ILO is a specialized agency of the United Nations

Summary of Discussion

Joint UN/ECE/ILO Meeting on Consumer price Indices (3-5 November 1999, Geneva)

Agenda item (a.3) - Information items

1. Under this agenda item Mr. D.Fenwick (ONS-UK) reported on the latest meetings of the Intersecretariat Working Group on Price Statistics (ISWGPS) and Technical Expert Group (TEG-CPI) on the revision of CPI Manual. He briefed participants on developments that took place since the latest meeting in 1997 and informed them about the work of the TEG-CPI on revision of the existing CPI Manual. The new Manual would be completed by early 2001. He also mentioned that invited papers to be discussed at the meeting were prepared with the idea of being used as inputs for the future Manual.

2. Mr. J. Greenlees (BLS-USA) informed the meeting about the latest Ottawa Group meeting held in August 1999 in Reykjavik. The topics at that meeting were: a) treatment of quality change in price indices for durable goods; b) different concepts of price indices serving different purposes and, c) the planned revision of the ILO CPI Manual. He mentioned that members of the Ottawa Group were generally supportive regarding the aims of the manual but had some reservations about the methods proposed. However, some of these reservations have been responded to during the TEG-CPI meeting. The next Ottawa Group meeting will be held in 2001 in Australia (hosted by ABS) and the seventh meeting will be held in 2002 in France.

3. Mr. J. Astin (Eurostat) reported on recent work by Eurostat in the area of Harmonised Indices of Consumer Prices (HICP). The work of Eurostat extends beyond the needs of member countries and involves all Accession Countries. They are expected to produce the HICP by January 2001 on fully comparable basis. Eurostat also continues to improve the quality of indices it produces and achieve the full comparability of national data. It also extends the coverage and will start producing monthly indices of monetary indicators. He also informed the meeting that an article on the HICP would be published in the next issue of the Statistical Journal of the UN/ECE.

Review of developments relating to biases in CPI - Invited paper by Mr. C. Obst (OECD)

4. The paper was drafted with the intention to be used as input for the future Manual on CPI. It is based on a thorough literature review and summarises the state of art of the theoretical knowledge about the bias issue covering many aspects such as: the definition of bias, the nature of bias, causes and possible solutions for various types of biases, etc. The study of CPI biases is very broad as it deals with practical issues of price collection and theoretical aspects of index numbers and utility theory. It was suggested that the issue of biases should be considered as inextricably linked to the actual choice of methods used in CPI compilation. An attempt was made to link together various arguments that relate to bias. In spite of a large body of work on this topic, the paper concludes that there remain some significant conceptual and empirical difficulties in defining the best price index.

5. In the discussion, it was pointed out that the approach taken by the United States in terms of evaluation of the components of bias is a positive one. France has tried to follow the same approach and provide figures for bias whenever it is possible. It was suggested that a hierarchy for biases should be established. The United States and Australia are using the geometric mean for aggregation, while other countries are using other price formulas. It was also mentioned that different parts of bias could depend on the formula used. Austria uses the arithmetic mean of price relatives. A calculation was made to compare the arithmetic mean with the geometric mean, which showed that CPI is biased upward by 0.1-0.2 percent. A quick survey of the method used was carried out among the participants in the meeting. The results were summarised and handed to interested participants.

6. Participants noted that there is a large variety of methods for calculation of price index in use and very few solutions for remedying biases intrinsic to each method of calculation. In this context it was pointed out that additional work on harmonisation is required in order to increase comparability among national CPIs.

7. At the end of extensive discussion, participants generally agreed that bias in CPI is inseparable from the type of index formula used. Related to this is a whole range of issues, including the methods for price data collection and data availability, sampling of items and outlets, changes in weights, etc. The formula effect on bias remains to be the most controversial. When deciding on the formula to be used, practitioners should make the best judgement and occasionally accept a compromise solution keeping in mind what exactly is being measured and what are the consequences of measuring if a particular formula is used.

Glossary of terms in CPI - Invited paper by Mr. R. Turvey

8. A draft of the glossary of terms prepared by Mr. R. Turvey was presented. The paper was prepared to provide an input for the future revised Manual on CPI and for the HICP Manual to be prepared by the Eurostat. It was stressed that it is very important to have a glossary of terms at the beginning of the revising of the CPI Manual. This will ensure that there is a consistency between different chapters being drafted by different authors.

9. The draft glossary consists of two parts. In preparing the draft the author took into account the glossaries from the United Kingdom, United States, Canada, Australia and the 1987 ILO CPI Manual. The author asked participants for comments and reactions in order to complete the work on the glossary.

10. Some new concepts have been proposed. For example, it was proposed that the concept of "variety", which was introduced in the 1987 ILO CPI Manual be replaced by term "sample product". It was noted that some of the proposals have arisen because of difficulties in translating certain concepts in other languages. For example, the term "item" is easily understood in English but not easy to translate in other languages. It was proposed that it could be replaced by the term "product". One possibility is to mention both types of terms by explaining their use in different countries.

11. The draft glossary proposes introduction of some completely new terms. An example was given with the term 'representative item' within which individual products are chosen for pricing. It was suggested that it could be called 'generic product', or 'product type'. However, a due consideration needs to be given to the question whether new proposed terms sound unambiguous. It was further suggested that 'acquisition approach' be replaced by 'transaction approach' in order to avoid ambiguity.

12. Participants welcomed the development of the glossary. It was pointed out that some terms needed further clarification, one example is 'imputations'. It was noted, however, that in the context of the HICP no imputations are allowed because its main aim is to measure actual inflation.

13. Participants noted that some terms were missing from the draft glossary, such as "unit value index" and "quality treatment". A particular area where consistency in terminology is needed is quality adjustments and terms used in this context, such as splicing. It was also noted that the issue of 'basket 'is essential in the context of price indices. Basket should not be defined just as a list of products. In describing the Laspeyres index, it was proposed to define it as 'a weighted average of price ratios'.

14. With regard to the use of different terms in English and Russian, it was mentioned that some confusion in communication could emerge if terms are used in different context, and therefore it was suggested to include synonyms in the glossary. This will allow various countries to find out the most appropriate term in their language. It was also suggested that some terms should have longer explanations, such as 'extrapolation' and 'retrapolation'. Terms related to sampling also need further elaboration

15. There was general agreement that the draft glossary needs to be enlarged. The fact that many definitions in the draft are concise was praised. However it was noted that for a glossary which is meant to be a user friendly, conciseness may contradict 'user friendliness'.

16. In concluding the discussion, the meeting was informed that a revised version of the draft glossary would be posted on the web site of the author. Participants were invited to send their comments and suggestions by electronic mail to Mr. R. Turvey.

Item 4 - Nature of indices (general vs. specific indices; cost of living index vs. pure price index, etc.)

Discussant: Mr. R. Lowe

Invited paper by: IMF, Mr. Peter Hill, UK and Contributed papers by: ILO, Russian Federation, and Mr. Erwin Diewert, Canada

17. The paper by IMF was prepared as an input to the future revision of the ILO CPI Manual. A reference was made to the establishment of the ISWGPS and its two technical expert groups (TEGs) - TEG-CPI and TEG-PPI. The objective of the ISWGPS and two TEGs is to develop two manuals (CPI and PPI) with a consistent set of concepts, methods and terminology for the two major price indicators.

18. A framework for price statistics was developed relating four principal price indices - PPI, CPI and Export and Import Price Indices (XPI and MPI) to the System of National Accounts - to economic aggregats to which these indices refer.

19. Two major definitions of a price index - conceptual definition and an operational definition could be distinguished. Conceptual definition relates the price index to the value aggregates while practical definition involves selection of items, determination of the item price and transactions to be included, sources of weights for selected formulas, and the type of formula to be used.

20. It was argued that the System of National Accounts (SNA) offers a systematic and consistent framework essential to the design of price indices. The SNA also gives classification of units to be surveyed. It recognises five institutional sectors resident in the economy and one non-resident institutional sector, the rest of the world. It was pointed out that the PPI applies to all institutional sectors, while the CPI applies mainly to the households. The SNA also offers precise and consistent terminology, standard language and standard set of definitions. It further ensures the geographical scope as national accounts are set up for national coverage. The SNA provides a standard nomenclature for the accounts by institutional sectors, from which the weights are taken.

21. The four main price indices and their associated national accounts aggregates are linked as follows: output of resident producers - PPI; individual consumption expenditure of goods and services (except consumption of own production, but including the imputed rent of owner-occupied dwellings) of the households - CPI; exports - XPI; imports - MPI.

22. The second issue associated with the nature of indices relates to the distinction between cost of living index (COL) and CPI. It was noted that it is recognised that there is not a very precise definition of what is meant by inflation. A notion of a "pure" price index has been recently discussed.

23. Mr. P. Hill's paper clarifies the relationships between CPIs intended to measure consumer price inflation and those intended to measure changes in the cost of living. It was noted that both objectives lead to the same kind of index formula in practice, provided that 'best practice' is followed. The similar best indices are both superlative indices. It is however, difficult to describe them as best practices, as it is very difficult to compute any type of superlative index in practice. They should therefore, perhaps be described as ideals to be aimed for. Indices that are 'biased' as measures of the cost of living indices can equally be regarded as 'biased' measures of inflation. At the same time, it was mentioned that there could be significant differences between the coverage, or domain, of indices intended to measure price inflation and those measuring changes in the COL. This could lead to significantly different results, especially over the longer term.

24. The issue of the scope of CPI was also discussed. It was noted that consumers' overall standard of living, or total welfare, depend not only on the quantities of goods and services consumed but also on the general environment in which those goods are consumed. 'Environment' is to be understood in a broad sense covering not only physical and climatic factors but also general economic, social and political factors affecting consumer welfare. A consumer may compensate for changes in these environmental factors by adjusting the quantities of goods and services consumed in order to maintain constant his total welfare. A COL index may therefore be defined as 'unconditional'' COL. It is defined as the ratio of the minimum consumption expenditures needed to maintain a constant level of total welfare under two different regimes between which both prices of goods consumed and also the 'environment' in which they are consumed may vary.

25. It was concluded that the appropriateness whether to broaden a COL index in this way to include factors other than price changes depends on the interest and purposes of the users.

26. The treatment of new goods in compiling pure price indices and COL indices was also discussed. In both cases treatment of new goods raises difficult conceptual and practical problems. In this context, definition of 'demand reservation price' was mentioned. It is defined as the hypothetical minimum price that would reduce demand to zero.

27. Several participants noted that the papers presented under this agenda item cover very important issue. Participants were informed that the Australian Bureau of Statistics (ABS) published an Analytical framework for price indices in 1997.

28. A second aspect in relation to the framework of price statistics is the PPI. For example, in Australia, at present, there is not a single PPI, but rather a whole range of producer price indices.

29. Some reservations were expressed regarding the use of the SNA as a framework for price statistics. It was noted that if the purpose of the index is deflation then national accounts offer a good frame. However, participants were reminded that CPIs existed well before the SNA was developed with a primary purpose - to compensate a group for price induced changes in their purchasing power. In this context, some participants feared that the national accounts framework is too restrictive.

30. Eurostat informed the meeting that in the case of the HICP, the national accounts conventions have to be followed unless there are good reasons to depart from them. One example, where there is a departure from the national accounts recommendations is that the HICP does not use imputations, which means that imputed rent for owner-occupied dwellings is not included. It was mentioned that the idea of using a national accounts framework for developing a system of price statistics is good as long as it is not used very rigidly.

Item 6 -Sampling and data collection

Discussant: Mr. G. O'Hanlon (Ireland)

Invited papers by: Canada, Austria and Netherlands; Contributed Paper by: Japan;

31. Mr. G. O'Hanlon (Ireland) opened the session by introducing three invited papers and one contributed paper. He noted that CPI suffers not only from an upper bias, which mainly concerned bankers, but also from a downward bias as perceived by most consumers. The three invited papers provided a very good contribution to the efforts of statisticians to improve the quality of the index. The papers addressed representativity issues and sampling problems in CPI production.

32. The Canadian paper identified areas in sampling that should be strengthened for a statistical office with limited budget. The recommendations made in the paper were the outcome from a specific situation which existed in Canada in 1990, which made it necessary to cut their sample size drastically. On the basis of the discussions from that time it was concluded that the most important issue in the quality of the index was not probability sampling, but treatment of quality changes.

33. The author referred to the need to identify the areas where the sample is least representative and introduced the idea of diversification of the sample of products, geographic localities and outlets, as well as issues of timing and frequency of collection. It was pointed out that emphasis should be put on product diversification because this was the biggest source of variation in price movement. This could be considered as a substitute for probability sampling in situation where funds were insufficient for real probability sampling.

34. It was pointed out that diversity of prices depends on whether decisions about prices are locally or centrally made. Although local selection has some advantages compared with central selection, some important factors might be missed, so it was recommended to have central selection to some degree and followed by local selection.

35. The Austrian paper looked at the practical aspect of sampling in terms of the trade off between representativity and continuity. It raised the question of the timing of payment and consumption, item specification (centralised and decentralised approach in defining item specification), training and personnel management.

36. The author mentioned two criteria of price collection that should be kept in mind:

- to collect in the present month what was observed in the previous month and

- to price what was seen in the shop, i.e. the most popular products in the shop by volume.

Sometimes these two criteria could be contradictory. In the Austrian case, if this happened priority was given to the latter criterion.

37. It was pointed out that the timing of price collection was not crucial in Austria although it might play a very important role in the CPI compilation in countries with high inflation.

38. Other issues discussed included the training of price collectors, the use of broad specifications and the greater importance of product variety over using a larger number of outlets and covering more regions. The use of electronic devices for price collection also contributed to higher quality data.

39. The paper from Netherlands presented results of the use of scanner data for assessing methods used in the selection of items and in the production of CPI. Cut-off sampling and three probability sampling methods were compared on the basis of Monte Carlo simulations from scanner data obtained directly from Dutch supermarkets. The methods compared were: simple random sampling, stratified sampling, cut-off sampling and sampling with probability proportional to size.

40. The empirical results showed that sampling with probability proportional to size did not necessarily decrease mean squared error. The research concluded that simple random sampling should not be used, sampling proportional to size was useful but that cut-off sampling led to better results. A second aspect of the research was comparing the computation of CPI from the scanner data, using different formulas.

41. As a result of this research, Netherlands has decided to use scanner data for computing CPI as from the next year. Concerning this, the question was raised whether to use all the data or to use only a sample. It was mentioned that with scanner data it was possible to compute not only Laspeyres but Fisher and other superlative indices as well.

42. The meeting discussed the use of scanner data for selection of items as well as for computing indices. Some participants were sceptical about the use of scanner data in CPI compilation. It was noted that the way the scanner data were used in the research, taking total value and dividing it by total quantities sold in order to get unit value, might not suit some statistical offices. In addition to this, the problem of discounts was discussed as well. The scanner data may have some hidden price changes that should not be included in the compilation of the index.

43. Participants noted that scanner data provided a tremendous possibility for comparing different methods in CPI production, using simulation. However, it was argued that it was not the right time to use them in current production of the index. Two reasons were given: coverage issues (extent to which scanner data cover retail sector) and quality of the scanner data. In addition to this it was noted that the expense involved in acquiring the data might be large if they have to be bought on the market. In order to decrease the cost it was suggested to collect the data directly from supermarkets. However some countries indicated that the quality of the data provided by the supermarkets did not suit their quality requirements.

44. Some other participants considered the Netherlands experience as most valuable and encouraged them to continue to work on it. It was concluded that there was a lot of scope for using scanner data but more work needed to be done. For the moment, most of the statistical offices had no expertise in using scanner data in the CPI compilation but they could start with combining them with data from other sources. Eurostat anticipates that in the long run (in 5-10 years) there would be no need for sending price collectors to the field as scanner data would cover the universe of prices. This would give the possibility to improve the quality of the index as a whole. A number of participants however, did not share this view.

45. It was also noted that there was a substantial difference between scanner data, which provided information on the prices of the items sold, and data collected by price collection, which provided information on the prices of the items available in the shops. In compiling CPI, the former was preferable. On the other hand, it was pointed out that scanner data could only cover goods and not services.

46. Observers from some developing countries indicated that because of the considerable difference between developing and developed countries the method applied in the Netherlands might not be convenient for their countries. It was, however, pointed out that there were other methods of electronic data collection which might also be useful to developing countries.

47. In the discussion it was noted that, in spite of the fact that in many statistical areas there had been a harmonisation of national practices, this was not the case with CPI. For example, concerning sampling methods, there were extreme situations. In some countries the sample was judgmental, whereas in some others it was fully random. For that reason, it was suggested that the new revised Manual should contribute to further harmonisation of price statistics by presenting best practices and assessing the consequences of decisions taken.

Discussant: Mr. R. Edwards (Australia)

Invited paper by: Netherlands; Contributed paper by: Slovenia, Azerbaijan and Netherlands

48. Mr. R. Edwards (Australia) introduced the paper from Netherlands dealing with sampling and non-sampling errors and commented on the three contributed papers.

49. The paper provided a general overview of the different types of errors that occurred during the compilation of CPI, methods used to measure these errors and procedures to minimise them. Two broad categories of errors were distinguished: sampling and non-sampling errors. Sampling errors consisted of selection and estimation errors and non-sampling errors of over-coverage, measurement, processing, under-coverage and non-response errors. Concerning the measurement of error, it was pointed out that it was difficult to estimate variance when non-probability sampling design was applied. Several estimation techniques were described and it was stressed that their application depended on the sampling design applied in practice.

50. The interrelationship between the sampling design, index bias and procedures to minimise errors was described at some length. The meeting discussed the possibility to speed up the process of introducing new items in the index in United States where probability sampling was used. It was noted that it was not possible to do so without violating the sampling method which was quite elaborate and rigid. It was however also not possible under cut-off sampling to bring in new items into the index immediately they were introduced in the market because their volume of sale would be small.

51. Some countries pointed out that they were more concerned with the representativity of the sample than with the introduction of new items.

52. The question of cost was raised again and the trade off between the number of outlets and number of items was discussed at some length. It was noted that the cost depends on the relative variance of outlets and items. If the variance of outlets is smaller than that of items, the number of items should be increased relative to outlets and vice versa. It was pointed out that the costs of price collection are closely related to sampling method and item descriptions. It was noted that an important type of bias, due to quality adjustment, which may arise from economic rather than statistical considerations should also be taken into account.

53. The discussion on sampling and data collection ended with the conclusion that:

- there was a lot of potential for use of the scanner data in the CPI production, but further investigation into it was needed,

- cost-effectiveness and practical constraints in the use of different sampling methods should be considered as a package,

- sampling methods were closely related with quality control.

Item 7- Issues of practical index work

Discussant: Mr. T. Lacroix (France)

Invited papers by: the United States, Finland, and the Netherlands; Contributed papers by: Switzerland, India, Italy and Mr. Erwin Diewert and Denis Lawrence

54. Mr.Thierry Lacroix (France) introduced and commented on the invited and contributed papers for this item, which dealt with theoretical and practical methods for quality adjustment. He pointed out that the methods for determining quality changes and the question of substitution effect when new product appears were closely linked.

55. During the discussion it was noted that there is a lot of controversy about bias, and the situation in countries varied. Hedonic method was identified as the best method for treating quality changes in various situations and scanner data may make it easier to do hedonics. It was pointed out that scanner data represent a very costly source of information and therefore international co-operation in the use of results from hedonic methods is very important. Eurostat is, for example, proposing this approach for computers.

56. The United States paper highlighted the main problems with quality and variety changes and described the major methods used by statistical offices to adjust for differences in the quality of a good or service when one item replaced another in the sample. The methods were categorised as either direct (or explicit), including the production cost and hedonic approaches, or indirect (or implicit), such as linking or class-mean imputation. The importance of quality adjustment for the accuracy of the index and its leading contribution to the bias in the index was highlighted.

57. The Finish paper presents different advanced methods as hedonics used for quality adjustment in different areas of CPI. It describes in particular new method for estimating the price effect of quality change for new cars that was implemented in the current Finish CPI. It is based on relatively simple hedonic models where the price of a car is expressed in relation to its size and power. Its main advantage is that there is no need for weights to compute the price index at the lowest level of aggregation. It was proposed as a substitute for random or stratified sampling. It was noted that it was not clear whether the value or quantity of the cars should be used for sampling and whether price could determine the quality of a car.

58. The Netherlands paper dealt with substitution and new/disappearing good bias. A modified price index was computed on the basis of all goods and services which were available in both periods and the relative elasticities of substitution. A simplified method of this procedure was demonstrated.

59. The general impression about the method was that it implied a large number of simplified assumptions which were not always reasonable. However, some participants argued that the assumptions made could accommodate a large part of real life situations.

60. Some participants noted that quality changes were mainly a sampling problem. It was, therefore, necessary to keep the sample up-to-date and to improve its representativity. Statisticians were forced to make many changes and to introduce new items. In doing so the real price change might be underestimated or overestimated. However, if the old items disappearing from the market were not replaced with the new ones, the index would be inevitably biased.

61. Participants encouraged the authors to continue their developmental work and pleaded that the authors should, in principle, write papers in a language that was understandable for economists as well. The discussion concluded that theorists and practitioners need to work together and come up with common solutions acceptable to both.

Item 8 -Quality management

Discussant: Mr. J. Astin (Eurostat)

Invited papers by: United Kingdom and Israel; Contributed paper by: Netherlands

62. In recent years many national statistical offices and international organisations have embarked on programs of quality management and quality assessment. These programs do not cover only the quality aspect of technical work and produced statistics but increasingly also relate to organisation, training and all other aspects of statistical work. The objective of these programs is ultimately to increase the quality, timeliness, accuracy and relevance of statistics produced with constant or decreasing financial and human resources. The user needs usually have a paramount role in defining the quality management programs.

63. The paper by Mr. D. Fenwick (UK) explained an ongoing program designed to improve the quality of the UK's Retail Price Index. It described different aspects of a quality management system and how the ONS addresses them. The ISO 9000 standard is a key element underlying the formal quality management system for the monthly production process, which also includes the use of the EFQM Excellence Model and benchmarking against other national statistical institutes.

64. The paper reviewed of the meaning of quality, the different approaches available to quality management, the potential these have to impact on business performance and how the latter can be measured. In addition, the paper discussed some of the benefits from using a formal Quality Management system and addressed some of the disadvantages that can be associated with it. Business performance is considered in the context of the ultimate quality of the consumer price index. Finally, the paper presented a number of case studies giving extra practical insight.

65. The paper from Israel (Mr. Y. Finkel) presented preliminary results of a five-year program aimed at re-designing the computation procedure of CPI in Israel that started in 1996. The objective of the program was to re-design the system architecture and basic principles of CPI production. Starting from re-definition of fundamental principles, terminology and the use of modern technology in the CPI framework, the training program was developed including elaborated training methods and testing procedures.

66. Starting from the definition of users, the cycle of CPI computation and an occupational analysis, a general training plan was devised. The program includes on-the-job training and specialised training. Furthermore, five categories of technology involved in CPI production were identified and suitable participants for training were selected. The training program itself includes four stages: training for operation of computers and other technological components; specific training for each role within the system; integration of the two mentioned and testing procedures. Although not completed, the program so far shows big potential and eventual benefits measurable in terms of production of a higher quality CPI.

67. In the ensuing discussion, participants generally agreed that quality management and training are very much related and that there is interaction between them. Adequate training will lead to better quality and would facilitate quality management, while well developed and defined quality management programs will ensure guidance and indicate which parts of the system may need additional training.

68. Another question that raised interest was the cost of training and its impact on the quality of CPI. Offices that have had more extensive experience with training reported that costs involved are justifiable when measured in terms of the improved quality of the CPI. It was, however, mentioned that it may be difficult to obtain an exact information about the extent of the improvement in a situation when a benchmark does not exists or when comparing an ex-ante (pre-training) situation to and an ex-post (post-training) because it may lead to erroneous conclusion.

69. Some offices have also gained a substantial experience with contracting out their data collection for CPI to professional companies. Both advantages and disadvantages of this option were emphasised. Contracting out data collection may cost less if organised by the statistical office itself. Data collection may also be easier carried out when detailed requirements are set out in the properly prepared contract with certain performance indicators. On the other hand, the data obtained in this way must often be thoroughly screened in order to ensure the desired quality.

70. Participants discussed with particular interest the issue of quality of CPI with regard to customers. Various experiences in dealing with customers and ideas how to improve the relationship were mentioned. There was a general agreement that a starting point in all plans for quality improvement and training should be related to identification of customers and their needs. CPI producers should be able to assess where customer requirements and needs are met and starting from this assessment develop their programs for quality improvement and additional training. While all type of users need high quality statistics, it was noted that statistical offices have to ensure that needs of both "small" (consultants, experts, academia) and "big" users (governments, financial institutions, international organisations) are met. In that respect it was mentioned that the feedback from "small" users is also important.

71. Some participants thought that CPI producers have to interact with customers in order to help them define their needs (what indices to produce; details of quality, which sub-indices, what level of accuracy of the index, etc.) It is believed that engaging customers in this way would help organisation to better meet sometimes conflicting needs of their customers and determine both which users should have priority and what are priorities of particular users. Some offices have also developed programs to involve various types of users in dialogue about the quality of statistics produced.

Updated by AL. Last update: 2 February 2000.