Dynamic multifactor clustering of financial networks

Area of Science:

• Quantitative Finance
• Financial Econometrics
• Network Analysis

Background:

• Financial instruments exhibit complex correlation structures influenced by multiple factors.
• Traditional clustering methods struggle to capture overlapping community structures in financial markets.
• Understanding these structures is crucial for portfolio management and risk assessment.

Purpose of the Study:

• To investigate the hierarchical clustering of financial instruments based on industry and geography.
• To develop and apply novel methods for disentangling sector and geographical influences on correlations.
• To analyze the impact of the 2008 financial crisis on geographical clustering in stock markets.

Main Methods:

• Utilized robust regression techniques to isolate and remove the effects of sector and geography from the correlation matrix.
• Employed advanced statistical analysis to detect and quantify clustering patterns.
• Compared pre- and post-2008 financial crisis data to assess changes in correlation structures.

Main Results:

• Identified a hierarchical clustering pattern where industry membership dominates, with geographical subclusters within sectors.
• Demonstrated the effectiveness of robust regression in revealing underlying correlation structures.
• Observed a significant increase in geographical clustering immediately following the 2008 financial crisis.

Conclusions:

• Standard clustering techniques are insufficient for capturing complex, multi-layered financial market structures.
• Robust regression offers a powerful tool for analyzing financial instrument correlations, disentangling sector and geographical influences.
• The 2008 financial crisis marked a shift, increasing the importance of geography in stock market clustering.