Harmonisation of Seasonal Adjustment Methods in EU and OECD Countries

1
Harmonisation of Seasonal Adjustment Methods in
EU and OECD Countries

• Ronny Nilsson
• Statistics Directorate

2
Contents

• Background
• Results of EU and OECD Surveys on SA
• - IT approach (EU survey)
• - Methodology (EU and OECD surveys)
• Future work

3
Background

• Monitor Euroland with early national data and
  facilitate comparison with other economies
• Software development for SA - DEMETRA
• Task Force, SA Co-ordination Group (CMFB) with a
  mandate - harmonisation of SA in EU
• - integration of X-12 ARIMA and Tramo/Seats
• - use of DEMETRA by NSI and NCB

4
Results of EU Survey

• Main conclusions
• X-12-ARIMA and Tramo/Seats only relevant methods
• Separation between research/production level
• Program version policy is urgent
• Integration of the two SA methods is important
  but not urgent
• DEMETRA can not fulfil the role of a standard

5
Priorities for harmonisation of SA procedures in
EU countries

• Use of single SA software integrating both X-12
  ARIMA and Tramo/Seats
• A single reference source code underlying the SA
  software (statistical and technological)
• Standardisation of the reporting of SA metadata,
• in particular quality aspects
• Definition of best SA practices via the European
  network of expertise

6
Results of EU- OECD surveys

• Survey characteristics
• EU countries Norway
• Sample16 NSIs inc. Eurostat,17 NCBs inc. ECB
• Response rate 94 for NSIs, 76 for NCBs
• OECD non-EU countries (15 countries)
• Sample 15 NSIs 15 NCBs and 5 other instit.
• Response rate 93 for NSIs, 50 for NCBs and
• 80 for other institutions

7
SA Methods To day ()

• TS X-11 X-12 TS/X-12 Other
• EU 19 50 0 23
  8
• Non-EU 0 43 35 13
  9
• OECD 10 47 16 19
  8
• OECD X-11 dominates the market
• EU TS in stand alone mode is widely used
  
• Non-EU X-11 and X-12 used by some 80

8
SA Methods In the future ()

• TS X-11 X-12 TS/X-12 Other
• EU 24 0 28 40
  8
• Non-EU 22 6 44 17
  11
• OECD 24 2 35 30
  9
• OECD X-12 and TS/X-12 in combination will be
  the two main methods
• EU TS/X-12 in combination will dominate
• Non-EU X-12 will be the main method

9
Methods Selection Process

• Others SEAABS, BV4, TESS, GLAS, STAMP
• Selection of one or several methods on
• - internal decision based on testing and
• evaluation phase
  (58 )
• - historical reasons or recommendations (60
  )
• Multiple methods are used because of
• - possibility of cross checking results
• - specific features of each method

10
SA Diagnostics

• Satisfaction level
• EU countries 88 , Non-EU countries 95
• OECD countries 90
• Main indicators/diagnostics used
• Graphical Inspection (81 )
• Result/Analytical Tables (79 )
• Diagnostic test for ARIMA models (75 )

11
Pre-adjustment features

• First priority
• Outliers detection and T-day correction/flow
  vari.
• Second priority
• Implementing national holidays
• missing obs. and forecast test for model type
• Third priority
• Level shifts, additive outliers, seasonal breaks,
  Easter effect, user defined/dummy variables

12
Direct vs Indirect adjustment

• EU countries
• Aggregation problem considered by 30 , but no
  method is predominant
• Non-EU countries
• Direct method is the most common method (58 )
• Indirect method is only used in large scale in
  the United States and Korea

13
Proj. factors vs Concurrent adj.

• EU countries
• 25 of the institutions are considering the two
  methods, but no one is predominant
• Non-EU countries
• Projected seasonal factors are used by 63 of
  the institutions as the regular method
• Concurrent adjustment is used by 32 on a
  regular basis

14
Update or Revision Policy

• Seasonal adjustment options
• fixed periodicity (Yearly) EU 58 , Non-EU 80
• after revisions in data EU 30 , Non-EU 36
  
• Model parameters
• fixed periodicity, EU no dominant pattern, but
  60 uses a yearly pattern in Non-EU
• for selection of fixed filters and ARIMA models a
  yearly periodicity is predominant

15
Metadata and Publication Policy

• Metadata on SA Method, SA Parameters and
  Working/trading day adjustment are stored by 85
  in the production database for internal usage
• Only metadata on SA Method and Working/trading
  day adjustment are stored by about 40 in the
  dissemination database for external usage
• Info on outliers is stored by 40 and other
  types of metadata by 20 of the institutions

16
Future Work

• A merge of X-12 ARIMA and Tramo/Seats would be
  welcomed by most institutions using both methods,
  EU and OECD support this development
• Eurostat proposal for a Reorganisation of SA
  Activities in the European Statistical System
  (ESS)
• - Steering Group (financial/human input)
• - User Group (evaluation/proposals)
• - Scientific Group (research and assessment)

17
Future Work

• OECD Expert Group on Short term Statistics
• - Harmonisation of SA methods across OECD
  countries - link to EU Activities
• - Metadata and publication policy of SA for
• external users
• - Presentation of data seasonally adjusted data
• vs trend-cycle (smoothed) data