ANALYSIS OF CAUSALITY BETWEEN BITCOIN AND ETHEREUM USING TRANSFER ENTROPY

Abstract

In the literature causality and information flow between time series are investigated. To analyze this relationship several causality tests are proposed. Mostly used causality tests are Granger, Toda-Yamamoto and Hatemi-J causality tests. These tests indicate whether there is causality and the direction of causality but do not measure the strength of causality. A new and information theory-based method transfer entropy can be used to analyze causality and information flow between time series. Unlike Granger, Toda-Yamamoto and Hatemi-J causality tests transfer entropy can detect nonlinear relationships between time series. Cryptocurrencies have gained significant importance in recent years. The most popular cryptocurrencies are Bitcoin and Ethereum. Demonstrating information flow and causality between these two cryptocurrencies are valuable for the investors. In this study we analyzed causality and information flow between Bitcoin and Ethereum by using transfer entropy method. We adopted a sliding window approach to demonstrate how transfer entropy between these two cryptocurrencies change through time. We used different window lengths to demonstrate whether obtained results are robust. In the analysis two datasets are used. These are daily and hourly data. In hourly data a significant causality is detected. However, in the daily data no evidence for causality is found.