Organisation de Coopération et de Développement Economiques
Organisation for Economic Co-operation and Development

REPORT ON THE JOINT WORLD BANK-OECD SEMINAR

Joint WORLD BANK – OECD SEMINAR ON PURCHASING POWER PARITIES
Recent Advances in Methods and Applications
WASHINGTON D.C.
30 January – 2 February 2001



Organisation de Coopération et de Développement Economiques
Organisation for Economic Co-operation and Development

PURCHASING POWER PARITIES:RECENT ADVANCES IN METHODS AND APPLICATIONS

REPORT ON THE JOINT WORLD BANK-OECD SEMINAR

(30 January 2001 - 2 February 2001, Washington, D.C.)

Introduction

This report summarises the papers presented at the meeting and the principal points made, though not necessarily generally agreed, after each presentation.

The papers presented at the seminar are available on the internet. Hyperlinks are provided in this report to each paper so it can be downloaded from the internet. The full set of papers can be found by clicking on the “meetings” button at and going to the heading “Joint World Bank OECD Seminar on Purchasing Power Parities”.

Opening statements

Opening statements were made by Shaida Badiee and Nicholas Stern of the World Bank and by Paul McCarthy of the OECD.

▪The last meeting of this type was held in 1989 at the OECD in Paris. Since then the 1990, 1993 and 1996 rounds of the European Comparison Programme (ECP) have been completed, the 1999 round will be concluded during 2001 and data collection for the 2002 round has already commenced. The 1993 round of the world wide International Comparison Programme (ICP) has also been completed. There have been no further rounds since then and the World Bank is currently promoting the rejuvenation and continuation of the ICP.

▪The ICP provides the data base for the core policies of international organisations relating to poverty alleviation and economic development. Yet the Programme faces tremendous difficulties in developing countries where such information is badly needed because the resources committed to the ICP have been declining. This decline is the result of a combination of factors. They include the fact that neither the PPPs themselves nor the implications of using exchange rates as an alternative are well understood by many policy-makers and analysts, the lack of confidence in the PPP estimates that are available, and the fact that the Programme is too demanding and resource intensive for any one organisation to take full responsibility for it.

▪Importantly, both the Castles Report and the Ryten Report reaffirm the importance of PPPs. Both also identify the areas on which to focus to improve PPPs. However, the resource problem requires international development agencies to recognise their vested interest in keeping the Programme alive. It also requires close co-operation and cost-sharing arrangements between regional and international partners of ICP.

▪The seminar provides a timely opportunity to review the outcomes of PPP research that had been undertaken since the 1989 meeting. The review serves a number of purposes. These include: keeping practitioners informed of theoretical developments and researchers up-to-date with practice, identifying the practical spin-offs from the research that could be used to improve the ECP and ICP, and illustrating further the usefulness of PPPs by describing their application in a variety of economic studies.

The opening statements were concluded by thanking Prasada Rao, who was responsible for the idea of having the seminar and getting it off the ground.

1




Organisation de Coopération et de Développement Economiques
Organisation for Economic Co-operation and Development

Topic A: Data issues

Sensitivity of results to the number of products priced

Paper: Optimisation of the number of basic headings and of items (Eurostat)

Eurostat has been investigating the impact of reducing the number of Basic Headings (BHs) from their present level and also the impact of reducing the number of products priced. Two simulations were carried out to attempt to measure the effects.

The first simulation was designed to evaluate the differences in results when using the same data (7 surveys, 1997 updated prices) and 3 different classifications: CHGS, COICOP 4 digit (COICOP4) classification with 199 BHs and COICOP 3 digit (COICOP3) classification with 117 BHs, also called classes).

The second simulation evaluated the differences resulting from using a smaller number of items within different surveys (4 surveys were tested), in conjunction with both the COICOP4 and COICOP3 classifications.

The results presented in this paper show that the migration from CHGS to COICOP has very little impact on the PPP results. Taking 1997 data for all surveys and calculating the differences in results based on fewer BHs (117 BHs - COICOP3) show that the they are very close to those resulting from the use of the present number of BHs (199 BHs – COICOP4). Also, reducing the number of products covered in the surveys by a very large amount has a generally insignificant effect on the overall PPPs. The simulation results cannot tell whether the PPPs calculated with fewer items are more or less “accurate” than the previous results: all they can do is to show that they do not differ much.

Paper: On reducing the number of basic headings and products priced (OECD)

This paper attempts to clarify a number issues raised in the Eurostat paper on the same issue. The Eurostat simulations are a response to the often repeated criticism that the expenditure classification is too detailed and product lists are too long. Experience shows that that many countries find it difficult to provide expenditure data at the five-digit level of CHGS-PPP. Because of this, countries co-ordinated by the OECD are currently only required to provide weights, and CPI sub indices, at the four-digit level. However, their selection of products takes place at the five-digit level to ensure a spread of products comparable to that of EU Member States. This suggests that, when the change to COICOP-PPP is effected, basic headings at different levels of aggregation could be identified for reporting prices, CPI sub-indices and expenditures.

Product lists are quite long given the preference for brand and model specifications, the need for each country to price at least one representative product per basic heading and the growth in the number of participating countries (from nine in 1975 to 31 in 1999). Product lists have grown not to increase product coverage at the basic heading level but, rather, to ensure country coverage at the basic heading level. Breaking down the participating countries into three groups has resulted in group product lists being more balanced, smaller and more manageable. Although the individual group lists are smaller, the combined list contains a similar number of products as before.

1




Organisation de Coopération et de Développement Economiques
Organisation for Economic Co-operation and Development

The paper concludes that the results of the Eurostat analysis are not as convincing as they may appear to be initially. In particular, the analysis is based on examining the effects of ex post reductions in the number of products priced and this cannot be readily carried across into an ex ante reduction, particularly given that survey baskets evolve over time with the introduction of new products and the elimination of out-of-date products. Also, reductions in the products priced and in the number of basic headings are not complementary aims. Finally, the analysis is very narrowly based, mainly covering western European countries. The Candidate Countries and other transition economies in the comparison have different consumption patterns which could give quite different results for your analysis. Moreover, the analysis draws conclusions on the basis of one year only, which is not necessarily representative.

Discussion on above 2 papers

  • In discussion, the point was made that decisions like these should ideally be taken against a background which fully took into account the requirements of users: against an understanding of the theoretical mechanisms at work: and with some view as to how likely the observed findings were to generalise to other situations. While the theoretical basis of the findings in the Eurostat paper were not fully understood, it seemed likely that they were heavily dependent on the relative homogeneity of the data-set studied: and that the results might not generalise to less homogeneous data sets.
  • There are no objective tests of the accuracy of PPPs so it is not possible to measure the effects on accuracy of reducing the number of items for which prices are collected. While it is relatively easy to make simulations of the effect of reducing items, such simulations merely show how smaller sets of items produce results which are different from those obtained using the full set of price data. It is important to note that such simulations can only be carried out after the pricing has taken place and so they do not form a good basis for making decisions on the extent to which the numbers of price specifications can be reduced.
  • There appears to be scope for reducing the number of items priced in ECP comparisons and, assuming no change in the resources devoted to price collection, this would allow price collectors to improve the quality of the prices collected by collecting more prices per item. Prices of identical items vary both between outlets and regions within countries. If the number of items were reduced it would make it possible to improve the outlet and regional coverage of the remaining, fewer, items that have to be priced.
  • Reductions should be made in consultation with the countries concerned. They should not go so far that they undermine the links between the sub-regions within the ECP or the links between Europe and the other regions covered by the ICP.
  • Aggregating basic headings will reduce the homogeneity of items at the basic heading level and this should be avoided if the number of items priced were to be reduced as well. An alternative would be to have different levels of disaggregation: the most detailed level for the collection of prices and less detailed levels for the provision of expenditure weights and price indices.

Editing procedures for aggregation below basic heading level

Paper: Editing procedures for aggregation below basic heading level (OECD)

This paper describes the editing procedures based on measures of central tendency and variability in the editing of price and expenditure data that are employed to edit the price data provided by countries participating in the Eurostat-OECD PPP Programme.

1




Organisation de Coopération et de Développement Economiques
Organisation for Economic Co-operation and Development

Two tables are prepared to edit the price data: one table covers basic headings, the other covers products. These tables are commonly called “Quaranta tables”. They are designed to provide an indication of consistency with respect to the basic headings, and to provide an analysis of the variability of the surveyed prices. The main aim of such quality control is to identify those prices which generate values of indices of price levels which fall outside given ranges.

The Quaranta tables are useful tools in identifying and correcting outliers. However, experience has also shown that the tables by themselves are not enough. The Quaranta tables need to be supplied to participating countries accompanied by a commentary which guides the countries around them, by pointing out to what they should be checking. When Quaranta tables have been circulated without such a commentary, their effectiveness as an editing tool has been noticeably blunted.

Discussion

  • Countries participating in the ECP are encouraged to carry out more credibility checking of their price data before submitting them to Eurostat and the OECD. This is done mainly by looking at the variance of different price quotations for specified items. Since only the collecting countries can validate their price data, it is important that they do as much editing as possible before sending them to the compiling agencies.
  • The OECD encourages countries to exchange data for mutual review prior to reporting them to the OECD. For example, Canada, Mexico and the United States circulate their prices for construction projects among themselves before submission to the OECD.
  • The main way in which Eurostat and the OECD check the price data from countries is to calculate “price-level indices” for basic headings with which to identify outliers and to query these with the countries concerned. There are no automatic rules for dealing with outliers as such an approach is not considered to be appropriate. Outliers are not necessarily wrong and whether or not they are can only be determined by verification by the reporting country.
  • The OECD is experimenting with comparisons of basic parities from one benchmark to the next, but time series comparisons remain problematic because of changes in specifications and in the numbers of countries participating in successive rounds.
  • The concept of “equivalent brands” is being used in the ECP. This can be seen as a compromise between generic specifications and brand and model specific specifications. Instead of specifying a single make and model of, say, TV sets, a group of brands is identified which are essentially the same in all important respects although they may be traded in different countries under different name and model numbers.
  • The OECD encourages countries to treat brand and model specific specifications as generic specifications when the brands and models specified are not available in their markets. Instead of pricing the brand and model specified, the country prices and an equivalent brand and model. Australia, for example, prices domestically produced motor vehicles which it has matched against the brands and models specified for the comparison.

Prioritisation of sectors for comparison

Paper: Prioritisation of sectors for comparison (John Astin)

  • PPPs must continue to be compiled for total GDP. It would not be useful for most purposes to compile PPPs only for PFCE as recommended by Castles and Ryten.

1




Organisation de Coopération et de Développement Economiques
Organisation for Economic Co-operation and Development
  • One way to improve overall reliability is to concentrate on improving the PPPs for PFCE because significant gains can be realised more easily there and because PFCE is such a large part of GDP. On the other hand there is considerable analytic interest in having good PPPs for GFCF and for non-market services such as health and education. A broad-based approach to improvements is preferable, at least in the medium term. In the short term it may be wise to concentrate more initially on PFCE.
  • The current method used by Eurostat to price equipment goods and construction projects for its annual benchmark comparisons is expensive since it involves private consultants. In the absence of an alternative approach, perhaps Eurostat should reduce the frequency of benchmarks, to say three yearly intervals. This should be done in a way that preserves the organisational structure within national offices. In other words, price collections for PFCE would continue to be spread over three years, but price collections for GFCF would take place only in the third year or benchmark year. PPPs for GDP for the years between benchmarks would be obtained by extrapolation. This is the method currently used for non-EU OECD countries.

Integration of CPI and ICP

Paper: Integration of CPI and PPP: methodological issues, feasibility and recommendations (D.S. Prasada Rao)

The principal objective of the present paper is to identify the possible sources for integrating the activities of gathering price and expenditure share data for the CPI and PPP computations. The paper elaborates on the similarities and differences in the conceptual and methodological foundations of these two activities, and argues that the long term viability of the ICP and its increased applicability and relevance to economists and policy makers in different countries will largely depend upon the integration of these activities in the near future. The paper starts with an overview of the conceptual framework and methods used in compiling CPI and PPP. It examines the price and expenditure share data requirements of CPI and PPP and elaborates on the aggregation methods used. The proposals for integration are based on a schematic representation of various price index number activities, undertaken at the national level, and it is argued that PPP activity can be considered as a natural extension of spatial or inter-area price index numbers. The paper also argues that a close alignment of PPP work with the CPI will have implications for the way ICP is implemented, and alludes to the possibility of identifying links between countries which have significantly overlapping CPI baskets of goods and services.

Paper: Integrating inter-area and international price comparisons with consumer price index compilation (Kimberly Zieschang, Paul Armknecht, Dale Smith)

The best available inter-area and international price comparisons for consumer goods and services use price and expenditure data collected from national statistical offices (NSOs). Generally, the price data are specially collected for the PPP Programme by the NSOs, which are also generally responsible for compiling consumer price indices (CPIs).