Investment Index Construction from Information Propagation Based on Transfer Entropy

Abstract

In recent years, the number of individual investors continues to increase in financial markets. However, the large information gap between individual and institutional investors unduly impairs individual investors, which may negatively influence the market. In this study, we propose a new investment index that focuses on the relationships among stocks to help manage the risk of individual investors. The relationships among stocks have often been analyzed by a cross-correlation matrix method. However, such methods are strongly influenced by irregular events, including drawdowns. Therefore, we employed a transfer entropy method to analyze the relationships among stocks. Transfer entropy is a sequence analysis method proposed by Schreiber that is robust for irregular events. First, we applied the partial correlation and transfer entropy methods to test the data and confirm robustness for unexpected events. Next, we generated stock-networks by transfer entropy that represents the relationships among stocks. Finally, we proposed an investment index that is calculated from stock-networks which are generated from transfer entropy. We compared the proposed investment index with long positions and obtained higher performance investment by our proposed method than with a long position.