A Dynamic Network for Cryptocurrencies

Abstract

Cryptocurrencies return cross-predictability yields information on risk propagation and market segmentation. To explore these effects, we build a dynamic network of cryptocurrencies based on the evolution of return cross-predictability and develop a dynamic covariate-assisted spectral clustering method to consistently estimate the latent group membership of cryptocurrencies. We show that return cross-predictability and cryptocurrencies' characteristics, including hashing algorithms and proof types, jointly determine the cryptocurrencies market segmentation. Portfolio analysis reveals that more centred cryptocurrencies in the network earn higher risk premiums.