Bayesian change point analysis of Bitcoin returns

Abstract

This paper studies existence of structural breaks in the average return and volatility of the Bitcoin price. We utilize a Bayesian change point model to detect structural breaks and to partition the time series into segments. We find that structural breaks in average returns and volatility of Bitcoin are very frequent. By merging segments with similar properties into regimes we identify several regimes with positive average returns and one regime with negative average returns. Across regimes, higher volatility is associated with higher average returns, with exception of the most volatile regime, which is the only regime with negative average returns.