Variance Decomposition and Cryptocurrency Return Prediction
利用日内数据研究已实现方差对加密货币横截面收益的预测能力，发现高方差加密货币后续周收益更低，且负向预测性主要由正向跳跃方差和跳跃稳健方差驱动，对小市值、低流动性、散户交易活跃的加密货币更显著。
Abstract This article examines how realized variances predict cryptocurrency returns in the cross section using intraday data. We find that cryptocurrencies with higher variances exhibit lower returns in subsequent weeks. Decomposing total variances into signed jump and jump-robust variances reveals that the negative predictability is attributable to positive jump and jump-robust variances. The negative pricing effect is more pronounced for smaller cryptocurrencies with lower prices, less liquidity, more retail trading activities, and more positive sentiment. Our results suggest that cryptocurrency markets are unique because retail investors and preferences for lottery-like payoffs play important roles in the partial variance effects.