A quantile approach to US GNP

In this paper we fitted a quantile self-exciting threshold autoregressive (QSETAR) time series model to the growth rate of real US GNP. We also presented a forecasting method for QSETAR models. This forecasting method makes it possible to obtain the predictive quantiles and predictive distribution function of xt+m given xt for m > 0, and hence any quantities of interest can be derived. Therefore, this new approach allows us to study the US GNP from a distribution point view, rather than from a mean point of view. The results obtained in this paper show that the method works very well in practice.