Hybrid Approach for Assessing Systemic Risks Generated by Cryptocurrencies in Stock Markets: A Combination of Dynamic Spillover Networks, Conditional Quantile Regression, and Extreme Value Theory

Abstract

Blockchain technology and the rise of cryptocurrencies have generated challenges for the stability of the financial system and its ability to measure and transmit risks due to their characteristics such as extreme volatility, non-linear dynamic systems, time-varying dependence structures, and presence of extremes. Therefore, this study suggests a hybrid methodology to assess the systemic risk caused by cryptocurrencies and the transmission mechanism of systemic risk from cryptocurrencies to stock markets. The proposed hybrid framework consists of three complementary components: a dynamic financial spillover network, conditional quantile regression (CQR), and extreme value theory (EVT). Empirical data were collected on the closing prices of Bitcoin and stock market indices from 2020 to 2024. Returns were calculated as daily logarithmic changes of the empirical series. First, a dynamic spillover index with a 60-day rolling window was applied to capture the time-varying dependence between the cryptocurrency and stock markets. Second, CQR was applied to determine the asymmetric dependence of Bitcoin returns. Third, EVT was used to model the tail dependence structure to capture the extreme loss events. Finally, the outputs from each step were integrated to construct a systemic risk index. The results evidence that the dependence structure between the cryptocurrency and stock markets varies over time. Indeed, estimates from the CQR show that the parameter associated with Bitcoin increases from 0.092 under a normal market environment to 0.171 at the 1% lower-tail quantile, suggesting an increase in the intensity of risk transmission in times of financial stress. The results from the EVT tests confirm heavy-tailedness and an increase in the probability of losses. Moreover, sensitivity analysis shows that the composite index performs better than the reduced-form models, as the omission of any component causes the systemic risk measure s performance to deteriorate

Keywords : Systemic risk; Cryptocurrencies; Dynamic spillover networks; Conditional quantile regression; Extreme Value Theory (EVT); Financial contagion