An Empirical Study of the Determinants of Non-performing Loan of State Owned Commercial Banks of Bangladesh

Abstract

The issue of nonperforming loan (NPL) is crucial in the context of Bangladesh. It is more crucial in case of the state-owned commercial banks (SCBs). As per annual reports of Bangladesh Bank (BB), the overall NPL ratio of the banking industry is about 9% during the research period, 20009-2020. But the NPL ratio of the SCBs is about 22% during the same period. Despite having valuable strengths like sovereign ownership, qualified manpower and enormous infrastructure, the performance of the SCBs is disappointing. According to the annual reports of the SCBs, the average share of the industry NPL of the SCBs was 47.44% as against 26.71% share of industry asset during 2009-2020. This high rate of NPL has become a matter of concern for all. Though the regulators and the government have taken several steps, the NPL is increasing at a progressive rate. The first step of the way out of this situation is obviously subject to the identification of actual causes of the NPL. Under this backdrop, this paper attempts to identify the determinants of the NPL of the SCBs of Bangladesh. For this purpose, I studied a novel panel data for a period of 2009-2020 along with the analysis of primary data to identify the impact of the qualitative factors. Note that it is not possible to go back beyond 2009 since one of the SCBs, BDBL, commenced its function as a commercial bank in 2009. The NPL is not a local issue only. In fact, it has become a global issue since the Asian crisis of 1997 and this issue has attracted increased attention after the global financial crisis of 2007 2008. Since then, researchers have carried out many studies to find out the determinants of the NPL of banks. Those studies are mainly quantitative analysis, in which static or dynamic panel data are regressed using estimator such as Pulled OLS, Fixed Effect (FE) and Randoms Effect (RE) and GMM either discretely or together. A few studies are, however, conducted using VAR model specially where the feedback effects of NPL on the independent variables are examined. Except for a very few cases, almost all the previous studies used panel data over a period of 5 to 12 years. Some of the previous studies estimate the impact of only macroeconomic variables while some estimate the impact of only bank-specific factors. On the other hand, many of the earlier studies investigate the joint impact of macro and bank-specific variables. The common macroeconomic variables used in the previous studies are GDP growth rate, inflation rate, unemployment rate, exchange rate, weighted average lending rate and share price index. On the other hand, commonly estimated bank-specific variables are credit growth, bank size, loan deposit ratio, management inefficiency, ROA, and ROE. However, some very important issues are missing in all the earlier studies. One very important missing issue is the study of the impact of supply of loanable fund on the loan quality. Loan is a major asset for a bank created using fund, which largely depends on the demand for loanable fund. This asset or use of fund must be impacted by the liability or source of fund i.e., the supply of loanable fund, a major portion of which comes in the form of deposits collected from the depositors by the banks. Another missing component is the analysis of the impact of ICT combining with other explanatory variables for the NPL. Modern lending business uses ICT all the time. The success of lending business largely depends on the proper management of ICT risk of a bank. Similarly, the examination of the influences of risk associated with asset liability management (ALM) is missing in the previous studies though bank business is basically a business of ALM. Previous iv studies have also not estimated the influence of forbearance as far as I know. Banks grant the forbearance, especially in the form of interest waiver, not as an endowment but for facilitating the recovery of loans and advances. So, studies that attempt to identify determinants of NPL should cover the impact of the forbearance too. Nevertheless, quantitative analysis is not enough for identifying the determinants of the NPL even if the impact of the supply of loanable fund, ICT, ALM and forbearance is considered. There are factors such external pressure, internal malpractice, poor corporate governance, aggressive banking etc., which are believed to have significant impact on the NPL but it is not possible to maintain time series data on such factors. To capture the impact of such factors, survey based qualitative study is needed. Though there are some qualitative analyses available in literature on the determinants of NPL, those are not done simultaneously with the quantitative studies. Moreover, previous qualitative studies suffer from two major shortcomings. The first one is the inadequacy of sample size in most of the cases and the second one is the exclusion of conversant respondents. All the previous survey-based studies collected opinion either from bankers or from borrowers or both. But the opinions of the respondents of two very important categories namely, regulators and researchers are not considered in the previous survey-based studies as far as I know. Specially, anonymous opinion of the officials working in the regulatory body is very important because those officials deal with loan performance and have the real clues about the causes of NPL. Therefore, the present dissertation adopts a mixed method of quantitative and qualitative analysis in the investigation of factors responsible for the high NPL of the SCBs of Bangladesh. In the quantitative analysis, a total of 14 hypotheses are tested. These propositions include 5 macroeconomic issues and 9 on bank-specific factors developed through comprehensive review of literature and gap analysis. The macroeconomic factors are real GDP growth rate, unemployment rate, change in real effective exchange rate (REER), stock price index and real lending rate. Note that ‘inflation rate’ has been dropped from the initial specification for its autocorrelation with REER and ΔREER has been used instead of REER for stationary purpose. The bank-specific variables are deposit growth, loan deposit ratio, management inefficiency, credit growth, bank size, capital adequacy ratio, growth of interest waiver, ALM risk management and ICT risk management. Note that I avoided ROA or ROE since SCBs are losing concerns and they depend on continuous budgetary support from the government. However, the quantitative analysis on the impact of supply of loanable fund (represented by deposit growth), growth of interest waiver (as proxy of forbearance), ALM risk management and ICT risk management combining with the impact of other exogeneous and endogenous factors is the first of its kind to the best of my knowledge. In the qualitative analysis, I collected the opinion of 394 respondents with a five-point LIKERT scale against a minimum sample size of 385 required for a large finite population. The survey is conducted among four categories of respondents namely, state-owned banker, borrowers of SCBs, officials of regulatory body involved with offsite and onsite supervision and regulation of the SCBs and researches who usually conduct researches on credit related issues. Survey based qualitative analysis accommodating opinion of regulators and researchers along with that of bankers and borrowers is also the first of its kind as far as I know. I collected primary data v on seven broad issues. These issues are macroeconomic conditions, corporate governance, credit risk management, loan administration, external pressure and internal malpractice, legal and regulatory framework, and competition in the banking industry. The parameters or variables of the issues of macroeconomic conditions, credit risk management and loan administration are designed in such a way that the outcome of the qualitative analysis on these issues can confirm or deny the outcome of quantitative analysis. As for econometric techniques of analyzing the quantitative factors, four estimators namely Pooled Ordinary Least Squares (OLS), Fixed-effect (FE), Random-effect (RE) and Generalized Least Squares (GLS) are used. Moreover, a comparison of the estimations of these four estimators are made through Breusch and Pagan Lagrange multiplier test and Hausman test to find the most efficient result. In the estimation process, model specification is validated though various diagnostic tests including panel unit root test, multicollinearity test, omitted variable test, heteroscedasticity test, autocorrelation test, and cross-sectional dependency test. The estimators mentioned above are used to regress a total of 3 (three) models. The first two models capture the stand-alone impact of only macroeconomic and bank specific variables. The third estimates the joint impact of the bank-specific and macroeconomic factors. For the robustness of the study, GMM estimation is also performed, but the GMM model itself is not significant at conventional significance level of <0.05 in terms of AR(1). On the other hand, technical analysis of qualitative factors is done through principal component analysis (PCA). In the process of analyzing the principal component, I conducted various tests for ensuring a precise result. These include the reliability test of survey data through Cronbach’s alpha, linearity test, sampling adequacy test through KMO Bartlett’s test, data reduction suitability test through Bartlett’s test of sphericity, significant outlier test, data distribution test, communalities test etc. According to the outcome of the panel data regression, a combination of macroeconomic and bank-specific factors explains the variation of NPL ratio of the SCBs of Bangladesh while standalone macroeconomic or bank-specific factors cannot explain the variation. In terms of the significance of individual factors, regression of Pooled OLS, FE, RE and GLS estimators suggest that unemployment rate, real lending rate and change in REER have significant positive correlation with the NPL ratio. This outcome confirms the finding of most of the previous studies. The unpredicted direct relation between NPL ratio and real GDP growth rate, though not unprecedented, is probably the result of business expansion by large borrowers during economic upturn withholding the repayment of current loan. Among the bank specific variables, deposit growth has been found to have an inverse association with the NPL ratio showing the significance of the supply of loanable fund in explaining the variation of NPL ratio, which has not been explored before. Similarly, ALRM and ICTRM are also found to have significant impact on the NPL ratio when assessed combinedly with other traditional macroeconomic and bank specific variables. Other bank specific variables sch as loan deposit ratio and management inefficiency have significant positive association with the NPL ratio as per the estimations. Conversely, credit growth and bank size are found to have significant negative correlation with the NPL ratio. These are all in conformity with the previous findings and the current model specification. Other factors included in the specification namely, stock price index, forbearance, and CAR cannot explain the variation of NPL ratio of the SCBs. vi GMM estimation, though not significant at the level of AR(1) = Pr<0.05, suggests that previous year’s NPL ratio rises current year’s NPL ratio. Regarding qualitative analysis, PCA of primary data reveals that poor corporate governance is the most influential component, which raises NPL. The second influential factor is the inadequacy of legal and regulatory framework and the third one is inefficient loan administration. Other principal components, which intensify the NPL are: adverse macroeconomic conditions; pressure from influential quarter and malpractice by the bankers; poor credit risk management; and unhealthy competition in the banking industry. The paper highlights the need for policy adjustment and operational modification in line with the findings of the quantitative and qualitative analysis. Both the quantitative and qualitative analysis reveals that macroeconomic condition impact loan performance of SCBs. Accordingly, this paper recommends for considering potential movement of exogeneous factors while formulating the lending policies and caring out lending operations. This dissertation also suggests for considering external competitiveness of economy while sanctioning large loans to the exporter borrowers. The discovery of a significant inverse relation between the supply of loanable fund (represented by deposit growth) and NPL ratio warrants the need for building low-cost core deposit base by the SCBs for easing the credit risk management accommodating appropriate risk appetite. Meticulous compliance of internationally best practiced rules and regulations can easily resolve the loan deposit ratio, management inefficiency etc. issues and reduce the NPL. Current study suggests for establishing a formal coordination channel between BB and MoF for improving the capacity of the SCBs specially for maintaining credit quality through improving corporate governance, and controlling external pressure and internal malpractice. Such coordination between these two authorities can help establish compliance and credit culture in SCBs. Nevertheless, there are some more complex issues such as composition of board, disposal of a growing number of pending cases with money loan court, maintaining accountability of the sanctioning authority for concerned NPL, raising bank size by merging losing small banks with larger ones, and ensuring enough punishment for willful defaults. Disposal of these issues requires the reform of legal and regulatory frameworks, which is not possible without strong political commitment.