Background
Data refer to end-of-month outstanding stocks as consolidated at the euro area level, and the associated flows, indexes
of adjusted stocks and growth rates.
Seasonal adjustment is the process of estimating and removing seasonal effects from a time series. The seasonal adjustment
procedures used by the ECB also cater for the trading-day adjustment.
The
general principles followed by the ECB in the seasonal adjustment of time series are laid down in
the ECB publication "Seasonal adjustment of monetary aggregates and HICP for the euro area" (August 2000),
pdf 881 kB,
en.
The
approach used to seasonally adjust the euro area monetary aggregates and counterparts relies on a
multiplicative decomposition using the X-12-ARIMA method. For internal purposes, the model-based approach of TRAMO-SEATS
is also used. As the series are collected as end-of-month stocks, trading-day adjustment is performed where applicable.
Outliers are taken into consideration in order to minimise distortions to the estimated seasonal and trading-day components.
(technical notes on the seasonal adjustment of monetary series:
pdf 13 kB,
en )
To ensure the
additivity of the seasonally adjusted components to the seasonally adjusted aggregates,
some of the seasonally adjusted series, in particular M3, are derived indirectly. From a numerical point of view, the
difference between direct and indirect estimates of euro area monetary aggregates is regularly monitored and has proved to
be generally negligible.
Forecast seasonal factors are used. In addition,
seasonal factors are revised whenever required, e.g. in
the case of large data revisions or the statistical inadequacy of the models used when new data are incorporated. For this
purpose, a concurrent adjustment is run on a monthly basis in order to assess the validity of the seasonal factors in use.
Specific issues are raised by the seasonal adjustment of relatively short time series
(technical note:
pdf 11 kB,
en ).
Compared to longer time series of, say, 20 or more years of monthly data, the seasonal adjusted results for time series of around
five years tend to be particularly sensitive to outliers and instability in the seasonal patterns.
