On the Application of the Cross-Correlations in the Chinese Fund Market: Descriptive Properties and Scaling Behaviors

Weibing Deng, Wei Li, Xu Cai, Qiuping A. WangAdv. Complex Syst 2011 · DOI

TL;DR

This paper investigates the complex interdependencies within the Chinese fund market by analyzing cross-correlations among fund returns. It reveals that these correlations exhibit power-law distributions and intermittency, suggesting underlying hierarchical structures and the influence of a few dominant funds.

What Problem Does This Paper Address?

Financial markets are complex systems with numerous interacting agents. Understanding the relationships and dependencies between individual assets (like funds) is crucial for comprehending market dynamics, pricing mechanisms, and potential systemic risks. While previous research has explored correlations in various financial markets, this paper specifically focuses on the Chinese fund market (CFM), a rapidly growing and significant market. The authors aim to uncover the descriptive properties and scaling behaviors of cross-correlations within the CFM, providing insights into its unique characteristics and potential universal features shared with other financial systems.

Methods

The researchers employed several advanced analytical techniques:

  • Cross-Correlation Matrix Construction: They calculated the cross-correlation coefficients ($C_{ij}$) between the relative daily logarithmic returns of 88 Chinese funds over a four-year period. The relative logarithmic return was defined to account for the average market movement.
  • Statistical Analysis of Coefficients: The distribution, mean, standard deviation, skewness, and kurtosis of the cross-correlation coefficients were analyzed to understand their basic properties.
  • Scaled Factorial Moment Method: This technique was used to investigate the intermittency and self-similarity of the cross-correlation dynamics across different time scales (monthly, quarterly, yearly). Intermittency suggests that the correlations can fluctuate dramatically over time.
  • Random Matrix Theory (RMT): RMT was applied to analyze the distribution of eigenvalues of the cross-correlation matrix. This helps in distinguishing between truly significant correlations and those arising from random chance, and can reveal underlying structures.
  • Correlation Strength Definition: A measure of "correlation strength" ($S_i$) was defined for each fund as the average of the absolute values of its cross-correlation coefficients with all other funds. This quantifies a fund's overall influence on the market.

Datasets

The study utilized daily closed prices of 88 different funds from the Chinese fund market, collected from June 2005 to October 2009. This dataset covers a significant period, including market fluctuations and potential "bubbles."

Key Insights and Ideas

  • Exponential Distribution of Returns: The absolute relative daily logarithmic returns of the funds follow an exponential distribution, a finding consistent with other emerging markets.
  • Time-Dependent Correlation Patterns: The distributions of cross-correlation coefficients change over time, with distinct behaviors observed during periods of market "bubbles." This suggests that investor behavior and fund interdependencies are not static.
  • Intermittency and Self-Similarity: The scaled factorial moment analysis reveals intermittency in the cross-correlation dynamics, implying that the market's interconnectedness can change abruptly. This also points to self-similarity, where similar patterns emerge across different time scales.
  • Dominant Funds and Hierarchical Structure: Random Matrix Theory analysis shows that while most eigenvalues of the cross-correlation matrix are small, a few are significantly larger. This indicates that a small number of funds have a disproportionately large impact on the overall market correlations, suggesting a hierarchical structure.
  • Power-Law Correlation Strength: The distribution of individual fund correlation strengths also follows a power-law. This reinforces the idea that a few funds exert a strong influence on the market, while many others have a more localized impact.

Key Results

  • The relative daily logarithmic returns of the 88 funds exhibit an exponential distribution.
  • The cross-correlation coefficients show time-dependent distributions, with noticeable changes during market bubbles.
  • Intermittency was observed in the cross-correlation dynamics across monthly, quarterly, and yearly time scales, suggesting scale invariance.
  • The eigenvalue distribution of the cross-correlation matrix exhibits power-law tails, with a few isolated large eigenvalues.
  • The distribution of correlation strengths for individual funds follows a power-law, indicating the presence of a few highly influential funds. The top 15 funds with the largest correlation strengths were identified.

Limitations and Future Work

The authors acknowledge that the dependence of cross-correlations on specific time spans is minimized by considering one-year-long data blocks, but it's an area that could be further explored. They suggest that future studies could focus on community structure analysis within the Chinese fund market to gain a deeper understanding of how funds group together and influence each other. The findings also suggest that the observed scaling invariance in correlations might be related to an ultimate hierarchical structure within the market.

Summary updated

BibTeX
@article{deng2011application,
  author = {Weibing Deng and Wei Li and Xu Cai and Qiuping A. Wang},
  title = {{On the Application of the Cross-Correlations in the Chinese Fund Market: Descriptive Properties and Scaling Behaviors}},
  journal = {Adv. Complex Syst.},
  year = {2011},
  doi = {10.1142/s0219525911002871},
}
Figure