank performance.
To systematize the theoretical basis of the factors affecting the bank profitability of the commercial banks in Vietnam
To analyze and evaluate the impact of each factor on the bank profitability
To give recommendations to help bank managers regulate appropriate policies.
RESEARCH QUESTION
To successfully complete the research objectives, the study must answer the following questions respectively:
What is profitability of commercial bank? How is it measured? How to measure factors influencing bank profitability, including liquidity creation?
How to build and test models to estimate the effect of each determinant affecting the bank profitability of commercial bank?
What solutions are offered so that banks can improve their profitability to ensure sustainable development of the banks?
RESEARCH SUBJECT AND SCOPE
The subjects of this study are financial determinants affecting bank profitability
The study collects financial indicators’ data from financial statements of 14 Vietnamese commercial banks during 2008 - 2020.
RESEARCH METHODOLOGY
This thesis adopts a quantitative research design and employs Stata software to conduct hypothesis testing and empirical analysis Data are collected using systematic data collection methods to align with and support the study’s objectives.
Quantitative analysis constitutes the official research approach used to draw conclusions about the topic In this method, the author tests the variables with a range of econometric models, including a regression framework based on balanced panel data and associated impact estimates using Pooled OLS, Fixed Effects, and Random Effects To ensure the analysis is suitable for testing, the thesis adopts a multivariate regression model with balanced panel data, enabling precise estimation of the relationships under investigation.
This study gathers research data from audited consolidated financial statements and the banks’ annual reports published on each bank’s website, as well as from VietstockFinance and the World Bank The analytical dataset integrates bank-level characteristics with macroeconomic indicators of the Vietnamese economy, based on a sample of 14 banks over the period 2008–2020.
EXPECTED CONTRIBUTION
The research results of the thesis have scientific and practical significance:
Firstly, the research contributes to identifying the factors having impact on the profitability of Vietnamese commercial banking system in the period 1 20 5 08-
92 1 020, especially liquidity creation In addition, the study contributes to supplementing empirical evidence for future scientific research references
Secondly, this study re-tests prior research findings using data from 14 Vietnamese commercial banks to provide readers with a clearer overview of the financial indicators that influence profitability in Vietnam's banking system, and to offer actionable recommendations for improving the operational efficiency of Vietnamese banks.
THE STRUCTURE OF RESEARCH
To achieve the research objectives, the author divides the content into five specific chapters:
Chapter 1 provides an overview of the research topic and explains the rationale for selecting this topic, emphasizing its relevance and significance It clearly states the research objectives and poses the core research questions that guide the study The chapter defines the research subjects and scope, delimiting the boundaries of the investigation to clarify what is included and excluded It outlines the research methodology, describing the chosen approach, data sources, and methods used to collect and analyze information Finally, it discusses the practical contribution of the research, highlighting how the findings can inform practice, policy, and future research.
This chapter presents the theoretical basis as well as practical evidence through previous research From there, the author mentions the research’s gaps
Chapter 3: Data and methodology of research
This section presents research methods, procedures, and construction methods, scale construction, sampling method, information collection process, processing tools data, and data analysis techniques used in the research process
Chapter X presents and interprets the results of formal quantitative research, with data collected and processed in Stata It reports the sample’s descriptive statistics, estimates from regression analyses, and the outcomes of hypothesis tests grounded in the theoretical framework The findings describe the impact of the independent variables on the dependent variable, including the direction and magnitude of effects, their statistical significance, and overall model fit The discussion ties these results back to theory, clarifies practical implications, and notes robustness checks and limitations for future research.
This final section presents the key conclusions drawn from the research problem and emphasizes their practical relevance for the banking sector The study offers actionable solutions for bank administrators to enhance risk management, operational efficiency, and customer engagement, while also outlining a set of concrete steps for implementation In addition, it identifies several directions for future research, including the study of technology adoption, data analytics, regulatory effects, and cross-context validation to broaden the applicability of the findings Together, these conclusions provide a concise, SEO-friendly roadmap for practitioners and scholars seeking to improve banking practices and advance academic inquiry.
Chapter 1 displays an overview of issues related to the research topic, including objectives and tasks of research, research problems and questions, research objects and scope research, research methods and data, and the contributions of the study From the above content, the topic "Factors affecting bank profitability: The case of commercial banks in Vietnam" is both scientific and practical This serves as the basis for further research steps It is an in-depth research on the theoretical basis, research methods, and data collection to produce research results and provide solution implications.
LITERATURE REVIEW
THEORETICAL BASIS OF BANK PROFITABILITY
According to Helhel (2015), the banking sector is the most essential component of the financial system, and a well-developed banking sector promotes stability and long-term economic growth Banks aim to maximize revenue while maintaining liquidity and safety margins to reduce risk, underscoring the importance of monitoring indicators that affect bank profitability Consequently, identifying these profitability determinants is crucial for informing banking policies that achieve the desired performance.
Bank profitability is a key indicator of bank performance because it represents the spread between interest earned on deposits and interest paid on loans (Demirgüç-Kunt et al., 2001) It can also be defined as a bank’s net after-tax income or net earnings (Rose, 2002) Economists rely on several measures to quantify profitability, including financial ratios such as return on assets (ROA), return on equity (ROE), and net interest margins (NIM) ROA and ROE have been used to assess bank profitability in studies by Naceur (2003), Mamatzakis and Remoundos (2003), Peters et al (2004), Staikouras and Wood (2003), Pasiouras and Kosmidou (2007), Athanasoglou et al (2008), Heffernan and Fu (2008), Rahman.
According to 2015 studies, ROA and ROE are the most frequently used metrics in academic research to measure operational efficiency Building on these findings, Rasiah (2010) notes that when analyzing the performance of any particular bank, it is often useful to consider a broader set of indicators—beyond ROA and ROE—including profitability, liquidity, asset quality, and risk exposure to gain a more comprehensive view of overall performance.
Profitability can be quantified in several ways, especially through return on assets (ROA) and return on equity (ROE) In today’s economic climate, understanding a bank’s financial metrics is essential for the success and growth of a developing bank ROA and ROE provide concrete measures to evaluate a bank’s profitability, and researchers Hassan (1999) and Samad (2001) identify these metrics as key performance indicators for banks.
Return on assets (ROA) is a key measure of a bank's operational efficiency and its ability to convert assets into net income It is calculated as operating profit divided by total assets, reflecting management's capacity to generate returns from the bank's resources Although external activity can influence ROA, it primarily represents how effectively management uses the institution's assets to produce revenue In short, ROA indicates how efficiently the company's resources are deployed to generate earnings and assesses management's effectiveness in turning all resources into net revenue According to Wen (2010), a higher ROA signals greater efficiency in resource utilization.
According to Nguyễn Minh Kiều (2009), ROE, or return on equity, is a financial ratio used to evaluate a company’s profitability It measures the net income returned to shareholders, sourced from the company’s income statement over a defined period—such as a month, quarter, semi-annual, or full year The denominator is common equity, so ROE reveals how much profit the company generates for every unit of common equity invested.
Return on equity (ROE) measures how effectively a bank uses its shareholders' equity to generate profits If this ratio is positive, the bank is profitable; if it is negative, the bank incurs a loss (Nguyễn Thị Ngọc Trang and Nguyễn Thị Liên Hoa, 2007) In other words, ROE reflects the effective use of equity and the return that shareholders receive when investing in the bank Therefore, ROE is one of the indicators that investors monitor and is often used as a basis for assessing a bank's operational efficiency and profitability, helping potential investors make informed decisions A high ROE indicates the bank is using its shareholders' money efficiently, as well as achieving a healthy balance between equity and debt.
In this study’s limitation, the author only uses the most two common financial indicators (ROA and ROE) to reflect the bank profitability.
FACTORS AFFECTING BANK PROFITABILITY
One of the determinants that have a great effect on profitability of bank is liquidity creation According to Berger and Bouwman (2009), Allen and Carletti
A fundamental function of banking institutions in any economy is liquidity creation, also known as liquidity transformation, whereby banks convert short-term liquid liabilities (such as demand deposits) into funding for long-term illiquid assets (such as business loans) to provide liquidity for their clients (Diamond and Dybvig, 1983) Deep and Schaefer (2005) describe a related concept called the liquidity transformation gap—the difference between the liquidity of a bank's liabilities and the liquidity of its assets A positive gap indicates that the bank is investing liquid liabilities into illiquid assets, thereby creating liquidity; conversely, holding more liquid assets than liabilities implies a reduction in liquidity creation Historically, banks mobilized short-term liabilities to finance longer-term assets to supply liquidity to the economy The measurement of a bank’s liquidity-creation ability followed Deep and Schaefer’s approach, which was applied to the 200 largest US banks from 1997 to 2001 Building on this framework, Berger and Bouwman (2007, 2009) developed four methods for calculating liquidity creation, each exploiting different on- and off-balance-sheet data to provide a more complete picture of liquidity creation across the banking system.
Empirical research on liquidity creation and bank profitability is still limited, as the study of liquidity creation has only recently gained attention Consequently, findings on its impact are mixed and vary across studies For example, Berger and Bouwman (2009) show that investors assign greater value to banks that generate more liquidity Other studies, including Duan and Niu (2020) and Vũ Hữu Thành et al., contribute to the growing evidence base, highlighting that the relationship between liquidity creation and profitability may depend on context and methodology.
A 2016 study confirms that liquidity creation positively influences bank profitability Building on this view, Berger and Bouwman (2009) argue that higher levels of liquidity generation lead to larger net surpluses that are distributed among stakeholders, including depositors, banks, and borrowers.
As a result, the influence of liquidity creation on bank profitability is positive
However, rare studies by Tran et al (2016) and Sahyouni and Wang (2019) find a negative relationship between bank profitability and liquidity creation Tran et al (2016), using data from all U.S banks between 1996 and 2013, show that the benefits of creating more liquidity and increasing capital to enhance a bank's financial stability are associated with lower profits Sahyouni and Wang (2019), analyzing Syrian banks from 2004 to 2016, demonstrate that liquidity creation negatively affects bank profitability as measured by ROA and ROE Fungáčová, Turk, and Weill (2015) argue that liquidity creation raises the risk of bank failure Chu Thị Thanh Trang et al (2021) also find that liquidity creation has witnessed the opposite trend against bank profitability In this thesis, the author estimates bank profitability affected by liquidity creation following the data collection method of Tran et al (2016), and hence expects an inverse correlation between liquidity creation and bank returns.
Based on the discussion above, hypothesis (H1) is proposed that liquidity creation has a negative impact on bank profitability
Bank size, measured by the natural logarithm of total assets, is a key determinant of profitability Generally, bigger banks can extend more loans and access markets that smaller banks cannot Empirical findings on the size–profitability link are mixed San and Heng (2013) and Alper and Anbar (2011) argue that bank size positively affects performance, as large banks benefit from economies of scale and can achieve higher profits In contrast, Lê Đồng Duy Trung (2020) finds only a small positive influence of size on ROA and no statistically significant impact on ROE, based on data from 30 Vietnamese banks in 2009 onward.
By 2017, bank growth in size brings advantages from economies of scale and scope through expanding networks, but it also introduces greater organizational complexity and administrative inertia These internal frictions can lead to inefficiencies in management, meaning that despite larger scale, profitability is not guaranteed to improve significantly.
Almazari (2014) finds that the size of total assets has a negative influence on profitability; as asset size grows, inefficiencies increase, compressing marginal profits and reducing average returns for developing banks.
Based on the discussion above, hypothesis (H2) is proposed that bank size has a positive impact on bank profitability
Capital adequacy is assessed by the proportion of total equity to total assets, indicating a bank’s ability to absorb losses Raising capital enhances loss tolerance for risks, especially credit risk, enabling greater credit expansion and higher returns (Berger, 1995) From a cost perspective, strengthening capital improves a bank’s credit ratings and lowers capital costs (Molyneux, 1993).
The capital-to-total-assets ratio serves as a hedge against financial distress and acts as a proxy for the funds held and owned by a bank During financial crises, it becomes clear how bank profitability is affected by poor-quality capital and unanticipated losses Beckmann (2007) argues that higher capital ratios can depress returns because banks with more capital tend to be risk-averse, forgoing opportunities and delivering lower expected returns in exchange for reduced risk In contrast, Lardic and Terraza (2019) find an inverse relationship between profitability and the capital adequacy ratio.
Empirical evidence suggests that while high capital strength may modestly reduce expected returns, well-capitalized banks incur lower bankruptcy costs and face reduced external funding needs—especially in developing economies where external borrowing is challenging Consequently, banks with strong capitalization tend to be more profitable than their undercapitalized peers Berger (1995) also clarifies that banks holding larger capital buffers generally bear less risk, although higher risk can be associated with higher returns A substantial capital base enables banks to absorb unforeseen losses, reduces the need to borrow, lowers funding costs, and provides sufficient resources to expand interest income and pursue other profitable investments (Gul et al., 2011; Tan and Floros, 2012).
Based on the discussion above, hypothesis (H3) is proposed that capital has a positive impact on bank profitability
Loans and leases to total assets ratio is a key indicator of a bank’s emphasis on traditional lending, calculated as loans and leases divided by total assets; while lending remains a principal income source for banks, in a volatile economic environment this activity strongly shapes profitability The ratio is used to investigate the effect of asset structure on bank performance (Syafri, 2012) According to Trịnh Quốc Trung and Nguyễn Văn Sang (2013), the loans-to-total-assets ratio tends to move in tandem with bank profitability, suggesting that higher lending shares can enhance profitability However, Nguyễn Việt Hùng (2008), in a study of 32 Vietnamese commercial banks from 2001 to 2005, finds an inverse relationship between the rate of lending and bank efficiency, arguing that increasing credit raises credit risk, with medium- and long-term loans carrying greater risk due to market and economic fluctuations Additionally, Adusei (2015) contends that this ratio can have a negative effect on bank stability as reflected in the ROA model.
Based on the discussion above, hypothesis (H4) is proposed that loans and leases to total assets ratio has a negative impact on bank profitability
Deposit factor, defined as the deposits-to-total-assets ratio, reflects a bank’s reliance on retail funding Kosmido et al (2008) note that this factor affects bank performance, with a low ratio suggesting underutilization of mobilized capital and a high ratio indicating more effective deployment of funds Deposits are the most common funding source and carry the lowest cost of funds, so higher deposits, when converted into loans through larger interest margins, can improve profitability (Alper & Anbar, 2011) However, Hashem (2016) and Abugamea & Gaber (2018) warn that banks with excessive deposits but poor loan utilization waste resources, potentially reducing profitability.
Based on the discussion above, hypothesis (H5) is proposed that deposits to total assets ratio has a negative impact on bank profitability
Noninterest income captures the variations in banks’ business models and is typically assessed by the share of noninterest income in operating income, serving as a key indicator of income diversification Banks have increasingly shifted away from traditional interest-based operations toward noninterest services, and noninterest income now represents a substantial portion of total revenue, helping to cushion interest-rate risk and broaden product lines Consequently, noninterest products are playing a growing role in bank revenue The literature on the impact of noninterest income on profitability is mixed: some studies show that expanding noninterest activities improves profitability, enhances operational efficiency, and reduces overall risk, while others warn that noninterest income can increase return volatility and reduce risk-adjusted profitability In specific contexts, such as a study of Liechtenstein banks, some evidence points to a negative impact, and research in OECD countries often finds only a modest or unclear effect on risk-adjusted returns, indicating that the influence of noninterest income is context-dependent Additionally, evidence suggests that large banks may benefit from noninterest income without becoming fully dependent on it for overall returns.
Based on the discussion above, hypothesis (H6) is proposed that noninterest income has a negative impact on bank profitability
To assess credit risk, this study relies on the loan loss provisions to total loans and leases ratio Beatty and Liao (2009) describe loan loss provisions as reserves that banks set aside to cover potential loan defaults, a policy designed to protect bank profitability and capital adequacy Across local and international institutions, loan loss provisions have been deployed to mitigate the adverse effects of credit risk Fueled by past crises and banking challenges, the loan loss provision to total loans ratio has become increasingly important for strengthening banks’ financial position The primary objectives of loan loss provisions include providing insight into the bank’s future performance and influencing regulatory capital management and earnings management, with potential tax implications.
2012); control the level of revenue volatility (Norden and Stoian, 2013); and prevent instabilities in risk-weighted assets that affect the bank's risk and profitability (Norden and Stoian, 2013)
Many empirical studies use loan loss provisions as a proxy for credit quality, i.e., the credit risk of commercial banks The conventional view holds that rising loan loss provisions or a deteriorating loan portfolio signal poorer credit quality and higher credit risk Yet, empirical results vary across countries, time periods, and performance measures For instance, Tahir et al (2014) find an inverse relationship between loan loss provisions and profitability (ROA, ROE) in Pakistan, implying that larger provisions erode bank profitability and financial stability Other researchers, including Mustafa et al (2012) and Alhadab and Alsahawneh, contribute to the mixed evidence surrounding how loan loss provisions relate to bank performance.
RESEARCH GAP
Through previous studies, the author found that there are gaps between the studies:
Fistly, studies of factors affecting bank profitability in Vietnam mainly focus on bank-specific factors Macroeconomic variables such as inflation, economic growth,… have not been included in many studies
Additionally, extensive research—both in Vietnam and globally—has explored the indicators influencing bank profitability However, results across studies are often inconsistent, which can lead to recommendations that are not suitable for Vietnamese commercial banks.
Chapter 2 provides the theoretical basis for bank profitability by outlining the concept of operational efficiency, presenting related theories, and describing how profitability is measured with ROA and ROE It also proposes the study’s hypotheses To identify the factors affecting bank profitability, the author uses both bank-specific and macroeconomic factors as dependent variables to ensure objectivity and a holistic view of their influence on the analysis The research model includes liquidity creation, bank size, capital, the ratio of loans and leases to total assets, the ratio of noninterest income to operating income, the ratio of loan loss provisions to loans and leases, as well as inflation and GDP.
METHODOLOGY
RESEARCH PROCESS
To analyze the factors affecting the efficiency of banking operations, the thesis uses a quantitative research method based on specific processes:
Step 1: Determining research objective of the topic
Step 2: Finding out theoretical basis and relevant empirical evidence that has been done before
Step 3: From the theoretical basis and relevant empirical evidence in step 2, the author selects variables to build the research model of the subject
Step 4: Collecting necessary data to estimate the model Specifically, the data utilized in the article is secondary data (collected from audited financial statements from 2008 to 2020 of 14 Vietnamese commercial banks officially published on the banks' websites and as well as VietstockFinance website in the period 2008 – 2020)
Step 5: From the model to export in step 3 and data collected in step 4, the author proceeds to use descriptive statistics, correlation analysis, and multicollinearity tests with support of Stata software
Step 6: After performing multicollinearity tests, the author regresses the model by OLS, FEM, REM methods and performs corresponding tests to find the most suitable model among the 3 methods
Step 7: After finding a suitable model, the author conducts testing of the model's defects (heteroscedasticity and autocorrelation) If the model has a defect, the author fixes it by FGLS
Step 8: The author analyzes the research results to determine which independent variables have statistical significance and how much impact they have on the dependent variables to ensure that the model used is proper , then evaluate the analytical results with the theoretical basis in Chapter 2 to see if the magnitudes and signs of the variables are as anticipated based on the theory given in Chapter 2 If yes, go to step 9; otherwise, go back to steps 3, 4 or 6 to see if the problem lies with the theoretical model or data collection or the choice of estimation method
Step 9: When the estimated model has been tested to be reliable, the estimated results of the model are used for regression analysis From the results, the author reaches the conclusion of the research problem and makes suggestions and recommendations
RESEARCH DATA
Due to limited information and time constraints, the study collects data from only 14 Vietnamese commercial banks The bank-specific characteristics of these banks are based on audited financial statements Macroeconomic indicators are sourced from the World Bank The author combines these datasets for comprehensive analysis.
Literature review Models and hypotheses
Collecting data Multicollinearity testing OLS, FEM, REM regression
Heteroscedasticity and autocorrelation testing Fixing model by FGLS method Analyzing research result
Solutions data with the synthesis and calculation according to the formula mentioned in Chapter
Using a secondary data source, the study includes 182 observations from January 2008 through December 2020 to evaluate the indicators affecting bank profitability The author employs panel data regression to statistically analyze the impact of the independent variables on the dependent variables, specifically ROA and ROE Panel data analysis is chosen because it offers advantages over pure time-series or cross-sectional approaches, such as its ability to control for unobserved heterogeneity and to exploit both cross-sectional and temporal variation, thereby improving estimation efficiency.
Panel data methods enable the interpretation of heterogeneity across cross-sectional observations by combining data over time into a single time-indexed dataset Cross-sectional data are often not identical, and panel data incorporate the time dimension to capture both between-unit differences and within-unit dynamics Through panel data analysis, researchers can estimate the individual-specific effects of cross-sectional units, revealing unique patterns that emerge over time (Phạm Thị Tuyết Trinh, 2016).
Panel data methods, by combining the time dimension with cross-sectional units, provide a larger number of observations and richer information In empirical research, understanding how relationships between variables evolve over time is a central concern Therefore, researchers can pool many cross-sectional units across a given time period, increasing observations, boosting degrees of freedom, and enhancing the power of statistical tests Additionally, this data fusion helps reduce multicollinearity, a common challenge in time-series models with numerous explanatory variables (Phạm Thị Tuyết Trinh, 2016).
Using the panel data method enables researchers to address broader and more complex problems by integrating cross-sectional data with time-series information This approach provides both time-driven analysis and insight into cross-unit differences arising from the data’s cross-sectional component, as noted by Phạm Thị Tuyết Trinh.
Panel data analysis enables researchers to construct and estimate more sophisticated models than those possible with pure cross-sectional or time-series data, by integrating observations across entities and over time; this approach supports technically efficient models, as demonstrated by Phạm Thị Tuyết Trinh (2016).
Panel data analysis helps reduce or eliminate biases arising from data aggregation across companies or individuals This approach yields more accurate variables than those derived from single-point collected data or macro-level measurements, thereby improving the reliability of empirical results (Phạm Thị Tuyết Trinh, 2016).
RESEARCH METHODOLOGY
Using a quantitative research method, this study determines the influencing trends and the levels of determinants that affect profitability in Vietnamese commercial banks It identifies how various profitability determinants influence bank performance and tracks their trends over time The analysis employs professional technical methods, including descriptive statistics to summarize data, correlation analysis to examine relationships, and panel data regression analysis to quantify the effects of determinants on profitability The findings provide robust insights into the drivers of profitability within the Vietnamese banking sector.
Descriptive statistics are utilized to provide general information about the variables in the research model Descriptive statistics consist of mean, minimum, maximum, standard deviation, and observations
Correlation analysis assesses the strength and direction of relationships in the research model, revealing whether correlations are strong or weak and whether they are positive or negative It also evaluates the presence of severe multicollinearity among predictors If the correlation coefficient between any pair of independent variables exceeds an absolute value of 0.8, the model may encounter severe multicollinearity, potentially biasing estimates and undermining interpretability.
Regression analysis of balanced panel data was employed to examine trends and the influence of factors on Vietnamese commercial banks’ profitability The study uses Pooled OLS, fixed effects (FEM), and random effects (REM) models, and conducts a Hausman test to choose between Pooled OLS and FEM with H0: Pooled OLS is preferred; and a Hausman test to compare FEM and REM with H0: REM is preferred To test hypotheses about the impact of indicators on profitability, F-tests are applied at the 1%, 5%, and 10% significance levels, and the sign and size of the slope coefficient β reveal the direction and magnitude of influence on profitability Multicollinearity is checked using the Variance Inflation Factor (VIF), with VIF > 10 indicating severe multicollinearity Heteroskedasticity is tested by the Lagrange multiplier test (H0: no heteroskedasticity), and autocorrelation is assessed via the Wooldridge test (H0: no autocorrelation) If heteroskedasticity is present, the model is re-estimated by feasible generalized least squares (FGLS). -**Sponsor**Looking to refine your article on regression analysis and its impact on Vietnamese commercial banks' profitability? We can help you highlight the core meaning and ensure it aligns with SEO best practices Consider emphasizing these key sentences: "Regression analysis of balanced panel data is used to test trends and levels of influence of factors affecting Vietnamese commercial banks’ profitability," and "The study uses F-test at 1%, 5% and 10% significance level to determine the confidence level of the influence of the independent and control variables." For further insights into financial analysis and investment strategies, explore [Freedom24 ETF - English](https://pollinations.ai/redirect-nexad/snaB2BWq?user_id=229098989) and discover how ETFs can diversify your investment portfolio With Freedom24, you can access expert market research and a wide selection of ETFs, all while managing trades in real-time.
RESEARCH MODEL
The study develops and expands the research model of factors affecting banking profitability It builds on and extends the prior work of Ali et al (2011), Alper and Anbar (2011), Ongore and Kusa (2013), San and Heng (2013), and Almazari to form a more comprehensive framework for analyzing the determinants of bank profitability.
A body of literature (2014), Rahman (2015), Abugamea (2018), Duan and Niu (2020), Trịnh Quốc Trung and Nguyễn Văn Sang (2013), and Lê Đồng Duy Trung (2020) supports a model to estimate the profitability of Vietnamese commercial banks The model evaluates profitability using indicators such as liquidity creation, bank size, capital, the loans and leases to total assets ratio, the deposits to total assets ratio, the noninterest income to operating income ratio, and the loan loss provisions to loans and leases ratio, along with macroeconomic factors of GDP growth and inflation The proposed research model integrates these bank‑specific and macroeconomic variables to explain profitability outcomes and guide empirical analysis of Vietnamese banking performance.
𝛽 5 DEP i,t + 𝛽 6 NII i,t + 𝛽 7 LLP i,t + 𝛽 8 GDP t + 𝛽 9 INF t + 𝜀 i,t
Y i,t : Profitability of bank i in year t
LC i,t : Liquidity creation / total assets of bank i in year t
SIZE i,t : Bank size of bank i in year t
CAP i,t : Total equity capital / total assets of bank i in year t
LOAN i,t : Loans / total assets ratio of bank i in year t
DEP i,t : Deposits / total assets ratio of bank i in year t
NII i,t : Noninterest income / operating income ratio of bank i in year t
LLP i,t : Loan loss provisions / total loans and leases ratio of bank i in year t
GDP t : GDP growth in year t
VARIABLE DEFINITION
Y i,t is the dependent variable is measuring the profitability of bank i in year t represents ROA and ROE
Return on assets (ROA) measures profitability per dollar of assets and indicates how efficiently a bank uses its assets to generate net income This key financial metric is calculated by dividing net income by total assets, i.e., ROA = Net income / Total assets By evaluating ROA, analysts assess a bank’s ability to convert assets into earnings, reflecting overall financial performance and asset utilization.
ROE is evaluated as the ratio of net income divided by average total equity The item of net income tax is taken from the income statement, and total equity is taken from the balance sheet The formula is shown as follows:
Berger and Bouwman (2009) propose four liquidity-creation measures: cat fat, cat nonfat, mat fat, and mat nonfat (LC) These measures classify all activities excluding loans by both product category and maturity, enabling a detailed assessment of liquidity creation across different dimensions However, due to data restrictions, the loan component is sorted by only one dimension, either category (cat) or maturity (mat).
To evaluate how much liquidity banks create on the balance sheet versus off-balance-sheet, this study follows Berger and Bouwman (2009) by classifying off-balance-sheet activities as 'fat' (included in liquidity measurement) or 'nonfat' (excluded) Due to data limitations on off-balance-sheet activities, the authors adopt a 'cat nonfat' measurement to estimate banks’ liquidity creation This approach builds on Berger and Bouwman’s framework and continues to quantify liquidity creation across both on-balance-sheet and off-balance-sheet activities.
(2009) methods to classify objects into different categories as presented in Table t 3.1
In the second step, all banks’ operations are assigned weights according to Table 3.1 In the final step, the author estimates the cat nonfat LC by utilizing the activities that have been classified and weighted in Steps 1 and 2 The measure of LC is presented as follows.
According to the liquidity creation theory, banks generate liquidity on their balance sheets by converting illiquid assets into liquid liabilities, which earns positive weights for both liquid liabilities and illiquid assets Conversely, illiquid liabilities and liquid assets receive inverse weights since liquidity is consumed when illiquid funding supports liquid assets Only half of the total liquidity created can be attributed solely to the origin or use of funds, which explains the application of weights 0.5 and -0.5 Semi-liquid assets and liabilities are assigned an intermediate weight of zero.
Table 3.1: Liquidity creation categorizing and weighting
Assets Level of liquidity Weight
Cash and cash equivalents Liquid -0.5
Balance with the State Banks Semi liquid 0
Balance with and loans to other credit institutions Semi liquid 0
Corporate and commercial loans Semi liquid 2 0 1
Capital contributions and long-term investments Illiquid 1 0.5 1
Liabilities Level of liquidity Weight
Borrowings from the State Bank of Vietnam Liquid 0.5 1
(Source: The author collected from previous studies)
Complete liquidity creation occurs when a bank funds illiquid assets with liquid liabilities, yielding a positive liquidity creation measure (LC) Conversely, when a bank funds liquid assets with illiquid liabilities, liquidity is destroyed, resulting in a negative LC At a two-to-one ratio, where two units of liquid liabilities fund one unit of liquid assets, no liquidity is created, indicating LC equals zero in that scenario.
LC = 1 0 2 2 5 e x 1 + - 1 0 1 5 x 1 = 0 Therfore, LC has a value between -1 and +1
SIZE is an independent variable calculated as the natural logarithm of a bank’s total assets (SIZE = ln(Total Assets)) Total assets data are drawn from the banks’ annual reports, ensuring consistent and auditable figures across institutions By applying the natural log to total assets, SIZE captures bank size and enables robust, scalable comparisons in empirical analyses.
SIZE = ln(Total assets) CAP
CAP is the independent variable defined as total equity capital divided by total assets Both total equity capital and total assets are taken from the balance sheet, and the calculation uses the formula CAP = total equity capital / total assets This approach directly links a company’s equity position to its asset base, providing a concise measure of capital structure based on balance sheet data.
Loans to total assets is an independent variable calculated by dividing loans by total assets, with both loans and total assets drawn from the balance sheet The LOAN metric is defined by the formula LOAN = Loans / Total Assets, expressing the proportion of a company’s assets financed by loans and serving as a key financial indicator in balance-sheet analysis.
Deposits to total assets ratio (DEP) is an independent variable defined as customer deposits divided by total assets Customer deposits are taken from the annual report and total assets are taken from the balance sheet The calculation formula is DEP = customer deposits / total assets, producing a ratio that reflects how much of a company’s assets are funded by customer deposits and helping assess liquidity and funding structure.
DEP is shown as follows:
Noninterest income to operating income ratio (NII) serves as an independent variable, measuring the relationship between noninterest income and operating income Noninterest income is sourced from the annual financial report, while operating income is taken from the income statement The formula is defined as noninterest income divided by operating income (NII / Operating Income).
Loan loss provisions to total loans and leases (LLP/TL&L) is an independent variable used in credit risk and financial analysis It is calculated by dividing the loan loss provisions by the total loans and leases, with both components taken from the balance sheet The resulting LLP to total loans and leases ratio provides insight into the level of reserves set aside for potential losses relative to the size of the loan portfolio.
GDP is an independent variable collected from Worldbank, representing the Gross Domestic Product growth rate
INF is an independent variable collected from Worldbank, representing the inflation rate
Symbol Variable name Measurement Expected sign in DEPENDENT VARIABLES
ROA Return on Assets Net income divided by total assets
ROE Return on Equity Net income divided by total equity capital
LC Liquidity creation / total assets ratio
Berger and Bouwman (2009) preferred “cat nonfat” liquidity creation measure divided by total assets
SIZE Size Natural logarithm of total assets (+)
CAP Capital / total assets ratio
Total equity capital divided by total assets (+)
LOAN Loans and leases / total assets ratio
Total loans and leases divided by total assets (-)
DEP Deposits / total assets ratio Deposits divided by total assets (-)
Noninterest income / operating income ratio
Noninterest income divided by total operating income ( 1 -)
Loan loss provisions / total loans and leases ratio
Loan loss provisions divided by total loans and leases (-)
GDP GDP growth Annualized growth rate of real GDP (+)
INF Inflation 12-month percent change in the consumer price index ( - + -1 )
In Chapter 3, the author outlines the research process and presents the study's methodology Drawing on both domestic and international models, the author develops a specific research model and defines the measurements for the related variables used in the thesis.
RESULT
DESCRIPTIVE STATISTICS
Research data were drawn from the financial statements of 14 commercial banks over a 13-year period (2008–2020), yielding 182 observations The descriptive statistics summarize each variable by reporting the mean, standard deviation, minimum, and maximum values, as shown in the accompanying tables.
Variable s Observations Mean Standard deviation Min i Max
(Source: The author’s summary from Stata)
Based on the results of Table ts 992 4 2 3 1, the author demonstrates the following conclusions:
The ratio of return on assets (ROA) of 14 commercial banks in Vietnam from
From 2008 to 2020, the mean ROA was 0.9%, indicating that commercial banks’ profitability remains below international benchmarks (ROA ≥ 1%) The standard deviation of 0.6% is relatively low, showing a high degree of similarity in asset efficiency across banks; Techcombank posted the highest ROA at 2.86% in 2020, while National Citizen Bank recorded the lowest at 0.0% in 2020 The ROE across 14 Vietnamese banks over the same period averaged 11.86%, still short of Moody’s financial capacity standards (12%–15%), with a standard deviation of 6.8%; the peak ROE was 26.82% for ACB Bank in 2011 and the trough 0.03% for National Citizen Bank in 2020 These figures suggest that bank business activities are not only inefficient but also that post‑tax earnings can be negative Nonetheless, profitability appears relatively less volatile, as shown by the ROA and ROE standard deviations of 0.6% and 6.8%, respectively, implying banks strive to minimize profit fluctuations to stabilize operations.
LC has a mean of -6.98%, standard deviation is 10.2%, the highest value of
MSB Bank recorded 51.61% in 2015, while VIB Bank posted the lowest value of -33.77% in 2020 Banks with strong capitalization, including BIDV, Vietinbank, and Vietcombank, exhibit LC values below 10%, suggesting they rely on a substantial mix of illiquid liabilities and liquid assets to finance their liquid liabilities and illiquid assets.
SIZE shows a mean of 32.80 with a standard deviation of 103.98%, indicating substantial variation in capital size among the selected banks BIDV has the highest SIZE value of 34.96 in 2020, while HDBank has the lowest value of 29.89 in 2008.
State-owned commercial banks such as BIDV, Vietinbank, and Vietcombank generally hold significantly higher value than private banks because they receive government capital to operate This advantage is supported by a long-standing asset base accumulated over many years of operation.
CAP has a mean of 8.23%, indicating that equity accounts for a relatively small portion of the bank's capital structure and signaling a higher risk potential when banks finance projects or investments with capital The standard deviation is 3.06%, showing substantial variation in banks' equity positions Eximbank recorded the highest value at 26.63% in 2018, while BIDV posted the lowest value at 3.82% in 2017.
LOAN has a mean of 55.39%, standard deviation is 12.81%, the highest value is
80.06% of BIDV in 2020, the lowest value is 80.06% 22.53% of Maritime Bank in
From 2008 to 2020, more than half of Vietnamese commercial banks' total assets were devoted to lending, making lending the primary source of bank income This indicates a high dependence on credit activities for Vietnamese banks and suggests that asset quality is largely driven by the quality of their loans.
DEP has a mean of 76.43% and standard deviation of 8.71% The highest value is 91.66% of VIB bank in 2008, the lowest value is 37.04 % of ACB Bank in 2011
NII has a mean of 9 8.40%, standard deviation of 6.08% The highest value is 29.62% of VIB bank in 2014, the lowest value is 0.2% of HDBank in 2009
LLP has a mean of 2 1.3%, standard deviation of 1 0.54% The highest value is 3.27% of Vietcombank in 2009, the lowest value is 0% of ACB Bank in 2020
From 2008 to 2020, Vietnam recorded an average GDP growth rate of 5.92% and an average inflation growth rate of 7.22% As a developing economy, Vietnam’s higher growth is typically associated with higher inflation, making low inflation unlikely.
CORRELATION ANALYSIS
To estimate the relationships among variables in the model, the study uses correlation analysis to measure the strength of association between the independent variables and the dependent variable The correlation coefficient, r, is a statistical measure of the degree of linearity between two variables and ranges from -1 to +1 A positive value of r indicates a positive correlation between the dependent and independent variables, while a negative value indicates a negative correlation Additionally, high absolute values of r may signal multicollinearity when r > 0.8 The linear relationship between variables can be identified by the magnitude and sign of r, according to Hoàng Trọng and Chu Nguyễn Mộng Ngọc (2008).
Correlation coefficient matrix using Stata is demonstrated in Table 3 4 5 2:
Table 4.2: Matrix of correlation between variables INF 1 (Source: The author’s summary from Stata)
ROA ROE LC SIZE CAP LOAN DEP NII LLP GDP INF
According to Table 7.4.2.8, the independent variables are not strongly correlated with one another, as the absolute values of their correlation coefficients fall below 0.8.
MULTICOLLINEARITY TESTING
Variance Inflation Factor (VIF) is used to determine whether a variable has a severe multicollinearity relationship with other predictors A variable is considered strongly correlated when its VIF exceeds 10 (Gujarati and Porter, 2004) The larger the VIF value, the greater the collinearity problem and the higher the risk of distorted regression estimates According to the results in Table ts, multicollinearity among the variables is evident.
94.3, coefficients of VIF are all less than 10, so it can be concluded that the model does not have serious multicollinearity
Table 4.3: Result of multicollinearity testing
(Source s : The author’s summary from Stata)
Descriptive statistics were computed to provide an overview of the study data, and correlation analysis was performed to assess relationships and detect multicollinearity among the variables The model was then estimated using three regression techniques—pooled OLS, fixed effects (FEM), and random effects (REM)—with the support of Stata software Finally, a model selection test was applied to identify the most suitable model for the data.
REGRESSION RESULT OF ROA
Table 4.4 shows the Pooled OLS regression results, with an R-squared of 0.380, indicating that the independent variables explain 38% of the variation in the dependent variable ROA The variables LLP and GDP are not statistically significant in the model The variables LC, SIZE, CAP, and NII are statistically significant at the 1% level (p ≤ 0.01, 99% confidence) LOAN and INF are significant at the 5% level, and DEP is significant at the 10% level.
Using the FEM method, Table 4.4 shows an R-squared value of 0.397, indicating that the independent variables included in the model explain 39.7% of the variation in the dependent variable ROA This finding demonstrates the predictive power of the chosen factors for ROA and underscores the portion of ROA changes captured by the model.
In the regression analysis, GDP does not show statistical significance in the model The independent variables LC, SIZE, CAP, and LPP are statistically significant with p-values ≤ 0.01, corresponding to 99% confidence DEP and NII reach statistical significance at the 5% level, and INF is significant at the 10% level These results indicate that, among the considered factors, GDP has no detectable impact, while LC, SIZE, CAP, and LPP demonstrate strong statistical support, with DEP, NII, and INF showing smaller but meaningful levels of significance.
Results from the regression using the Random Effects Model (REM) as reported in Table 4.4 show that GDP is not statistically significant, while the independent variables LC, SIZE, CAP, and LLP are statistically significant, with their coefficients associated with p-values indicating significance These results imply that, under the REM specification, GDP does not explain the dependent variable, whereas LC, SIZE, CAP, and LLP do.
≤1% (99% confidence level) DEP, NII, INF have statistical significance with 5% significance level
Regression results of ROA by three methods OLS, FEM, REM are summarized in Table 9 4.4 9 :
Table 5 4.4: Summarizing regression models of ROA
(Source: The author’s summary from Stata)
Notes: *, **, *** indicate significance at 10%, 5%, 1%, respectively Values in ( ) indicate error
4.4.2.1 Pooled OLS and REM regression
The author performs the Breusch-Pagan Lagrange test (LM test) to choose between the REM model and the Pooled OLS model, with the following hypothesis:
The result is shown as follows:
Table 4.5: Result of Lagrange test on ROA
Breusch-Pagan / Cook- Weisberg test for heteroscedasticity
Null hypothesis H 1 0: Constant variance chi2(1) 7.88
(Source s : The author’s summary from Stata)
Based on Table 4.5, the test yields a P-value of 0.005, which is below the 5% significance level, allowing the researcher to reject H0 and favor H1, meaning the REM regression model provides a better fit than Pooled OLS for ROA The analysis then proceeds to compare the REM model with the FEM model to determine the most appropriate specification for modeling ROA.
The author performs the Hausman test to find a more appropriate regression model between the FEM regression analysis model and the REM regression analysis model with the hypothesis:
H 1 0: REM regression model is appropriate
H 2 1: FEM regression model is appropriate
The result is shown in Table t 2 4 1 6:
Table 9 4.6: Result of Hausman test on ROA
H 2 0: difference in coefficients not systematic chi 3 2(9) = 4 3.2 7 7
(Source: The author’s summary from Stata)
Table 9.4.6 reports a Hausman test P-value of 0.9525, which is well above the 0.05 significance level, so we fail to reject the null hypothesis This result indicates that the random effects model (REM) is the appropriate specification for the ROA regression, and thus REM is more suitable than the fixed effects model (FEM) for the dependent variable in the ROA model.
To conclude, REM regression model is the most proper model to use for regression analyzing on ROA
The author implements the heteroscedasticity testing using Breusch-Pagan test with the hypothesis:
If p-value ≤ (α = 0.05) accepts hypothesis H1, rejects hypothesis H0 The result of the analysis is shown as follows:
Table 4.7: Result of heteroscedasticity testing on ROA chibar2 (01) Prob > chi2 Result (compare to significance level 5%) Defect
(Source: The author’s summary from Stata)
The result from Table 9 4 1 7 1 shows that P-value = 0.0000 < 0.05, so the regression model includes heteroscedasticity phenomenon
The author uses the Wooldridge test for autocorrelation of residuals in the regression model, implemented with the xtserial command, to detect first-order serial correlation and identify an optimal solution to fix the model defect The test evaluates the null hypothesis of no autocorrelation (no first-order serial correlation) against the alternative that autocorrelation is present, guiding the selection of the most appropriate corrective action to improve model specification and inference.
The result is shown as follows:
Table 4.8: Result of autocorrelation testing on ROA
Result (compare to significsignificanceane level 5%)
(Source: The author’s summary from Stata)
The testing result from Table 4.8 shows that autocorrelation is detected in regression model due to Prob = 0.0000 F value of the F-test being less than the 0.05 significance level, the result is statistically significant, leading to the selection of the fixed effects model (FEM) for regression analysis The author then proceeds to compare the FEM with the random effects model (REM) to determine which specification provides the better fit and more reliable inferences for the data.
The author performs the Hausman test to find a suitable regression model between the FEM and REM regression analysis models with the hypothesis:
H 2 0: REM regression model is appropriate
H 2 1: FEM regression model is appropriate
The result is shown in Table t 9 4.1 5 2:
Table 9 4.1 5 2: Result of Hausman test on ROE
H 2 0: difference in coefficients not systematic chi 4 2(9) = 7.40
(Source: The author’s summary from Stata)
In Table 4.12, it can see that the P-value coefficient of the Hausman test is
30.5960 > 5%, so the author can accept H0 For the regression model FEM and REM of the dependent variable is ROA, REM model is more suitable
The author implements heteroscedasticity testing through the Breusch-Pagan test with the hypothesis:
If P-value ≤ α (α = 0.05) accepts hypothesis H1, the author rejects hypothesis H0 The result of the analysis is shown as follows:
Table 4.13: Result of heteroscedasticity testing on ROE chibar2(01) Prob>chibar2 Result (compare to significance level 5%) Defect
(Source s : The author’s summary from Stata)
The result in Table t 8 4.13 shows that P-value = 0.0000 < 0.05, so the regression model includes heteroscedasticity phenomenon
The author assesses residual autocorrelation in the regression model using Wooldridge's test to identify the optimal solution for correcting model error, implemented through the xtserial command under the specified hypothesis.
The regression result is shown as follows:
Table 4.14: Result of autocorrelation testing on ROE
F( 1,13) Prob > F Result (compare to significane level 5%) Defect
(Source: The author’s summary from Stata)
The testing result from Table 4.14 shows that autocorrelation is detected in regression model because Prob > F = 0.0000 < 0.05
The regression model shows heteroscedasticity and autocorrelation; to correct these defects in the REM when ROE is used as the independent variable, the author applies Feasible Generalized Least Squares (FGLS) This approach yields a clearer model that isolates the key determinants of profitability for Vietnamese commercial banks, with the detailed findings reported in Table 7 4.15 8.
Table 8 4.1 4 5: FGLS regression analysis of ROE
Variable Coefficient Standard error P-value
(Source: The author’s summary from Stata)
From the research result, the regression model is demonstrated as follows:
ROE i,t = - 0.862 – 0.12 LC i,t + 0.037 SIZE i,t – 0.298 CAP i,t - 0.204 DEP i,t
Table 4.16 presents the combined results of factors affecting bank profitability with 2 models of dependent variables ROE and ROA:
Table 4.16: Summarizing suitable regression models of ROA and ROE
(Source: The author’s summary from Stata) Notes: *, **, *** indicate significance at 10%, 5%, 1%, respectively Values in ( ) indicate error
Liquidity creation (LC) negatively affects bank profitability in both ROA and ROE analyses, with a regression coefficient of -0.0159 in the ROA model significant at the 1% level and -0.120 in the ROE model significant at the 5% level These results are consistent with the findings of Tran et al (2016), Sahyouni and Wang (2019), and Chu Thi Thanh Trang et al (2021) The study thus supports the bankruptcy cost hypothesis, indicating that higher liquidity creation by banks can raise bankruptcy costs and reduce profitability.
Research shows that greater illiquidity risk, indicated by rising liquidity creation, can increase default risk and negatively affect bank performance, implying that higher liquidity may reduce bank profitability Since elevated liquidity creation raises the likelihood of bank failure (Fungáčová et al., 2015), regulators may adopt policies to limit excessive liquidity creation by banks The findings support hypothesis H1: liquidity creation has a negative impact on bank profitability in both ROA and ROE models.
Bank size (SIZE) has a positive impact on banking performance, evidenced by regression coefficients of 0.00310 for ROA and 0.0366 for ROE, significant at the 1% level for ROA and the 7% level for ROE, respectively This confirms that larger banks tend to achieve higher profitability and overall performance The results align with prior studies by San and Heng (2013), Anbar and Alper (2011), and Duan and Niu.
According to a 2020 study by Lê Đồng Duy Trung, the positive effect of bank size on profitability indicates that when commercial banks expand their scale, grow assets, and broaden their branch networks, their operational efficiency rises This improvement can be explained by microeconomic efficiency gained from allocating fixed costs over a larger production base Vietnamese commercial banks with large scale and extensive branch networks enjoy advantages in mobilizing capital, developing products and services, and reaching more customers, which enhances profitability Furthermore, the competitiveness of large-scale banks is stronger than that of smaller banks, and increasing bank size tends to raise profits Thus, the evidence supports the hypothesis H2: size has a positive effect on bank profitability.
Capital (CAP) has a positive impact on ROA with a regression coefficient of
Results indicate that a higher capitalization ratio positively influences return on assets (ROA) and negatively affects return on equity (ROE) The ROA effect arises because greater capital improves loss absorption, enhances asset management effectiveness, and allows banks to rely less on external funding, thereby lowering funding costs and boosting ROA Conversely, a higher equity base reduces financial leverage and overall risk, limiting upside returns and lowering ROE, in line with the risk–return trade-off; the ROE coefficient is negative (approximately -0.298) and statistically significant at the 10% level These findings align with previous studies by Ongore and Kusa (2013), San and Heng (2013), Le Dong Duy Trung (2020), and Trịnh Quốc Trung and Nguyễn Văn Sang (2013) The results support Hypothesis H3 that capital has a positive effect on bank profitability in the ROA model but not in the ROE model.
Loans and leases to total assets
Loans and leases to total assets (LOAN) ratio exerts a negative impact on the profitability of Vietnamese commercial banks, as indicated by its relationship with return on assets (ROA) The regression coefficient is -0.00927 and is significant at the 5% level, meaning that increases in the LOAN ratio are associated with decreases in ROA and overall profitability This finding highlights the trade-off between lending activity and profitability in Vietnam’s banking sector and has implications for credit risk management and capital allocation.
ROE was not statistically significant, indicating that expanding banks' lending activities does not automatically improve operational efficiency This result aligns with the studies of Alper and Anbar (2011), Duan and Niu (2020), and Lê Đồng Duy Trung (2020) In fact, as credit volume increases, credit risk also rises, with medium- and long-term loans bearing higher risk due to market and economic fluctuations, which negatively affect performance as reflected in ROA The findings support hypothesis H4: the loans and leases to total assets ratio has a negative impact on bank profitability through ROA.
The deposits to total assets ratio (DEP) has a converse effect on both ROA and ROE at 5% and 1% significance level with regression coefficients of -0.00960 and -
Results indicate that a one-unit increase in the deposits-to-total-assets ratio is associated with a 0.00961-unit decrease in ROA and a 0.204-unit decrease in ROE, consistent with prior studies by Alper and Anbar (2011), Abugamea (2018), and Duan and Niu (2020) This suggests that when banks attract excess deposits, debt-servicing pressure rises while current credit demand remains low As a result, revenue from credit activities declines while operating expenses stay unchanged, leading to lower bank profitability However, the magnitude of profit reduction across commercial banks is not substantial Therefore, this study supports hypothesis H5: deposits have a negative impact on bank profitability.
The ratio of noninterest income to operating income (NII) has a converse influence on bank profitability, with regression coefficients -0.0200 and - 1 0.287 for
RESEARCH DISSCUSION
Building on the findings from Chapter 4, this study outlines the main conclusions and offers actionable recommendations based on the factors that affect the profitability of commercial banks in Vietnam Chapter 5 also discusses the study’s limitations and provides directions for future research to deepen understanding of profitability drivers and enhance methodological approaches.
This study analyzes the determinants of bank profitability by estimating two models with ROA and ROE as dependent variables, respectively, and a set of explanatory variables including liquidity creation (LC), the ratio of noninterest income to operating income (NII), deposits to total assets (DEP), bank size (SIZE), capital (CAP), loans and leases to total assets (LOAN), loan loss provisions to loans and leases (LLP), GDP growth (GDP), and inflation (INF) Using panel data from 14 Vietnamese commercial banks over 13 years (2008–2020), totaling 182 observations, the analysis relies on secondary data collected from banks’ annual financial statements as well as the VietstockFinance website and the World Bank.
The thesis employs OLS, FEM, and REM models and finds that the REM specification is the appropriate regression analysis for both ROA and ROE To address heteroscedasticity and autocorrelation, the author uses feasible generalized least squares (FGLS) to analyze factors such as liquidity creation, bank size, capital, loan loss provisions to loans and leases ratio, deposits to total assets ratio, loans and leases to total assets ratio, noninterest income to operating income ratio, GDP growth, and inflation The results show that these factors most appropriately explain changes in the dependent variables (ROA and ROE).