Banks play a crucial role in financing economies, particularly through debt. The core function of banks, according to financial intermediation theory, is to provide liquidity and manage risk. Liquidity is a critical input for banking operations, while credit is a highly profitable asset. However, low liquidity and poor loan quality are major threats to bank profitability and stability. The 2008 global financial crisis highlighted the importance of liquidity risk (LR), leading to increased attention from policymakers and academics. Similarly, non-performing loans (NPLs) are a key determinant of bank failures and crises. While numerous studies examine the determinants of CR and LR individually or their relationship with bank profitability/stability, fewer studies explore their complex reciprocal relationship. This study addresses this gap by investigating the linear and non-linear causal linkage between NPLs and LR in both directions using data from the most significant Tunisian banks (2000-2018). Tunisia provides a suitable case study because its banking sector is the primary source of investment finance, with a less developed stock market. This work distinguishes itself by analyzing the relationship in both linear and non-linear frameworks and identifying optimal thresholds for both risks, enabling targeted regulatory reforms.

Existing literature predominantly focuses on either the determinants of credit and liquidity risk or their connection to bank performance and stability. Studies on the reciprocal relationship between CR and LR present conflicting findings. Many studies find a positive correlation, suggesting that high NPL ratios hinder a bank's ability to meet withdrawal demands. Some research indicates a negative correlation, particularly in the context of low interbank competition. Other studies find no significant relationship. Existing studies examining the non-linear relationship mostly focus on the impact of LR on NPLs, but do not investigate the reciprocal relationship or combine both linear and non-linear analyses. This study aims to address these limitations.

The study uses data from 10 major Tunisian banks from 2000 to 2018, obtained from annual reports and the World Bank database. Two empirical methodologies are employed: 1. **Seemingly Unrelated Regression (SUR):** This model is used to analyze the linear reciprocal relationship between LR (measured by the LTD ratio) and CR (measured by the NPLs ratio). The model includes bank-specific variables (bank diversification, equity-to-total assets ratio, bank size, net interest margin) and macroeconomic variables (GDP growth rate, inflation rate). The SUR model accounts for the potential correlation between the error terms of the two equations (one for LR and one for CR). 2. **Panel Smooth Transition Regression (PSTR):** This model investigates the non-linear reciprocal relationship. The PSTR model allows for different relationships between LR and CR depending on whether a transition variable (either LTD or NPLs) exceeds a certain threshold. The model uses the same bank-specific and macroeconomic variables as the SUR model. The logistic function is used to model the smooth transition between regimes. The analysis includes tests for linearity and regime determination to identify optimal thresholds for both LTD and NPLs ratios.

**Linear Analysis (SUR):** The SUR model shows a positive and significant relationship between CR and LR in both directions. Bank-specific factors such as high capital ratios (EQTA), larger bank size, and higher profitability (NIM) significantly decrease both NPLs and LTD ratios. Macroeconomic factors show that higher GDP growth decreases both risks, while inflation increases NPLs. **Non-linear Analysis (PSTR):** The PSTR model confirms the presence of a threshold effect. The optimal threshold for NPLs is 9.87%, while the optimal threshold for LTD is 102%. Below these thresholds, the relationship between NPLs and LTD is negative and significant. Above these thresholds, the effect is positive, but only significant for the impact of CR on LR. This suggests that exceeding these thresholds increases the risk of adverse effects. **Additional findings:** Bank diversification does not significantly affect either credit or liquidity risk in either linear or non-linear analyses.

The findings highlight the complex and reciprocal relationship between credit and liquidity risks. The linear analysis confirms the generally accepted positive correlation, while the non-linear analysis reveals important threshold effects. The identified thresholds (9.87% for NPLs and 102% for LTD) provide crucial benchmarks for Tunisian banks and regulators. Exceeding these thresholds significantly alters the risk dynamics, indicating that exceeding certain levels of NPLs or LTD ratios can amplify the risk of adverse consequences. The impact of bank-specific and macroeconomic factors is consistent across both models, emphasizing the importance of sound capital management, efficient risk management strategies, and macroeconomic stability for maintaining bank resilience.

This study contributes to the understanding of the complex interplay between credit and liquidity risk by employing both linear and non-linear modeling approaches and analyzing the relationship in both causal directions. The identified thresholds offer valuable insights for Tunisian policymakers and banking institutions. Future research could expand the sample size, include additional proxies for LR (like LCR and NSFR), and explore the relationship in other contexts.

The study's limitations include a relatively small sample size (10 banks) that may limit the generalizability of the findings to other banking sectors and contexts. The use of only the LTD ratio to measure LR may also affect the robustness of the results; future studies could incorporate additional measures of liquidity risk. The focus on Tunisian banks also restricts the broader applicability of the findings.