Title: Establishing guidelines for creating long time series for short-term economic statistics
1Establishing guidelines for creating long time
series for short-term economic statistics
OECD Short Term Economic Statistics Working Party
2Current methods used for linking and backcasting
the IIP
- Given differing lengths of time series the lack
of international guidelines OECD had concern over
the consistency of practices across countries - Initial analysis based on metadata reported to
OECD, IMF and Eurostat is that a large variety of
methods are used and terminology is not
consistent or well understood - Concerns for comparability of data, and many
versions of full historical series most likely
exist
3Differences in published series length
4Brief review of OECD countries linking practices
for IIP
- On basis of available metadata
- 16 reported the use of chain linking using a
variety of methods (10 annual, 6 five yearly) - 5 reported Laspeyres index with base year 2000
weights and time series starting in 2000 (i.e.
true Laspeyres) - 6 reported Laspeyres index with base year 2000
weights but time series starting before 2000
(i.e. this is not a true Laspeyres index but a
simple fixed weight index) - 2 reported the use of some other method
5Policy and practices for producing long time
series
- Differs a lot between EU and non EU countries
(reflected in time series length) - Not much reference to user needs, number of
references to EU regulation - Bank of Italy comment seemed representative of
user needs as viewed by OECD - Importance of reliable long time series for STES
for a full appraisal of business cycle both in
historical perspective and for forecasting
purposes
6Common themes in comments
- Backcasting is resource intensive
- Especially when classifications change
- US Retail trade presentation is a good example of
the detailed work involved just to produce data
back to 1992 (OECD linked index at aggregate
level goes back to 1959) - Changes in classification / methodology mean that
historical series on old basis no longer relevant
/ cannot be made comparable - Regulations in EU seen as a proxy for user needs?
7User needs
- Had comments from Bank of Italy stressing the
importance for empirical analysis - OECD Economics Department demands longest time
series as possible for their empirical analysis - Hence the policy of the Statistics Directorate as
supplier - If we dont provide the data, ECO do their own
linking - ECB have remarked on the important of data from
the Main Economic Indicators (compared to New
Cronos) because of the longer time series
8User needs
- Analytical users need long time series if the
NSO doesnt provide them they will source them
from someone who does or recreate their own - Regardless of methodological issues or costs to
NSOs - Multiple versions of historical series exist,
using different methodologies for backcasting /
linking - Lack of knowledge of good practices (e.g. linking
methodology first common period, first common
year), lack of knowledge of methodology of
different segments of old series - Impact on seasonal adjusted series needs careful
examination (change in seasonal factors between
versions)
9User needs
- Is the National Statistical Office (or compiling
authority) best placed to create one official
version of longest time series possible? - If not them, then who is?
- Could a distinction be made between backcasting
at a detailed level and at an aggregate level? - How to define the aggregate level?
- What are the real user needs?
10Guidelines on backcasting and linking
- Nothing comprehensive exists for short-term
statistics - IMF manual for quarterly national accounts
- PPI manual
- Eurostat Methodology of short-term business
statistics - Most comprehensive source for STES but guidelines
for backcasting / linking need more detail and
examples (no EU country uses the method
recommended in the current guidelines)
11Guidelines on backcasting and linking
- About half countries commented on importance of
STESWP establishing recommendations - Many commented on the need for this to be done in
conjunction with Eurostat, who have already
established some task forces for NACE
implementation - How to represent the interests / needs of non EU
countries - Especially those also planning the introduction
of new classification systems in the coming years
12Eurostat task force on NACE implementation
- Producing a handbook on back-casting
- Questionnaire
- 3 methods
- Detailed working on micro data FR, NL, Estat-B5
- Proportion methods STS(Estat-D3), AT, DE, UK
- Extrapolation between bench markings
ind(Estat-D1), ? - First draft mid September 2006 TF meeting Sep
21st - Handbook November 2006 on the website
- Detailed application of handbook to short-term
statistics - Proposal to WG in December 2006 following release
of first draft of handbook
13Possible scope of guidelines
- Terminology (goal to have harmonised metadata)
- Index types, advantages and disadvantages and
frequency of weighting updates define
preferred, alternative, poor practices - Linking methods, advantages and disadvantages,
interaction with seasonally adjusted series
(issue for SA handbook?) - Backcasting in conjunction with classification
changes - Possible distinction between detailed and
aggregate levels, where goal at aggregate levels
is longest time series possible - Consistency in practices across set of related
indicators and range of short-term economic
statistics - Assessment of user needs (how to focus this?)
14Discussion points
- Scope of proposed guidelines, do they cover the
correct issues - Should STESWP get involved in this work, how
should it be co-ordinated with Eurostat, who
wants to be involved - Is STESWP an appropriate forum for discussing
policy on the production of long time series - Can we have any influence on national practices?
- General discussion on the role of NSOs for
compiling long time series given wide range of
comments on the paper - Interested to hear Eurostat views and especially
the views of user organisations present at this
meeting