Testing for mean reversion in Bitcoin returns with Gibbs-sampling-augmented randomization

Abstract

In the present paper, we attempt to verify whether the Bitcoin log-returns are mean reverted in the presence of heteroskedastic disturbances driven by a mixture distribution. To tackle this problem, we use the autoregression test of mean reversion based on the Gibbs-sampling-augmented randomization methodology. In general, our results indicated that Bitcoin is mean averting for different returns horizons, model specifications and for sub-sample periods, which show the explosive characteristic of the Bitcoin in the period of analysis from 2010 to 2019.