Page 228 - Contributed Paper Session (CPS) - Volume 6
P. 228
CPS1904 Tímea B. et al.
hand-in-hand by providing in-house architectural services for the subject-
matter statisticians.
3. Details on the seasonal adjustment process
Seasonal adjustment (Mazzi, G. L. (2018), HCSO (2017), EUROSTAT
(2015a)) is an integrated step of the (Hungarian) Generic Statistical Business
Process Model in official statistics, therefore a proper system that guarantees
the uniform and high quality results is important for every National Statistical
Institute.
In general we can say that seasonal adjustment needs two types of
knowledge: the mathematicalstatistical knowledge which is extraordinarily
important for time series analysis, and the sound knowledge of the specific
time series. In an optimal world one person owns these knowledge but
commonly this is not the case.
In Europe there are two main scenarios to handle seasonal adjustment.
One is when one person actually owns the required knowledge (mathematical-
statistical and domain-specific) and can conduct the statistical business
process by himself/herself from the beginning until the publication. This
solution does not guarantee standard, accepted solutions within the entire
institute. The other one is the scheme of having experts of the subject-matter
domain fields who know without a doubt their time series and experts who
know everything about time series analysis (they are generally methodologist).
At the Hungarian Central Statistical Office the experts of the Methodology
Department work firmly together with the experts from the subject-matter
domain departments during the whole seasonal adjustment process (second
scenario above). Methodological experts own the mathematical, statistical
knowledge on time series analysis and the subject-matter experts know every
deep detail on the subject matter and their time series. This means that the
Hungarian Central Statistical Office performs a centralized internal system in
case of seasonal adjustment process.
The division of labour has two big runs in every year. At the beginning of
a given year, before the first publication of seasonally adjusted data of the
year, the whole annual revision of the models is made in case of every single
time series, with the help of software JDemetra+. The Methodology
Department is responsible for this task but key information is needed from the
subject-matter domains on the time series. During the year the subject-matter
domains update the data and models, however, this part of the work is also
supported (checked, and if needed, modified) by the Methodology
Department.
Figure 2 is an understandable flow chart representing our system during
a year in case of one time period where blue means tasks for the subject-
matter domain departments, green means tasks for the Methodology
217 | I S I W S C 2 0 1 9