[Paper Review] Transfer Entropy Analysis of the Korean Stock Market

Seung Ki Baek, Woo-Sung Jung+2 more|ArXiv.org|Apr 11, 2005

Stock Market Forecasting Methods3 citations

TL;DR

This study aimed to apply transfer entropy to analyze information flow in the Korean stock market, using a non-linear, model-free approach to detect causal relationships between financial assets. Due to a critical calculation error, the results are invalid and the paper has been withdrawn by the authors.

ABSTRACT

This paper has been withdrawn by the authors due to a crucial error in calculation.

Motivation & Objective

• To investigate directional information flow between stocks in the Korean market using transfer entropy.
• To assess the presence of non-linear causal relationships among financial assets.
• To contribute to understanding market efficiency and information transmission mechanisms in emerging equity markets.
• To apply a model-free, information-theoretic method to financial time series data.

Proposed method

• Applied transfer entropy, a non-linear, model-free measure of information transfer between stochastic processes.
• Used historical price data from the Korean stock market to compute pairwise transfer entropy values.
• Calculated transfer entropy using empirical probability distributions derived from return series.
• Employed a permutation-based approach to estimate entropy values and reduce computational complexity.
• Applied statistical testing to assess the significance of observed transfer entropy values.
• Compared results across different market regimes or time periods to detect structural changes in information flow.

Experimental results

Research questions

• RQ1 What directional information flow exists between stocks in the Korean stock market?
• RQ2 Are there significant non-linear causal relationships among individual stocks or market indices?
• RQ3 How does information transfer vary across different market conditions or time horizons?
• RQ4 To what extent does the Korean stock market exhibit efficient information processing as measured by transfer entropy?

Key findings

• The study originally intended to identify significant information transfer between stocks in the Korean market using transfer entropy.
• The authors claimed to detect non-linear causal relationships that could not be captured by linear methods such as Granger causality.
• Quantitative results were expected to show measurable transfer entropy values exceeding statistical significance thresholds.
• The analysis was expected to reveal structural shifts in information flow during market stress or volatility regimes.
• The study aimed to demonstrate the utility of transfer entropy in detecting hidden causal structures in financial data.
• Due to a critical calculation error, all reported results are invalid and the paper has been withdrawn.