Abstract

We set out to present the recent advancements of DSA2 and the automatic filter selection in X-11.
The development of the fractional airline model and the flexibilisation of several seasonal adjustment methods, have enabled the redesign of the DSA approach. Version 2 of DSA now incorporates both X-11 and SEATS-type decompositions, automatic choices for seasonal filters, and new diagnostics. We will delve into the specifics of the DSA2 procedure, including the testing and determination of design choices and default settings.
The classic X-11 seasonal adjustment method for monthly and quarterly economic data, known for its data-driven selection of Henderson trend filters and 3×k seasonal filters (where k ∈ {3, 5, 9}), has a long-standing tradition dating back to 1960. However, these routines have not been adopted in the recent JDemetra+ implementation of a modified X-11 method tailored to higher frequency time series, which often feature multiple seasonal patterns with potentially fractional periodicities. To address this gap, we extend the legacy selection concept based on the moving seasonality ratio by allowing for a broader family of 3×k seasonal filters and incorporating a common scheme for obtaining their asymmetric variants. This generalisation facilitates the derivation of indifference intervals, where the moving seasonality ratio is recalculated from suitably shortened observations. We will demonstrate the automatic selection routine through three examples using quarterly, monthly, and daily macroeconomic time series for Germany.