While the convergence of sustainable finance and rapid technological innovation is changing the way portfolios are constructed, there is a lack of clear understanding of how risk moves between green bonds and high-growth technology assets, especially during periods of extreme volatility. In this study, we examined return spillovers among green bonds, major US technology equities, and AI/cybersecurity ETFs. We applied a novel Time-Varying Parameter Frequency Quantile VAR (TVP-FQVAR) framework, which decomposed connectedness across investment horizons and market states, capturing dynamics that traditional linear and static models cannot detect. We further employed SVM and XGBoost machine learning algorithms to forecast systemic connectedness using macro-financial indicators, including a novel gold-to-Bitcoin ratio. Three key findings appeared. First, overall system connectedness was high but concentrated in the short term, while long-run structural spillovers remained weak. Second, green bonds were consistently net receivers of shocks, acting as portfolio stabilizers, while technology and AI/cybersecurity ETFs were net transmitters. Third, our time-varying extension revealed that market distress (the lower return quantile) is characterized by substantial parameter instability: NVIDIA and Tesla emerged as the system’s dominant net shock transmitters only once time variation was properly modeled, a dynamic concealed by static models. The strongest machine learning predictors of tail-risk connectedness were lagged VIX, inflation expectations, and the gold-to-Bitcoin ratio. These results carry direct implications for policymakers and investors looking to adopt dynamic, regime-aware hedging strategies in the rapidly evolving sustainable-tech financial landscape.
Time-Varying Tail connectedness between green bonds, US tech equities, and AI/Cybersecurity ETFs: A TVP-FQVAR and machine learning approach
Arturo Leccadito
2026-01-01
Abstract
While the convergence of sustainable finance and rapid technological innovation is changing the way portfolios are constructed, there is a lack of clear understanding of how risk moves between green bonds and high-growth technology assets, especially during periods of extreme volatility. In this study, we examined return spillovers among green bonds, major US technology equities, and AI/cybersecurity ETFs. We applied a novel Time-Varying Parameter Frequency Quantile VAR (TVP-FQVAR) framework, which decomposed connectedness across investment horizons and market states, capturing dynamics that traditional linear and static models cannot detect. We further employed SVM and XGBoost machine learning algorithms to forecast systemic connectedness using macro-financial indicators, including a novel gold-to-Bitcoin ratio. Three key findings appeared. First, overall system connectedness was high but concentrated in the short term, while long-run structural spillovers remained weak. Second, green bonds were consistently net receivers of shocks, acting as portfolio stabilizers, while technology and AI/cybersecurity ETFs were net transmitters. Third, our time-varying extension revealed that market distress (the lower return quantile) is characterized by substantial parameter instability: NVIDIA and Tesla emerged as the system’s dominant net shock transmitters only once time variation was properly modeled, a dynamic concealed by static models. The strongest machine learning predictors of tail-risk connectedness were lagged VIX, inflation expectations, and the gold-to-Bitcoin ratio. These results carry direct implications for policymakers and investors looking to adopt dynamic, regime-aware hedging strategies in the rapidly evolving sustainable-tech financial landscape.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


