Forecasting Adverse Financial Markets Regimes with Graph Theory

Turbulent financial market periods are characterized by low mean returns, high volatility and often by the disappearance of the diversification effect. Therefore, accurate predictions of financial turbulence can significantly improve the risk-reward-ratio of investment strategies and is therefore a lively field of academic and professional research. This paper augments the work of Baitinger and Flegel (2021a) that focused on persistent homology for the detection of turbulent financial market regimes. Precisely, we introduce an alternative turbulence indication that is based on graph theory applied to financial networks, i.e. network-based analysis. With the help of out-of-sample studies, we compare the performance of the alternative turbulence indicator with "classical" regime indicators and the persistent homology-based indicator. The results indicate that turbulence indication relying on network-based analysis is superior to existing approaches. Furthermore, the performance analysis reveals that the novel indicator successfully navigates the Covid-19 selloff and the stock market turbulence of 2022