1. The Singapore Department of Statistics and the Organisation for Economic
Co-operation and Development (OECD) jointly organized an international
workshop on the Consumer Price Indices (CPI) in June 2001. The objective
of the workshop was to review draft chapters of the revised CPI manual
and obtain feedback from national statistical offices before it was finalized.
It also aimed to discuss issues of concern to CPI compilers in Asian countries
and share experiences of OECD countries. It was the first time that the
draft CPI manual was being discussed in an international gathering comprising
representatives from the developing countries.
2. Some 35 representatives from 6 ASEAN countries (Indonesia, Lao-PDR,
Malaysia, Singapore, Philippines, Vietnam) as well as China, Hong Kong
SAR, India and Korea participated. Expert speakers from international
agencies like OECD, International Monetary Fund, International Labour
Office as well as Australian Bureau of Statistics presented the draft
chapters and led the discussions at the workshop.
3. More than 14 chapters of the draft manual were presented and intensively
discussed. Participants shared their experiences on CPI construction,
as well as their concerns and difficulties encountered during the compilation
process. The detailed report of the workshop is given
in Annex I.
and Discussions on Issues
4. The Workshop discussed and concluded on the
- The current draft manual
is more useful to developed country compilers
since it covers more theoretical materials
with more focus on the conceptual issues rather
than practical ones. It should aim to
serve as the standard to both developed and
- The manual should be user-friendly
and it should provide more practical and clearer
guidelines on the treatment of specific products.
- Quality adjustment - some
form of flow-diagram with different adjustment
methods, and the conditions under which they
are applicable, should be provided.
- The manual should provide
more information on COICOP and outline the
pros and cons of switching to the international
- The manual should deal
with practical issues of index compilation
and present approaches applicable in developing
- Developing countries should
be invited to join in the drafting process
to reflect their concerns and issues relevant
- The draft manual was much
longer than the 1987 edition and hence its
presentation should be restructured.
5. The following issues were discussed at the
I) Uses of CPIs
- It was acknowledged
that CPIs are commonly used as
tool for the indexation of wages, social
security payments and contractual payments;
- a measure of price inflation
for use by governments for inflation-targeting.
- Component indices are also
used extensively to deflate national account
6. These uses will have an impact on coverage
of CPIs in terms of target population as well
as the basket of goods and services.
II) Concepts and Coverage
7. The Workshop discussed and agreed that the manual should address the
differences between a consumption CPI (or the cost of living CPI) and
an inflation CPI (or the transaction CPI), as well as a CPI which measures
pure price change. On the coverage of CPIs, the participants
generally felt that the manual should recognise that the coverage of a
CPI depends on its main purpose and it should provide clear guidelines
on how coverage is related to use in practice.
8. Though the participants recognised the importance in the harmonization
of concepts and terminology, they were of the view that the CPIs differ
significantly from national accounts data in the way in which they are
used. Thus, the workshop suggested that the manual should note the linkages
between CPI and SNA uses and concepts, and acknowledge the need for differences
and explain them. The participants felt that the CPI cannot simply
be "slotted in" to the SNA 93 framework. The general view was that
the CPI should not be a "subordinate" of national accounts since its main
purpose is not for national accounts deflation. However, as
switching a CPI framework to that of the SNA 93 is a long process which
requires much effort, the participants expressed that the manual should
suggest working towards integration gradually.
9. On the three valid conceptual bases on which a CPI can be compiled,
the participants suggested that the manual should give equal recognition
to "acquisition", "use" and "payments" approaches as they felt that the
manual gave the impression that the CPIs should adhere to the "use" concept
without giving sufficient weight to the other two approaches.
III) Sampling and Price Collection
10. The draft manual delves on the advantages and disadvantages of probability
sampling methods and explains the practical reasons for the adoption of
non-probability sampling methods in most countries. The Workshop expressed
that it would be useful for the manual to include practical guidance
on how countries could apply the various sampling methods in the collection
of data for CPI and the possible practical problems e.g. high cost involved
in probability sampling.
IV Treatment of Specific Products
11. Problems which most countries encountered, are highlighted below:
- Cross-border shopping
- Second-hand goods
- Extreme seasonal variations
in prices and availability
- Changes in prices of goods
and services that arise from changes in subsidies
- Fashion clothing
- Price indicator for gambling
- Quality changes in household
appliances, telecommunication products and
- Measurement of rents for
- Quality adjustment in pricing
of cooked food.
- Education abroad
12. The treatments may vary according
to countries' practices. The participants
were of the view that these could be addressed
in detailed in the manual.
V) Quality Adjustment
13. While the Workshop agreed that the subject on quality change deserved
comprehensive treatment in the manual, it felt that the focus should be
shifted towards practical guidance, and away from conceptual discussion.
It was brought up that developing countries have a strong need for procedures
that are defensible, reproducible and are seen to be objective.
The participants felt that the manual should provide some form of strategy,
or decision flow-diagram, covering the procedures for identifying quality
change and appropriate action, choice of adjustment technique.
14. It was suggested that the chapter on quality adjustment should be
clear with real examples, about what is considered quality change and
what is not. The participants felt that hedonic regression has been
heralded as the "correct approach" in the manual while other approaches
as inferior. They expressed that more emphasis should be given to
the difficulties and limitations of the hedonic approach.
VI) Index Numbers
15. While it was agreed that it is convenient to have theoretical background
in the manual, the Workshop noted that the treatment of index number theory
was too detailed and extensive. The draft manual recommends that
elementary price indices should be calculated using geometric mean formula,
or ratios of average prices if arithmetic mean is used. The participants
however felt that the most appropriate method is dependent on the situation
of each country.
VII) Weighting and Linking
16. The workshop noted that the manual presented the following ways for
"weights" to be considered during the revision of CPI:
- the period to which the
expenditure data relate (as weighting base);
- the denominator of the
price relatives (price reference base).
17. Participants felt that the manual should
also highlight the importance of aligning the
weighting base period with the price reference
period - one way is to update the
weights to reflect price change between the
two periods. It was stressed that
the manual should provide more guidance on the
linking of two historical series when price
collection for both old and new baskets had
not been conducted in an overlapping period,
and the way to link series where there are significant
changes in the item structure. It was
suggested that the manual should explain the
advantages of using the modified Laspeyres and
provide more numerical examples, especially
on the imputation of missing prices using both
fixed base and modified Laspeyres.
VIII) Sampling and Non-Sampling Errors
18. The participants expressed that it would be useful to have a table
showing the different types of sampling and non-sampling errors.
They also discussed the topic on formula bias and arrived at a conclusion
that different formulae for computing elementary aggregate indices may
be appropriate depending on the degree of homogeneity, and the elasticity
of substitution of the items. It was noted that the assumption that
substitution will lead to upward bias in the index may not be valid in
IX) Users and dissemination
19. There was a general agreement that users should be informed of the
limitations of the CPI and important methodological changes, weight revisions
or re-referencing of the index. It was proposed that advisory
groups of main users and experts be created, and seminars to educate users
20. The workshop provided a good opportunity
for participants to discuss technical issues
on the compilation of the CPI and to obtain
feedback on the draft international CPI manual
before it is finalised. Constructive suggestions
were also raised on ways to make the ILO CPI
manual more relevant and useful to the Asian
on Consumer Price Indices
4 - 8 June 2001, Singapore
Participants attended from Australia, China, ESCAP, Hong Kong SAR, ILO,
IMF, India, Indonesia, Korea, Laos, Malaysia, OECD, Philippines, Singapore,
The workshop had two main aims:
- to review draft chapters
of the revised CPI manual and suggest changes
needed to make it relevant and useful to Asian
- to discuss issues of
concern to CPI compilers in Asian countries,
and share experiences of OECD countries.
The workshop was structured using half-day
sessions studying particular topics, with discussants
introducing the draft chapters of the manual
relevant for each topic.
General conclusions on the draft manual
Gaps - All participants stressed that the manual should be user-friendly,
and it should provide more practical advice and clear guidelines on how
to treat real problems, with more examples, numerical and otherwise, instead
of stating the options available. More detailed practical guidelines
are needed on the treatment of specific products (cars, computers, mobile
phone services, purchases via the internet, seasonal products, gambling,
airfares, purchases abroad, etc.) rather than general guidelines.
It was suggested that an electronic version of the manual might be produced,
with numerical examples of treatments in spreadsheets to enable readers
Balance - It was felt that in the current draft, conceptual
issues dominate over practical issues. Although it was agreed that
it is convenient to have theoretical background material in the manual,
several participants felt that the current balance made the manual more
useful to developed country compilers. In fact, several people felt
that the main purpose of the manual should be to transfer the knowledge
from developed to developing countries, and that its main users will be
developing countries. In which case, the manual should be more prescriptive,
and should suggest options which are applicable in developing countries.
On the other hand, participants from more developed countries said that
they envisage using the manual as a textbook for educating compilers and
users, and as a tool for justifying compilation methods and warding-off
ill-informed criticism. In other words, the manual would be used as an
international standard in all countries, developed and developing. Also,
by explaining the complexity of the conceptual and theoretical issues,
the manual might demonstrate why it is not always possible to be prescriptive
about best practice.
Another issue highlighted was ensuring that the discussion is balanced
by providing better indications of the potential importance or magnitude
Length - Several participants noted that the manual, as a whole,
was already much longer than the present ILO manual, and it was suggested
that it should be reduced to a size only slightly longer than the existing
ILO manual - say ten chapters compared with eight in the existing manual.
Other material could be presented as annexes. Another possibility
would be to publish the manual in two or more volumes with the first volume
containing all material covering day-to-day practical issues of index
compilation, and the conceptual and theoretical material in subsequent
volumes. This applies particularly to index number theory and hedonic
quality adjustment techniques. It was also suggested that the manual
should be simple, practical and easy to follow.
Concepts and coverage - Many participants were concerned that
chapter 4 gave the impression that CPIs should be fully consistent with
the SNA93. While it was recognised that the harmonisation of concepts
and terminology is generally to be encouraged, most participants made
it clear that they, and their users, considered that some of the more
important uses for which CPIs were required were not directly satisfied
by reference to SNA93 aggregates. The current manual and ILO resolution
recognises three equally valid frameworks for CPIs based on the concepts
of "acquisition", "use" and "payments". The new manual should continue
to give equal recognition to these three alternative concepts. The
manual should recognise the linkages between CPI and SNA uses and concepts,
and acknowledge the need for differences and explain them. It was
also suggested that the manual should explicitly address the question
of producing indices of core or underlying inflation.
Formula bias was discussed at some length, and it was argued that
the assumption that substitution bias (both lower and higher level) will
generally be upwards may be invalid in developing countries. This
should be acknowledged, and discussed, in the manual.
Quality adjustment - In order to make the chapter more applied
and user-friendly, it was agreed that some form of decision flow-diagram
should be provided, covering the procedures for firstly identifying quality
change and then taking appropriate action, including the choice of adjustment
technique. A table of the different adjustment methods, and the
conditions under which they are, and are not, applicable, was requested.
Similarly, a clear list of what is, and is not, considered quality change
with real examples was requested. Developing countries have a strong
need for procedures that are defensible and seen to be objective, and
it was felt that the chapter is too heavily weighted towards hedonic techniques
at the moment.
The drafting process was discussed and several participants expressed
disappointment that developing countries had not been invited to join
either the Inter-Secretariat Working Group on Prices, or the Technical
Expert Group. They felt that the manual addressed the issues that
are more relevant to the developed countries than the developing countries.
Uses of CPIs
PIs are used for a wide range of purposes and
the principal use, or uses, of CPIs will vary
from country to country. The commonest
- CPI as a tool for the
escalation (or indexation) of wages,
social security payments and other contractual
- CPI as a measure of price
inflation for use by governments for inflation-targeting.
It was noted that the main purpose of the all
items CPI has never been for deflating national
accounts (SNA93) aggregates although component
indexes are used extensively for this purpose.
One use mentioned by participants, which is not considered in the manual,
is for market research - companies are interested in price movements of
the goods and services they produce.
In some large Asian countries, regional CPIs are at least as important
as national CPIs because of the wide disparities of prices and living
standards in different regions.
It was noted that in many countries, the overall (headline) CPI may not
be perfectly suited to any of its main uses. Many CPIs started life
years ago as so-called cost-of-living indices designed to monitor the
change in expenditure needed to maintain a basic standard of living.
In most cases the coverage was limited to the "basic" goods and services
needed for survival, but this has now been widened to be representative
of the expenditure of an "average" household. Thus, these CPIs have
their origins in income escalation (indexation). Many, however,
are now used as tools to monitor monetary inflation, and as such require
a different coverage. It may be the case, therefore, that many headline
CPIs are in fact hybrids, having neither the ideal coverage for an income
escalator for the average household, nor for inflation monitoring.
It is for this reason that many countries are moving towards publishing
a range of CPIs, with coverages suited to their different uses.
The manual distinguishes between indices designed to measure the change
in price of a fixed basket of goods and services ("pure price change"
indices), and indices which reflect the substitution behaviour of
consumers in response to different relative price changes. It was
noted that over the short-term, and under conditions of moderate inflation,
a CPI designed to measure pure price change (inflation) will be very similar
to one intended to measure changes in the cost of living.
Concepts and Coverage
The principal use of a CPI should
determine the coverage of the index i.e.
- the goods and services to be included and
the population group whose expenditures are
used for weighting the index. For inflation
monitoring, the coverage both in terms of goods
and services and of population should be as
wide as possible, while an index used for escalation
or for regional purposes will have a narrower
coverage. Similarly, certain users might
need indices with more limited coverage for
specific analytical purposes, e.g. "core" or
underlying inflation - the precise definition
is a matter of choice for the user, but these
are usually CPIs excluding seasonal and other
volatile commodities, and/or excluding any sales/excise
tax effects, subsidies, controlled prices, etc.
It was suggested that the manual should explicitly
address the question of producing indices of
core or underlying inflation.
On the note of the treatment of home-produced goods, some participants
disagreed with the exclusion of these items as these goods formed a significant
portion of the households' food consumption in their countries. It was
also suggested that more examples be cited on the coverage of these goods.
The 1993 SNA contains definitions of household consumption, namely final
consumption expenditure of households and actual final consumption
of households. The meeting spent much time discussing the draft
of chapter 4 (A System of Price Indices) which suggests that CPIs should
be consistent with these SNA93 definitions (coverage) of household consumption
and thus with SNA93 valuation principles and terminology.
Many participants were concerned that chapter 4 gave the impression that
CPIs should be fully SNA compliant, i.e. adhering to the use
concept, without giving sufficient weight to the concepts (and thus
coverage) of acquisitions and payments, as required for
inflation monitoring and income escalation purposes. While it was
recognised that the harmonisation of concepts and terminology is generally
to be encouraged, most participants made it clear that they, and their
users, considered CPIs to differ significantly from national accounts
data in the way in which they are used, and thus cannot be simply "slotted
in" to the SNA93 framework. They also expressed the view that changing
a CPI framework to that of the SNA93 is a long process, requiring much
effort for both compilers and users, and they felt that the manual should
suggest working towards integration gradually.
It was concluded that the manual should clearly recommend that the coverage
of a CPI depends on its main purpose, and should provide clear guidelines
on how coverage is related to use in practice. It should also spell out
the links and the differences between CPI concepts and SNA93 concepts,
and the reasons for the differences.
Sampling and price collection
The draft chapter on sampling explains the
advantages and disadvantages of probability
sampling methods, with reference to the practical
reasons why non-probability sampling is widely
used. The meeting discussed the trade off between
representativity and continuity, and the related
question of item specification (loose or broad),
was discussed. The ideal approach, of random
sampling of outlets and items, based on the
relative variance of outlets and items, within
a budgetary constraint, was described.
It was pointed out that there are many practical
ways in which random sampling can be strengthened,
within a limited budget.
The conditions under which different elementary aggregate formulae are
unbiased (see discussion of Index Quality, below) when using random sampling
with probability of selection proportional to size were explained.
While some felt the chapter dealt with sampling theory in a way that
was inaccessible to the non-specialist reader (too many formulae), others
felt that the level of technical detail was about right.
There was also a comment that Chapter 8 was drafted based on the assumption
of perfect conditions. It was suggested that more practical problems such
as the high cost involved in probability sampling should be mentioned.
In addition, the chapter should also include practical guidance on how
countries could apply the various sampling methods in the collection of
data for CPI.
ABS introduced the draft chapter on price collection, and all countries
gave short presentations of their price collection methods focussing on
the issues and problems that they face, and that they would, therefore,
like the manual to address. These issues include:
- Cross-border shopping,
including purchases over the internet, and
any purchase of foreign currencies that might
- Whether to include purchases
of foreign currency as a hedge against inflation;
- Whether to reflect the
price changes associated with the elimination/introduction
of taxes or subsidies, including situations
where the price before or after was zero (free);
- Whether to include sales
prices where the sales last only a few days;
- Treatment of tuition
fees for overseas universities paid by local
- Treatment of gambling
- what could be used as a price indicator?
- Where bargaining exists,
most countries buy the commodities in question
in order to obtain a true selling price, but
if the NSO is not in a position to purchase
commodities, is it reasonable to ask purchasers
what they have paid?
- How to deal with outlets
which are not fixed, e.g. in semi-fixed market
places, stall-holders often move around;
- Whether to record a product
as missing if it is not physically displayed,
but is available from the store-room or can
- How to deal with seasonal
items which are sold per piece, not by weight,
but where the size/weight of the pieces varies
from season to season.
Treatment of specific
Participants were asked to identify specific
goods and services that cause particular problems
in their own countries and which should be dealt
with in the manual. They mentioned:
- Second-hand goods;
- Extreme seasonal variations
both in prices and availability;
- Changes in prices of
goods and services that arise from changes
in subsidies and taxes;
- Fashion clothing;
- Quality changes in household
appliances, telecommunication products (e.g.
mobile phones) and cars;
- Measurement, through
surveys, of rents for owner-occupied dwellings.
- Cooked food e.g. set
- Education abroad
Several participants raised questions about
cross-border shopping, which is important in
several countries of the region, such as Hong
Kong SAR and Singapore. "Cross-border
shopping" is also a problem in countries that
compile regional indices and where residents
in one region may cross into another region
to buy goods and services at lower prices.
It was noted that if a CPI is intended to measure
the " price experience" of the resident population,
then prices paid for goods purchased abroad
should, in principle, be reflected in the CPI.
If the CPI is intended to measure changes in
domestic price levels (for monetary policy purposes),
cross-border shopping by residents is not relevant
to the CPI, but purchases by foreign visitors
is. The European HICP has adopted the
Internet shopping, when the purchaser and seller are in different countries,
also has some features in common with cross-border shopping. There
is, however, an important difference in that purchasers make their purchases
from within their own boundaries. Participants agreed that the manual
should deal specifically with the treatment of e-commerce.
It was agreed that more detailed practical guidelines are needed on the
treatment of specific products (cars, computers, mobile phone services,
purchases via the internet, seasonal products, gambling, airfares, purchases
abroad, etc.) rather than general guidelines. It was suggested that
an electronic version of the manual might be produced, with numerical
examples of treatments in spreadsheets (it was suggested that these could
be adapted by compilers for their national CPIs).
Some participants felt that the manual would be voluminous if it were
to consider all the problems encountered by different countries.
It was suggested that the international agencies could set up a web-site
providing specific guidelines to deal with such problems. The web-site
could also include case studies on how countries might resolve the problems.
While it was unanimously agreed that quality
change is one of the most important challenges
facing CPI compilers, and it was generally felt
that the subject deserved comprehensive treatment
in the manual, there was concern that the emphasis
of the chapter should be shifted towards practical
guidance, and away from conceptual discussion.
On the other hand, some participants did not
favour shortening the chapter as they felt that
they would benefit from having all the arguments
presented in the manual. This would also
demonstrate to critical CPI users that index
compilation is a complex and evolving science.
A compromise solution might be to split the
chapter into a conceptual and an applied section.
In order to make the chapter more applied and user-friendly, it was agreed
that some form of strategy, or decision flow-diagram should be provided,
covering the procedures for firstly identifying quality change and then
taking appropriate action, including the choice of adjustment technique.
Similarly, a table of the different adjustment methods, and the conditions
under which they are, and are not, applicable, was requested.
It was agreed that developing countries have a strong need for procedures
that are defensible, reproducible and are seen to be objective. It could
also be pointed out that quality adjustment is an extremely difficult
area, in which the more developed countries are not much more advanced
than their colleagues - in other words, the lack of progress is not necessarily
due to a lack of resources.
It was suggested that the chapter is unbalanced in its treatment of the
different methods of quality adjustment - hedonic regression appears to
be presented as the "correct" approach, with all others as inferior.
More space should be given to the difficulties and limitations of the
hedonic approach. Besides the hedonic method, the manual should also consider
other more robust and straightforward methods which are easy to follow.
The chapter also needs a clearer listing of what is considered quality
change, and what isn't, with real examples. Several participants
explained that under conditions of tighter competition, they are increasingly
seeing producers reducing quality in order to hold prices flat, and in
other cases are providing "free" services, e.g. free delivery of groceries,
after-sales care. Other participants raised the issue of changing
quality of housing as city-dwellers move to more comfortable out of town
housing estates. The more usual examples of household appliances,
PCs, cars, food away from home, and mobile phones were also given as examples
of problem areas, with the "bundling" of goods and services common with
mobile phones causing particular concern.
Aggregation and index
It is presently intended that chapter 3, which deals with characteristics
of index numbers, will be split into several separate chapters.
Parts B through E will become the new chapter 3, and will contain most
of the material that is considered to be immediately relevant for CPI
compilers. The other parts of the present chapter 3 will follow
as chapters 4 and 5 and the worked examples in the chapter will become
Although it was agreed that it is convenient to have theoretical background
material in the manual, several participants felt that the treatment of
index number theory was much too detailed and extensive. Some participants
suggested that the material could be made more easily accessible to compilers
and users of CPI.
Participants noted that Parts B through E describe several types of indices
in addition to the Laspeyres indices which, as the chapter itself acknowledges,
is the only index formula that can be used for CPIs because of the current
lack of current-period quantity data (although scanner data may provide
current weights in the future, albeit with limited coverage).
The presenter explained that this was done to demonstrate the limitations
of the Laspeyres index and the biases that it may introduce into the CPI.
This could be useful for CPI compilers confronted with questions about
CPI bias such as have resulted from the Boskin Report (see section
on Quality of CPIs, below). Also, these other types of indices
are commonly encountered in research studies. A list of the countries
currently using these formulae could be included in the manual for reference.
Chapter 11 on index compilation recommends that elementary (unweighted)
price indices should be calculated using a geometric mean formula, or,
if the arithmetic mean is used then as ratios of average prices rather
than as averages of price ratios. At the present time most countries use
arithmetic averages, and some countries use the arithmetic average of
price ratios which is identified as an inferior method in the manual.
However, the meeting noted that the most appropriate method is dependent
on the situation of each country.
Weighting and linking
The close links between weighting issues and
coverage questions were acknowledged. Whilst
it was agreed that a Household Expenditure Survey
(HES) is usually the main source for CPI weights,
the need for supplementary sources for certain
commodities was discussed - notably to correct
for HES under-recording of alcohol and tobacco,
where customs and excise data and/or national
accounts household final consumption expenditure
(HFCE) estimates may be used (a commodity flow
approach is commonly used), and national accounts
estimates of owner-occupied housing services.
In the context of HFCE it was noted that, where
CPIs are used in the deflation of HFCE, this
is done at a disaggregated level using group
or product sub-indices of CPIs (this gives an
approximation to a Paasche deflation of HFCE,
as required by national accountants).
It was also noted that users often compare the
movement in CPIs with movements in the implied
deflator of HFCE, and question the differences
- the US Bureau of Labor Statistics periodically
publishes explanations of the differences.
Turning to the CPI reweighting exercise, it
was agreed that the manual should establish a standard terminology for
the different types of base or reference periods encountered in the price
index world. These types of base are:
- the period to which the
expenditure data relate, usually 12 months
- the denominator of the
price relatives, one or 12 months (price reference
- the period chosen to
be 100 in published indices, usually a year
(publication reference base?).
Whilst it was agreed that where a set of weights
will be used for several years it is important
that the weighting base year should be typical
in terms of expenditure patterns, it was noted
that this may be impossible to achieve in practice
where an HES is planned several years in advance,
and may therefore take place when expenditure
patterns are affected by unusual events such
as drought, financial crises, etc. It
was stressed, however, that the weights can
be corrected to some extent using information
from other sources. Of course, annual
reweighting based on a continuous HES is a luxury
which makes atypical years less of a problem
(weights can be smoothed over consecutive years
discussed the importance of aligning the weighting base period with the
price reference period. One way is to update the weights to take
account of price change between the two periods. The latter procedure
overcomes the need to pre-judge which extra new items might be introduced
into the basket, and which therefore have to be priced during the weighting
base period. A suggestion was made to include a section on the implications
of not aligning the weighting base period with the price reference period
in the manual.
Questions were raised about how to link two historical series when price
collection for both the old and the new baskets had not been conducted
in an overlap period, and how to link series where there had been significant
changes in the item structure (classification changes). It was suggested
that the material in the manual on linking should be expanded to provide
practical solutions to such questions and it should outline the pros and
cons of changing the classifications.
More numerical examples were requested, especially on the imputation
of missing prices using both the fixed base and the modified Laspeyres
index. The advantages of using the modified Laspeyres in these circumstances
should be explained.
Classification of Items
It was suggested that more information be provided
on COICOP. Another suggestion was to outline
the pros and cons of switching to the standard
international classification. Some participants
commented that the classification of items should
meet the needs of users rather than follow the
international standard strictly.
Organisation and management
Some participants felt that the draft for
chapter 15 should be re-ordered so that training
is given a much higher priority. It was
felt that the material on quality management
systems and processes is too long and that the
descriptions of general quality management systems
is unnecessary. The section on training
(particularly training of price collectors)
could be much longer and come before the quality
management material. Of course, the links
between the two areas should be clear -
well developed and defined quality management
programs will indicate which parts of the compilation
system need improvement and additional training.
chapter discusses the use of benchmarking between different national statistical
offices (i.e. critical evaluation of a country's methodology and
systems, by comparing them with those used in other countries), but some
participants argued that benchmarking is not always the best approach;
or an effective way to improve on the quality of the index. Hence
it should not be recommended above other techniques.
Users and dissemination
There was general agreement that a single CPI will not be applicable for
all purposes, although the publication of more than one index may be confusing
for users. Thus, every effort should be made to inform users about
the need for, and the relationships between, the published indices.
They should also be informed about the limitations of the index, any important
methodological changes, and weight revisions or re-referencing of the index.
Advisory groups of main users and experts should be created, and seminars
should be organised to educate users.
Quality of CPIs
The draft chapter provides a general overview of the different types
of bias, and errors, that may occur during the compilation of CPI, their
causes, measurement, and procedures to minimise them. The chapter distinguishes
two broad categories of error - sampling and non-sampling errors.
It was suggested that a table showing the different types of sampling
and non-sampling errors would be useful. Also, the terminology used should
be made consistent with the terminology used elsewhere.
Formula bias was discussed at some length, and it was agreed that different
formulae for computing elementary aggregate indices may be appropriate
depending on the degree of homogeneity, and the elasticities of substitution
of the items in the elementary aggregate. In particular, it was
noted that the manual recommends use of the geometric mean (which assumes
elasticity of substitution of 1), over the average of relatives (which
assumes elasticity of zero).
However, some participants argued that the assumption that substitution
bias (both lower and higher level) will generally be upwards may be invalid
in developing countries. Research on consumer behaviour in countries
where incomes are rising rapidly shows that it may be an oversimplification
to assume that consumers will substitute whenever the price of one product
rise faster than the price of another. Cultural and religious habits
should be taken into account.
IMF quality assessment
The IMF explained the new data quality assessment framework (DQAF), and
in particular the DQAF for CPIs. The linkages to the IMF's special
data dissemination standard (SDDS), and general data dissemination system
(GDDS) were explained.
ILO draft resolution on
The ILO presented the revised version of the new ILO resolution regarding
international standards for the compilation and dissemination of CPIs.
The presentation covered the historical background of the international
standards, recent developments in the area of CPI theory and practices
and reasons for revising the existing resolution on CPIs. The procedure
for amending the resolution was also described. Finally, the content of
the draft proposals for the new international standards and the major
differences between it and the existing resolution were pointed out.
The participants would like the draft resolution be circulated for comments
before its submission for discussion at the Meeting of Experts on Labour
Statistics in October 2001. They also called for a standardisation
of terminology used in the draft resolution, the manual as well as in
other relevant publications.