Bitcoin in Portfolio Selection: A Multivariate Distribution Approach

Abstract

This study presents a multivariate study regarding Bitcoin and its interactions with other financial assets of different classes. This is done by adjusting a multivariate semi heavy-tailed distribution to portfolios containing indexes, currencies, and commodities and one cryptocurrency. Later, a rolling window is deployed to obtain the dynamic parameters of the distribution in a weekly basis. With a Markowitz specification problem, the optimal portfolio weights are computed dynamically using the parameters of the multivariate NIG distribution as inputs. The results provide evidence that correlations of Bitcoin with other assets may provide certain degree of diversification to portfolios; nevertheless, the high volatility of this asset makes it unpractical to employ in significant weights. This paper is relevant for researchers and practitioners as it provides a new tool to manage portfolios with cryptocurrencies and more reliable weights to the asset allocation.