Seasonal and calendar related effects prevent observing the general tendency of data because of their temporary characteristics. Therefore, identification of seasonal patterns of short term indicators plays crucial role in order to make reliable comparisons between consecutive periods.
The seasonal adjustment of Turnover Indices is carried out using TRAMO-SEATS methodology. The software that is used for the application of this method is JDemetra+ developed by the National Bank of Belgium (NBB) in cooperation with the Deutsche Bundesbank and Eurostat in accordance with the Guidelines of the European Statistical System.
Process of the seasonal adjustment
Seasonal and calendar adjustment process of Turnover Indices begins at the end of each year with the identification of the models of next year. This specified model structure is kept fixed throughout the year to adjust seasonally and/or calendar effects. At the end of the year, just like the previous year, specification of econometric estimation models for the following year is done. The identified process repeats itself in a cyclical manner each year.
While 199 subtitles of Turnover Indices contain both seasonal and calendar effects, 14 subtitles contain only calendar effects, 32 subtitles contain only seasonal effects and 13 subtitles contain neither of these effects.
Direct or indirect seasonal adjustment
Seasonally and calendar adjusted figures of Turnover Indices are produced by indirect approach. Firstly the three digit NACE Rev2 classification individual subseries of Turnover Indices are seasonally adjusted and then aggregated to derive seasonally and calendar adjusted totals. The advantages of indirect approach are calculation of contributions to growth from components of main aggregates and retaining additivity.
Seasonal adjustment procedure is subject to revisions over time because of the re-estimation of seasonal component as new observations are added. These revisions are implemented on the data of the last three years excluding the current year.
Adjusted data is published in 3 different ways.
- “Calendar adjusted” data is derived from unadjusted data by removing calendar and holiday originated effects. Calendar adjusted data should be used in comparisons regarding the same month/period of the previous year.
- “Seasonally adjusted” data is derived from unadjusted data by removing effects originating from seasonal effects. Seasonally adjusted data should be used in comparisons regarding the previous month/period.
- If unadjusted data contains both calendar and holiday, and seasonal effects, “seasonally and calendar adjusted” data is derived by removing these effects. Seasonally and calendar adjusted data should be used in comparisons regarding the previous month/period