Aurora Borealis
10 September 2020 CFA Institute Journal Review

Exogenous Drivers of Bitcoin and Cryptocurrency Volatility—A Mixed Data Sampling Approach to Forecasting (Summary)

  1. Antony Jackson, CFA

CFA Institute Journal Review summarizes "Exogenous Drivers of Bitcoin and Cryptocurrency Volatility...," by T. Walther, T. Klein, and E. Bouri, from Journal of International Financial Markets, Institutions & Money, November 2019.

Within a GARCH-MIDAS framework, the authors use prediction quality to identify key drivers of volatility in cryptocurrencies, such as bitcoin, Ethereum, Ripple, and Stellar. The framework allows volatility to be split into short- and long-term components and identifies the Global Real Economic Activity as the key driver of long-term cryptocurrency volatility.

What Is the Investment Issue?

Cryptocurrency markets tend to have higher and more persistent episodes of volatility than traditional asset classes. Regulators, policymakers, and investors will benefit from volatility frameworks that look beyond modeling past daily returns and trading volume. The GARCH-mixed data sampling framework (GARCH-MIDAS) allows for the inclusion of variables that are available only at monthly frequencies and for volatility to be decomposed into short- and long-term components. This is important for investors with longer-term investment horizons.

How Did the Authors Conduct This Research?

The authors use the daily time series of various cryptocurrencies starting from May 2013 and ending in July 2019, as well as the CRIX cryptocurrency index. In order to explain the long-term component of volatility, the authors consider the monthly time series of various financial and economic variables, such as the Global Economic Policy Uncertainty index (GEPU) and the trade-weighted US dollar index.

The authors use a rolling out-of-sample window, which starts in May 2015 for bitcoin and August 2016 for the CRIX index, with forecast horizons of 1 day, 7 days, and 30 days. The forecasts are evaluated using heteroskedasticity-adjusted mean squared errors (HMSEs) and mean absolute errors (HMAEs). The authors identify outperforming models by deriving Model Confidence Sets at the 90% and 75% confidence levels.

What Are the Findings and Implications for Investors and Investment Professionals?

The monthly economic variables that appear most frequently in Model Confidence Sets are

  • Global Real Economic Activity (GREA)

  • Global Financial Stress Index (GFSI)

  • Chinese Economic Policy Uncertainty (CEPU)

  • S&P 500 Realized Volatility

The authors note that the benchmark GARCH (1, 1) model is rarely included in Model Confidence Sets. They are also unable to support previous findings that the Cboe Volatility Index (VIX) is important for forecasting the volatility of bitcoin.

Overall, the GARCH-MIDAS model with the GREA as an exogenous economic driver is the best model.

Model averaging is the second best approach and is a promising avenue for future research.

We’re using cookies, but you can turn them off in Privacy Settings.  Otherwise, you are agreeing to our use of cookies.  Accepting cookies does not mean that we are collecting personal data. Learn more in our Privacy Policy.