Equity Premium Forecasts with an Unknown Number of Structural Breaks

Estimation of models with structural breaks usually assumes a pre-specified number of breaks. Previous models which do allow an endogenously determined number of breaks require a simple structural model, and rarely allow for information transfer across the break. We introduce a methodology that allows the number of breaks to be determined endogenously and including an economically motivated model of transition regimes between each break. We demonstrate the usefulness of our approach for forecasts of the equity premium. We find the demonstrated success of the historical average can be improved upon by an economic model with theory informed priors estimated using our methodology.