Revision of the ESS Guidelines on seasonal adjustment By Gian Luigi Mazzi and Rosa Ruggeri Cannata Seasonal Adjustment User Group 16 th October 2012.

Revision of the ESS Guidelines on seasonal adjustment

By Gian Luigi Mazzi and Rosa Ruggeri Cannata

Seasonal Adjustment User Group16th October 2012

Background

ESS guidelines on seasonal adjustment were endorsed by the ESSC and the CMFB in 2007

They constitute the first attempt to further harmonize the PEEIs production process following requests formulated at several occasions by the ECOFIN Council (Recommendations EFC SRs)

Progress in the last 5 years

ESS guidelines on seasonal adjustment have represented a very important reference for both Eurostat and member states

Thanks to their implementation the harmonisation of seasonal adjustment practices has increased:

– Differences among sectors and countries have considerably decreased

– Tangible benefits in terms of comparison of indicators

Need for updating

during the global economic and financial crisis, it appeared clear that the guidelines did not provide sufficient guidance concerning how to proceed in periods of time characterised by high level of uncertainty and big shocks

some relevant aspects were covered only partially or not covered at all by the guidelines

They were less directive for data producers than revision policy guidelines

A task force to revise the guidelines

Eurostat decided to bring to the attention of the SASG the possibility of producing a revised version of the seasonal adjustment guidelines

In July 2011 the SASG has decide to convey a small task force to revise the guidelines

TF participants: Mark Boxall (ECB), Gary Brown (ONS), Dario Buono (Eurostat – B1), Robert Kirchner (Deutsche Bundesbank), Dominique Ladiray (INSEE), Agustin Maravall (Bank of Spain), Gian Luigi Mazzi (Eurostat – C4), Pilar Rey Del Castillo (Eurostat – B1), Rosa Ruggeri-Cannata (Eurostat – C4)

Main lines for the revisions of the guidelines

ESS guidelines on SA are widely accepted and implemented; they should not be re-written but only updated where necessary; continuity should be preserved

ESS guidelines on SA could be reviewed in order to be more harmonised with those on revision policy

Some options and alternatives should be clarified and made clearly distinct, taking care of being more directive when necessary

A detailed analysis of the foreseen changes

Section 0: Maintain section 0 with some shortening

A New section 1 will be added with:

General policy at European and Member States level Domain specific SA policies Need of consistency within each specific domain between Eurostat

and MSs and between synthetic and base statistics domains Dissemination of general and domain specific policies (website,

documents, etc) Stability over time

Current section 1 – Pre-treatment: Item 1.1 Mentioning methods should be favoured with respect to mentioning

software; software could be referred to in the introduction, as officially agreed tools for the implementation of the guidelines

It should be stressed that pre-treatment has to be based on RegARIMA models

It should be clarified why alternative B is better than alternative C

A reference to the rest of the section can be added to clarify what is meant by detailed pre-treatment

Item 1.2 has to become more directive

Item 1.3 The treatment of bridge effects should be added The issue of quarterly versus monthly calendar correction for quarterly data can be addressed 1.3 and 1.3.1 should be merged In 1.4, impulse outliers should be called additive outliers; ramps can be added together with a reference to seasonal outliers which will be treated in the relevant section; outlier detection in time of crises has to be added or, more in general, in times of economic changes; a clear distinction between outliers and errors should be introducedIn 1.5, guidance on model selection should be improved and reference to diagnostics made more concrete In 1.6 the expression: "non-positive series …" should become "series with zero or negative values"

Current section 2 – Seasonal adjustment In 2.1, references should be to "model based" and "filter based" methods

and neither to software nor to interfaces; options should be extended In 2.2, some guidance on the benchmarking methods to apply could be

added; "unadjusted" is preferred to "raw" data; the need for benchmarking in the presence of multi-frequency data can be mentioned, stressing the importance of avoiding breaks in the series and preserve the series' pattern

In 2.3, some relation between the size of discrepancy and the use of multivariate benchmarking techniques could be added; advantages and drawbacks of direct and indirect approaches could be better analysed; some more details on the optimal level of disaggregation should be added

A new item 2.3.2 can be added on the seasonal adjustment of aggregates when only a subset of components are seasonal

The treatment of seasonal outliers has to be added

Current section 3 – Revision policies

Major revisions have to be included; even if this section could seem hard to read and rewording can be considered, it has been agreed that improvement would be difficult to achieve without losing the completeness of the descriptions

Current section 4 – Quality of seasonal adjustment

The use of the standard ESMS has to be included. Desirable properties of seasonal adjusted series could be clarified.

Current section 5 – Specific issues on seasonal adjustment

The section will be restructured to contain:seasonal adjustment of short time series

chain-linking

seasonal adjustment of long time series

trend-cycle estimation

an additional item concerning other frequencies than monthly and quarterly could be added if a significant proposal can be drafted. Finally the item 5.2 on seasonal adjustment of problematic time-series will be removed

Current Section 6 on Dissemination issues will not change

TIMETABLEData Activity

End 08/2012 First draft of the revised sections and of the new section

10/09/2012 Discussion of the first draft sessions

End 11/2012 Revised versions of sections

End 01/2013 Approval of the new sections

End 03/2013 Harmonisation of style and check of the consistency among sections 06/2013 Approval by the SASG

End 12/2013 Presentation to Eurostat DM and to the ESSC