Deterministic versus stochastic seasonal fractional integration and structural breaks

摘要

This paper considers a general model which allows for both deterministic and stochastic forms of season-ality, including fractional (stationary and nonstationary) seasonal orders of integration, and also incorporating endoge-nously determined structural breaks. Monte Carlo analysis shows that, in the case of a single break, the suggested procedure performs well even in small samples, accurately capturing the seasonal properties of the series, and correctly detecting the break date. As an illustration, the model is estimated using four US series (output, consumption, imports and exports). The results suggest that the seasonal patterns of these variables have changed over time: specifically, in the second subsample the systematic component of season-ality becomes insignificant, whilst the degree of persistence increases.