5. Concluding
remarks In this paper we have introduced a STAR-type time series model for inflation, where regime switches are based on the relative size of the forecast of the underlying series. The specification allows low and high inflation forecasts to have different impacts on future values of inflation. The model is applied to GDP deflator-based US inflation rate, where we use the Michigan Inflation Expectationeries as inflation forecasts. Since the level of inflation changes over time, we include a time-varying threshold parameter in the L-STAR specification, such that the relative size of the forecast determines regime changes. Our empirical results show that (i) forecasts lead to regime changes, and have an impact on the level of inflation; (ii) forecasts seem to signal regime switches better than lagged inflation in economically stable periods, and similarly well in other periods; (iii) a positive (negative) shock to the inflation forecast results in actions that increase (decrease) the inflation rate, which is in line with the expectations trap literature; (iv) the absorption time of shocks in the forecast of inflation is about four quarters; and (v) a counterfactual scenario where forecasts during the financial crisis in 2009 were assumed to be correct would have resulted in a lower level of inflation in the subsequent quarters.
