American Journal of Business Management, Economics and Banking ISSN (E): 2832-8078 Volume 43, December - 2025 P a g e | 13 www.americanjournal.org ECONOMETRIC ANALYSIS OF THE IMPROVEMENT OF CREDIT SERVICES Zafar Absamatovich Umarov Tashkent State University of Economics umarov_zafar@@tsue.uz Nilufar Khikmatullayevna Sharipova Tashkent State University of Economics sharipova_nilufar@tsue.uz A B S T R A C T K E Y W O R D S The quality of credit services and the level of non-performing loans (NPLs) are key determinants of financial stability and banking sector soundness. This study analyzes the main factors influencing the volume of NPLs in the banking sector of Uzbekistan through advanced econometric modeling. Using monthly data from January 2020 to January 2025 obtained from the Central Bank of the Republic of Uzbekistan, we examine the impact of capital adequacy, term deposits, liquidity indicators, and other key variables on NPL dynamics. Pearson correlation analysis is applied to identify significant relationships, followed by multiple regression estimation. Diagnostic checks reveal the presence of autocorrelation and heteroskedasticity in the static model, which is addressed through dynamic modeling using ARMAX and ARIMA specifications. The ARMAX(1,1) model demonstrates that capital adequacy has a negative and statistically significant impact on NPLs, whereas term deposits and liquidity ratios exert positive effects. The final ARIMA(0,1,0) model provides a robust baseline for short-term forecasting, indicating an expected monthly increase of 300.27 units in NPL volume. The findings emphasize the importance of strengthening capital adequacy, managing liquidity prudently, and ensuring controlled deposit growth to mitigate credit risk and enhance the quality of credit services in Uzbekistan’s banking sector. Non-performing loans (NPL); Credit risk; Banking sector; Econometric modeling; ARIMA; ARMAX; Capital adequacy; Term deposits; Liquidity ratio; Uzbekistan. Introduction Credit risk remains one of the most critical challenges to the stability of the banking sector. A high level of non-performing loans (NPLs) not only undermines banks’ balance sheets and profitability but American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 14 www.americanjournal.org also weakens the overall financial system, limiting credit growth and slowing economic development. In many countries, NPL dynamics are closely linked to macroeconomic conditions and banking system indicators, including capital adequacy, deposit structure, and liquidity ratios. Strengthening credit quality is therefore essential for maintaining financial stability, especially in emerging economies. In recent years, Uzbekistan has experienced rapid financial sector development characterized by expanding credit portfolios, growing deposit bases, and evolving liquidity management practices. These changes have increased the importance of monitoring credit risk and understanding the factors driving NPLs. Drawing on international best practices, this study applies econometric modeling techniques to evaluate the relationship between NPLs and key financial stability indicators in the Uzbek banking sector. By integrating both static and dynamic modeling approaches, the research seeks to provide evidence-based recommendations for strengthening credit service quality and supporting sustainable banking sector growth. 2. Literature Review / Analysis The concept of credit has long been recognized as a fundamental economic category. Abdullaev Yo. et al. define a loan as a relationship arising from the borrowing and repayment of temporarily available funds with remuneration over a specified period. This relationship involves two parties: the lender (creditor) and the borrower (debtor), forming the foundation of modern credit markets [1]. Rashidov U. et al. further describe credit as a set of economic relations tied to the borrowing and repayment of funds, emphasizing its structured, economic—not merely social—nature [2]. L. Krolevetskaya highlights that credit emerged historically after the introduction of money, accelerating economic activity by enabling the time-shifting of consumption and investment [3]. Abdullayeva Sh. and Azizov underscore the contractual basis of credit transactions: borrowers are obliged to repay borrowed funds within a fixed period along with additional remuneration [4]. At the macro level, O. Sattarov notes that the loan-to-GDP ratio is a critical indicator of investment orientation and financial sector depth, with higher ratios typically associated with stronger economic growth and financial intermediation [5]. Together, these perspectives establish credit as both a microeconomic and macroeconomic driver of development. Understanding its dynamics is crucial for analyzing credit risk, NPL trends, and the stability of banking systems—particularly in developing economies such as Uzbekistan. 3. Data and Methodology This study is based on monthly time-series data for the period January 2020 – January 2025, collected from the Central Bank of the Republic of Uzbekistan. ⎯ Dependent variable (Y): Volume of non-performing loans (NPL) ⎯ Independent variables: ⎯ X1: Capital adequacy ratio (%) ⎯ X2: Volume of term deposits (billion UZS) ⎯ X3: Fixed capital (billion UZS) ⎯ X4: Highly liquid assets (billion UZS) ⎯ X5: Instant liquidity ratio (%) American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 15 www.americanjournal.org The methodological framework consists of the following steps: 1. Correlation Analysis — Using the Pearson correlation coefficient to identify statistically significant linear relationships between NPLs and explanatory variables. 2. Multicollinearity Check — Exclusion of variables with strong mutual correlations (e.g., X2X_2X2 and X3X_3X3 correlation of 0.9872). 3. Static Regression Analysis (OLS) — Estimation of a baseline multiple linear regression model. 4. Diagnostic Testing — Application of the Durbin–Watson statistic to test for autocorrelation and Breusch–Pagan test for heteroskedasticity. 5. Dynamic Modeling (ARMAX) — Inclusion of lagged dependent variables to correct autocorrelation and capture temporal dynamics. 6. Forecasting (ARIMA) — Identification of the optimal ARIMA specification for NPL forecasts based on AIC and residual tests. 4. Results and Discussion 4.1 Correlation Analysis Taking into account the specific characteristics of Uzbekistan’s banking system, this study employs econometric methods to analyze the impact of several key indicators on the volume of non-performing loans (NPLs). The indicators considered include the overall capital adequacy ratio, the volume of term deposits, fixed capital, highly liquid assets, and the instant liquidity ratio. The analysis is based on official data provided by the Central Bank of the Republic of Uzbekistan, which were used to assess how these economic indicators influence changes in the volume of NPLs in the banking sector. Based on the empirical results, corresponding forecasts were also developed. Since improving the quality of credit services is directly linked to reducing the volume of non- performing loans, the monthly volume of NPLs for the period 2020–2025 was selected as the dependent variable, while the banking system stability indicators were chosen as independent variables. To evaluate the relationships between these factors and NPL dynamics, the Pearson correlation coefficient (r) was applied to measure both the strength and direction of the correlations. This approach made it possible to identify the most statistically significant indicators affecting the level of non- performing loans in Uzbekistan’s banking sector Table 1. Analysis of the Correlation Between NPL and Banking System Indicators 1 Correlation Relationship Probability NPL (Y) Capital Adequacy Ratio (X₁) Volume of Time Deposits (X₂) Fixed Capital (X₃) Highly Liquid Assets (X₄) Instant Liquidity Ratio (X₅) NPL - Y 1.000000 Capital Adequacy Ratio (X₁) -0,7764 1.000000 Volume of Time Deposits (X₂) 0,8081 -0,5708 1.000000 Fixed Capital (X₃) 0,8010 -0,5533 0,9872 1.000000 Highly Liquid Assets (X₄) 0,8253 -0,6979 0,8905 0,8694 1.000000 Instant Liquidity Ratio (X₅) 0,8451 -0,7674 0,6932 0,6864 0,8778 1.000000 1 Calculated using Gretl software. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 16 www.americanjournal.org Based on the statistical table at the 5 percent significance level, the coefficients of the Pearson correlation (r) were found to be statistically significant. The correlation coefficients between NPLs and the selected indicators were as follows: capital adequacy ratio – –0.7764, volume of term deposits – 0.8081, fixed capital – 0.8010, highly liquid assets – 0.8253, and instant liquidity ratio – 0.8451. When selecting explanatory variables for regression analysis, it is essential to ensure that these variables are not excessively correlated with each other. High levels of multicollinearity can distort estimation results and undermine the statistical reliability of the model. In particular, when the correlation between independent variables exceeds their correlation with the dependent variable, the precision and interpretability of the estimated coefficients may be compromised. The correlation matrix reveals a clear multicollinearity problem among the selected indicators. Specifically, fixed capital (X₃) ) exhibits a stronger correlation with the volume of term deposits (X₂) — 0.9872 — than with NPL itself. Likewise, highly liquid assets (X₄) ) display stronger correlations with term deposits (0.8905) and fixed capital (0.8694) than with NPL. Therefore, to mitigate multicollinearity and improve the robustness of the regression model, the indicators for fixed capital and highly liquid assets were excluded. Pearson correlation analysis was then repeated using the remaining variables to examine their relationship with the volume of non- performing loans. The updated results are presented in the following table. Table 2 Analysis of the Correlation Between NPL and Banking System Indicators 2 Корреляцион боғлиқлиги Эҳтимоллилик NPL(Y) Capital Adequacy Ratio (X₁) Volume of Time Deposits (X₂) Instant Liquidity Ratio (X₅) NPL - Y 1.000000 Capital Adequacy Ratio (X₁) -0,7764 1.000000 Volume of Time Deposits (X₂) 0,8081 -0,5708 1.000000 Instant Liquidity Ratio (X₅) 0,8451 -0,7674 0,6932 1.000000 Based on the statistical table at the 5 percent significance level, the coefficients of the Pearson correlation (r) were found to be statistically significant. Accordingly, the correlations between NPLs and the respective indicators were determined as follows: capital adequacy ratio – –0.7764, volume of term deposits – 0.8081, and instant liquidity ratio – 0.8451. 4.2 Static Regression Model (OLS) The general form of the multiple linear regression equation can be expressed as: Y = α + β1x1 + β2x2 + β3x3 + β4x5 + ⋯ + βnxn + εn where: 2 Calculated using Gretl software. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 17 www.americanjournal.org • α − — the constant term, representing the initial value of the dependent variable when all explanatory variables are zero; • β1, β2, … βn − — the regression coefficients that measure the magnitude and direction of the impact of each independent variable on the dependent variable; • x1, x2, … xn — the explanatory (independent) variables; • εn − — the error term capturing unobserved factors that may influence the dependent variable. Using the factors that exhibited significant correlations with NPLs, a probabilistic regression analysis was carried out to estimate the relationship between non-performing loans and the selected explanatory variables. This step allowed for quantifying the marginal effects of key banking indicators on NPL dynamics while controlling for potential noise in the data Table 3 Regression Statistical Analysis of NPL and Its Influencing Factors (Model 1))3 Dependent Variable: NPL Method: Least Squares Sample (adjusted): 2020:01-2025:01 (T = 61) Included observations 61 Variables Coefficient Std. Error t- Statistic Prob. C 16552,0 6464,73 2,560 0,0131 ** Capital Adequacy Ratio (X₁) −877,229 267,157 −3,284 0,0018 *** Volume of Time Deposits (X₂) 0,0578580 0,0104818 5,520 8,67e-07 *** Instant Liquidity Ratio (X₅) 98,1541 26,2621 3,737 0,0004 *** R- squared 0,839672 S.D. of Dependent Variable 5683,958 Sum of Squared Residuals 3,11e+08 S.E. of Regression 2335,040 F(3, 57) 99,50674 Adjusted R-squared 0,831233 Log-likelihood −557,5895 Prob(F-statistic) 1,24e-22 Schwarz Criterion (SC) 1131,623 Akaike Information Criterion (AIC) 1123,179 ρ (rho) parameter 0,881792 Hannan–Quinn Criterion (HQC) 1126,488 Mean of Dependent Variable 14712,60 Durbin–Watson Statistic 0,240215 Based on the regression statistical analysis of NPL and its influencing factors, the regression equation can be constructed as follows: y = 16 552−877,229 ∗ x1 + 0,0578 ∗ x2 + 98,1541 ∗ x5 + ε 3 Calculated using Gretl software. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 18 www.americanjournal.org In this regression model, all the variables are statistically significant at the 5 percent significance level: • Capital adequacy ratio (p = 0.0018 < 0.005), • Volume of time deposits (p = 8.67e–07 < 0.005), • Instant liquidity ratio (p = 0.0004 < 0.005). This indicates that each of these factors has a significant impact on the volume of non-performing loans (NPL). Specifically: • A 1% increase in the capital adequacy ratio reduces the volume of non-performing loans by approximately 877.229 billion UZS; • A 1 billion UZS increase in the volume of time deposits leads to an increase in non-performing loans by about 57.8 million UZS; • A 1% increase in the instant liquidity ratio results in an increase in non-performing loans by approximately 98.1541 billion UZS. Hence, the results suggest that while strengthening capital adequacy contributes to a reduction in credit risk, the growth in time deposits and higher liquidity levels may, under certain conditions, be associated with an increase in non-performing loans in the banking sector. To determine the statistical reliability of the above model, the Akaike Information Criterion (AIC) and the Schwarz Criterion (SC) were examined, along with tests for the presence of first- and second-order autocorrelation in the residuals. Based on the 5 percent significance level from the Durbin–Watson statistical table, the condition for the absence of autocorrelation is considered to hold when 1.54 < DW < 2.46. From the regression results above, the Durbin–Watson statistic (DW) is 0.240215, which is far below the lower threshold of 1.54. This indicates the presence of strong positive autocorrelation between the residuals of the model. In practical terms, this means that the residuals (errors) from one period are highly correlated with those from the next period — suggesting that important time-dependent dynamics may not have been fully captured by the model. To address this issue, methods such as including lagged variables, applying autoregressive models (AR(1)), or using the Cochrane–Orcutt or Generalized Least Squares (GLS) correction procedures can be employed to reduce autocorrelation and improve the reliability of the regression estimates. To test for the presence of heteroskedasticity in the residuals of the selected regression model, the Breusch–Pagan (BP) test is applied. In this test, the null hypothesis (H₀) states that heteroskedasticity is not present — meaning that the residuals have constant variance (homoskedasticity). The interpretation of the test results is as follows: • If p-value > 0.05, we fail to reject H₀, indicating that heteroskedasticity is not present, and the model’s random errors are homoskedastic. • If p-value ≤ 0.05, we reject H₀, which implies that heteroskedasticity is present — the variance of the residuals is not constant, and this may affect the reliability of the estimated coefficients. Thus, the Breusch–Pagan test helps to confirm whether the assumption of equal error variance holds for the selected regression model, ensuring the statistical robustness of the estimation results. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 19 www.americanjournal.org Table 4 Breusch–Pagan Test Results for the Selected Regression Model of NPL and Its Influencing Factors 4. Breusch–Pagan Test Results for the Selected Regression Model (Second-Order Heteroskedasticity) Method: Least Squares Sample (adjusted): 2020:01-2025:01 (T = 61) Dependent Variable: uhat Variables Coefficient Std. Error t-Statistic Prob. const 2,55504 3,26232 0,7832 0,4368 Capital Adequacy Ratio (X₁) −0,120962 0,134816 −0,8972 0,3734 Volume of Time Deposits (X₂) −1,60018e-05 5,28944e-06 −3,025 0,0037 *** Instant Liquidity Ratio (X₅) 0,0232335 0,0132527 1,753 0,0850 * Sum of Squared Residuals = 15,572 Test Statistic: LM = 7,785975, Prob.= P(Хи-square (3) > 7,785975) = 0,050648 According to the results of the Breusch–Pagan test, since P < 0.13, it can be concluded that heteroskedasticity is present in the model, indicating that the residuals of the selected regression are not homoskedastic. Therefore, considering that the chosen model did not pass the regression diagnostic tests satisfactorily, the forecast indicators derived from this model cannot be considered reliable. 4.3 ARMAX Model (1,1) To ensure the statistical robustness of the model and to eliminate autocorrelation, the next step is to modify the model. For this purpose, a dynamic regression approach is applied by including the lagged dependent variable Y(t−1) as an explanatory variable, and the ARIMA model is employed. If the initial model is… Yt=β0+β1X1t+β2X2t+β3X5t+εt If the initial model was expressed in the following form, the new model — incorporating the lagged dependent variable — will take the following form: Yt=α0+α1Yt−1+α2X1t+α3X2t+α4X5t+ut As a result, the residuals ut will exhibit less autocorrelation. 4 Calculated using Gretl software. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 20 www.americanjournal.org Table 5 ARMAX Model (Model 2) for NPL and Its Influencing Factors5 Dependent Variable: NPL Method: ARMAX Sample (adjusted): 2020:01-2025:01 (T = 61) Included observations 61 Variable Coefficient Std. Error z P-value const 17875,7 6559,43 2,725 0,0064 *** phi_1 0,870335 0,0695245 12,52 5,92e-036*** theta_1 0,152754 0,134101 1,139 0,2547 Capital Adequacy Ratio (X₁) −768,170 289,136 −2,657 0,0079*** Term Deposits Volume (X₂) 0,0654652 0,0241074 2,716 0,0066 *** Instant Liquidity Ratio (X₅) 50,1475 25,5143 1,965 0,0494 ** Standard Deviation of Innovations 1030,483 Log-likelihood −510,6050 Standard Error of the Model 5683,958 Schwarz Criterion: 1049,986 Akaike Information Criterion 1035,210 Mean of Innovations: 20,55437 Hannan–Quinn Criterion 1041,001 Mean of Dependent Variable: 14712,60 Real Part Imaginary Part Modulus Frequency Root (AR Part) 1,1490 0,0000 1,1490 0,0000 Root (MA Part) -6,5465 0,0000 6,5465 0,5000 Based on the regression statistical analysis of NPL and its influencing factors, we can construct the regression equation according to the above model. The autoregressive model with a mean and exogenous variable component, ARMAX(1,1) (AutoRegressive Moving Average with eXogenous variables), takes the following form: Yt=17 875,7 + 0,8703⋅Yt−1 + 0,1528⋅εt−1 − 768,17⋅X1t + 0,0655⋅X2t + 50,15⋅X5t + εt where: Yt — the level of non-performing loans at time t; Yt−1 — the volume of non-performing loans in the previous period X1t — capital adequacy ratio at time; X2t — volume of term deposits at time; X5t — instant liquidity ratio at time; εt — the stochastic error (shock) at time. In this regression model, all explanatory variables — capital adequacy ratio (p = 0.0079 < 0.05), volume of term deposits (p = 0.0066 < 0.05), and instant liquidity ratio (p = 0.0494 < 0.05) — are statistically significant at the 5% significance level. 5 Calculated using Gretl software. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 21 www.americanjournal.org These results allow us to draw the following conclusions: • A 1% increase in the capital adequacy ratio leads to a decrease in the volume of non-performing loans by approximately 768.17 billion UZS. • A 1 billion UZS increase in the volume of term deposits results in an increase in non-performing loans by around 65.5 million UZS. • A 1% increase in the instant liquidity ratio raises the volume of non-performing loans by approximately 50.15 billion UZS. This implies that while stronger capital adequacy has a stabilizing effect by reducing credit risk, increases in term deposits and higher liquidity levels are associated with rising NPL volumes during the observed period. The statistical robustness of the ARMAX model is supported by several diagnostic results: ⎯ Improved log-likelihood: The log-likelihood increased to –510.6, compared to –557 in the static model, indicating a better overall fit. ⎯ Akaike Information Criterion (AIC): The AIC declined to 1035, which is lower than the lag- free model’s value (~1123), confirming superior model specification. ⎯ Model stability: The autoregressive root (AR = 1.149 > 1) lies outside the unit circle, demonstrating stationarity of the model. ⎯ MA(1) term exclusion: The MA(1) coefficient was not statistically significant (p > 0.25) and was therefore removed, simplifying the model without compromising explanatory power. Taken together, these diagnostics show that the dynamic ARMAX specification provides a more robust and statistically reliable framework for analyzing NPL dynamics and their determinants compared to the static regression model. Moreover, the ARX(1) component of the model successfully captures the temporal dependence of the series. The inclusion of the lagged dependent variable Yt−1Y_{t-1}Yt−1 effectively eliminated residual autocorrelation, as confirmed by the Durbin–Watson test. This confirms the model’s suitability for further analytical applications and short-term forecasting of NPL trends. Subsequently, the study evaluates alternative ARIMA models using Stata to further refine the forecasting process. Three specifications were constructed: ARIMA(0,1,0), ARIMA(1,1,0), and ARIMA(0,1,1). A comparative analysis of these models is presented below, with detailed computations provided in the appendix. Table 6 Comparative Analysis of the Models 6: Модель AIC (≈ 2k – 2LL) Log- likelihood AR/MA Value Conclusion ARIMA(0,1,0) 1014.14 –506.072 — Drift (constant): Statistically significant ARIMA(1,1,0) 1013.34 –505.672 AR: Not statistically significant AR(1): Redundant / not necessary ARIMA(0,1,1) 1013.41 –505.705 MA: Not statistically significant MA(1): Redundant / not necessary Note: AIC ≈ –2 × LogL + 2k By comparing the estimated models, the following conclusions can be drawn: 6 Calculated using STATA software. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 22 www.americanjournal.org • In all three models, the Akaike Information Criterion (AIC) differences are very small (< 1), indicating that the models are equivalent in terms of fit quality. • Both AR(1) and MA(1) terms are not statistically significant (p > 0.3). • Only in ARIMA(0,1,0) is the constant (drift) statistically significant at the 5% level. drift=300,27; p=0,044. 4.4 ARIMA Model for Forecasting Based on the calculations above, we constructed the simple ARIMA(0,1,0) model (Model 3): Yt=Yt−1 + 300,27 + εt The results of the Ljung–Box test confirm that the residuals of the model are free from autocorrelation: Q=18.13Q = 18.13Q=18.13, p=0.1118p = 0.1118p=0.1118. Therefore, the model is adequate. This implies that the ARIMA(0,1,0) model can be confidently used for future baseline forecasting based on the time series data. Figure 1. One-Year Forecast of Non-Performing Loans Based on the ARIMA(0,1,0) Model 7 Based on the forecast generated by the ARIMA(0,1,0) model, it can be observed that the volume of non-performing loans is expected to increase steadily by approximately 300.27 units per month. Maintaining a high level of capital adequacy (X₁) plays a critical role in mitigating the risk of rising NPLs. In contrast, expansion of the deposit base (X₂) may contribute to the accumulation of non- performing assets if credit quality is not subject to strict monitoring and control. Although a high liquidity ratio (X₅) generally indicates a stable banking system, it also requires prudent and balanced management. Relying on aggressive lending strategies solely on the basis of high 7 Calculated using STATA software. 50 00 10 00 0 15 00 0 20 00 0 25 00 0 2020m1 2022m1 2024m1 2026m1 2028m1 month_date Y forecast_only American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 23 www.americanjournal.org liquidity can increase credit exposure and heighten systemic risk. Overall, the ARIMA(0,1,0) forecast underscores the importance of robust capital management, prudent deposit expansion, and carefully balanced liquidity-driven lending strategies as key measures to contain the upward trajectory of NPLs. The analysis was based on 61 monthly observations from January 2020 to January 2025. Initially, a correlation analysis was conducted to identify statistically significant relationships among variables. A linear regression model was then estimated, with the volume of non-performing loans (Y) as the dependent variable and capital adequacy ratio (X₁), volume of term deposits (X₂), and instant liquidity ratio (X₅) as explanatory variables. Although the coefficients were statistically significant, the model did not pass the autocorrelation diagnostic tests, necessitating the use of a dynamic modeling approach. To address this issue, the lagged dependent variable Yt−1Y_{t-1}Yt−1 was introduced, and an ARIMA (0,1,0) model was developed. This specification proved to be statistically reliable, capturing the underlying trend in the dependent variable, with NPL volumes rising at a stable monthly rate of approximately 300.27 units. Subsequently, an ARMAX model was estimated to incorporate both exogenous factors and lagged dynamics. This model successfully passed diagnostic and stability tests, confirming its suitability for robust analysis and short-term forecasting. The final equation is expressed as: Yt=17 875,7 + 0,8703⋅Yt−1 + 0,1528⋅εt−1 − 768,17⋅X1t + 0,0655⋅X2t + 50,15⋅X5t + εt This model highlights the key drivers of the observed upward trend in non-performing loans: ⎯ A 1% increase in capital adequacy reduces NPLs by approximately 768.17 billion UZS. ⎯ A 1 billion UZS increase in term deposits increases NPLs by approximately 65.5 million UZS. ⎯ A 1% increase in the instant liquidity ratio raises NPLs by around 50.15 billion UZS. 5. Conclusion The results of this study provide strong empirical evidence that capital adequacy, term deposit volumes, and liquidity levels play a significant role in shaping the dynamics of non-performing loans in the banking sector. While higher capital adequacy contributes to a reduction in credit risk, increased deposit volumes and excessive liquidity may, under certain conditions, lead to a rise in NPL levels. These findings underscore the importance of proactive and balanced risk management strategies. To address these challenges and strengthen financial stability, several policy measures are recommended. First, banks should strengthen oversight and control over the use of term deposits to prevent a deterioration in credit quality. Second, maintaining and supporting adequate capital buffers is crucial to enhancing the resilience and stability of the banking sector. Finally, the development and implementation of sound regulatory measures for balanced liquidity management and effective credit risk control will help ensure the sustainable and secure functioning of the financial system. These measures, when implemented in an integrated manner, can significantly mitigate credit risk, improve the quality of credit services, and contribute to long-term banking sector stability in Uzbekistan. American Journal of Business Management, Economics and Banking Volume 43 December- 2025 P a g e | 24 www.americanjournal.org References 1. Abdullaev, Yo., Qoraliev, T., Toshmurodov, Sh., & Abdullaeva, S. (2009). Banking operations: A textbook. Tashkent: IQTISOD-MOLIYA. 2. Abdullaeva, Sh. Z., & Azizov, U. O. (2019). Banking operations. Tashkent: Iqtisod-Moliya. 3. Krolevetskaya, L. P., & Tikhomirova, E. V. (2009). Banking: Credit activity of commercial banks. Moscow: KNORUS. 4. Rashidov, O. Yu., Alimov, I. I., Toymuhamedov, I. R., & Tojiev, P. P. (2008). Money, credit and banks. Tashkent: Tashkent State University of Economics. 5. Sattarov, O. B. (2018). Improving the methodology for ensuring the stability of the banking system of the Republic of Uzbekistan (Doctoral dissertation abstract). Tashkent. 6. Pancotto, L. (2024). The evolution and determinants of the non-performing loan. Pacific-Basin Finance Journal, 84, 102221. https://doi.org/10.1016/j.pacfin.2024.102221 7. Isakov, O. (2024). Factors affecting non-performing loans: Empirical evidence from commercial banks in Uzbekistan. Journal of Economics and Financial Analysis, 3(2), 45–58. 8. Rustamov, J. (2023). Trends and patterns of accumulation of non-performing loans in the banking sector of Uzbekistan. Journal of Eurasian Economies, 14(1), 75–88. 9. Central Bank of the Republic of Uzbekistan. (2025). Official banking statistics. Retrieved from https://cbu.uz 10. Umarov Z.A. Tijorat banklarida kreditlash jarayonini boshqarishda buxgalteriya hisobining ahamiyati // Yashil iqtisodiyot va taraqqiyot ilmiy jurnali 2025, 4-son, 2409-2412 11. Umarov Z.A. Bank kredit portfeli ichki auditining o’ziga xos xususiyatlari // Aktuar moliya va buxgalteriya hisobi ilmiy jurnali 2025, 5(02), 66-73 12. Umarov Z.A. Tijorat banklarida kreditlash jarayonini boshqarishda buxgalteriya hisobining ahamiyati // Yashil iqtisodiyot va taraqqiyot ilmiy jurnali 2025, 4-son, 2409-2412 https://cbu.uz/