Source: MIL-OSI Submissions
14 April 2021
Below is an update on our approach to seasonal
adjustment in light of COVID-19-related disruptions to activity, which
have affected our usual methods. We are conducting a review of some of
our seasonally adjusted series and will introduce any recommended changes
in our upcoming March 2021 releases.
Impact of COVID-19 on seasonal adjustment
The COVID-19 lockdown and restrictions have
caused abrupt changes to actual values for many time series.
Our standard seasonal adjustment method (X13
ARIMA SEATS), was not able to handle the extreme movements in data points
resulting from COVID-19-related disruptions in activity. X13 ARIMA allocated
too much of the unusual value to the seasonal and trend components instead
of the irregular component.
Untreated, this would have significantly
altered the seasonal factors and caused large and undesirable revisions
to the seasonally adjusted time series. As the COVID-19 disruption in the
June 2020 quarter was an abrupt shock, it should be reflected in the irregular
component and not affect historical seasonal patterns; allowing this would
be incorrect and misleading.
To remedy this, we have identified and treated
unusual data points for affected time series using additive outliers so
that they were attributed to the irregular component.
For further information, see Impact
of COVID-19 on seasonally adjusted and trend series.
Visit our website to read this methods
on COVID-19 and seasonal adjustment