Improving the predictability of stock returns with Bitcoin prices

Abstract

Abstract This paper examines the role of Bitcoin prices (BTC hereafter) in stock return predictability of G7 countries. For completeness, the existing predictive models for stock returns that account for both country-specific and common factors are singly and jointly compared with BTC-based predictive model. The analyses are conducted for both in- and out-of-sample forecasts with multiple forecast horizons. The following results are evident in the paper. First, the paper finds evidence in favour of BTC in terms of individual in-sample forecast performance albeit the reverse for interest rate. However, the common factors appear to compete fairly with Bitcoin as the former consistently outperform the latter in three countries. Second, in terms of forecast combination, the stock returns of the G7 countries are better predicted by BTC-based model than their respective macroeconomic variables combined except for Japan. Third, while the out-of-sample forecast results for short forecast horizon reflect those obtained for the in-sample period; the performance however seems to diminish over longer forecast horizon. This is underscored by the fact that investors in the Bitcoin market often speculate over short periods due to the market tendency for higher volatility (risk). Fourth, we also test whether accounting for structural breaks matters in the BTC-based predictive model and the results largely suggest otherwise except for the out-of-sample forecast of USA. Overall, the predictive power of Bitcoin can be exploited when modelling stock returns particularly over the period coinciding with high volumes of Bitcoin transactions.