Momentum and market volatility: a Bayesian regime-switching model

Abstract

Our study finds that momentum is a persistent phenomenon that exhibits great variability in its strength in the UK stock market. Inspired by psychological evidence that cognitive biases can shift overtime, we conjecture that there may be two stock market states, namely, the calm and the turbulent market state, and that the switch between these two market states is governed by market volatility. Using Bayesian estimation methods, our results confirm the role of market volatility as the critical switching variable, which is also found to have additional predictive power for momentum returns in the turbulent market state. Somewhat contradictory to the findings in cross-sectional studies, we find that past returns have a negative impact on momentum profits. We also find that both winners and losers tend to perform better in the turbulent market state than in the calm market state and that losers’ outperformance is responsible for large momentum losses in the turbulent market state. Investment strategies that take advantage of the predictability of momentum dynamics outperform momentum strategies. Our findings are not readily reconciled with risk-based explanations but can be loosely explained in a behavioural framework.