Using Principal Component Analysis on Crypto Correlations to Build a Diversified Portfolio
28 Pages Posted: 10 Sep 2021
Date Written: July 30, 2021
Abstract
A simple look at cryptoassets’ historical can lead us think that in recent years most have followed Bitcoin’s wake. If so, it would be very difficult to build an exposure to this market without being highly exposed to Bitcoin, and on the other hand a portfolio with many cryptos poses a great operational risk due to the lack of institutional custody. To this aim, this paper presents an updated correlation analysis of 31 crypto assets, among them and with some equity and gold indices. Furthermore, we conduct a PCA to identify the group of cryptos that present different correlation patterns and may help us build a diversified portfolio.
The correlation update shows that these cryptoassets, which account for aprox. 80% of the market, have been positively correlated since 2017 and Ether has been the asset with the highest results. These correlations increase during bear markets, especially in the current bear period started in April 2021. When analyzing Bitcoin against equity markets, we confirmed that correlation is very volatile and swings from positive to negative continuously, which makes it very difficult to use Bitcoin as an equity hedge. As a closing, we have observed that the only times that Bitcoin presented negative correlation with equity indexes coincides with times when gold also showed negative correlation, which could reveal the use of the digital asset as a store of value.
Finally, the PCA show a great number of assets from different category, size and design around a single, highly concentrated cluster. This confirms the great speculation that exists in the market, which moves all the assets en masse. When using the PCA to build a diversified portfolio we achieved better results in terms of return, risk-adjusted return and with a lower correlation to Bitcoin.
Keywords: crypto market, Bitcoin, correlation, principal component analysis, portfolio construction, portfolio optimization
JEL Classification: G11, G12, G14, C13
Suggested Citation: Suggested Citation