Modeling Systemic Risk Dynamics in Iranian Financial Markets: A Multi Regime Framework Integrating ARIMA, DCC GARCH, GMM, and Markov Switching Models

Purpose: This study develops an advanced hybrid framework to model the dynamics of systemic risk in Iran’s financial markets. In structurally fragile economies characterized by acute political and economic volatility—as in Iran—the interdependencies among key assets and the intensity of risk spillovers vary significantly across different market behavior regimes. Consequently, the primary objective of this research is to analyze both the dependency structure and the transmission mechanisms of systemic risk across three major assets—gold, the free-market USD/IRR exchange rate, and the overall index of the Tehran Stock Exchange—within distinct behavioral regimes. To achieve this, we integrate multiple modeling approaches, enabling rigorous identification of latent regimes and precise measurement of systemic risk under varying market conditions.

Method: The proposed framework consists of five sequential stages. First, daily return series for the three assets were filtered through ARIMA models to remove linear trends and autocorrelations, isolating non-linear residuals. Second, GARCH(1,1) was employed on each series to estimate conditional volatilities. Third, DCC-GARCH was applied to capture time-varying conditional correlations between asset pairs. To detect regime shifts and structural breaks in market behavior, two complementary techniques were utilized: the Markov Switching model, which classifies observations into low- and high-volatility regimes based on transition probabilities, and the Gaussian Mixture Model (GMM), which probabilistically clusters observations into three behavioral regimes—tranquil, semi-unstable, and crisis—according to their statistical features. Finally, the Conditional Value-at-Risk (COVAR) metric was computed for each asset pair within each identified regime, providing a comprehensive assessment of systemic risk transmission intensity.

Findings: Analysis of daily data from March 21, 2015 to March 20, 2024 reveals that the intensity and structure of systemic risk transmission in Iran’s financial markets are highly regime-dependent. In the “tranquil” regime, conditional correlations between gold and the USD/IRR exchange rate, as well as between gold and the Tehran Stock Exchange (TSE) index, fall below 0.30, and average CoVaR values remain under 0.15. As volatility rises into the “semi-unstable” regime, these measures increase to roughly 0.40–0.50. During the “crisis” regime, gold–USD/IRR correlation exceeds 0.80 while gold–TSE correlation surpasses 0.30; COVAR for the gold⇄USD pair peaks at 0.78 and for gold⇄TSE at 0.30, reflecting flight-to-safety flows into currency and precious metals at times of acute stress. Notably, exchange rate⇄equity and exchange rate⇄gold linkages in intermediate regimes not only serve as early warning indicators but occasionally exhibit higher COVAR values in ostensibly tranquil periods than in semi-unstable ones—an outcome indicative of latent negative expectations or erratic policy shifts. Overall, integrating DCC-GARCH, Markov Switching, and GMM enabled a more granular regime classification and precise measurement of risk spillovers, underscoring the necessity of accounting for nonlinear, multi-layered contagion effects when analyzing systemic risk in volatile economies.

Conclusion: This study emphasizes that in economies marked by structural fragility and severe political–economic volatility, employing regime-sensitive and dynamic frameworks is essential for systemic risk analysis. This framework not only enables the early identification of high-risk and intermediate-risk periods but also serves as an effective tool for policymakers, regulatory authorities, and professional investors in designing macroprudential policies, optimizing capital allocation, and managing systemic risk.