Predicting Value at Risk for Cryptocurrencies Using Generalized Random Forests

Abstract

We study the estimation and prediction of the risk measure Value at Risk for cryptocurrencies. Using Generalized Random Forests (GRF) (Athey et al., 2019) that can be adapted to specifically fit the framework of quantile prediction, we show their superior performance over other established methods such as quantile regression and CAViaR, particularly in unstable times. We investigate the small-sample prediction properties in comparison to standard techniques in a Monte Carlo simulation study. In a comprehensive empirical assessment, we study the performance not only for the major cryptocurrencies but also in the stock market. Generally, we find that GRF outperforms established methods especially in crisis situations. We further identify important predictors during such times and show their influence on forecasting over time.