World Comparison Real Gross Domestic Product and Purchasing Power

IV. DETAILED RESULTS OF THE GLOBAL COMPARISON

A. COUNTRY COVERAGE AND ORGANIZATION

In phase V of the ICP a total of 64 countries and areas participated in one or another of the regional comparisons conducted with reference year 1985, as described in section III. After the regional comparisons were completed in the European Communities, the OECD, Eastern Europe (Group II), Africa, ESCAP and the Caribbean, arrangements were made to generate global results by linking the above regional comparisons. From the outset, the global programme in phase V was designed to be based on the set of regional comparisons with reference year 1985. As all regional comparisons follow basically the same principles of the general ICP methodology, it helped to establish consistency among regional data. The database compiled and processed in the regional exercises was used in the global computation. The processing yielded globally linked results for 56 countries.

At the world level, the Statistical Division of the United Nations Secretariat (UNSTAT) played an overall coordinating role in respect to the global comparison effort. The Statistical Office of the European Communities (EUROSTAT) undertook the calculation of the global results.

An overview of the global framework of the 1985 ICP is provided in figure I.

In phase V, regional comparisons were linked to form a global comparison by the core country method. Core comparisons are either bilateral comparisons between two countries belonging to different regions or the result of a country's participation in two different regional comparisons simultaneously. In phase V it was intended that in each region one or more countries, designated as "core" countries, would price some specifications for another region as well as their own. The choice of a country to play the role of core country may have a significant influence on the quantity indices between countries belonging to different regions. It is thus important to have a sufficient number of core comparisons so that this source of error, which may be likened to sampling error, is reduced.

One of the greatest difficulties in phase V has been the insufficient number of core comparisons. Although about 20 core comparisons were envisaged when phase V of the ICP was launched in 1984, one year later, when the actual work began, it became clear that resources would not permit more than 10. Soon, even this expectation turned out to be over-optimistic. The United Nations Statistical Commission, at its twenty-fifth session, in 1989, recognized that unforeseen difficulties were encountered in respect to the linking of the regional comparisons because the number of core comparisons carried out proved to be significantly lower than originally planned. Some core comparisons were discontinued; work on others was not proceeding well. Finally, only the United Kingdom-Kenya comparison and the two implicit core comparisons (involving Austria and Japan, by virtue of their participation in the work in two regions simultaneously) were satisfactorily completed in phase V.

There are several reasons why the core comparisons have proved to be so difficult. One important reason is that the core comparisons constitute a substantial additional burden on the countries participating in them. This arises because of the limited comparability between countries belonging to different regions. At the same time, participation in the core comparison brings little direct benefit to the core countries themselves, since, however interesting their results may look in the overall comparisons, they may not be particularly interested in the results of the specific bilateral comparison with the partner country. Another important factor was the shortage of financial resources, which restricted the travel of country experts to meet their counterparts to match commodity specifications in the partner countries. The entire system of the regionalized world comparison was based on the assumption that, for the sake of the reliability of the world comparison, there would be willingness on the part of a number of countries to accept the extra burden of the core comparison work. In practice, however, the extent of cooperation fell short of anticipations.

In spite of the many drop-outs, EUROSTAT was able to provide the necessary minimum links among most regions. As agreed by the coordinating international organizations involved in the ICP, the data were processed to present results at a common aggregated classification level concerning 53 categories, or detailed expenditures on GDP. The EEC countries were linked with the other OECD countries through the common EEC-OECD comparison. Hungary, Poland and Yugoslavia were linked by binary comparisons through Austria at the level of 53 categories. For the ESCAP countries, the 53 categories of the regional comparison results were linked through Japan's dual participation in the OECD and ESCAP comparisons. Africa was linked with the EEC in the following manner. For consumption, the link was obtained by comparing prices of all the countries involved and the comparison was reinforced by additional pricing carried out by the United Kingdom and Kenya. For the other aggregates, the link was made through a binary United Kingdom-Kenya comparison. The linking of the African comparison was achieved at the level of 53 categories by keeping for each aggregate the volume ratio between the EEC total and the African total. Overall, globally linked data on GDP and its main components were generated for 56 countries at the 53-category level. Tables 1-14 contain data according to this GDP breakdown that constitute the most detailed available results of the programme at the global level with reference year 1985.

Working on the processing of the global results, EUROSTAT faced outstanding technical difficulties due to problems of correspondence of classifications and differences in the methods applied by the various regions. Regarding the linking of the Caribbean region to the rest of the world, these efforts have greatly delayed the completion of the global processing. In spite of the lack of a core comparison of the Caribbean countries with other regions, EUROSTAT attempted to establish a link for the Caribbean with Africa for the final consumption of households by using the prices of all the countries of the two regions; however, the difficulties encountered could not be resolved entirely and the comparison of aggregates other than consumption could not be achieved. In addition to this, the Caribbean regional data themselves had not been approved by the time of the preparation of this report due to difficulties in convening a regional meeting of the participating countries. To avoid further delay in releasing the global results, the organizations involved reluctantly gave up the intention to incorporate the seven Caribbean countries in the final tables. Therefore, in this publication comparison results for the Caribbean countries are not presented. In the case of Nepal, the participation of which in the ESCAP programme was limited to the comparison on consumption in phase V, results are presented only in the regional context.

During the processing of the final tables, EUROSTAT made every effort to keep the regional results unchanged. Since this was not always possible, readers will find that the global results in respect of some countries may not be the same as those obtained in the regional context. This phenomenon is the consequence of the system of international average prices applied, namely, that the price structure and, accordingly, the PPPs and the quantity ratios depend on the country coverage for which they are calculated. In other words, this means that the parities and real quantity ratios between a given pair of countries (say, France and Italy) would be different if calculated bilaterally, or within the frame of the European Communities (EEC), or of OECD, or of Europe, or, let us say, in the context of the 56 globally linked countries. To avoid the proliferation of the results, the principle of "fixity" has been imposed. The practice of "fixing" the results implies that comparison results between countries belonging to a certain grouping are supposed to remain unchanged even if their results are later incorporated in any other comparison involving a larger group of countries. In phase V, however, in spite of EUROSTAT's best efforts regarding the global processing, the fixity principle could not be honoured for all countries and regions in this study, partly because the expenditures used for the same country in two different regional comparisons diverged. It should be noted that in the global comparison, figures for Austria and Japan are those used in the common EEC-OECD comparison.

Having said this, users of the global data should be aware of the following considerations when analysing 1985 results worldwide. The linking was carried out by starting from the regional results available at the 53-cate- level. While most of the intraregional results were originally computed by using an aggregation method that guarantees additivity, this property is not guaranteed for global (interregional) results. As a consequence of the method of linking applied in phase V, fixity of results for countries in some regions (mainly ESCAP and Group 11) might have been slightly affected after the global processing was completed, although the intention was to keep fixity for all regions throughout the linking. Furthermore, one should keep in mind that the only real values that are additive are those of the United States; that is, in the case of the United States the real value is equal to the nominal value for each aggregate. However, this is not the case for other countries in the comparison.

For detailed information on the method of linking the regional results in phase V, readers may direct inquiries to EUROSTAT, Luxembourg.

In the preceding sections some thought was already given to potential uses of international comparisons of macroeconomic aggregates, such as GDP and its components. In this section some further aspects of the advantages and limitations of ICP results are presented.

In general, the level of accuracy and reliability of overall ICP results is believed to be high; however, the limitations described here deserve attention with a view to the correct interpretation of the results.

1. Some uses of ICP data

In an increasingly interrelated world economy, countries-whether large or small-need to review their economic attainments and prospects in an international context. Perhaps the most direct use of ICP results is the use of the data for gaining new perspectives for studying a country's economic situation in respect to its price structure, its real output, and its utilization of real output for various purposes. Naturally, prices and quantities could be studied solely in the context of the national information of the country concerned. On the other hand, the ICP approach can permit the evaluation of the development of price levels and the structure of prices in an intercountry perspective in real terms, in addition to the information obtained within the given economy over time. Thus, the insights gained from studies limited to available national data may be augmented, if the evidence for similar economic topics in other countries is taken into account. Such issues as the relationship of food prices within a country in comparison to the prices of industrial goods or services can be much better analysed if the equivalent relationships in other countries can also be studied.

In the regional and global comparisons in the ICP framework, the data of 64 countries are available to engage in research of this sort, permitting a wide scope of comparisons. The principal advantage of the ICP lies in achieving real term comparability for the economic data of nations by converting them into quantitatively meaningful magnitudes internationally. This is the consequence of using PPPs as conversion factors which eliminate price-level differences between countries so that the PPP-converted expenditures reflect only differences in the quantities. Low quantities times high prices can just as easily produce a certain total value (and percentage share in the national total expenditure) as high quantities multiplied by low prices can. However, if the quantity of, say, food consumption were to be estimated in both countries, and the same prices were applied to it in both countries, the level of food consumption in the two countries would become comparable in real terms and the shares of food consumption within total expenditure could be more meaningfully compared between them.

For example, in comparing the food consumption levch in country A and country B, if unadjusted national data are used, it is not possible to tell whether an identical share of food consumption in the two countries (say, 15 per cent of total expenditures in both cases) reveals a true identity of those shares or not. The ICP results, in contrast, reveal the level and the share of food consumption in the two countries in real terms, which transforms the nominal national data to a comparable quantitative basis.

The ICP, in the process of comparisons, produces comparable quantitative information by providing an international analysis of national price systems. Departing from the simple notion that data about prices multiplied by the relevant quantities equal the values for the same category, the use of international prices not only can establish the quantities in real terms for each country, but also can throw light on the national prices in each country in every expenditure category. Hence, it can lead to the evaluation of national price levels and price structures in the international context. Moreover, it can lead to insights into areas where a country's prices are higher or lower than they are internationally. Regarding price structure and price-level analysis, global ICP data provided in table 6 show the effects of macroeconomic policies that may move the exchange rates independently of relative price movements, while table 7 shows the effects of different cost and pricing structures within a country.

The use of internationally comparable quantity and price data for any given country can be manifold. As mentioned earlier, it can enlighten the study of the domestic structure of prices and may lead to the better understanding of the underlying causes of national price peculiarities. Furthermore, it can lead to a more comprehensive evaluation of consumption levels, as it can enhance the research of domestic consumption structures and may inspire decisions based on their probable evolution.

In respect to capital formation, if only national currency data are used, the problems faced are similar. Only by recalculating the national capital formation data into internationally feasible quantity and price magnitudes can a comprehensive and thorough evaluation of the data be performed. Indeed, the problems of economic growth and development can be studied in a much more extensive way if the database expressed in national currency is augmented by internationally comparable data, such as those produced by the ICP.

Overall, the data from the ICP permit the review of a number of issues in an international economic perspective that cannot be effected by relying on national data alone. In addition to macroeconomic analysis, countries often use the vast information base of the ICP for the evaluation of domestic development projects in specific areas.

In the context of the uses of ICP data, it would be erroneous to suggest that either countries or international organizations can limit their interest to the results of just the regional or just the world ICP data. Typically, most users can benefit from the availability and simultaneous analysis of both.

Since most countries have intensive economic interactions within their own regions, comparative data about countries with which they share a common interest can be most helpful. The regional ICP comparisons provide important details for such reviews. Likewise, organizations responsible for economic matters within a region need data about the composition of production and consumption by the countries that form the region. Also, the evaluation of regional prospects can be enhanced by having internationally comparable data about capital formation, the role of government consumption in the countries and similar information that is also available from the ICP.

Nevertheless, countries usually have important international economic ties outside their regions as well. For example, more and more economic transactions take place between countries from various continents, like economic relations between developed market economies and dynamically emerging countries of Asia or the transition economies of Eastern Europe. In order to understand the background to economic processes in the countries in question, the availability of internationally comparable world-level data can be extremely helpful. The worldlevel comparisons, in addition to their use to countries, can be very well utilized by global business entities (such as transnational corporations) and by world-level international organizations as well. In fact, global ICP data have been widely used in all kinds of research projects and their analytical uses seem to be boundless. These considerations, however, should in no way curtail the validity of the agreement reached by the producers of ICP data that the results would not be used for administrative purposes at the global level. The same applies to regional ICP results unless there is a consent of the participating countries concerned e.g., in the case of the EEC. This position is supported by the United Nations Statistical Commission and is reflected accordingly in its documents.

2. Limitations

In the following, some limitations of the ICP data are discussed. Understanding the limitations of methodology and other statistical factors, indeed, is a prerequisite for the proper use of the ICP results. Although the limitations do not necessarily indicate shortcomings of the ICP, as they may simply reflect properties of the methods applied, analysts who are not aware of or disregard the limitations of the data can reach faulty conclusions. Moreover, by attempting to reduce unfunded expectations towards the programme, one can hope to avoid disappointment with the results among users. Also, for the sake of fairness, there should be an emphasis not only on the advantages of having ICP results available but on the limits of their uses as well.

The enumerated benefits from the ICP do not accrue to the countries or international bodies without costs. Although these costs are probably not very high if compared to the importance of studying the economic issues at stake, they are not negligible in the reduced framework of ever-shrinking statistical budgets available at national and international levels. On these funds alone the ICP effort could not have been materialized in phase V, had it not been for the cooperative spirit and contributions by governmental and international agencies as well as the unselfish enthusiasm and willingness of participating countries to incur most of the expenses involved in reporting data for the comparison. A costly and time- and labour-consuming venture, the ICP had to be cautiously designed and carried out in phase V to satisfy user needs by producing meaningful results with the resources available.

Restricted availability of resources, naturally, has made an impact on the outcome of phase V, particularly in respect to the global results. The coordinating agencies often decided to choose solutions that could realistically be accomplished even if they had preferred a more favourable option from a strictly technical point of view. The way of linking the regional results of phase V is an example of such compromises, as described in section IV,B.

Care was taken to avoid practices that would seriously damage the reliability of results. However, the most obvious limitations of the data, which should be kept in mind when analysing ICP results, are described in the following paragraphs, supplemented by selected references to other characteristics that have already been taken up elsewhere in this report.

1. Possibility of multiple results

Users should not be perplexed by the fact that international comparisons can lead to somewhat divergent results. One source of differences can be the method of conversion applied. The reasons why results obtained by using official exchange rates for conversion are different from PPP-based data, like those generated in the ICP, are given in section 1. The chances of the incidence of multiple results for certain countries grow if a number of com parisons with overlapping country coverage are conducted, as has been the case with the ICP in phase V It is important to recognize that in PPP-based comparisons, even if no statistical inaccuracies play a role, comparisons of countries can yield different results depending on the number of countries involved in the comparisons, that is, whether they are carried out in a binary or in some multilateral (regional, continental, world-wide) context. The application of the fixity principle aims at reducing the consequences of the possible proliferation of results, as explained in section IVB.

2. The impact of aggregation procedures

This report cannot review the technical details pertaining to different aggregation procedures; these can be found in the Handbook of the International Comparison Programme and other ICP literature. In actual practice, when prices for many hundreds of commodities, classifications with over 250 expenditure categories and scores of countries are involved in the work, the aggregation procedures are vastly more complex than the illustrative examples in the brief methodological references indicate.

Statisticians have developed a number of procedures in answer to various problems encountered in comparing countries. Differences in the aggregation procedures apacross regions have arisen because alternative methodologies typically offer various advantages for carrying out the calculations and for interpreting their results. The results of the comparisons are influenced by the choice of the aggregation method, such as the Geary-Khamis formula and the EKS method. Generally, statisticians do not claim absolute superiority for one method over the other, in the sense that any one would give positively "true" or preferable results for all comparisons under all circumstances. The principal advantage of the G-K method is that it produces additive results in accord with standard national accounting practice. Having the property of matrix consistency, the G-K results can be compared across countries for any basic heading and will add up to the total at any level of aggregation, unlike EKS or other non-additive formulas which do not meet this requirement. As a matter of fact, the matrix consistency of the G-K method must be given up at the world level if the fixity principle is applied. In phase V, for instance, fixity has been applied in a way that the numbers do not add up to GDP, and comparisons across the rows of tables I and 2 for countries of different regions may have only limited analytical value. Regarding the choice of aggregation methods, those experts who prefer the EKS method generally argue that it uses a weighting system that accords the same importance to each country, which may have more basis in consumer theory. As a consequence, the use of the EKS system leads to a less narrow range of real values than the one produced by the G-K method. This may appear a desirable feature to some; nevertheless, it should be preferred only by those who do not consider additivity, which would make international comparison results analogous to national accounts, an indispensable property. In phase V the organizers of the various regional comparisons have chosen either of these two formulas for aggregation and, as a matter of fact, in most regions the G-K method was applied in the 1985 round; however, this may not necessarily be their choice for future rounds of the comparison.

In a similar manner, the choice of the index formula (CPD or EKS) used in computing basic parities will generate different estimates at the level of the basic heading, though after the aggregations of the basic headings the overall results are unlikely to be affected by which method was selected. In summary, as the choices made in the comparison work are considered judgements, users of the data need to recall that the results could be influenced by these choices. This influence not only adds to, but also intertwines with, the complexities arising from the particular survey framework utilized.

3. Accuracy of the comparisons

No matter how sophisticated and meticulous the mathematical techniques used for aggregation or applied in the treatment of data are, the results of international comparisons, at any rate, depend on the quality of the underlying national data sets collected. This holds equally for the regional and the global comparisons, since the processing work of the latter was based on data compiled in the first place for the purposes of regional studies.

It should be kept in mind that the quality of national statistical information can vary. Economic measurements themselves, such as GDP estimates, are subject to some measurement error. Statisticians toil at reducing the inaccuracies, which is a long and arduous process. Even though there has been much progress in introducing international statistical standards, such as the System of National Accounts (SNA), into the national practices of countries, adherence to these recommendations is not at the same level in all countries. In addition, the ICP requires a breakdown of GDP expenditures that is much more detailed than either the classification applied by the SNA or the national practices of most countries. Considering that countries need to estimate many of the expenditure details which they report for ICP purposes, chances are that the principles of methodology in reclassifying national data across ICP expenditure groups might not be Strictly followed. If this happens, it may introduce a certain degree of incomparability in the basic data and, consequently, in the results. While organizers of the regional comparisons had more possibilities to correct this source of error, resources for eliminating such incomparability among the regions for data at the global level were extremely limited in phase V.

In responding to data-reporting needs, many participating countries face a variety of economic and social conditions within their borders. This is true of huge countries, such as India or Canada; moreover, medium-sized and smaller countries can also be characterized by internal regional diversity (for example, the north and the south of Italy) that should not be ignored when analysing ICP results. Uniform ICP principles require each participating country to report national average data; in general, this requirement is well observed, but adherence to it may depend on the extent of development of the national statistical system of individual countries.

In spite of the care taken in the regional work in the selection and matching of items to be priced, as well a

as the reliance on close item specifications in the actual pricing procedure, the data collected might yield implausible conversion rates for certain commodities. Most of the time regional organizers had a chance to reveal and correct such inconsistencies. For some important expenditure categories (typically, for services), the quality dif- across countries could not always be captured; however, there were such attempts, as in the case of the Group 11 comparison. Similarly, other conditions, like the density of the network of commercial facilities and the assortment or variety (the number of sorts, colours, styles, sizes) of commodities, may remain hidden behind the price differences between two countries.

One of the greatest challenges in the application of ICP methods regarding PPP estimation is how to balance the conflicting requirements of representativeness and comparability in the item selection for pricing. Price data collection in most countries is based on a large set of specifications that countries usually maintain for compiling their national consumer and producer price indices. The number and composition of items or the detail of description applied in the national specifications can be very different from country to country, and against the backdrop of the ample specification lists developed for regional comparison purposes may turn out even less complete. The organization of comparisons in the regional context offered a better opportunity to satisfy both the requirement of representativeness and that of comparability by means of thorough matching of representative items among countries. In contrast, the global comparison effort has suffered from the lack of direct communication among country experts from different regions and coordinating staff in phase V.

Overall, apart from the potential limitations arising from the quality of the national data, comparing very different economies can constitute a further source of limited comparability. Many experts believe that the regionalization of be programme in phase V has contributed to reducing this source of distortion, at least as far as regional results are concerned. However, one should mention that comparing very different economies within one region (like, for example, Turkey and Sweden in the OECD, or Japan and India in ESCAP, or Poland and Austria in Group 11) may present difficulties because of a narrow range of comparable goods available for pricing. Yet there are even more extensive differences regarding consumption habits, expenditure patterns or institutional arrangements among countries belonging to different geographical areas, or among those that have reached very different levels of economic development; therefore, more efforts are needed to ensure improved interregional comparability.

4. The effect of regionalization on timeliness

The regionalization of the ICP is believed to have positively contributed to a more timely completion of the regional comparisons; on the other hand, it implied a drawback lengthening the time-span of finalizing the overall world-wide comparison. Some of the regional results (e.g., EEC, OECD) became available very quickly, as early as two years after the reference year of the comparison. To complete comparisons in some other regions (e.g., ESCAP, Caribbean) took a longer time and their work ended not much before 1992. The approach adopted for linking regions to arrive at the global comparison in 1985 did not allow for global processing to begin before all regional results were finalized. In addition, as described in section IVB, a number of anticipated, as well as unexpected, difficulties in establishing the links came up, causing long delays in generating globally interlinked results. All in all, unfavourable conditions in phase V did not help to break the tradition of the previous two phases of the ICP, in which the world results became published about seven years following the reference year. However, the organizers learned many lessons that could be utilized in conducting future rounds. The next round of the global ICP with reference year 1993 and a preliminary participation of about 80 countries is under way.