Does interest rate affect non-performing loans? Empirical evidence from commercial banks in Uzbekistan

Аннотация

This study investigates non-performing loans (NPL) and credit risk in Uzbekistan's commercial banking sector, focusing on the period post-2016. Using dynamic panel approach, bank-specific and macroeconomic factors have been included in the econometric model to estimate their effects on NPL. According to STATA results, loan-deposit ratio, GDP growth and loan interest rates have positive impact on NPL. On the other hand, banks with foreign ownership experience lower rate of NPL compared to local banks. Since the beginning of 2017, the volume of loans have been increasing significantly. As of July 1st, 2024, the total outstanding loans in commercial banks of Uzbekistan amounted to 494 trillion Uzbek soums [1] which comprises approximately 44% of the GDP of Uzbekistan. Significant expansion of loans to greater population and improvements in financial inclusion are associated with higher NPL ratio in credit portfolios. While this ratio was fluctuating between 2% and 3% until the end of 2020, it started increasing during the pandemic period reaching as high as 6.2% during mid 2021 (See Picture 2.5). This can be partially explained by the fact that many households faced financial troubles during 216 “MILLIY IQTISODIYOTNI ISLOH QILISH VA BARQAROR RIVOJLANTIRISH ISTIQBOLLARI” mavzusida xalqaro ilmiy-amaliy konferensiyasi pandemic lockdowns and they had to postpone their loan payment obligations. Picture 1 illustrates the dynamic change of NPL for the beginning periods of 2020 – 2024 along with three financial indicators of commercial banks: total assets, total deposits and total credit. According to the graph, all three indicators have experienced significant growth while NPL has shown sharp increase during the 2021-2022 period. Picture 1. Comparative illustration of NPL and main financial indicators of commercial banks in Uzbekistan. This study examines the main factors affecting the non-performing loans of commercial banks, both bank-specific factors, such as leverage, size, loan-to- deposit ratio, state ownership, as well as macroeconomic factors, such as annual GDP growth rate, weighted average interest rate on loans and exchange rate. Using panel data of 35 commercial banks of Uzbekistan on quarterly basis over the period 2019-2023, we estimate the impact of aforementioned variables on NPL. The brief description of each variable and its expected impact on NPL are provided in Table 1. Considering a dynamic panel data analysis in this study helps us to examine the relationship between variables over time while taking into account individual-specific effects and potential endogeneity issues. It also allows for the inclusion of multiple time periods and considers the dynamic nature of the data, capturing both the cross-sectional and time-series dimensions. This technique became more prominent with the introduction of the Generalized Method of Moments (GMM) estimators by Arellano and Bond in 1991 [2]. They introduced a GMM estimator for dynamic panel data models, addressing issues like autocorrelation, heteroskedasticity, and the presence of lagged dependent variables. Later, Blundell and Bond [3] proposed a system GMM estimator, which combines equations in levels and first differences to improve efficiency, especially in cases with persistent data. 217 “MILLIY IQTISODIYOTNI ISLOH QILISH VA BARQAROR RIVOJLANTIRISH ISTIQBOLLARI” mavzusida xalqaro ilmiy-amaliy konferensiyasi Table 1. The description of variables being used for the analysis. Variable Description Expected Sign Dependent variable Non-performing loans (NPL) Troubled loans/Total loans Independent variables Bank-specific Loan-deposit ratio (LDR) Total loans/ Total deposits Negative/Positive Size Natural logarithm of total assets Negative/Positive Leverage Total liabilities/Total assets Negative/Positive State-ownership = 1 if state ownership, 0 otherwise Positive Foreign-ownership = 1 if foreign ownership, 0 otherwise Negative Macroeconomic Interest rate Weighted average interest rate on loans, % Positive Exchange rate Change in UZS/USD exchange rate, % Negative GDP growth rate Annual GDP growth rate, % Negative The specified regression model used in the study is as follows: NPL =  NPL + ’X +  (1) i,t 0 i,t-1 i i,t i,t with  =  +  i,t i i,t here the subscript i denotes the cross-sectional (banks) and t denotes time dimension of the panel sample. NPL is non-performing loan ratio, NPL is its i,t i,t-1 lagged value, ’ is a 1xk vector of parameters, X is a vector of independent i i,t variables including their lagged values and  is the error term.  has two i,t i,t orthogonal components:  are the unobserved individual effects and  are the i i,t observed specific errors. Table 2. Estimation results using system GMM. Source: Author’s calculations System GMM Variable Coefficient Std. Error NPL 0.7528*** 0.0300 (i-1) LDR 0.0035* 0.0019 Size -0.0115 0.0118 Leverage 0.0716 0.0534 Exchange 0.0077 0.1420 GDP 0.8799*** 0.1704 ROE 0.0021 0.0018 State-own 0.0440