INTRODUCTION
Motivation
Over the past several decades, numerous empirical studies have examined the impact of credit risk on bank performance, particularly in commercial banks where credit operations pose significant risks Credit risk arises when borrowers default on loan repayments, potentially jeopardizing the bank's operations and leading to bankruptcy This risk serves as a critical indicator of a bank's financial health (Duttweiler, 2011) and poses a threat not only to individual institutions but also to the stability of the entire banking system (Eichberger and Summer, 2005).
Numerous studies globally have explored the connection between credit risk and bank performance, revealing mixed results While some research indicates a positive correlation between credit risk and bank profitability, others suggest a negative impact Notably, regression analysis of panel data demonstrates that credit risk positively influences bank performance in Ghana, highlighting that local banks remain profitable even with elevated credit risk levels Boahene (2012) examined this relationship among six commercial banks in Ghana from 2005 to 2009, using return on equity (ROE) as the key dependent variable.
A study conducted by Huynh Thi Huong Thao in 2018 analyzed financial risk factors impacting the performance of 35 Vietnamese commercial banks from 2008 to 2017, utilizing Data Envelopment Analysis (DEA).
The study utilizes the ratio of bad debt to total outstanding loans and the ratio of provision for credit risk to total outstanding loans as independent variables Findings indicate that the credit risk measurement variable positively influences the profitability of Vietnamese commercial banks.
Numerous studies have explored the efficiency of banking operations, including works by Nguyen Thanh Dat (2021), Alshatti (2015), and Saeed & Zahid (2016), as well as Hamza (2017) However, there is a notable gap in research concerning the impact of credit risk on banking operations across Southeast Asia Consequently, it is essential to verify research findings to enhance their reliability.
The global financial system is primarily dominated by banks, which play a crucial role in the economies of countries worldwide The financial crisis that began in late 2007 in the US significantly impacted the global economy, severely damaging the reputation of commercial banks Consequently, numerous banks faced bankruptcy as they prioritized immediate profits over the operational safety of the banking system This situation has raised concerns about the effectiveness of risk management mechanisms, particularly the underestimation of credit risk.
In response to the financial crisis of 2007-2008, the Basel Committee on Banking Supervision (BCBS) introduced the Basel III accord to improve coordination, supervision, and risk management within the banking sector This framework presents innovative approaches to capital, leverage, and credit, aimed at enhancing regulatory measures for better oversight and risk management in the industry.
Credit risk and effective credit risk management are vital for the stability of financial markets and the banking industry They help mitigate potential losses from borrower defaults, ensuring the soundness of financial institutions By assessing creditworthiness, banks can make informed lending decisions, which enhances overall market confidence Furthermore, robust credit risk management practices contribute to regulatory compliance and promote financial stability Effective management of credit risk also enables banks to optimize their capital allocation, ultimately supporting economic growth Overall, understanding the significance of credit risk is essential for the sustainable operation of the financial sector.
Southeast Asian banks, while not severely impacted by the recent financial crisis, raise questions about their safety and efficiency, potentially benefiting from limited integration into international finance In Vietnam, the crisis prompted notable changes in the banking sector, enhancing internal governance, organizational structure, technology application, and the development of modern banking services However, the broader economic instability has led to significant losses across the banking system, ultimately affecting the economy and resulting in serious consequences.
2012) In the integration trend, the requirement to manage credit risks and ensure the efficiency of banking performance is more necessary than ever
Vietnam, despite being one of the Southeast Asian countries with low per capita income, has a high number of banks but lacks a leading pillar bank to compete regionally (Nguyen Cong Tam and Nguyen Minh Ha, 2012) This study aims to analyze the impact of credit risk on bank performance in Southeast Asia from 2010 to 2021, thereby assessing how credit risk influences the efficiency of banking operations in Vietnam.
This study investigates the "Impact of Credit Risk on Commercial Bank Performance" in Southeast Asian countries from 2010 to 2021, aiming to address existing research gaps By analyzing and comparing case studies across the region, the author seeks to provide valuable insights into the relationship between credit risk and the performance of commercial banks.
Vietnam plans to propose policy recommendations aimed at enhancing banking performance by addressing credit risk This dissertation offers empirical evidence and valuable insights into the relationship between credit risk and banking efficiency, thereby supporting informed policy development.
Research objectives and questions
This thesis aims to identify the factors influencing credit risk and assess its impact on bank performance, focusing on Vietnam from 2011 to 2021.
On that basis, the specific objectives of the study are determined to be:
Analysis of the influence of credit risk on performance of commercial banks, case studies of Southeast Asian nations and Vietnam
Suggest policies on credit risk management to perform banking in Vietnam more effective
The study focuses on answering the following questions:
Is there any difference in the results of research on factors affecting bank credit risk in Vietnam and other Southeast Asian countries?
How does credit risk affect banking performance in Southeast Asian countries?
What are the policy recommendations related to risk management and ensuring the efficiency of banking performance in Vietnam?
Object and scope of research
Object subject: Impact of credit risk and banking performance, a case study of Southeast Asian countries
This research expands its scope beyond individual countries, analyzing nine Southeast Asian nations—Cambodia, Indonesia, Laos, Malaysia, Myanmar, the Philippines, Singapore, Thailand, and Vietnam—over the period from 2010 to 2021.
The research utilizes data from BankFocus and Osiris, two reputable global banking data sources, ensuring uniformity and enhancing reliability in assessing the impact of credit risk on banking performance.
Southeast Asia, including Vietnam, is experiencing rapid international economic integration, necessitating swift reforms within the banking system To enhance the role of banks in driving economic transformation and preparing for financial liberalization, it is crucial to improve their operational capacity and competitiveness Additionally, refining the policy framework for the banking sector during this transformative period is essential for achieving these goals.
Recent studies have explored the relationship between credit risk and the profitability of commercial banks However, there is a lack of research specifically examining how credit risk affects various financial indicators of commercial banks in Southeast Asia from 2010 to 2021 This gap highlights the importance of conducting a comprehensive analysis of the overall performance of these banks and the significant impact of credit risk on their financial outcomes in the region.
9 results will support the bank administrators in making decisions to manage credit activities of Vietnamese commercial banks in the integration process.
Research Methodology
To accomplish the research objectives, the thesis has inherited the research model of Serwadda
(2018) and Dat, N T., & Duyen, T T M (2021) to analyze the impact of credit risk on banking performance
The model employs estimation methods such as Ordinary Least Squares (OLS), Random Effects Model (REM), and Fixed Effects Model (FEM) to analyze the impact of various factors on profit ratios, specifically Return on Assets (ROA), Return on Equity (ROE), and Net Interest Margin (NIM).
Collected data includes the following datasets:
The data source for banking information in Southeast Asia is derived from Osiris and Banksfocus, which includes banking data from nine countries in the region Notably, Singapore's financial reporting and macro data were excluded, along with Brunei and East Timor due to the absence of banking and financial statements Consequently, there are a total of 58 commercial banks across these nine Southeast Asian nations.
Dissertation structure
The content of the dissertation consists of 5 main parts, specifically as follows:
This chapter provides a comprehensive overview of the study, highlighting the necessity of research to pinpoint existing gaps It outlines the research objectives, questions, and scope, while also detailing the data sources, research methods, and overall structure of the thesis.
LITERATURE REVIEW
Credit risk
Credit risk is a critical concern for commercial banks, as effective management of this risk enhances both the sustainability and profitability of their operations By addressing credit risk, banks play a vital role in promoting economic stability and ensuring efficient capital distribution within the economy (Psillaki et al., 2010).
According to the Basle Committee on Banking Supervision and Bank for International Settlements
Credit risk, as defined in 2000, refers to a customer's capacity to borrow funds while failing to meet their repayment obligations to the bank According to Boffey and Robson (1995), credit risk represents the most significant threat to bank performance, highlighting the importance of credit value as a key indicator of a bank's financial health.
According to Circular 02/2013/TT-NHNN issued by the State Bank of Vietnam on January 21, 2013, the categorization of assets, deduction levels, and risk preparation mechanisms for credit institutions and foreign bank branches are defined This regulation highlights that credit risk in banking refers to potential losses arising from customers' failure or inability to fulfill their financial obligations.
Credit risk is evaluated using the bad debt ratio, which compares bad debt to total outstanding debt (Ongore and Kusa, 2013; Pham Huu Hong Thai, 2014) Research, including the work of Heffenan and Fu (2008), has also utilized the ratio of credit provisions to total loans to gauge credit risk, highlighting the prudence of banks in managing overdue loans.
The ratio of provision expense for credit risks to total loans serves as a critical indicator for evaluating a bank's credit risk in relation to its profitability throughout the year (Pham Huu Hong Thai, 2014).
High non-performing loan (NPL) ratios in commercial banks negatively impact the banking industry, as identified by Nair and Fissha (2010) Elevated bad debt ratios, which indicate increased credit risk, can detrimentally affect banks' financial performance and raise the likelihood of financial crises A significant level of credit risk correlates with higher default risk, ultimately jeopardizing depositors' interests (Bizuayehu, 2015) Thus, effective credit risk management is essential for banks, enhancing both sustainability and profitability while ensuring efficient capital allocation and overall economic stability (Psillaki et al., 2010) Additionally, factors such as capital efficiency (CEE), human resource efficiency (HCE), and capital structure (SCE) positively influence banks' financial performance (Le Hong Nga & Nguyen Thanh Dat, 2021).
2.1.2 The factors affecting credit risk
The traditional method of measuring credit risk involves various indicators, including overdue debt coefficients, bad debt ratios, and capital loss risk coefficients, with bad debt being the most prevalent measure According to Decision No 493/2005/QD-NHNN, bad debts, classified as non-performing loans (NPLs), fall into groups 3, 4, and 5 Group 3 consists of sub-standard debts that are overdue between 90 and 180 days, while Group 4 includes doubtful debts that are delinquent from 181 to 360 days Group 5 encompasses potentially losing debts that have been overdue for more than 360 days.
Credit risk is assessed not only through traditional measurement methods but also by evaluating loan provisions, which represent a bank's expense.
The risk provision, set at 13, is allocated to cover potential losses from the bank's loans, utilizing a comprehensive classification of debt groups that extends beyond just bad debts This method enhances the overall measurement of risk From a research standpoint, the loan provision ratio offers a more reliable indicator for assessing credit risk, as it is prominently reflected in the bank's financial statements, making it a more dependable figure than the bank's reported bad debt ratio.
Research indicates a positive correlation between loan loss provisions and non-performing loan (NPL) ratios, suggesting that higher provisions indicate lower loan quality and increased credit risk (Gueyie & Ortiz, 2000) This metric has gained traction in credit risk studies, with Tsolas and Charles (2015) highlighting its relevance Knaup and Wagner (2012) evaluated credit risk using various criteria, including loan provisions and debt ratios, finding that loan provision costs significantly affect banking performance compared to other risk measures However, some researchers, like Podpiera and Weill (2008), criticize the use of provision ratios as an accurate risk indicator, noting that it relies on estimates and varies according to a bank's risk management policies.
Previous studies (Bonfim and Kim, 2014; Trenca, Petria, and Corovei, 2015) have assessed credit risk through the credit provision ratio relative to the total loan amount Research indicates that banks with a significant proportion of loans, lower liquidity, or weaker capital structures face heightened credit risk (Bonfim and Kim, 2014) This dissertation aims to expand on existing findings by utilizing three key variables to measure a bank's credit risk: the non-performing loan ratio (NPLR), loan loss provision ratio (LLPR), and loan deposit rate (LDR).
Performance of commercial banks
Banking performance is primarily assessed through profitability, with studies focusing on bank profitability grounded in two key theories: market power theory (MP) and structural efficiency theory (ES).
2.2.1.1 The theory of market power
The theory of market power (MP) encompasses two main approaches: the Structure-Behavior-Performance (SCP) theory and the relative market power theory The SCP theory posits that market structure influences enterprise behavior, which in turn affects market outcomes such as profits, technological advancement, and growth In concentrated sectors, this often leads to negative economic effects, including reduced output and monopolistic pricing (Bain, 1951) According to the SCP hypothesis, a more concentrated banking industry results in higher loan rates and lower deposit rates due to diminished competition.
Meanwhile, RMP (Relative market power) theory implies that enterprises with substantial market shares and distinctive goods might exert market power and make non-competitive profits (Berger,
In concentrated industries, such as banking, major brands can leverage their market power to increase prices on products and services, ultimately enhancing profitability Research indicates a positive correlation between market dominance and profits, suggesting that as market power increases, so does the efficiency and scale of operations within companies Consequently, larger banks tend to experience greater profitability due to these scale efficiencies.
The efficient structure theory (ES), proposed by Demsetz in 1973, suggests that banks achieve greater profitability and market share through enhanced governance rather than merely competitive market forces Anyanwaokoro (1996) emphasizes the importance of profitability in attracting depositors, noting that strong profit figures can reassure stakeholders—including investors, borrowers, and employees—while also mitigating the risk of financial crises Consequently, higher profits lead to more stringent cost risk controls within banks.
Efficient structure theory (ES) posits that the relationship between market structure and firm performance is shaped by the firm's performance itself, indicating that banks are more profitable due to their efficiency (Olweny and Shipo, 2011) ES theory is often presented in two distinct ways, depending on the type of performance being evaluated, suggesting that more efficient firms typically achieve better outcomes.
Implementing the X-Efficiency strategy allows firms to achieve larger market shares and increased profits by reducing production costs across all output levels (Al-Muharrami and Matthews, 2009) This concept is closely related to the Scale-Efficiency method, where economies of scale enable larger banks to operate with lower production costs, ultimately leading to enhanced profitability (Olweny and Shipo, 2011).
The market power theory (MP) posits that a bank's profitability is influenced by external market forces, while the efficient structure theory (ES) asserts that a bank's performance is primarily driven by internal factors, including management decisions and operational efficiency.
2.2.2 The concept of commercial bank business efficiency
Classical economists assert that firms aim to maximize profits in perfectly competitive markets, while in monopolistically competitive markets, businesses focus on revenue maximization and cost minimization to enhance profitability Profitability serves as a key indicator of business efficiency, with Farrell (1957) defining efficiency as a unit's capacity to maximize output relative to input costs Systems theory further elaborates on efficiency in two dimensions: the ability to manage profitability and reduce costs for improved competitiveness among financial institutions, and the optimal combination of inputs to generate output effectively.
When evaluating the business efficiency of an enterprise, it can be based on two criteria that are absolute efficiency and relative efficiency
Absolute efficiency is defined as the difference between business results and the costs incurred to achieve those results, reflecting the scale, volume, and profit under specific conditions, time, and place However, this indicator may not be easily comparable among similarly sized enterprises with long-term strategies, as it does not effectively demonstrate the absolute utilization of resources when assessing long-term business performance across organizations.
Relative efficiency is determined by the comparative ratio of output to input factors, defined as Efficiency = output/input or Efficiency = input/output This metric is particularly useful for comparing organizations of varying sizes, spaces, and timeframes.
According to Rose (2002), evaluating a bank's business performance involves both theoretical foundations and the unique characteristics of commercial banks In a narrow perspective, this performance is defined by the bank's ability to generate profits while maintaining operational safety Conversely, a broader view encompasses not only profit generation but also the structure of liabilities and assets, alongside a consistent trend of profit growth Key resources, including labor, facilities, and financial assets, play a crucial role in assessing the efficiency and factors influencing banking performance, particularly in core activities such as deposit acceptance, lending, and investment.
In banking performance studies, various approaches are utilized to analyze banks' roles Some researchers adopt a production perspective, viewing banks as production units (Benston, 1965; Shaffnit et al., 1997), while others consider them as financial intermediaries (Sealey and Lindley, 1977; Maudos and Pastor, 2003) Additionally, a modern approach recognizes that banks fulfill both functions (Denizer et al., 2000; Athanassopoulos and Giokas, 2000).
The productive approach, as proposed by Benston in 1965, views banks as service providers to their customers, where inputs encompass physical elements such as labor, materials, and information systems The output of this model is defined by the financial services offered, quantified through metrics like the volume and types of transactions, as well as the number of documents processed over time.
In efficiency studies of specific banks, transaction flow data has been substituted with metrics on the number of deposit and loan accounts, serving as an alternative indicator of service levels provided.
The intermediary approach, as outlined by Sealey and Lindley (1977), posits that commercial banks serve as vital intermediaries between savers and investors by mobilizing capital and investing in productive assets like loans and securities This perspective considers operating costs and interest as inputs, while categorizing loans and large assets as outputs However, there is ongoing debate regarding the classification of deposits, with Mester (1987) asserting that bank assets are outputs, whereas deposits, capital, and labor function as inputs Interest income emerges as a crucial component of a bank's profitability, highlighting the importance of credit development When viewed as a product, bank loans are priced based on the loan interest rate, with deposits from capital owners serving as the primary source of borrowed capital, thereby acting as inputs in the loan production process.
RESEARCH METHODOLOGY
Research Methodology
This study utilizes Panel data analysis through Stata software to investigate the impact of credit risk on bank performance across Southeast Asian countries, including Vietnam, from 2010 to 2021 It analyzes data from 58 banks in the region, comprising a balanced panel that integrates cross-sectional and time-series components This dual data structure offers unique advantages and challenges in assessing changes over time and differences among the banks To address these challenges, the research employs various econometric models, including Pooled OLS, Fixed Effect Model (FEM), and Random Effect Model (REM).
3.1.1 Pooled OLS classical regression model with panel data
Research worldwide, including in Vietnam, typically utilizes cross-sectional data and the multiple regression method based on ordinary least squares (OLS) to evaluate the relationship between dependent and independent variables This reliable approach is widely used to estimate linear connections Previous studies have investigated credit risk and banking performance in Europe using the standard pooled OLS regression model (Roman and Sargu, 2015; Bassey and Moses, 2015).
Cross-data OLS estimation is limited by constant coefficients, leading to inadequate model outputs in real-world scenarios Current research on liquidity, capital, and banking performance predominantly utilizes panel data over cross-sectional data For instance, Petria et al (2015) employed Random Effects Model (REM) and Fixed Effects Model (FEM) to examine data from 27 European countries between 2004 and 2011, revealing a positive relationship between capital and banking efficiency Similarly, Goddard et al (2004) found a favorable correlation between capital/assets and banking performance through a dynamic panel data model, analyzing data from 665 banks across six European nations from 1992 to 1998 Furthermore, earlier studies (Iannotta et al., 2007; Berger and Bouwman, 2013) demonstrate that banks with higher equity enhance safety, maintain market share, and increase profits during crises.
3.1.2 Fixed effects model (FEM) and random effects model (REM)
The Fixed Effect Model (FEM) focuses on individual changes that influence the model, eliminating autocorrelation This approach utilizes fixed factors to analyze the model's impact, functioning as an Ordinary Least Squares (OLS) model with dummy variables.
28 dummy variable serves as the fixed factors This strategy, however, has the disadvantage of lowering the model's degrees of freedom, especially when the number of dummy variables is significant
The Random Effects Model (REM), also referred to as the Error Components Model (ECM), focuses on individual differences over time and their impact on the analysis While autocorrelation can be a potential concern, this model effectively addresses variable variance, making it a valuable tool in statistical analysis.
Panel data research enhances observation numbers and addresses multicollinearity by incorporating more information and analyzing cross-unit behavior over time This study utilizes panel data alongside two estimating models: the fixed effects model (FEM) and the random effects model (REM), which effectively mitigate the limitations of cross-sectional data By examining the impact of various variables on credit risk, both models accommodate variations among different banks Previous studies by Sufian and Chong (2008) and Anbar and Alper (2011) employed OLS, FEM, and REM to analyze research data, building on established findings.
Model selection using the F test: The F test is used to decide between the Pooled Regression model (POOLED) and the Fixed Effect model (FEM)
The Hausman model selection test: Following estimate, use the Hausman test to select between the fixed effect model and the random effect model
To decide between the Pooled Regression model (POOLED) and the Random Effect model (REM), the Breusch-Pagan test was conducted to assess whether the individual effects are random This test helps in determining the most suitable model for the data analysis.
ESTIMATION RESULTS AND MAJOR FINDINGS
Descriptive statistics
Descriptive statistics in research encompass various aspects such as data collection, summarization, and presentation, aimed at accurately reflecting the studied subjects The accompanying table outlines key results for specific variables, including the total number of observations, variable names, mean values, standard deviations, and the minimum and maximum values, providing a comprehensive overview of the study's variables.
Variable Obs Mean Std Dev Min Max
The average return on equity (ROE) for commercial banks in Southeast Asia is 14.37%, with a minimum of 0.04% and a maximum of 51.88% The standard deviation of 8.02% indicates significant variability in ROE among these banks This disparity can be attributed to differences in operational scale, equity, and other factors among the commercial banks in the region.
The average Net Interest Margin (NIM) stands at a low 3.71%, with a standard deviation of just 1.85%, indicating a similarity in performance among banks The non-performing loan ratio (NPLR) averages 4.95%, exceeding the State Bank of Vietnam's target of 3%, suggesting that while bad debt levels vary across countries, they are not excessively high In Southeast Asia, the bad debt management, particularly in Vietnam, remains a challenge.
The loan provision ratio (LLPR) averages 0.98%, with a standard deviation of 1.09% The provision rate for credit risk ranges from a minimum of 0% to a maximum of 13.13%.
The Loans to Deposits Ratio (LDR) stands at 72.91%, indicating that banks maintain a lower loan ratio compared to their deposits Additionally, a standard deviation of 18.65% reflects a significant variation in LDR ratios across different banks.
Correlation matrix
ROA ROE NIM NPLR LLPR LDR
The correlation coefficient matrix reveals a significant positive correlation between the independent variable LDR and the dependent variable, with a coefficient of 0.1371 and a p-value of 0.0003 Conversely, the variables NPLR and LLPR exhibit negative correlations with the dependent variable, but these results do not achieve statistical significance.
The analysis reveals significant correlation coefficients among the independent variables, with the strongest positive correlation of 0.3395 between LLPR and NIM, followed by a positive correlation of 0.3280 between LDR and NIM Additionally, there is a positive correlation of 0.1280 between LDR and ROE, while a negative correlation of -0.1310 exists between LLPR and ROE To further investigate the presence of high correlation relationships among these independent variables, the subsequent section will address the multicollinearity test for the research model.
OLS, FEM and REM model regression results
In this thesis, the author uses "three typical regression models of panel data, namely Pooled OLS, FEM and REM
Table 4: Regression model OLS, FEM, REM dependent variable ROA
Note: FEM - Fixed effects model, REM - Random effects model;
Significance of the symbol*: *p 0.05 at the 5% significance level, so we accept the hypothesis H0 and conclude that there is no autocorrelation in the ROA model
Table 15: Wooldridge test for ROE
The results from Table 15 indicate a significant finding, with Prob > F equaling 0.0000, which is less than the 0.05 threshold at the 5% significance level Consequently, we reject the null hypothesis (H0) and accept the alternative hypothesis (H1), confirming the presence of autocorrelation in the ROE model.
Table 16: Wooldridge test for NIM
The test results from Table 16 show that, Prob > F = 0.000 < 0.05 at the 5% significance level, so the hypothesis H0 is rejected, the hypothesis H1 is accepted: there is autocorrelation in the NIM model
After conducting relevant tests, the author concludes that the research model's dependent variables—Return on Assets (ROA), Return on Equity (ROE), and Net Interest Margin (NIM)—do not exhibit significant multicollinearity However, the analysis reveals issues with variable variance and autocorrelation To address these challenges, the author will employ the Random Effects Model (REM) and utilize Feasible Generalized Least Squares (FGLS) estimation methods to effectively mitigate autocorrelation and variance in the research model.
FGLS test (Feasible generalized least squares)
Variance and autocorrelation are common issues that can bias model results, prompting researchers to utilize the Feasible Generalized Least Squares (FGLS) method to identify optimal models By addressing these errors, the FGLS model enhances the accuracy of the impact of independent variables, leading to improved estimated results after correcting regression violations.
49 model with the REM and FEM estimation methods by the FGLS method are presented in the following table:
Table 17: FGLS test for ROA, ROE and NIM
Significance of the symbol*: *p|t| [95% Conf Interval]
Total 2123.09866 695 3.05481821 Root MSE = 1.7291 Adj R-squared = 0.0213 Residual 2068.89125 692 2.98972725 R-squared = 0.0255 Model 54.2074026 3 18.0691342 Prob > F = 0.0005 F(3, 692) = 6.04 Source SS df MS Number of obs = 696 reg ROA NPLR LLPR LDR
F test that all u_i=0: F(57, 635) = 42.20 Prob > F = 0.0000 rho 77863647 (fraction of variance due to u_i) sigma_e 8248692 sigma_u 1.5470324
_cons 8321748 1954917 4.26 0.000 4482863 1.216063 LDR 0133957 0026078 5.14 0.000 0082747 0185167 LLPR -.0981423 0359125 -2.73 0.006 -.1686638 -.0276207 NPLR -.0038086 0023813 -1.60 0.110 -.0084848 0008677 ROA Coef Std Err t P>|t| [95% Conf Interval] corr(u_i, Xb) = 0.0010 Prob > F = 0.0000 F(3,635) = 12.29 overall = 0.0254 max = 12 between = 0.0174 avg = 12.0 within = 0.0549 min = 12 R-sq: Obs per group:
Group variable: BankID Number of groups = 58Fixed-effects (within) regression Number of obs = 696 xtreg ROA NPLR LLPR LDR,fe
ROE Model rho 78394114 (fraction of variance due to u_i) sigma_e 8248692 sigma_u 1.5712337
_cons 8331254 2815847 2.96 0.003 2812294 1.385021 LDR 0133884 0025626 5.22 0.000 0083658 018411 LLPR -.0986417 0356153 -2.77 0.006 -.1684464 -.028837 NPLR -.003794 0023448 -1.62 0.106 -.0083897 0008018 ROA Coef Std Err z P>|z| [95% Conf Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(3) = 38.00 overall = 0.0254 max = 12 between = 0.0174 avg = 12.0 within = 0.0549 min = 12 R-sq: Obs per group:
Group variable: BankID Number of groups = 58Random-effects GLS regression Number of obs = 696 xtreg ROA NPLR LLPR LDR,re
_cons 10.73935 1.22489 8.77 0.000 8.334406 13.1443 LDR 0641345 0162316 3.95 0.000 0322654 0960036 LLPR -1.109574 2766371 -4.01 0.000 -1.652723 -.5664255 NPLR 010297 0153591 0.67 0.503 -.0198591 0404531 ROE Coef Std Err t P>|t| [95% Conf Interval]
Total 44753.1854 695 64.3930725 Root MSE = 7.8837 Adj R-squared = 0.0348 Residual 43009.4487 692 62.1523825 R-squared = 0.0390 Model 1743.73669 3 581.245563 Prob > F = 0.0000 F(3, 692) = 9.35 Source SS df MS Number of obs = 696 reg ROE NPLR LLPR LDR
F test that all u_i=0: F(57, 635) = 11.44 Prob > F = 0.0000 rho 49561603 (fraction of variance due to u_i) sigma_e 5.7804324 sigma_u 5.7299702
_cons 9.157392 1.369947 6.68 0.000 6.467219 11.84757 LDR 0950007 0182748 5.20 0.000 0591144 130887 LLPR -1.657164 2516636 -6.58 0.000 -2.151357 -1.16297 NPLR -.0162787 0166877 -0.98 0.330 -.0490484 0164911 ROE Coef Std Err t P>|t| [95% Conf Interval] corr(u_i, Xb) = -0.1568 Prob > F = 0.0000 F(3,635) = 23.54 overall = 0.0364 max = 12 between = 0.0013 avg = 12.0 within = 0.1001 min = 12 R-sq: Obs per group:
Group variable: BankID Number of groups = 58Fixed-effects (within) regression Number of obs = 696 xtreg ROE NPLR LLPR LDR,fe
NIM Model rho 46912612 (fraction of variance due to u_i) sigma_e 5.7804324 sigma_u 5.4338727
_cons 9.461249 1.479492 6.39 0.000 6.561498 12.361 LDR 0896415 017282 5.19 0.000 0557695 1235136 LLPR -1.592808 2460242 -6.47 0.000 -2.075007 -1.110609 NPLR -.0114744 015894 -0.72 0.470 -.0426261 0196774 ROE Coef Std Err z P>|z| [95% Conf Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(3) = 67.51 overall = 0.0370 max = 12 between = 0.0015 avg = 12.0 within = 0.1000 min = 12 R-sq: Obs per group:
Group variable: BankID Number of groups = 58Random-effects GLS regression Number of obs = 696 xtreg ROE NPLR LLPR LDR,re
_cons 1.131451 2592003 4.37 0.000 622538 1.640365 LDR 0286669 0034348 8.35 0.000 021923 0354107 LLPR 5190042 0585395 8.87 0.000 4040679 6339406 NPLR -.0023983 0032502 -0.74 0.461 -.0087797 0039831 NIM Coef Std Err t P>|t| [95% Conf Interval]
Total 2403.34113 695 3.4580448 Root MSE = 1.6683 Adj R-squared = 0.1952 Residual 1925.93177 692 2.78313839 R-squared = 0.1986 Model 477.409367 3 159.136456 Prob > F = 0.0000 F(3, 692) = 57.18 Source SS df MS Number of obs = 696 reg NIM NPLR LLPR LDR
F test that all u_i=0: F(57, 635) = 44.20 Prob > F = 0.0000 rho 80076451 (fraction of variance due to u_i) sigma_e 78140792 sigma_u 1.5665594
_cons 1.600065 1851915 8.64 0.000 1.236403 1.963726 LDR 0278413 0024704 11.27 0.000 0229901 0326925 LLPR 1124194 0340203 3.30 0.001 0456136 1792253 NPLR -.0043427 0022559 -1.93 0.055 -.0087726 0000871 NIM Coef Std Err t P>|t| [95% Conf Interval] corr(u_i, Xb) = 0.0976 Prob > F = 0.0000 F(3,635) = 48.58 overall = 0.1469 max = 12 between = 0.1427 avg = 12.0 within = 0.1867 min = 12 R-sq: Obs per group:
Group variable: BankID Number of groups = 58 Fixed-effects (within) regression Number of obs = 696
xtreg NIM NPLR LLPR LDR,fe
Model ROA rho 75991894 (fraction of variance due to u_i) sigma_e 78140792 sigma_u 1.3902167
_cons 1.570622 2602324 6.04 0.000 1.060576 2.080668 LDR 0280218 0024508 11.43 0.000 0232183 0328253 LLPR 1284414 034105 3.77 0.000 0615968 1952859 NPLR -.004228 0022431 -1.88 0.059 -.0086244 0001685 NIM Coef Std Err z P>|z| [95% Conf Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(3) = 153.46 overall = 0.1518 max = 12 between = 0.1500 avg = 12.0 within = 0.1864 min = 12 R-sq: Obs per group:
Group variable: BankID Number of groups = 58Random-effects GLS regression Number of obs = 696 xtreg NIM NPLR LLPR LDR,re
Test: Ho: difference in coefficients not systematic
B = inconsistent under Ha, efficient under Ho; obtained from xtreg b = consistent under Ho and Ha; obtained from xtreg LDR 0133957 0133884 7.31e-06 0004835
Test: Ho: difference in coefficients not systematic
B = inconsistent under Ha, efficient under Ho; obtained from xtreg b = consistent under Ho and Ha; obtained from xtreg LDR 0950007 0896415 0053592 0059414
Test: Ho: difference in coefficients not systematic
B = inconsistent under Ha, efficient under Ho; obtained from xtreg b = consistent under Ho and Ha; obtained from xtreg
ROA[BankID,t] = Xb + u[BankID] + e[BankID,t]
Breusch and Pagan Lagrangian multiplier test for random effects
ROE[BankID,t] = Xb + u[BankID] + e[BankID,t]
Breusch and Pagan Lagrangian multiplier test for random effects
4.5 FGLS test (Feasible generalized least squares)
H0: sigma(i)^2 = sigma^2 for all i in fixed effect regression model
Modified Wald test for groupwise heteroskedasticity
Wooldridge test for autocorrelation in panel data
xtserial ROA NPLR LLPR LDR
Wooldridge test for autocorrelation in panel data
xtserial ROE NPLR LLPR LDR
Wooldridge test for autocorrelation in panel data
xtserial NIM NPLR LLPR LDR
_cons 7237457 1136324 6.37 0.000 5010303 9464611 LDR 0114673 0014543 7.88 0.000 0086169 0143178 LLPR -.0512662 0261939 -1.96 0.050 -.1026053 0000729 NPLR -.0034589 0008363 -4.14 0.000 -.0050981 -.0018197 ROA Coef Std Err z P>|z| [95% Conf Interval]
Prob > chi2 = 0.0000 Wald chi2(3) = 94.45 Estimated coefficients = 4 Time periods = 12 Estimated autocorrelations = 0 Number of groups = 58 Estimated covariances = 58 Number of obs = 696
Cross-sectional time-series FGLS regression
xtgls ROA NPLR LLPR LDR,panels(hete)
_cons 12.9293 1.502884 8.60 0.000 9.983699 15.8749 LDR 0424065 0189338 2.24 0.025 0052969 079516 LLPR -.9012707 1859082 -4.85 0.000 -1.265644 -.5368973 NPLR -.0165163 012167 -1.36 0.175 -.0403632 0073306 ROE Coef Std Err z P>|z| [95% Conf Interval]
Prob > chi2 = 0.0000 Wald chi2(3) = 32.11 Estimated coefficients = 4 Time periods = 12 Estimated autocorrelations = 1 Number of groups = 58 Estimated covariances = 1 Number of obs = 696
Correlation: common AR(1) coefficient for all panels (0.7510)
Cross-sectional time-series FGLS regression
xtgls ROE NPLR LLPR LDR,panels(i) corr(a)
_cons 2.281363 141576 16.11 0.000 2.003879 2.558847 LDR 0201164 0018904 10.64 0.000 0164113 0238215 LLPR 1117007 0221088 5.05 0.000 0683682 1550332 NPLR -.0032436 0013523 -2.40 0.016 -.0058941 -.0005931 NIM Coef Std Err z P>|z| [95% Conf Interval]
Prob > chi2 = 0.0000 Wald chi2(3) = 140.36 Estimated coefficients = 4 Time periods = 12 Estimated autocorrelations = 58 Number of groups = 58 Estimated covariances = 58 Number of obs = 696
Cross-sectional time-series FGLS regression
xtgls NIM NPLR LLPR LDR,panels(h) corr(p)