Forecasting Bitcoin volatility: A new insight from the threshold regression model

Abstract

Asset returns, especially negative returns, represent the leverage effect and are found to be informative for forecasting financial market volatility. The purpose of this paper is to dig out more useful information in Bitcoin returns when we predict Bitcoin volatility. We use the threshold regression model to differentiate positive and negative returns. The threshold regression results suggest that not only a decrease in normal returns but also an increase in extremely positive returns would lead to an increase in future Bitcoin volatility. To further capture and address the leverage effect of large positive returns in the Bitcoin market, we propose a model switching method. We empirically demonstrate the significant predictive ability of large positive returns when forecasting Bitcoin volatility both in‐ and out‐of‐sample. And, our findings still hold after considering long‐horizon forecasts, additional leverage effects, jump components, and a wide series of extensions and robustness tests. In an asset allocation exercise, the proposed volatility forecasting models, which highlight the predictive ability of large positive returns, can generate larger economic gains.

Which factors drive Bitcoin volatility: Macroeconomic, technical, or both?

“…First, numerous existing studies explore the forecasting ability of relevant exogenous drivers in equity (Paye, 2012), foreign exchange, bonds, and commodities markets (Christiansen et al, 2012). With a focus on identification of the superior predictors, exploration of the forecasting ability of relevant exogenous drivers has recently become a hotpot in academia, especially in the period when Bitcoin prices and trading increase rapidly (Balcilar et al, 2017;Fang et al, 2019Fang et al, , 2020Kraaijeveld & De Smedt, 2020;Li et al, 2021;Su et al, 2020;Umar et al, 2021;Walther et al, 2019;Yin et al, 2021;Zhang et al, 2022). For example, Fang et al (2019) found that considered factors including global economic policy uncertainty, the news-based implied volatility, and realized volatility can successfully predict long-term cryptocurrency volatility.…”

Forecasting and trading Bitcoin with machine learning techniques and a hybrid volatility/sentiment leverage

“…This paper adopts daily data rather than intraday data, which may not fully capture the price movements. Zhang et al (2021) suggest that BTC returns can be used to predict BTC volatility by using aggregation of intraday information. In addition, Süssmuth (2022) provides evidence that Baidu–Google search statistics forecast BTC price dynamics at relatively high frequencies.…”

“…Due to the highly volatile property of cryptocurrency and decentralization of BCH, prediction in BTC thereby becomes the most challenging target. Prior studies focus on the traditional statistical models (e.g., GARCH), which are well documented (Gourieroux et al, 2020; Katsiampa, 2017; Zhang et al, 2021). However, only limited work has been done in cryptocurrency prediction using ML algorithms.…”

Electricity and cryptocurrency mining: An empirical contribution

“…Besides, they are prone to external shocks like other markets [ 15 , 16 ]. Despite these, the return & volatility of cryptocurrency markets, like the stock markets, are predictable given the right factors and methods [ [17] , [18] , [19] ].…”