Predicting Cryptocurrency Returns Using US Macroeconomic Variables

Abstract

In this article, we undertake a comprehensive empirical investigation to evaluate the predictive ability of 26 US macroeconomic indicators for the returns of six cryptocurrencies. The results of the in-sample test show that several macroeconomic variables, such as the three-month Treasury bill rate, default yield spread, term spread, and economic policy uncertainty index during the full sample period, as well as the term spread, VIX, three-month Treasury bill rate, dividend-price ratio, and economic policy uncertainty index during the monetary policy intervention period, show the ability to forecast returns for a minimum of two cryptocurrencies. We also find enhancements to these predictive capabilities during the monetary policy intervention period. The out-of-sample test validates the effectiveness of our in-sample findings and unveils additional macroeconomic variables that also have predictive ability. Our results show superior forecasting ability when utilizing these combination methods. Our findings also suggest that incorporating global factors, such as the MSCI All Country World Index, can improve forecasting ability. During the cryptocurrency bubble period, we find that the 10-year US government bond yield, dividend yield ratio, capacity utilization rate, and US economic policy uncertainty index can forecast at least two cryptocurrency returns. Furthermore, we observe that in addition to US macroeconomic variables, certain trade-related variables from other countries, such as the annual growth rates of export values for the United States, United Kingdom, and China, as well as the annual growth rates of import values for Japan and Italy, can forecast returns for at least two cryptocurrencies.