Level shifts in stock returns driven by large shocks

This paper employs a parametric model of persistent (level) shifts in the conditional mean of stock market returns which are endogenously driven by large positive or negative return shocks. These shocks can be taken to reflect important market announcements, monetary policy regime changes and/or changes in business conditions affecting stock market. The model assumes that both the timing and size of breaks are stochastic. The last property of the model distinguishes it from other nonlinear models of the literature employed to capture level shifts in stock returns. Implementation of the model to the US stock market indicates that it can successfully capture level shifts in the mean of the aggregate return of this market which follow a cyclical pattern. Most of these shifts are triggered by negative large return shocks. The latter can be of smaller magnitude than that of the positive ones. Finally, the paper shows that the model can be employed to successfully forecast future expected stock returns.