Spillover and predictability of volatility of 50 major cryptocurrencies : evidence from a LASSO-regularized quantile VAR

Abstract

Previous studies examine spillover effects across the volatility of several cryptocurrencies in the mean or across quantiles without addressing the issue of high dimensionality. Using a large dataset of 50 cryptocurrencies, we employ a LASSO-regularized Quantile VAR framework and show that spillover effects differ across low, medium, and high volatility regimes, especially when evaluated dynamically over time, with sharp increases around tail events such as the war in Ukraine. Importantly, we demonstrate that the LASSO-QVAR model delivers statistically significant forecasting improvements over its univariate counterpart, underscoring the role of interconnectedness in enhancing volatility prediction across cryptocurrencies.

HIGHLIGHTS • Consider a large dataset of 50 cryptocurrencies. • Use LASSO-regularized Quantile VAR model to tackle the issue of high dimensionality. • Show that the dynamic spillovers differ across low, medium, and high volatility regimes. • They exhibit sharp increases around tail events such as the war in Ukraine. • LASSO-QVAR model delivers significant forecasting improvements.