Liquidity Risk Dynamics in the Iranian Interbank Network under Macroeconomic Stabilization Policies: A Support Vector Machine (SVM) Approach

Abstract

This study aimed to investigate liquidity risk dynamics in the Iranian interbank network and evaluate the impact of macroeconomic stabilization policies on systemic liquidity risk using a Support Vector Machine (SVM) framework. The study employed a quantitative and applied research design based on daily interbank market data from 2013 to 2023. The sample consisted of 32 banks and credit institutions operating within Iran’s interbank network. Systemic liquidity risk was calculated using network-based and macro-financial indicators and classified into low, moderate, and severe risk categories. Ten explanatory variables, including network centrality, Herfindahl–Hirschman Index (HHI), repo rate deviation from the policy corridor, overdraft volume, interbank interest rate volatility, illiquid asset ratio, inflation rate, exchange rate, monetary policy measure, and liquidity contagion index, were selected as model inputs. Linear, polynomial, and radial basis function (RBF) kernels were evaluated, and optimal hyperparameters were identified through grid search and five-fold cross-validation. The RBF-SVM model with C=10 and γ=0.05 demonstrated superior predictive performance. During the stabilization period, the model achieved an overall accuracy of 94.3%, an F1-score of 0.941, and an ROC-AUC value of 0.97 for severe liquidity risk classification. Comparative analysis revealed that stabilization policies reduced interbank interest rate volatility by 43.6%. However, liquidity risk persistence increased substantially, with the first-order autocorrelation coefficient rising from 0.52 to 0.78, while market concentration increased as the HHI rose from 0.24 to 0.41. Feature importance analysis identified network centrality, HHI concentration, and repo rate deviation from the policy corridor as the strongest predictors of severe liquidity risk. Although macroeconomic stabilization policies successfully reduced short-term market volatility, they simultaneously increased liquidity risk persistence and concentration within the interbank network. The proposed SVM model provides a highly effective early-warning mechanism for detecting systemic liquidity stress and can support policymakers and central banks in strengthening financial stability and liquidity risk management.