Generalizing Smooth Transition Autoregressions

We introduce a new time series model capable to parametrize the joint asymmetry in duration and length of cycles - the dynamic asymmetry - by using a particular generalization of the logistic function. The modelling strategy is discussed in detail, with particular emphasis on two asymmetry tests and relative diagnostics, whose power properties are explored via Monte Carlo experiments. Several case studies illustrate the high versatility of the new model, which is able to characterize the dynamic asymmetry in the cycle in different fields. In a rolling forecasting exercise our model beats its linear and conventional nonlinear competitors in point forecasting, while this superiority becomes less evident in density forecasting, specially when relying on robust measures. Finally, dynamic asymmetry is an important feature to take in account in uncertain environments.