A moving-window bayesian network model for assessing systemic risk in financial markets.

Area of Science:

• Financial economics
• Network science
• Time series analysis

Background:

• Systemic risk is a critical concern in financial systems, often linked to interconnectedness.
• The "too connected to fail" concept highlights the importance of network structures in financial stability.
• Existing methods for measuring systemic risk may not fully capture dynamic network changes.

Purpose of the Study:

• To develop a novel approach for quantifying systemic risk using dynamic network analysis.
• To introduce a new metric, "order distance," for assessing changes in stock market network topology.
• To evaluate the predictive power of network statistics and order distance for extreme market events.

Main Methods:

• Time series Bayesian networks were constructed to model stock return dependencies.
• Topological orders of stocks within these networks were analyzed.
• An "order distance" metric was developed to quantify changes in topological order.
• LASSO regression was employed to predict extreme absolute returns using network statistics and order distance.

Main Results:

• Network statistics derived from time series Bayesian networks significantly enhance predictability.
• The novel "order distance" metric proves effective in forecasting extreme market risks.
• The methodology provides valuable insights into the assessment and prediction of systemic risk.

Conclusions:

• Dynamic network analysis, particularly using Bayesian networks and order distance, offers a powerful tool for understanding and predicting systemic risk.
• The findings suggest that monitoring network topology changes can serve as an early warning system for financial instability.
• This approach improves the assessment of systemic risk beyond traditional financial metrics.