A common practice nowadays for a National Statistical Institute (NSI ), when dealing with systems of time series collected on sub-annual basis, is to perform seasonal adjustment (SA ) in order to help users to interpret published statistics. By separating the non-seasonal part from the seasonal and calendar effects a user is likely to obtain a refined picture about the underlying movement from the time series observations. Hence, the SA -procedure eliminates the estimated seasonal and calendar effects from the original time series and obtain the SA estimates. Such estimates are likely to reveal what is new in a time series, which is a crucial issue related to seasonal adjustment. Hence, SA may be viewed as an aid in decision making, usually used for comparisons between different regular periods in time (month-to-month, quarter-to-quarter, etc.) but also for forecasting purposes and for model-building. For example, SA of macroeconomic indicators is useful for policy makers and other users because of the need for understanding repetitive fluctuations in economic activity (business-cycles) as well as the short-term and the long-term movements in time series. These effects are in a SA -procedure regularly expressed in terms of a unified trend-cycle component (see e.g. Statistics Canada, 2009; ABS , 2008).
Since SA is a modelling procedure which transforms the original data in order to obtain the estimates a natural question is how reliable these estimates are. Further issues usually associated with SA are reliability and quality with respect to benefits and costs associated with the procedure in question. Some other issues, such as revisions, outlier treatment, aggregation and data presentation are also common to different domains of statistical production which necessitates standardized, coherent and consistent treatment of SA -procedures.
A NSI should also take care about the needs of both the internal and external users, which typically implies shifting focus from a pure methodological aspect to some other (perceived) quality aspects. Balancing between these two aspects is recommended since statistics should be of high quality but also easily interpretable for users.
A vast and very detailed literature about issues related to SA is already available to the public. See e.g. IMF (2001), ECB (2003), Dagum and Cholette (2006), European Communities (2001) etc. The websites of some prominent statistical offices and developers of statistical software offer detailed information about the related procedures (e.g. Statistics Canada, 2009; ABS , 2008; U.S. Census Bureau, 2012; Bank of Spain, 2012; Koopman and Lee, 2010). Also, the European Statistical System (ESS ) developed a set of guidelines on seasonal adjustment (Eurostat , 2009) and the new software Demetra+ (Eurostat , 2012). Although the ESS Guidelines provide a set of recommendations for the best practices, this document by its nature does not give a comprehensive introduction to seasonal adjustment for the non-specialists and typical users at a NSI .
Hence, in this and related modules the main focus is put on aggregating information about SA from different sources and experiences in order to assist the users at NSIs with relevant easy-to-read information and references to the more detailed technical and methodological description.
To read the entire document, please access the pdf file (link under "Related Documents" on the right-hand-side of this page).
Your feedback is appreciated. Please send your remarks, suggestions for improvement, etc. to email@example.com.