A Markov Switching Approach to Herding

Existing models of herding suffer from the drawback that conventional measures assume it is constant over time and independent of the state of the economy. This paper proposes a Markov switching herding model which supports the view that herding is not constant. By means of time-varying transition probabilities, the model is able to link changes in herding behavior with proxies for sentiment, the VIX, and the term structure. For the US stock market our estimates reveals that during periods of high volatility, investors disproportionately rely on fundamentals rather than on market consensus. Existing theory supports such a conclusion. Some policy implications are also drawn.