|
Abstract:
|
Economic time-series can be or become short due to a variety of reasons: production of new indicators, changes in methodologies, definitions, nomenclatures, etc. While users always ask for seasonally adjusted figures, their production can be very problematic and their quality not necessarily in line with the requirement of official statistics. The problem does not arise when series are very short (three years or less) since most used SA methods cannot be applied. For series longer than three years, SA data can be obtained but results can be volatile and subject to relevant revisions. The aim is to present a comparative study based on the use of the most known SA methods on a selected group of macro-economic indicators. For each indicator we select three sample timespans (short, medium and long) and we compare the quality of the SA data on the common period to the three timespans. A variety of parametric and non-parametric SA quality measures is used and synthetic comparative tables are derived. The attention is focused on the stability of the SA pattern when changing the timespan, as well as on the relative performance of alternative SA methods when dealing with short time-series.
|
