MINISTRY OF EDUCATION HO CHI MINH, SEPTEMBER 2021 THE STATE BANK OF VIETNAM AND TRAINING BANKING UNIVERSITY HO CHI MINH CITY NGUYỄN HOÀNG PHƯƠNG UYÊN Major Banking and Finance FACTORS AFFECTING THE NE[.]
INTRODUCTION
Research Questions
The study attempts to answer and clarify the following questions based on the aforementioned objectives:
Firstly, which are theorical basis of the net interest margin of commercial banks from previous studies in Vietnam as well as all over the world ?
Secondly, which is the most suitable analysis model of factors affecting the net interest income of commercial banks in this study ?
Thirdly, what factors affect the net interest margin of joint stock commercial banks in Vietnam ? How much that factors affect the net interest margin of joint stock commercial banks in Vietnam ?
Finally, which solution is suitable for improvement of the net interest margin of joint stock commercial banks in Vietnam ?
Research’s contribution
The Vietnamese banking system is currently challenged with numerous difficulties and obstacles, including global crises, national economic weaknesses, and shortcomings in banking operations At a time when the banking industry has been exposed for many flaws, it is critical for Vietnamese commercial banks to increase efficiency in all of their business activities in order to stay afloat in the unstable money market As a result, the most pressing issue for all Vietnamese commercial banks is how to maintain profitable growth, long-term development, and survival in the face of international integration Both financial stability and economic growth are dependent on a stable level of bank profitability On the one side, financial health is maintained by suitably funded financial institutions and interest income, which make up a significant portion of bank capital and are dependent on the ability of a bank to reserve profits On the other hand, bank profitability has ramifications for the actual economy, because a prosperous banking sector is required to maintain access to credit for businesses and consumers, hence boosting long-term economic growth Furthermore, when the Covid-
19 pandemic affects businesses and the economy, banks are the most directly impacted; particularly revenue from credit activities, which amounts for a substantial portion of a bank's total revenue The influence of the Covid-19 outbreak on banks' operations has been exposed in certain elements such as daily operations, credit balance, profit, and non-performing loans as the epidemic grows more intricate and unpredictable The managements of banks should assess their operational flaws and devise strategies to address problem that pandemic has brought out.
The net interest margin (NIM) evaluates a bank's health and efficient management of deposit storage and loan making The characteristics which indicate bank's ability to set the interest income from loans above interest expenditures, have been studied extensively in the literature With the recent intensification of banking deregulation in the Vietnamese system, banks have broadened their range of activities in quest of long- term profitability In recent times, Vietnamese commercial banks is considered to maintain interest rates loans are too high, causing difficulties for businesses business is thirsty for capital Although the ceiling interest rate mobilization has reduced gradually according to the decision of the State Bank, but interest lending rates have not yet decreased to the extent that the business can accept.
In terms of scientific significance, the study contributes to the consolidation and extension of earlier studies in defining the factors impacting the net interest margin of commercial bank and the direction of influence of these factors on NIM in theVietnamese commercial bank market Internal and external influencing elements are covered in the influencing factors, which help to define and encompass the aspects that affect profit In past, previous research illustrated many factors affecting the net interest margin including bank’s specific variables such as bank’s size, capitalization,liquidity, customer deposit,… and macroeconomic indicators such as inflation, gross domestic product, etc This study not only remains variables mentioned in previous researches but also applies the impact of non-interest income and market share of selected commercial banks ratio Simultaneously, the period author using in this research is 12 years (from 2009 to 2020), which is the update suggesting the higher accuracy of study In terms of practical meaning, the research results will show the group of factors affecting the net interest margin of Vietnamese commercial banks and the level of influence of those factors Accordingly, this study points out the factors affecting the net interest income ratio of commercial banks and makes recommendations for the interest rate policy of the State Bank Based on the results of this study, the State Bank can use more effective tools (instead of administrative measures) to achieve higher net interest margin.
Objects and scope of study
Research’s object is the net interest margin and factors affecting the net interest margin of Vietnamesse Joint Stock Commercial Bank.
The research scope mentioned is in the period 2009 – 2020 The author chooses the scope of the study is the period from 2009 to 2020 to increase the reliability of the research paper because this is the period when Vietnamese banks are in the recovery phase after the crisis global financial crisis and the Covid-19 pandemic.
Structure of study
Beside table of contents, appendices and bibliography, thesis structure consists of 5 chapters:
Chapter 2: Theoretical framework and literature review Chapter 3: Research method
Chapter 4: Empirical result and discussion
Chapter 5: Conclusion and management interpretation
THEORETICAL FRAMEWORK AND LITERATURE REVIEW
Theoretical Framework
There are a variety of definitions about commercial banks depending on different perspectives When it comes to legal perspective, there are several definitions of commercial banks from several countries.
Commercial banks are enterprises whose regular place is to receive money from the public in the form of deposits, or in other forms and to use that resource for themselves in mining operations credit and finance (Banking Act of France, 1941)
Commercial bank is a type of credit institution that is entitled to carry out all banking activities and other related business activities” In which, banking activities are monetary business and banking services including: capital mobilization, short, medium and long-term loans, discounting of valuable papers, factoring, financial leasing, overdraft,, consumer loans, and providing all other banking services (Law No. 47/2010/QH12 of the Law on Credit Institutions of Vietnam)
From the above perspective of definition, commercial bank is an organization doing business in the monetary sector, receiving deposits from people in the economy, and then performing lending operations and providing financial services, credit, payments, At the same time, commercial banks also carry out investment activities in other earning assets In other words, a commercial bank is a financial intermediary act as a bridge between the capital surplus and the shortage source need to use capital for their business activities.
In terms of commercial bank functions, it is apparent that commercial banks are one of the financial institutions that play a critical role in the economy and are regarded as the lifeblood of the economy.
Firstly, commercial banks serve as an intermediary, coordinating and rotating deposits and savings into credits for individuals, organizations, corporations, and other financial institutions in order to respond to consumer demand, production, and business as well as investment activities Other positions include credit operations, payment, guarantee, and commercial bank agent, which are becoming increasingly significant in our country's process of international economic integration.
Secondly, commercial banks play a part in payment by making payments on behalf of clients who are purchasing goods and services, such as paying by check, offering an electronic payment network, and so on.
Third role is guarantee, as a guarantor, commercial banks commit to pay debts back to customers when customers lose their solvency.
Commercial banks' fourth role is position of agent, in which they oversee and underwrite securities issuance on behalf of their customers Simultaneously, commercial banks provide credit to individuals and households to suit a growing range of needs, including consumer credit, real estate credit, and so on Furthermore, commercial banks are one of the most prominent participants in the bond market, which is used to support public initiatives by state and municipal governments.
Finally, commercial banks play a vital role in the implementation of government macroeconomic policies, assisting in the regulation of economic growth and the pursuit of social objectives.
The net interest margin (NIM) evaluates a bank's health It indicates a bank's ability to set the interest income from loans above interest expenditures, have been studied extensively in the literature.Through this ratio, the bank can control the assets generated profit and evaluate which source of capital has the lowest cost NIM is determined by total interest income minus total interest payment expenses (net interest income) on total assets This ratio is displayed as a percentage, the higher the NIM the better.It can be written in the formula below: ô Interest income-Interest expense
In basic terms, NIM could be tell how much interest earnings were generated from invested capital (assets) Net interest income measures efficiency as well as profitability The higher the NIM value, the more efficiency, because the company is earning more income on less investment Because of the balance sheet accounting equation, note that total assets are also the sum of its total liabilities and shareholder's equity.
Well-performed banks reflect the high circulation and efficiency of capital flows in the economy, thereby creatingfinancial resources for the state, jobs for the people, and close cohesion between economic actors Therefore, the analysis and measurement of profit and related influencing factors is necessary to assess the overall performance and business situation, thereby proposing policies and measures to improve the performance of thebusiness improve the economy of banks in particular and the country in general 2.1.4 NIM generating financial services:
Research results of Ho and Saunder (1981) show that commercial banks act as intermediaries between borrowers and lenders The net interest margin generates from the difference of capital mobilization (bank capital channel) and credit extension which create interest expense and interest income In this section, the study explains and classifies these factors generating NIM so as a base to review previous studies and choose suitable variables in next chapters.
According to credit extension, commercial banks make a variety of credit to difference purposes of customers Forms of credit extension can be classified into 3 groups based on the manifestation of assets in the transaction (Peter Roses, 2013):
• Cash credit group: includes loan, discount and factoring Loan is a service that the lender delivers or commits to hand over a sum of money to the customer for a specified purpose within a specified period of time a certain period of time by agreement, with the principle of a full refund enough principal and interest Besides, discount service is the purchase with a term or a purchase with recourse to the beneficiary's negotiable instruments and other valuable papers before maturity Finally, in factoring service, banks grant credit facilities to the seller or the purchaser through the acquisition reserves the right to access claim receivables or payables arising from the purchase and sale of goods and provision of services under the goods sale and service contract.
• Sign credit group: includes bank guarantee and other foreign exchange commitments Bank guarantee is a form of credit whereby the bank commits to the guarantee recipient that it will perform financial obligations on behalf of the customer when the customer fails to perform or does not fully perform the committed obligation,the customer must accept the debt and return it to the Bank as agreed Besides, foreign exchange commitments are commitments to pay, repay, extend credit, etc., contracts between the bank and customers that generate exchange rate in the future that are off the balance sheet Because they are committed but not yet implemented, they are only recorded on the off-balance sheet.
• Leasing is an asset credit that financial leasing company representing the lessee will purchase an asset from the Supplier and lease it back to the lessee according to the payment schedule shown in the Lease Contract.
Concerning about mobilize capital, it generates from two source including deposit and non-deposit borrowing:
Literature review
2.2.1 Some typical studies in Vietnam:
Thu & Huyen (2014) determined the key factors that affect the net interest margin of banks from 2008 to 2011 banking industry and to what extent Quantitative research results found that the bank's risk aversion, credit risk and implied interest payment are positively related and is statistically significant with the net interest margin. Meanwhile, management quality is related inversely and statistically significant with net interest margin.
Khanh & Tra (2015), from the OLS regression result (firm estimate) with 175 observations of the period from 2008 to 2012, pointed out that NIM is significantly influenced by operating costs, managerial quality, risk aversion, andinflation rate, while concentration ratio have a negative effect on NIM.
Linh & Huong (2015) studied the factors affecting the net interest margin of joint stock commercial banks in Vietnam The study used data from 27 joint stock commercial banks in the period 2008-2013, by means of least squares method (Pooled OLS), FEM and REM methods, general least squares method (GLS) The research results showed that the factors that have a positive impact on marginal interest income include: Credit risk, interest rate, equity size, lending activity scale, bank size, at the same time, factors such as GDP growth, management efficiency have a negative impact on marginal interest income.
Dien & Nga(2018) examined the effect of the Lerner index, the HHI index and the opportunity cost of reserves on the rate of net interest margin (NIM) of commercial banks in the period 2011 - 2015 The study used the adjusted standard error estimation model (PCSE) for balance sheet data from 27 joint stock commercial banks in Vietnam in the period 2011 - 2015 in order to determine the factors affecting the revenue ratio. marginal interest income of Vietnamese commercial banks Research results showed that factors such as Lerner index, opportunity cost of reserves, operating costs have a positive relationship with the rate of profit margin.
2.2.2 Some typical studies in the world
In the period 1996-2009, Hamadi & Awdeh (2012) examined the determinants of commercial bank net interest margins in Lebanon, utilizing bank-specific, industry- specific, monetary policy, and macroeconomic data The result indicated that the net interest margins varied between domestic and international banks Size, liquidity, and efficiency of domestic banks, as well as capitalisation, credit risk, concentration, dollarization, and economic growth, all have a detrimental impact on interest margins.
On the other hand, net interest margin was influenced by deposit growth, lending, inflation, the central bank discount rate, national saving, domestic investment, and, to a lesser extent, the interbank rate For foreign banks, the study discovered that size, liquidity, capitalization, and credit risk have no substantial impact.
Sentürk, B (2016) analyzed the factors affecting the Net Interest Margin (NIM) in the banking industries of Russian and Japanese throughout the years from 2005 to 2014.With macroeconomic and banking variables, the thesis apply multi-way cluster estimation approaches that account for cross-sectional and time-series dependence In Russia, NIM is influenced by capitalization, liquidity risk, inflation, economic growth, and private and public debt In Japan, on the other hand, loan and deposit market concentration, as well as bank size, reign supreme The substitution effect, cost efficiency and profitability are all relevant characteristics in both countries.
Angori, G., Aristei, D., & Gallo, M (2019) discovered the factors that influenced net interest margin from 2008 to 2014 within the Euro Area The study used the main bank- level drivers affecting the net interest margin such as market power, capitalization, interest risk and the level of efficiency, besides, the research explicitlyinvestigated for the effects of regulatory and institutional settings.
Additionally, Cruz-García, P., & Fernández de Guevara, J (2020) conducted a research on the key factors affecting the net interest margin by using a sample of banks from 31 OECD countries (Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Italy, Japan, Korea, Latvia, Luxembourg, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States) over the sample period 2000-2014 The study used GMM technique to deal withthe model of Ho & Saunders (1981) The researcher considered that internal determinants respectively : implicit payments, efficiency, average operating costs, the intensity of competition, the deposit insurance premium and capital stringency have strong impacts on NIM.
Yuksel, S., & Zengin, S (2017) defined the factors effecting NIM in Turkish banking sector in the period between 2003 and 2014 To demonstrate the association, the research developed a model using the multivariate adaptive regression splines method.The study's finding is that non-interest income, non-performing loans, total assets, and exchange rates are all negatively connected to net interest margin.
RESEARCH METHODS
Research model
In order to satisfy the research’s objective, multi linear regression model with balance panel data is used in this study Demirguc-Kunt (2014), Nassar (2014), Hanweck and Ryu
(2005) selected this panel data method to research and background of those research are suitable for Vietnam Beside, some advantages of regression model with panel data are presented such as:
• Panel data allows for interpretation of heterogeneity or heterogeneity of cross units.
• Panel data analysis can take into account each feature of each cross unit.
• Larger and more informative number of observations of panel data thanks to combination of time factor and cross unit Besides, the phenomenon of multi-addition linearity will be reduced when the data are combined in the research models multivariable.
• Problems are solved in multiple ways thanks to panel data used Panel data allows both time dynamic analysis and analyze the difference between the cross-units by cross- component in the data.
• Panel data will reflect research results more accurately and optimally when researched from a macro perspective because to minimizing the biases when synthesizing,collect and measure data.
For the purpose of compensating for the deficiency of previous study, tests of multivariable regression models and the selection of between Pooled OLS, FEM and REM models will be implemented to increase accuracy valid for the study.
Industry and market factors Bank specific factors
Source: Author's own calculation and visualization
The equation relating the net interest margin to the set of explanatory variables including banking factors and macroeconomic factors is therefore:
NIM i,t = β 0 + β 1 SIZE i,t + β 2 DEP i,t + β 3 CAP i,t + β 4 NI + β 5 NPL + β 6 CI + β 7 LIQ + β 8 IP + β 9 MS + β 10 INF + β 11 GDP + ε i
Firstly, in order to determine the effect of bank size on interest rate margins, the natural log of assets (SIZE) is used to detect the effect of bank-specific characteristics on bank
NIM The influence of deposit growth (DEP) will also be investigated The equity-to- asset ratio is used to determine the relationship between bank capitalization and NIM (CAP) We'll also look at how bank liquidity (LIQ), efficiency (measured by the cost- to- income ratio – CI), and non-interest factors such as non-interest income (NI), implied interest payment (IP) affect to NIM Finally, we will examine the influence of non- performing loan (NPL) on interest margins in this study.
Secondly, in terms of industry-specific variables, we use the market share (MS).
Finally, we use real GDP growth (GDP) and end-of-period inflation rates (INF) to determine the link between macroeconomic parameters and bank NIM.
Defining and Measuring Variables
3.2.1 Dependence variable: The net interest margin
The dependent variable used in this research is the net interest margin (NIM) The key factors of NIM, which measures a bank's ability to function with greater interest rates than interest expenses, were investigated This formula, which is calculated as the ratio of the difference between financial income and expenses to total assets, is commonly applied in last studies including Angori & Gallo (2019), Dien & Nga (2018), Demirgỹỗ (1999). Concerning to financial income, this proxy used interest income that represents how efficiently the bank converts its interest bearing liabilities (deposits) to interest earning assets (loans).
In this research, the effect of bank size is computed by the natural log of total asset Bank size is used to assess how internal factors affect the net interest margin The larger the scale is, the more advantages the bank has in business activities such as easier access to large capital sources at low cost, meeting the borrowing needs of customers and contributing to improve profits of the bank, from which the bank's operating system will be improved In addition, the positive relationship is also explained by economies of scale because when the scale grows to a certain limit, it will bring advantages to the bank in competition as well as in operational efficiency.
However, in previous studies state that the interest margins of larger banks are much lower than those of smaller banks According to their findings, the former pays higher interest on deposits and/or charges lower interest rates on loans As a result, it appears that larger banks generate less on interest income than their smaller counterparts, as they may offer more fee-based services Furthermore, major banks may offer greater deposit rates to take advantage of cross-selling and economies of scale Therefore, result of Hamadi & Awdeh (2012) provides negative effect on the relationship between net interest margin and bank size.
Hypothesis 1: There is a negative correlation between bank size and the net interest margin (H1).
The impact of the financing structure on the net interest margin of commercial banks is calculated by the ratio of customer deposit to total asset.
Deposits are considered a form of official capital mobilization for banks Banks will use this mobilized money to convert into credits for the needs of economic entities At maturity, the bank will have to pay interest based on the total deposits of customers, so it can be considered that the deposit source is a part of the "debt" of the bank The more deposits are converted into loans, the more profits will be earned However, if the deposit is too much, the bank also has to pay the corresponding interest rate, this issue indicates that the bank pays a high rate of interest on these deposits The net interest margin will fall as interest expenses rise.
In previous empirical studies of authors such as Yuksel & Zengin (2017), Memmel & Schertler (2011), used this ratio as an independent variable to measure the impact on the net interest margin of banks and as a result, this ratio has a negative effect.
Hypothesis 2: There is a negative correlation between deposit and the net interest margin (H2).
Capitalization is calculated by equity capital to total asset.
Equity capital plays an important role in banks, it is the bank's own capital, which is initially contributed by the owner and supplemented in the course of business operations. Equity shares perform a number of irreplaceable functions, which is to provide initial resources for the bank to maintain operations when it was first established, to be the basis for creating trust for customers to make transactions and to prevent business risks Equity includes initial capital, additional capital during business operations and funds Equity size is considered as a valuable tool to show the capital status, safety and financial soundness of a bank Well-capitalized banks are considered less risky and are better able to raise – uninsured – funds in order to compensate the drop in deposits (Van den Heuvel, 2002) The existing literature on size effects is quite contradictory While Angori & Gallo (2019) made the result that capitalization significantly and positively affects the net interest margin, the result of Hamadi & Awdeh(2012) illustrated that capitalization is negatively correlated with NIM From previous studies, author confirms the positive effect to NIM of capitalization This is consistent with ideas claiming that well-capitalized banks can charge higher interest rates on loans and pay less interest on deposits because they are less risky of bankruptcy. Hypothesis 3: There is a positive correlation between capitalization and the net interest margin (H3).
NI is calculated by non-interest income to total asset.
Non-interest income and non-interest expenses are calculated from income and expense of business activities other than credit activities Bank also can generate income from non- interest source Some sources of non-interest income are deposit and transaction fees, insufficient funds fees, annual fees, monthly account service charges, inactivity fees, check and deposit slip fees, investment from other financial firms and so on When the net interest margin is reduced, the market's volatility decreases as well As a result, it is thought that when the net interest margin is low, banks aim to generate revenue from non- interest sources.
Therefore, Yuksel & Zengin (2017) concludes that it is projected that NIM and non- interest income will have a negative relationship.
Hypothesis 4: There is a negative correlation between non-interest income and the net interest margin (H4).
NPL is measured as Total non-performing loan to Total Loan Balance.
Non-performing loans includes sub-standard loan, doubtful loan and loan loss (Regulations on classification of assets, level of deduction, method of making provision for risks and use of provisions to handle risks in operations of credit institutions, foreign bank branches, 2013)
This variable is in the CAMELS analysis framework, representing the letter A (Asset
Quality) and is one of the core indicators in the International monetary fund's set of financial strong cold indexes Bad debt is one of the indicators reflecting credit risk in banking operations Bad debt is an inherent disease that has not been resolved and has a potential risk of collapse of the whole credit system The impact of bad debt is quantified by the ratio of bad debt to total outstanding loans, if this ratio is high, it can push the bank to bankruptcy Because non-performing loans are defined as loans that are unable to be repaid by consumers, banks' interest revenue will be lower when the NPL amount is high.
As a result, a negative relationship between net interest margin and NPL amount is projected.
In the study of Yuksel & Zengin (2017), the result show that the bad debt ratio has a negative impact on the net interest margin of commercial banks.
Hypothesis 5: There is a negative correlation between Non-performing loan ratio and the net interest margin (H5).
This ratio is measured by interest cost to interest income is used to analyze the impact of the bank expense efficiency on the net interest margin of commercial banks.
One of the most significant factors of bank expense efficiency is the cost-to-income ratio. Its key components are personnel salaries and overhead expenses incurred in the day-to- day operations of the bank This ratio is used to determine the impact of management efficiency on a bank's profitability A higher ratio shows that management is inefficient. This is the ratio that shows the bank's operating costs, with salaries and benefits being the most important component A lower score suggests greater spending control, which translates to increased bank profitability In previous researches, Hamadi & Awdeh (2012) show the negative impact between cost efficiency and the net interest margin So that this study expect the negative effect between CI and NIM.
Hypothesis 6: There is a negative correlation between cost efficiency ratio and the net interest margin (H6).
Liquidity ratio is measured as high liquidity asset to total asset High liquidity asset includes cash, deposit at SBV, and demand deposits at other credit institutions, trading account securities, valuable papers, and government bonds.
The relationship between liquidity and NIM is computed by the proxy which is the ratio of cash and cash equivalent to total asset Cash and cash equivalents are highly liquid, so the higher the ratio the more the bank's liquidity will be guaranteed, the bank's profit revenue is increased Thus, banks have to pay an extra cost which is interest rates to attract deposits Therefore, the author expects liquidity ratio will have a negative effect on NIM. This expectation is consistent with previous thesis of Hamadi & Awdeh (2012).
Hypothesis 7: There exists a negative correlation between the liquidity and the net interest income (H7).
Implied interest payment calculated by taking non-interest expense minus non-interest income, then divide for total assets.
Non-interest income and non-interest expenses are calculated from income and expense of business activities other than credit activities, such as: service activities, foreign exchange trading, trading activities, trading securities, forex trading activities, investment securities trading activities, other activities, Banks can pay implied interest for customers to encourage customers to deposit money at the bank Implied interest payment can be in the form of bank transactions when banks offered at a cheaper price marginal fees, or promotions savings deposit trade This study also uses the same method to measure implied interest payment In author’s expectation, the cost of interest implicitly has a positive relationship with NIM, because banks will increase NIM to compensate for interest implied interest paid to the customer.
Hypothesis 8: There exists a positive correlation between the implied interest payment and the net interest income (H8).
In the study, this factor is calculated by the formula which is the ratio of bank’s total asset to total asset of 28 selected banks.
There has been a variety of theoretically and empirically research on the market share– profitability relationship According to scholars, market share indicates a financial firm's current competitive position in the marketplace, and firms with large market shares are thought to better serve customers' requirements and hence have a competitive edge over smaller competitors Based on the relative market power hypothesis that an increase in market share can increase the interest rate margin, a positive relationship between market share and the NIM ratio was found in McShane and Sharpe (1985).
Hypothesis 9: There exists a positive correlation between market share and the net interest income (H9).
Inflation is an inherent category of the market economy, it occurs when the requirements of the commodity economic law are not respected, especially the law of money circulation In terms of inflation, higher rates lead to higher inflation expectations and, as a result, a larger inflation risk premium on loans Because banks, like investors, are concerned with rate returns, they will tend to raise lending rates in response to rising inflation As a result, banks will want to hedge against inflation risk by raising their NIM. Therefore, author expects a positive autocorrelation between inflation and NIM This expectation is consistent with Hamadi & Awdeh (2012) and López-Espinosa, Moreno & de Gracia (2011).
Hypothesis 10: There is a positive correlation between inflation and the net interest income (H10).
Method of data collection
The research sample of the thesis is commercial banks in Vietnam Secondary data is collected from the consolidated financial statements, the consolidated audit the annual report in accordance with accounting standards and these reports are presented on the official website of the bank Besides the data, the micro variables are mentioned above, data on macro variables such as GDP growth rate and inflation rate are affected fakes obtained from the Asian Development Bank (ADB) report.
The research is based on the data of 28 commercial banks trade in the period 2009- 2020.
Number Bank’s name Stock Symbol
2 Asia Commercial Joint Stock Bank ACB
3 Vietnam Bank for Agricultural and Rural
4 JSC Bank for Investment and Development of Vietnam
5 Vietnam Export Import Commercial Joint
6 Hochiminh City Housing Development Bank HDB
7 Kien Long Commercial Joint Stock Bank KLB
8 Lien Viet Post Joint Stock Commercial Bank LPB
9 Military Commercial Joint Stock Bank MBB
10 Vietnam Maritime Joint Stock Commercial
14 Orient Commercial Joint Stock Bank OCB
15 Joint Stock Commercial Petrolimex Bank PGB
16 Sai Gon Thuong Tin Commercial Joint
17 Sai Gon Commercial Bank SCB
18 Sai Gon Thuong Tin Bank SGB
19 South East Commercial Bank SeABank
20 Sai Gon – Ha Noi Commercial Joint Stock
23 Vietnam International and Commercial Joint
26 JSC Bank for Foreign Trade of Vietnam VCB
27 Vietnam Joint Stock Commercial Bank for
Source: Author's own calculation and visualization
The selection of commercial banks is based on basis presented below:
• Joint venture banks, foreign banks will not be counted from the date of score
2020 according to the published financial statements.
• The independent and dependent variables must be based on publicly available data and the most complete in the period 2009-2020 by each year.
Some banks weren’t chosen in the thesis due to a number of integers being considered presented below:
• Data from joint venture banks and foreign banks will mostly not be disclosed fully and widely, often the structure will be very different from the financial statements of domestic banks due to the influence of foreign currency overseas parent bank The organization and way of operating are also not consistent with the banks selected for the study.
• Incomplete data of independent and dependent variables of commercial banks will distort with the results of the thesis.
Method of data analysis
In order to be suitable with the research objectives, the analytical steps are performed according to a detailed procedure Stata 13 is the support software as well as the right analysis tool to perform the proposed analysis steps Detailed steps are as follows:
Source: Author's own calculation and visualization
The mathematical operations and statements will be used by the author in Stata 13 software to conduct the most descriptive statistical analysis such as: maximum value, minimum value, mean value, mean position and standard error of the variables mentioned in the model Thereby the author can make appropriate decisions and also filter the research data if necessary through those statistical criteria.
3.4.2 Testing Pool-OLS, FEM, REM models
Panel data regression uses three main methods, namely Pooled-OLS method, fixed effects method (FEM) and random effects method (FEM) The Pooled-OLS method is essentially the use of panel data for analysis in the form of using all the data in a stacked manner and regardless of individual cross-units This is the most common and simplest method, similar to normal OLS analysis, regardless of the spatial and temporal dimensions of the data The Pooled OLS model is detailed as follows: y it =^ 1 + Ị 1 X 1it + /? 2 X 2it + + P k X kit + U it
Where: y it is the dependent variable of observation i in period t, X kit is the independent variable of observation k in period k.
This model has some disadvantages, such as incorrect identification shown in Durbin – Watson (DW) and too tight constraint on cross units, which is unlikely compared to reality Therefore, to overcome the above disadvantages, FEM and REM models are used.
To show the specific effect of each cross-unit on the dependent variable, let the slope change for each unit but the slope coefficient does not change That method is known as the fixed-effects model (FEM) regression method, which means that the intercept can vary across cross units but does not change over time.
With the assumption that each unit has distinct characteristics that can affect the explanatory variables, FEM analyzes this correlation between the residuals of each unit and the explanatory variables, thereby controlling and separating the effects of the individual (time-constant) characteristics out of the explanatory variables so that we can estimate the net effects of the explanatory variable on the dependent variable The FEM model has the following form: y it = C i + Ịx it + +U i
Where y it is the dependent variable of observation i in time t, x it is the independent variable of observe i in time t, C i is the intercept coefficient for each study unit, is the slope coefficient for factors x and U it are residuals.
The difference between the random effects model and the fixed effect model is shown in the variation between units If the variation between units is related to the independent variable - the explanatory variable in the fixed effect model, then in the random effects model the variation between the units is assumed to be random and not correlated. regarding the explanatory variables.
Therefore, if the difference between units has an effect on the dependent variable, then REM will be more appropriate than FEM In which, the residual of each entity (not correlated with the explanatory variable) is considered as a new explanatory variable The idea of REM model also starts from the following model: y it = C i + f)X it +_ +U i
Instead of in the above model, C i is fixed, in REM it is assumed that it is a random variable with mean C 1 and the intercept value is described as follows C i = c + S i (ĩ = 1, …, )
Where: si is a random error with mean 0 and variance ơ 2 Model becomes: y it = c + ^X it + S it + U it hay y it = c + ^X it + W it và W it = S it + U it
Where: £ it is the component error of different objects (different characteristics of each enterprise) and U it is the error of other combined components of both individual characteristics of each object and over time.
Compared with the FEM method, the REM method can overcome all the disadvantages of the FEM method, but REM considers each individual characteristic of the si units to be uncorrelated with the independent variables Therefore, if this happens, the estimated REM is no longer accurate.
First, unnecessary variables will be removed from the model through redundancy tests. Variables that are not statistically significant from the estimation results of the Pooled OLS, FEM and REM models will be selected to be excluded to make the model more suitable Wald test will be used to check the necessity of selected variables with the model. After the variables are eliminated, the model will be regressed by the author with the remaining independent variables, then test the parameters The T-test (T-test) will be conducted to check the fit of the regression coefficients Statistical significance levels at 1%, 5%, 10% will be selected to fit the model.
Heteroskedasticity means that the variances of the residuals are not constant, that is, they differ in different observations Its consequences will lead to problems such as: estimates of variance will be biased, OLS estimates will remain unbiased but no longer efficient. The results will make the regression coefficient tests ineffective Breusch – Pagan test will be conducted for Pooled OLS or FEM model If the model has variable variance, the research model will be overcome by re-estimating the selected model by GLS method If the Random effect model is selected, the topic only tests for multicollinearity and autocorrelation because the Random Effect model does not have a way to test variance. Autocorrelation is the correlation between the components of a chronologically ordered series of observations in time series data, or spatially ordered, for spatial data Some consequences can occur if autocorrelation occurs such as: the variances and standard numbers of the prediction are not efficient, the OLS estimate is still an unbiased linear estimate, sometimes too low compared to with true variance and standard error, leading to exaggeration of the t-ratio but no longer being an efficient estimate, possibly unreliable coefficients of determination and likely to receive high estimates, the tests t and F are unreliable, the usual formula for calculating the variance of error is a biased estimate of the true variance and in some cases seems to be an underestimation of the true variance. Tests based on Durbin – Watson test rules will be conducted in the study If there is an autocorrelation phenomenon, the author decided to choose the remedial variable to estimate based on Durbin - Watson statistics.
Multicollinearity means that two or more explanatory variables in a regression expression have a linear relationship with each other If the variables have a linear relationship, the estimated coefficients and the T-statistic will no longer be reasonable Multicollinearity can lead to the following consequences: OLS estimates and standard errors become very sensitive to changes in the data, the sign of the estimates of the regression coefficients can be wrong, adding or subtracting variables that are collinear with other variables, the coefficients of the remaining variables can vary greatly and change both their variance and covariance of large OLS estimators, about the confidence is large, the T-ratio loses significance, the coefficient of determination is high, but the T-ratio loses significance The two-way multicollinearity test will be carried out in two ways The first way is through correlation coefficient analysis to test for multicollinearity of each pair of independent variables The correlation coefficient (Pearson) is calculated by dividing the covariance of the variables by the product of their standard deviations If the correlation coefficient between the independent variables is greater than 0.8 (also known as the high correlation coefficient), we have high multicollinearity The second way is to test for multicollinearity between one independent variable and the other independent variables by using VIF exaggeration If multicollinearity occurs, the authors will overcome it by removing the independent variable with multicollinearity, this is the simplest way because after removing the independent variable with multicollinearity, the regression coefficients The reduction of the remaining variables from non-zero and non-statistically significant can become statistically significant non-zero.
EMPRICAL RESULT AND DISCUSION
Description statistics
Data for research including 28 banks in the commercial banking system in Vietnam are extracted from financial statements, annual reports, audited management reports, published in the period 2009- 2020 After cleaning the data with EXCEL software, the data will be entered into STATA 13 software for research processing The results of descriptive statistics are shown in the table 4 below:
Table 4: Descriptive statistics of research variables
From chart 1, it can be seen that the average NIM of commercial banks during the study period is approximately 3.07% In addition, the maximum value of NIM is 3.63% in 2012, the smallest value of NIM is -0.79% belongs to Tien Phong Bank in 2011 In the period of
2009 - 2020, the net interest margin of commercial banks in Vietnam has gradual growth rate Specifically, in the period 2009-2012, NIM increased from 2.78% to the highest level of 3.63% and then decreased slightly and from 2012-2014, this rate decreased sharply to a minimum of 2.67% From 2015 onwards, this rate grew again to 2.92% in 2020.
Chart 2: Yearly average of LIQ
Cash and cash equivalents are among the most liquid assets available today, used directly for payment, circulation and storage Through descriptive statistics, it can be seen that cash and cash equivalents of Vietnamese commercial banks in the period2009-2020 reached an average of 4.21% showing that commercial banks reserve assets with little liquidity, leading to an urgent need to arise However, it shows that commercial banks have effectively use and invested capital instead of storing highly liquid asset Besides, the highest value was 5.42% in 2009 and the lowest value was
Chart 3: Yearly average of capitalization
Equity ratio of Vietnamese commercial banks in the period 2009-2020 reached an average value of 9.75%, standard deviation reached 5.48% In general, the capitalization ratio of Vietnamese commercial banks has decreased steadily year by year, from the highest level of 16.61% in 2009 to 7.92% in 2020 Part of the reason is that Vietnam is increasingly integrating with other countries In the world, large commercial banks increased their capital at the same rate too quickly while small-scale commercial banks did not keep up, leading to a slight decrease in the average equity ratio among commercial banks in recent years.
Chart 4: Yearly Average of size of bank
In the period 2009-2020, the largest average value of SIZE ratio is 8.323 in 2020 and the smallest value is 7.495 in 2009, indicating that there has not a large difference between the scale of Vietnamese commercial banks Besides, SIZE variable has mean value of 7.91 and with a standard deviation of 52.62%, it represents for a high dissimilarity in size between banks This is also the reason why banks are now increasingly expanding the scale of operations, changing the image to customers and easily identifying the brand, thereby improving competitiveness and increasing operational efficiency.
Chart 5: Yearly average of non-performing loan
It can be seen that the average NPL of commercial banks during the study period is quite low, only approximately 1.25%%, showing that most commercial banks have a fairly high level of bad debt, with a standard deviation of 1.6%, most commercial banks have no difference in solving bad debt In addition, the maximum average of NPL is 1.6% in 2012, the smallest value of NPL is 1.04% belongs to 2009 In the period of
2009 - 2020, the non-performing loan of commercial banks in Vietnam has a slight growth rate Specifically, in the period 2009-2012, NPL increased from 1.04% to the highest level of 1.6% and then decreased slightly and from 2012-2020, this rate decreased gradually to 1.15% This decrease of NPL rate was good signal of performing of commercial banks in Vietnam.
Chart 6: Yearly average Non-interest income
Chart 6 shows the average NI of commercial banks during the study period is quite low, only approximately 0.17% In addition, the maximum average of NPL is 0.93% in
2019, the smallest value of NPL is 0.55% belongs to 2015 In the period of 2009 -
2011, the non-performing loan of commercial banks in Vietnam has a unstable growth rate Specifically, in the period 2009-2011, NPL increased from 0.59% to 1.6% and then decreased slightly and from 2012-2015, this rate decreased gradually to 1.15% before increase to 0.87% in 2020 This increase illustrated the development of banks services that generate fees and charges nowadays.
Chart 7: Yearly average of Cost efficiency
In the period 2009-2020, the average cost to income ratio of commercial banks in Vietnam reached 64.95% Besides, the largest value is 74.05% in 2011 and the smallest value is 57.41% in 2015 In general, the bank's size through total assets has fluctuate between the average ratio.
Chart 8: Yearly average of implied interest payment
From chart 8, it can be seen that the average implied-interest payment of commercial banks during the study period is approximately 0.65% In addition, the maximum value of IP is 0.86% in 2020, the smallest value of IP is -0.54% belongs to Viet A Bank in
2015 In the period of 2009 - 2020, the IP of commercial banks in Vietnam has slight growth rate Specifically, in the period 2009-205, NIM decreased from 0.84% to the lowest level of 0.39% and then increased and from 2012-2014, this rate increased sharply to a maximum of 0.86% in 2020.
Chart 9: Yearly average of deposit
In the period 2009-2020, the average rate of customer deposits in Vietnamese commercial banks is 62.87% In which, the largest average value is 70.83% in 2016 and the smallest value is 45.26% in 2011 Chart 9 shows the total deposits of customers in commercial banks in the period 2009 – 2020 In general, it can be seen that the ratio of customers' deposits is slightly increasing During the period 2009 – 2011, due to the global financial crisis, customers lacked confidence in commercial banks and that led to a decrease in customer deposit ratio from 57.79% to the lowest level of 48.22% in
2011 However, since 2011, the amount of deposits from customers into the bank increased again and reached 70.24% in 2020.
Chart 10: Yearly average of Inflation
Based on descriptive statistics, the average inflation rate of Vietnam reached 5.91% during the studied period In which, the highest inflation rate reached 18.68% in 2011 and the lowest inflation rate reached 0.88% in 2015.
After the global crisis in 2009, in the situation of the domestic and global economies facing many difficulties, the inflation level was well controlled in our country and was in line with the national economic growth rate This is considered a great achievement in the direction of the macro-economy of our country From 2009 to 2011, domestic inflation continued and increased, reaching 18.13% in 2011 and this is considered the highest growth rate The reason is due to inefficient investment in special services in the state-owned enterprise sector, excessive dependence on credit of the weak financial system, which is dominated by the State Bank From the reasons mentioned above, in the following year, the SBV took the initiative in operating and flexibly using monetary policy tools along with the coordination of fiscal policies dragged down to 6.81% in
Correlation analysis between variables and multicollinearity test
The correlation relationship over time is the relationship between the series of numbers that fluctuate over time, in which there are some series showing the fluctuations of the causal indicators and a series of numbers showing the fluctuations of the indicator result (its fluctuation depends on the fluctuations of the causal indicators) The results of the correlation analysis are shown in Table 5.
According to table 5, most of the correlation coefficients between variables are quite low. The results show that NIM has the strongest correlation with CI (-0.5994) and the weakest correlation with NPL (-0.0058).
For the independent variables, the absolute values of the correlation coefficients of the independent variables are all less than 0.8 According to Farrar & Glauber (1967), if the correlation coefficient between the independent variables is greater than 0.8, the relationship between the pairs of variables will be very close, have mutual influence and the model is likely to have multicollinearity Besides, the smallest correlation coefficient is 0.0001 between GDP and MS The highest correlation coefficient is - 0.6242 betweenCAP and SIZE.
Table 5: Matrix of correlation coefficients between variables
NIM LIQ CAP SIZE NPL NI CI IP DEP INF GDP MS
However, this criterion is often inaccurate and there are cases where the correlation coefficient is quite low but multicollinearity still occurs Therefore, to limit errors as well as ensure the robustness of the model, the thesis will further test it by analyzing the variance exaggeration factor (VIF) in STATA.
Table 6: Test of variance exaggeration factor VIF
In table 6, the results of the VIF indexes are shown all independent variables
chi2 = 0.0092
Breusch – Pagan test is conducted to test FEM model or Pool-OLS model, which is more appropriate for researching factors affecting the net interest margin.
In table 11, Breusch – Pagan test concludes that the Pool-OLS method is not the optimal for NIM model because it has heteroscedasticity error (test value chi2 has p- value = 0.0000 < 0.05) So that FEM is the more suitable model.
Table 11: Breusch – Pagan Test result
Value Chibar2 = 117.12 p-value Prob>chi2 = 0.0000
(Note: YES, NO represent for having and without defects respectively)
After Hausman test and Breusch – Pagan test, FEM model is selected from three models include FEM, REM and Pooled-OLS So that, the Ward test is used to detect heteroskedasticity for FEM model.
In table 12 the heteroscedasticity test result shows that the chi2 test value has p-value chi2 = 0.0000
(Note: YES, NO represent for having and without defects respectively)
From table 13, testing the autocorrelation phenomenon of FEM model, the results of the F test have p-value< 0.05, so it is concluded that FEM model has autocorrelation defects.
Through the above partial test results, it could be concluded that the research model (FEM model) does not have multicollinearity defects However, FEM models occur with autocorrelation and homoskedasticity defects, which will make the estimates obtained by conventional regression on panel data ineffective and the tests no longer valid Therefore, the general least squares (GLS) is used to estimation method to overcome the defects in order to ensure that the obtained estimates are stable and efficient (Wooldridge, 2002).
4.3.8 Regression results of research model according to GLS:
Table 14: Regression results of research model NIM
Table 14 is the result of regression model explaining the impact of factors affecting the net interest margin of Vietnamese commercial banks measured by GLS method.
Regarding model result, the variables including NPL (Non-performing loan), NI (Non- interest income), IP (Implied interest payment), GDP (Gross domestic product) has no statistical significance explaining the impact on the change of NIM.
NIM i,t = 0.0355 + 0.0293 LIQ + 0.0364 CAP + 0.0036 SIZE - 0.0785 CI – 0.013DEP + 0.0973 INF - 0.0631 MS+ ε i,t