Large cryptocurrency-portfolios: efficient sorting with leverage constraints

Abstract

Using daily data of the 100 largest cryptocurrencies, we construct the efficient sorting portfolios and the quantile-based sorting portfolios based on ten factors. We find two price factors that can well predict cryptocurrency returns. The efficient sorting portfolios outperform the traditional quantile-based portfolios and the naive 1/N portfolios. The outperformance is largely due to the use of DCC-NL estimator, which captures the dynamic of covariance matrix and meanwhile addresses the curse of dimensionality. In addition, leverage constraints are important for cryptocurrency portfolios to control their risks.