Title: Unit roots in periodic time series
Authors: Domenico Cucina - Roma Tre University (Italy) [presenting]
Francesco Battaglia - University La Sapienza, Rome (Italy)
Roberto Baragona - University La Sapienza, Rome (Italy)
Abstract: Many time series are subject to seasonal fluctuations that might not be constant over time. It has been shown that such series may be well described by Periodic AutoRegressive (PAR) models, in which each season of the year follows a possibly different AR process. When seasonality also contains a stochastic component, the problem of seasonal unit roots, associated with changing seasonality, arises. The study of seasonal unit roots in periodic autoregressive models merits attention because it allows a more complete description of the seasonal component. There exist many tests to examine if a monthly time series has seasonal unit roots for autoregressive processes. The so-called HEGY method, for example, tests the presence of unit root and considers different combinations of constant, trend and seasonal dummies. We propose an empirical method to examine whether a periodic model has one or more seasonal unit roots, and to detect them.