Practical Applications of What Information Variables Predict Bitcoin Returns? A Dimension-Reduction Approach

Abstract

In What Information Variables Predict Bitcoin Returns? A Dimension-Reduction Approach from the Spring 2023 issue of The Journal of Alternative Investments, Sang Baum Kang of the Illinois Institute of Technology, Yao Xie of Morningstar, and Jialin Zhao of St. Mary’s University found that blockchain technology, stress level, and investor sentiment play important roles in predicting bitcoin returns. The authors use 25 theoretically motivated explanatory variables falling within these three categories and two others, including macroeconomic variables and other assets, such as gold, to predict bitcoin returns with several dimension-reduction techniques. These techniques are used to eliminate variables with redundant information and avoid problems associated with multicollinearity and noise. Given the recent growth in data availability, dimension reduction is an increasingly relevant issue. The importance of variables within the five categories varied over time. Interestingly, macroeconomic variables and variables in the “other assets” category were unimportant, except for during the Covid-19 period. Simulating dynamic trading strategies based on predictions, the authors show that the “three-pass regression filter” performed best relative to their other dimension-reduction approaches.