Linear and nonlinear market correlations: Characterizing financial crises and portfolio optimization

Area of Science:

• Quantitative Finance
• Network Science
• Financial Econometrics

Background:

• Stock market dynamics exhibit complex interdependencies.
• Understanding these relationships is crucial for risk management and investment strategies.
• Market turbulence periods reveal significant shifts in stock interconnections.

Purpose of the Study:

• To construct and analyze complex networks of U.S. S&P500 stock returns.
• To investigate the nature and evolution of stock market dependencies.
• To identify potential early warning indicators for market anomalies and optimize investment portfolios.

Main Methods:

• Construction of complex networks using Pearson correlation and mutual information.
• Analysis of network properties during different market conditions, including crises.
• Derivation of a measure for nonlinear dependencies using surrogate data.
• Application of Markowitz mean-variance portfolio optimization incorporating nonlinear dependency measures.

Main Results:

• Complex networks effectively capture qualitative differences in stock interconnections, especially during turbulent periods.
• Nonlinear dependencies significantly increase during financial crises, contrary to expectations of reduced dependencies to linear correlations.
• Network centrality measures show potential as early warning indicators for abnormal market behavior, demonstrated during the 2008 subprime mortgage crisis.
• Integrating nonlinear dependency measures into portfolio optimization leads to significant outperformance compared to traditional methods.

Conclusions:

• Stock market interdependencies are dynamic and exhibit significant nonlinear characteristics, particularly during crises.
• Network analysis provides valuable insights into market behavior and potential risk indicators.
• Incorporating nonlinear dependency measures into portfolio optimization enhances investment performance.