Learning Cross-dependency of Cryptocurrencies from Multivariate Time Series Models
8 Pages Posted: 18 Feb 2020
Date Written: January 22, 2020
The ever-growing volume of cryptocurrency transactions including those from major banks indicates the importance to understand the new cryptocurrency market. We analyse several cryptocurrencies simultaneously to study their cross-dependency while allowing for their wide volatility, high kurtosis and strong persistence. The vector autoregressive moving average model with Student's t innovations is proposed to capture these features. We consider four cryptocurrencies, namely Bitcoin, Ripple, Litecoin and Dash, which have top market capitalisation and estimate the model using the computational efficient expectation/conditional maximisation algorithm. We interpret the results in relation to their technological setups.
Keywords: Student's t distribution, vector ARMA model, persistence, ECM algorithm, cryptocurrencies
JEL Classification: C01
Suggested Citation: Suggested Citation