Workshop on Standardisation Brussels, 14-15 October 2010
Comparative review of SDC and SA
standardsJean Marc MUSEUX – Eurostat
Workshop on StandardisationBrussels, 14-15 October 2010
Seasonal Adjustment and Statistical Disclosure Control
• Why SA and SDC ?
• State of the art of standards
• Process leading to standard
• Learnings and conclusions
Workshop on StandardisationBrussels, 14-15 October 2010
Why SA and SDC ?
Statistical Disclosure Control
X-sectional tabular output and micro data files
Seasonal Adjustment
Time-series and Infra-annual statistics
“Historical” domains where Eurostat central methodological unit developed some
expertise in the 90’s
Important steps for Eurostat business process
5. Process - 6. Analyse
Specific expertise independent of statistical domains
Workshop on StandardisationBrussels, 14-15 October 2010
Why SA and SDC ?
Statistical Disclosure Control
Committee on Statistical Confidentiality since 1997
Working group on Statistical Confidentiality since 2009
Seasonal Adjustment
Informal Working Group on Seasonal Adjustment 1999-
2002 SA Steering Group since 2007
Dedicated working group for ESS coordination
Workshop on StandardisationBrussels, 14-15 October 2010
Why SA and SDC
Statistical Disclosure Control
Coherence required because secondary confidentiality in tables
Disclosure risk increases if uncoordinated release at EU and MS level
Strong impact on EU data utility Sensitivity of breach of
confidentiality in the ESS
Seasonal Adjustment
Seasonally adjusted data: reference key indicators for analysis and forecasting exercises
Reliability and comparability EU aggregate derived from MS
series – need for coherence Non linear process with
propagation of error
Striking need for harmonisation
Workshop on StandardisationBrussels, 14-15 October 2010
SA standards – state of the art (1/3)
ESS Guidelines on Seasonal Adjustment
– Endorsed by CMFB and SPC in 2008– SA process decomposed in substeps (pre-treatment, signal
extraction, revision, release, metadata, …)– For each elementary steps, the guidelines lists three alternatives
• A, the best approach to be aimed at;• B, acceptable and viable if A proved to be too costly• C, practice to be avoided.
– Provide a open framework to • design SA process (guidance), • Improve SA process (clear preference)• benchmark several processes
Workshop on StandardisationBrussels, 14-15 October 2010
SA standards – state of the art (2/3)
ESS Guidelines Implementation
– Seasonal Adjustment Steering Group (SASG) in charge of overseeing the implementation of the guidelines
– SASG is high level group bringing together Eurostat, ECB and SA experts in MS and CB
– Main barriers for implementation• 1) Lack of institutional recognition• 2) Organisational issues• 3) Cost of option A (human resources)• 4) Methodological issues • 5) Knowledge, skills• 6) IT infrastructure
Workshop on StandardisationBrussels, 14-15 October 2010
SA standards – state of the art (3/3)
ESS Guidelines Implementation
– Three main strands for fostering implementation decided by SASG
• Information in sectoral WG, Scientific conferences
• Cooperative (re)development of a software tool (Demetra+) in line with the guidelines (A and B options can be implemented)
• Training, workshops (with experts) for spreading knowledge and exchange of experience
– Further difficulties
• No global review and impact assessment
• Need to continuously refine, going beyond guidelines (crisis, …)
Workshop on StandardisationBrussels, 14-15 October 2010
inp
ut
ou
tpu
t
Implementation
Knowledge Generation
(R&D/innovation)
Good-practiceGeneration
(ESSnets)
& Tools
Knowledge Formalisation
CompetenceBuilding
OperationalGovernance
Production Strategy
Quality
Methodology
resources
products
Framework for analysing process leading to standard
Workshop on StandardisationBrussels, 14-15 October 2010
Good-practiceGeneration(ESSnets)
32
4 QualityProduction
Strategy
Knowledge Formalisation
Methodology
Implementation& Tools
Knowledge Generation
(R&D/innovation)
4
Competence & Capacity
Building3
3
31
2
1
1
input
RE
SO
UR
CE
S
Operational gouvernance
PR
OD
UC
TS
output
Tracking back SA standardisation process
2
2
2
1:Eurostat initial studies on SA:
1995-1999
2: Eurostat development of Demetra1999-2005
3: SASG and guidelines2006-2009
3
4: Demetra + cooperative development2009-2010
Workshop on StandardisationBrussels, 14-15 October 2010
SA standards process – key elements
Early expertise development
First Demetra tool– Did not achieve harmonisation– Distanciation from expertise
SASG breakthrough– Technical expertise– Governance– Methodology
Demetra+– After methodology– Cooperative and open source
Workshop on StandardisationBrussels, 14-15 October 2010
SDC standards – state of the art (1/4)
Tabular data
– Standard tool for protecting tables • Tau-Argus developed by Statistics Netherlands • Financial support from Commission since 2001• Difficulty to integrate in standard production processes
– Handbook on SDC (last edition 2010) – glossary and review of options
– No standard (SDC depends on national perception and rules)– EU table protection at the border of feasibility – Lack of standardisation hampers release of EU figures
(suboptimal solution)
Workshop on StandardisationBrussels, 14-15 October 2010
SDC standards – state of the art (2/4)
Tabular data
– First step towards standardisation• Confidentiality Charter (SBS, Prodcom, Animal production)
– Objective rules for protection of cells
– Practical rules at domain level ensuring consistency between Eurostat and MS processing
– Flexibility in primary confidentiality documented (flags)
– Methodology for SDC of EU aggregate
Workshop on StandardisationBrussels, 14-15 October 2010
SDC standards – state of the art (3/4)
Micro data protection
– Baseline methodology and corresponding software (Mu-Argus – Statistics Netherlands) – suitable for one off application
– Domain in constant development (computer science, Web, ..)– Anonymisation of EU micro data– Output harmonisation through input harmonisation
• Same global recoding for all MS datasets• Micro aggregation for all records• Little flexibility• Least common denominator effect • Low information content (almost public use files)
Workshop on StandardisationBrussels, 14-15 October 2010
SDC standards – state of the art (4/4)
Micro data protection – way forward
– ISTAT model Q2008 – Community Innovation Survey PSD2010 – Harmonisation of SDC
• Bounded flexibility• Core common disclosure scenario and risk assesment methodology• Flexibility among a common set of methodologies for protecting
records – adapting to country specificities – transparent parametrisation – good balance between global methods (recoding, top coding) and local
methods (perturbation)– Data utility target (threshold) using common measures and constraint
on comparability (benchmarking using key research statistics)
Workshop on StandardisationBrussels, 14-15 October 2010
QualityProduction
Strategy (CVD)
Knowledge Formalisation
(LDF)
Methodology
Implementation
& Tools (IT)
Good-practiceGeneration(ESSnets)
Knowledge Generation
(R&D/innovation)
4
4
4
Competence & Capacity
Building
2
3
3
33
31
2
21
1
input
RE
SO
UR
CE
S
Operational gouvernance
PR
OD
UC
TS
output
Tracking back SDC developments
2
2
2
4
1:CASC FP5 research project
1997-2004
2: CENEX project
2004-2006
3: ESSnet SDC II
2007-20094 Next steps
Workshop on StandardisationBrussels, 14-15 October 2010
SDC standards – key elements of the process
Importance of research – early steps towards tool development – centric development – difficulty of integration
Importance of ESSnet for sharing good practices
Lack of technical governance– To identify best practices
– To set up priorities
Next steps– Technical governance (TF)
– Open source development
– More flexibility – stronger metadata
Workshop on StandardisationBrussels, 14-15 October 2010
Conclusions
Review of two cases studies with difference outcomes
Main differences• political context • technical governance• type of guidelines
Main communalities• Need for harmonisation• Need for expertise – research• Process steps relatively independent of the statistical domain• Need for flexibility• Need for appropriate – sustainable software tools• Living standard – need for maintenance