A1238
Title: New proposal for seasonal adjustment of long time series
Authors: Cheyenne Amoroso - University of Coruna (Spain) [presenting]
Abstract: A common economic task is the seasonal adjustment of time series, which involves removing the seasonal component from the data. Currently, at the National Statistics Institute (INE), this task is performed using the Tramo-Seats methodology. The time series currently being processed extend over many years, which generally complicates the identification of a single reg-ARIMA model that adequately describes the behavior of the entire series. Moreover, the data suggest that the behavior of the series has changed following the 2008 crisis. Motivated by the aforementioned issues, new general methodologies are suggested to perform seasonal adjustment in a long time series with two identified models and a transition period. The series before and after the event are considered modellable using ARIMA models, while the transition period is modeled as a weighted average of the other two events through a time-dependent weighting function. The proposals are assessed through an exhaustive simulation study aimed not only at verifying the gains compared to classical methodologies but also at evaluating their robustness in unfavorable scenarios.
