ESCWA/UNSD Expert Group Meeting on National Accounts 12-14 May 2009, Cairo, Egypt U N Statistics Division

1. ISIC and CPC Implementation ESCWA/UNSD Expert Group Meeting on National Accounts 12-14 May 2009, Cairo, Egypt U N Statistics Division 1

2. Implementing New classifications Presentation Outline Introduction Implementation process - Steps Structuring national classifications Adaptation of business register Sample designs and weights for surveys Backcasting Points for discussion

3. Introduction • Since its creation in 1948, ISIC had two goals: • Provide a tool for international comparison • Provide guidance to countries for a national activity classification structure • ISIC periodically updated to keep it relevant with economic developments. • Side effect: Increasing need for detailed data has lead to more detailed versions of ISIC

4. Implementing New classifications Structuring national classifications Creation of a new national version of national classification (NSIC) according to ISIC Rev.4 for national use Should involve other stakeholders / users/ data providers Recoding Business Register Switchover in the Business Register to new NSIC Statistics update process Co-ordinate simultaneous switchover in statistics to new NSIC and the control of the quality of time series

5. Structuring national classification How can national classifications be structured? 1. Using ISIC as a starting point 2. Based on historical national versions 3. Starting completely from scratch

6. Using ISIC as a basis Countries that use ISIC as a basis for their national classification, can do this to varying degrees: Adopt ISIC “as-is” Use the complete ISIC and add subdivisions to reflect nationally important industries (but maintain the ISIC coding structure) – can be “numerically truncated” back to ISIC [Example] As above, but with changes of the coding structure (example: NACE) – requires correspondence table [Example] Elevating lower level ISIC categories to higher national levels, (e.g. combine ISIC categories at 2- or 3-digit level) [Example]

7. Using ISIC as a basis The first three methods maintain full comparability with ISIC at all levels Option 4 limits internationally comparability to a higher aggregation level only Is use of the same coding as ISIC 4 a requirement? - No, but it makes comparisons easier.

8. Using historical national classifications as a basis Using non-ISIC based classifications always creates difficulties for international comparison Correspondence tables are necessary May limit data conversion due to splits Efforts are encouraged to line the historical versions up to ISIC At detailed level (without considering aggregation structures) or By lining up individual sections

9. What detail should be considered? (1) The United Nations Statistical Commission recommends that detailed categories of a national classification can be rearranged and aggregated so that they correspond with the 2-digit level of ISIC without loss of data. However, most statistics and users will require more Countries might want to add detail for industries of particular importance to the national economy. Countries might want to remove detail for reasons involving size and relevance, confidentiality or homogeneity

10. What detail should be considered? (2) Classification for collection may be more detailed than for distribution of data Using more detail for collection allows for future adjustments if individual industries are growing Level of detail for publishing depends on type of statistics anyway No fixed guidelines exist for the proper choice of detail

11. “Formal approach” “Pragmatic approach” Homogeneity based ISIC Possible approaches • Based on variables like: • # of statistical units • # of employees • value added • Based on: • Input from data users • Special concerns (confidentiality, extra burden, growing industries) Based on: Homogeneity ratios (as described in ISIC Rev. 3) Compress or expand classification based the value of these ratios • With chosen variable, compute ratio R between what is found within a category and average among “siblings”: • R < 0.5 => delete • R  [0.5, 1.5] => keep • R > 1.5 => split • New subclasses only created if user demand • Data users must justify their needs for splits, and estimate number of affected units and turnovers • Take confidentiality and extra burden into account in advance • No strict thresholds • Drawbacks: • Not enough usable data • No definitive and mutually exclusive definition of activities by products • Drawbacks: • Hard to define levels/weights • Level of detail influence outcome • Ignore dynamic aspects • Drawbacks: • Challenging and intense discussions, not all user needs can be met

12. Options to consider or avoid (1) ISIC structure and definition are based on few criteria (input, process, output, use of outputs) Should other criteria be added for national purposes, such as private vs. public entity, manufacturing by hand (crafts) vs. manufacturing by machines? What are the applications? Generally, avoid unnecessary addition of detail

13. Options to consider or avoid (2) When following the ISIC structure and coding system, don’t renumber codes if you want to skip a code number Regardless of legality, size or other concerns, it is still within conceptual scope, and must be accounted for in the SNA. Renumbering makes the ISIC link less intuitive Rules for good housekeeping If a category at level n is not further subdivided, the code at level n+1 should be the same code with a “0” appended Use digit “9” for residual categories

14. Implementation of ISIC and CPC Expert Group Expert Group discussed a proposal for the implementation of ISIC and CPC at its meeting in April 2007 Main recommendations relate to: Timetable Documentation and tools Regional workshops Technical cooperation

15. Implementation Timetable EG recommended two milestones: Countries should adapt their national classification to ISIC Rev.4 by 2009 To meet population census target of 2010 International data reporting in ISIC Rev.4 format should begin in 2012 No specific dates have been recommended for completion of: Adaptation of business register Sample designs and weights for surveys Backcasting Reasons: The timing and organization of surveys varies by country Scope (length) of backcasting differs by country

16. Documentation and tools UNSD will develop: A Companion Guide to ISIC and CPC that helps to understand: the concepts and structure of the classifications Application of the classification Correspondence tables ISIC and CPC indexes UNSD will also set up an open discussion forum and expand the classifications website into a larger knowledge platform

17. Documentation and tools UNSD will develop documents, based on European and other experiences, for: Establishing a national classification Recoding of business registers Sampling design and weight estimation Backcasting

18. Implementation of ISIC Rev.4 Major tasks and challenges: Defining a new classification and associated tools Reclassification of all units on the business register according to the revised classification Maintaining two classifications for an interim period Sampling and weighting under the new classification Simultaneous estimation and results assessment under both new and old classification Construction of industry weights for short term statistics Construction of back series in terms of the revised classification Handling of the national accounts move to the revised classification

19. NSIC setup process

20. Register update process

21. Statistics update process

22. Coordination of work Groups involved in the three processes need to be coordinated Need for a programme manager Communication with users of the classification (potential data providers) Need to get other users involved in setup and keep them updated on work and impact according to a fixed schedule

23. Recoding the business register Based on known correspondence table, can use different methods: Direct (straight) recoding Use of additional info from NSO or external registers Surveys Profiling Probabilistic models

24. Recoding the business register Dual coding should be kept for several years Problem: some units may cease operation Use of reference dates may help Can be implemented by using a transition code Reflect correspondence table Can computer-assisted coding help?

25. Points for discussion Has the process for implementing new classifications been set up by the ESCWA countries? If Yes – Status No – Plan for implementation

26. Thank You

27. Regional workshops Regional workshops have been supported in two phases: Workshops on final versions of ISIC and CPC in 2007/8 Workshops on specific elements of implementation*, such as: Adaptation of business registers Statistical collection programmes Use of the classification for administrative data * (carried out most likely in connection with industry workshops)

28. Technical cooperation EG supported regional partnership for technical cooperation Proposal: US, Canada, selected S. American countries support ECLAC EU members support ECA, ECE, ESCWA (where relationships already exist) Australia, Japan, India support ESCAP UNSD will serve specific sub-regions First steps of cooperation will be organized through the open discussion forum

29. Sample timetable for transition (1)

30. Sample timetable for transition (2)

31. Additional guidance More detailed guidance for developing new national classification, recoding of business register, sampling design and weight estimation, backcasting is being prepared in the implementation materials