79541 PER CAPITA INCOME Estimating Internationally Comparable Numbers January13, 1989 InternationalEconomics Department - U~~-0 TABLE OF CONTENTS SUMMARY OF FINDINGS AND RECOMMENDATIONS ...... .............. i Quality of National Data . ................................................. Measuring Income in a Common Numdraire ......... i ........................ i Multi-year Averaging and Denominating in SDRs ............................. ii A Special Case: The "High-Income" Benchmark .............................. ii INTRODUCTION ...................................................... 1 METHODOLOGY ..................................................... 1 Measuring Income ...................................................... 2 Purposes ........................................................ 2 Limitations ........................................................ 3 Reliability, Comparability, and Improvements of Basic Statistics ............... 3 Finding a Common Num6raire ............................................ 3 International Comparison Project ........................................ 5 Official Exchange Rates ............................................... Review and Adjustment ............................................. 6 6 Estimating Alternative Conversion Factors .............................. 6 Smoothing Fluctuations in Conversion Factors with SDR Rates ............. 7 Exchange Rate Averaging ........................................... 7 COUNTRY CLASSIFICATION ........................................... 8 Adjusting Operational Guidelines .......................................... 8 Box 1: SDR deflator in US dollar terms - present and proposed formulas . .. .9 The Need for Reasonably Stable Categories ................................. 10 High-Income Countries. A Special Case .................................... 10 The Proposed Country Categories ......................................... 12 ANNEXES Annex 1: National Accounts Issues and Alternative Measures of Income .......... 14 Concepts and Definitions ............................................. 14 Box Al.1: SNA concepts: measurement and coverage .................... Domestic and National Product - GDP vs GNP ........................... 15 16 Improvements to Data: Bank Actions .................................... Box A1.2: Cameroon: Bank staff adjustments to official estimates 17 .......... 17 A Major Source of Difficulties: Informal Economic Activities ................ Growth and Growth Rates: Analytical Uses and Relationship to 18 "Benchmark"-based Comparisons .................................... 18 Benchmark Estimates ................................................ 20 Some Illustrations of GNP and GNY ............... .................... 20 Other Concepts, Measures and Variants ............. .................... 22 Table Al.1: Per capita GNP for 1983: Atlas method and selected altematives .......................................... 24 Annex 2: Environment And Resource Accounting ........ ................... 27 Annex 3: The Intemational Comparison Project ......... .................... 29 Methodological Issues ............................................... 29 Box A3.1: The International Comparison Project: an historical perspective ........................................... 30 The Consistency of Quinquennial Benchmark Findings ...... .............. 31 Box A3.2: PPP rate related to GDP per capita ......... ................... 32 Transivity and Fixity: The Relationship between Methodology and Country Rankings ............................................ 33 Annex 4: The Search for Better Conversion Factors ........ .................. 35 Box A4.1: Countries with conversion factors estimated by IECSE for Atlas per capita GNP ............... ..................... 36 Annex 5: Country Classification: Intemational Organization Practices ..... ...... 37 Table A5.1: Country classification of World Bank and some other intemational organizations .......... ..................... 38 Box A5. 1: Operational guidelines and country classification ..... ......... 39 Box A5.2: Changes in lending categories: FY89 operational guidelines (SecM88-1028) ........... .................... 40 Box A5.3: 1987 GNPper capita ................. .................... 41 SUMMARY OF FINDINGS AND RECOMMENDATIONS 1. In 1983, in consultation with an external panel of and, resources permitting, increase its efforts to im- experts, Bank staff reviewed the methodological issues prove these data. This will involve continued efforts in relevant to the calculation and use of per capita gross IEC, increased priority to basic data issues by the Bank's national product (GNP) for operational purposes. The ExternalResearch Supportprogram,and, incountrypolicy report embodying this review was endorsed by the panel dialogues, increased emphasis on the need for improving (SecM 83-1120, hereafter called the 1983 study) and dis- national accounts (including joint work with UN agencies cussed at a seminar of the Executive Directors on January on resource accounting) and other economic statistics. 5, 1984 (SD84-2). At that time, it was agreed that Bank Meanwhile, for both operational and analytical purposes, practice would again be reviewed after five years. This it is proposed that the Bank should continue to use report has been prepared in response to that decision. country information on national accounts, comple- mented and amended in light of other types of Bank 2. The 1983 study began, "Per capita GNP figures ... staff estimates. are used by the Bank and others as important inputs into the determination of eligibility for various programs." It MeasuringIncome in a Common Nume'raire acknowledged (paragraph 3) that there had been little discussion about why per capita GNP should serve such a 5. major decision-making purpose, but suggested that "an Forreducing numbers to a common base, the Bank has little choice atpresentbut to use exchange-rate-based implicit consensus exists as to what (per capita GNP) does measuresforinternationalcomparisonsofper capita GNP and does not represent." The present report concludes that and other macroeconomic indicators. One alternative, the available options remain balanced in favor of the reliance on the International Comparison Project (ICP), methodology adopted after the last seminar. It reviews the was extensively discussed in the 1983 study. ICP offers underlying issues and proceeds to an explicit discussion of conceptually valid expressions of GNP volumes for dif- the link between measurement of per capita income and ferent economies at uniform international prices, but is not country classification for operational and analytical pur- yet adequate with respect to country coverage or timeli- poses. It proposes that staff continue work on alternative ness. It also faces a set of index number problems: ICP methods of estimating per capita income converted to a country rankingsare dependent on various methodologi- common numdraire, while recognizing that the near-term cal choices. return on such research may not be great. 6. All methods of transforming GNP denominated in Quality of NationalData national currencies to a common basis suffer from a variety of conceptual difficulties. The use of official exchange 3. Since the 1983 study, the International Economics rates in particular suffers from the problem that exchange Department (EC, which is used throughout this report to rate changes can shift countries' relative incomes in a denote both the present department and its predecessor) manner inconsistent with any reasonable view of relative has increased its activities to identify weaknesses and changes in real incomes in these countries. However, use inconsistencies in country data and to correct them. It of official exchange rates has clear advantages, notably in systematically reviews country statistics, in cooperation terms of country coverage, timeliness, and, (generally) in with operational economists, and undertakes some country being relatively unambiguous. Staff therefore propose to missions and other technical assistance activities. How- continue using this method while they pursue the ever, the increase in such activities at best barely compen- development of alternative approaches. sates for the widespread deterioration of national statistical efforts and institutions that has occurred over the period. 7. As recommended in the 1983 study, when condi- tions are egregiously different from those prevailing under 4. The reliability and comparability of national ac- free trade, alternative conversion factors are estimated that counts data remain subject to serious limitations. For that are deemed to reflect the actual rate at which foreign reason, it is proposed that the Bank should continue transactions take place. Such estimates seek to take into trading on the grounds that its per capita GNP had long exceeded account the nature and restrictiveness of the that of Ireland, the country with the lowest income tradi- relative prices, and the evolution regime, information on in tionally classified as "industrial." On the same grounds, of real exchange rates, etc.. This is a difficult process, some oil-producing countries have been listed as "high-in- part judgmental, and its use has been kept to a minimum. es- come oil exporters." However, the Bank has continued to Since 1984, a substitute conversion factor has been include in the "middle-income" category several those other year. From timated for no more than six countries in any countries whose incomes have long been higher than time to time, for several non-member and a few member of Spain and Ireland. countries, the official exchange rate has been deemed no inappropriate, but for lack of adequate information, 12. This has no direct operational implication for the It is proposed to continue alternative has been estimated. Bank since these countries are no longer Bank borrowers, the practice of systematically reviewing and occasional- but it has analytical consequences, and it may have some ly estimating alternative conversion factors. bearing on trade relationships and financial obligations. country Meaningful analysis can only be achieved if the Multi-year Averaging and Denominatingin SDRs groups are composed of reasonably homogenous currently countries. The middle-income group, however, 8. To smooth the year-by-year fluctuations of the per includes countries whose debt, trade, and social indicators Atlas capita GNP numbers used for both the World Bank has as well as per capita income differ significantly from the and operational purposes, conversion into U.S. dollars resulting analysis majority of countries in that group; the of real traditionally been by means of a multi-year average can therefore be weakened. A seven-year average was used bilateral exchange rates. The a three-year average thereafter. from 1970 to 1974; it smoothes out 13. As discussed with the Executive Directors in longer the averaging period, the more 1984, stated in each WDR, and reaffirmed in this paper, rates; by the same year-to-year changes in real exchange per capita income constitutes the Bank's main criterion for have token, the longer it takes to reflect changes that im- classifying countries. The presence of certain high-income occurred. Ideally, lasting changes should be reflected which out; countries in the middle-income group is an anomaly, mediately, and cyclical fluctuations should be ironed one there is no reason to continue. In future, it is proposed unfortunately, it is not always possible to distinguish prices, main- that the per capita GNP level of $6,000, in 1987 from the other. As a compromise, it is proposed to it be the benchmark for separating "middle-income" tain the three-year averaging period, while keeping lead to from "high-income" economies. This would to under active review. classifying as "high-income" economies, in addition Is- the present group of industrial market economies, the 9. Some of the fluctuations in exchange rates and the rael, Singapore, and Hong Kong, and to keep in to the resulting dollar-denominated GNP numbers are due high-income category, Kuwait, Qatar, Saudi Arabia itself. To mitigate the effects fluctuating value of the dollar (subject to confirmation of preliminary estimatescon- for of these fluctuations on operational categories, operational 1987), and the United Arab Emirates. Subject to deflator guidelines have been updated by means of a three- rirmation of*preliminary estimates of 1987 per capita rates in the G-5 countries and a reflecting inflation GNP, Libya would be classified as a middle-income exchange rate, converted back year average dollar-SDR the economy. In addition to these income groups, there would into the U.S. dollar. We propose to continue with highly indebted remain analytical subdivisions (e.g. change methods now used, subject to a minor technical countries) as well as operational, income-based sub- in the way we update the guidelines to be consistent capita divisions. with the method for computing the Atlas per GNP. 14. In general discussions, the term "developing the set economies" may continue to be used for denoting A Special Case: The "High-Income" Benchmark of low andmiddle-income countries. However, it will be term clearly stated (notably in WDR) that the use of the to 10. While the Bank has a policy for "graduation" does not imply either that all the economies belonging by from lending, and this policy is periodically reviewed of the group are actually in the process of developing, nor that the Board, it has no formal policy on the classification reached some those not in the group have necessarily countries as "developing" or "industrial." preferred or final stage of development. 11. Only one country, Spain, was ever reclassified by was done the Bank from "developing" to "industrial." This INTRODUCTION 1. A staff report on estimating per capita GNP for its predecessor in considering issues on the borderline of operational purposes (SecM83-1120) was discussed at a traditional national accounting, notably bearing on the seminar (SD84-2) of the Executive Directors on January sustainability of an estimated income level, including 5, 1984. The discussion referred briefly to the possible use depreciation, depletion, and environmental accounting. of other indicators of welfare, particularly physical quan- The report also discusses in somewhat greater depth issues dties of certain products consumed, and health and similar relating to converting GNP from national currencies to a considerations, but noted that most of the other indicators common numeraire and some uses for per capita income are qualitative and not amenable to national accounting measures, including country classification (without techniques of the type used to estimate gross national pretending to resolve the issue of what constitutes a product (GNP), gross domestic product (GDP) and their developing country). Where possible, details on methodol- components. ogy, technical discussions, and country examples have been relegated to annexes. 2. There is little to add to these comments and this report does not attempt to do so. The report does go beyond METHODOLOGY 3. The Bank compiles intemationally comparable per propriate for this purpose: countries using their economic capita income for its own operatdonaland analyticalpur- output in ways that do not enhance welfare presumably do poses and for presentation to the wider public. To do so, it not deserve more lenient treatment than those that do. The must first measure income in national currencies, which criteria used must be practical. As noted in the 1983 study raises various national accounting issues and the choice (paragraph 43): among altemative measures of income (detailed in Annex 1). Having measured income, there are additional, if often "The main consideration is the ordinal ranking of related, issues in finding a common num6raire. Given the countries. The important point is whether or not one well-known limitations of exchange rates as conversion country is more eligible for certain programs than factors, work continues along the lines of the International another, it is much less important to determine with Comparison Project (ICP, detailed in Annex 3), and more precision what exact fraction of one country's per capita generally insearch of better conversion factors (see Annex GNP isequivalent to the per capita GNP of another." 4). 5. Analytical purposes refer to review work and re- 4. The Bank's operational purposes relate to search on developments in member countries and on the decisions about member eligibility for beneficial borrow- efficacy of policies. They include inter-country com- ing terms and other advantages. These purposes tum on the parisons and require, at some stage, the conversion of view that poorer countries deserve better conditions from indicators expressed innational currencies into a common the Bank and thus imply a search for comparative estimates numeraire. Analysis also entails the study of a country over of economic capacity. Standards of living and welfare are time and therefore must recognize, and sometimes ex- relevant concepts. Howe.ver, even if they were directly clude, effects of changes in price levels. Because of the measurable, they would not necessarily be the most ap- product (NNP) are additional measures of a country's so-called index number problem, it is widely accepted that income that serve specific purposes and are discussed in one cannot establish an unequivocal measure of changes Annex 1. over time in complex aggregates (e.g. national income or price levels). Intercountry comparisons face similar, but Certain concepts must be particularly stressed. 10. less well-known issues. The index number problem arises Only goods and services produced are counted in GNP. because the weighting pattern for aggregation seldom Production requires human intervention. Mushrooms and remains unaltered over time, or between countries.The orchids that grow wild in the forest and the natural increase methodological challenge is to construct standards that are in the wild elephant herds do not count; but when they are meaningful, well-adapted to specific uses and unbiased, picked or hunted for sale, or even for consumption, their while maintaining awareness of their limitations. total value is attributed to the picker and hunter. 6. For presentationof data to the general public and 11. No distinction is made between the aims and thus to a wider, less-specialized audience, the Bank needs ultimate uses of the product according to whether or not a methodology that is clear and easily understood. It should they merely offset some natural or other obstacles, or harm, be emphasized that the Bank is not the only intemational or indeed whether or not they contribute to welfare. For organization reporting national accounts; in fact, United example, other things being equal, GNP is higher in cold Nations has formal responsibility in this area. countries, with expenditures on heating and warm clothes For all purposes, it is most important to achieve to keep people from freezing, than in balmy climates where 7. people are comfortable wearing light clothes in the open broad country coverage and to have current information. air. GNP is higher if people commute by train over long There is no obligation to use the same methodology for distances than if they live near their workplaces; GNP is operational, analytical, and presentational purposes. How- higher in countries with larger police forces than Utopia. ever, the staff's working premise is that methodological variants should be minimized to avoid confusion among Furthermore, GNP is gross: no deduction is made 12. users, and to keep down the costs of compiling and defend- for the wear and tear of physical capital. Conceptually, net ing similar series. national product (NNP) reflects this adjustment, but data on it are particularly patchy and unreliable. Nor does GNP Measuring Income (or even NNP) reflect the evolution of natural resources. Although GDP measures production in an No correction is made for pollution or the depletion of 8. mineral sources. Such adjustments present conceptual and economy, and is a core macroeconomic aggregate, the practical difficulties, such as assigning a "value" to cancer- Bank has traditionally used GNP as its basic measure of causing pollution, accounting for resource discoveries, income for comparing economies. The difference between assigning proper economic values to resources which the two measures is net factor income from (or to) abroad. markets do not yet perceive as "scarce," allocating costs For instance, the repatriated profits of foreign-owned firms which are essentially global to a national accounting operating in a country, and interest payments on foreign framework. debt, are deducted from GDP to arrive at GNP. For some countries, as explained in Annex 1, the difference between 13. Yet there is a growing recognition that a quantita- GDP and GNP is significant. As stated in the 1983 study, tive framework is urgently needed for a better under- "Because these flows [net factor incomes] truly constitute standing of these issues. Expert opinion seems to be resources available to the residents of the countries under moving towards the elaboration of satellite accounts (see consideration, or resources taken away from them, expert Annex 2). opinion holds that they should be taken into account when considering eligibility for certain programs. It is conse- Satellite accounts would, to a large extent, take 14. quently proposed to continue to base the Bank's opera- care of important issues like so-called "defensive expendi- tional decisions on per capita GNP," tures." These are costs incurred to protect the environment 'hile GNP is a broad measure, it is not an all-en- and to combat, either privately or at a collective, public 9. level, those undesirable aspects of economic activity that compassing one. Per capita GNP does not, by itself, con- lead to environmental decay, resource degradation and stitute or measure welfare or success in development. For depletion, and pollution. At present, when defensive ex- most analytical purposes, it needs to be complemented by penditures are incurred privately, they are treated as other other considerations and indicators. However, it remains costs and they are not normally part of GNP. The contribu- the best single indicator of economic capacity and tion to GNP of, say, a steel plant is equal to the value of the progress. Gross national income (GNY), and net national 2 steel produced minus all costs incurred, whether to buy Reliability, Comparability, and Improvements of iron ore or to abate pollution. However, when pollution Basic Statistics abatement or environment protection is incurred publicly, 18. The 1983 study and subsequent Board seminar it is deemed to contribute to GNP without any correspond- stressed the importance of strengthening the basic statisti- ing deduction having been made necessarily elsewhere. cal and national accounting capabilities in member countries. There was broad agreement that the Bank should 15. Satellite accounts can reveal more clearly how cautiously proceed in activities aimed at improving statis- the exploitation, production, and sale of scarce subsoil and tical capabilities at national level. The 1983 study proposed other non-renewable natural resources impinge on the that systematic review, evaluation, and adjustment of na- broader, longer-term viability of an economy. The overall tional accounts and conversion factors should be done surpluses and distributable financial rewards determined within the overall framework of country economic work by corporate accounting methods, which are officially and policy dialogue. It recognized that, in the long term, incorporated into the subsequent statistical assessment of improvement depended on concerted efforts at the national output, value added, and income in the national accounting level, with technical assistance from international agen- context, may impart a false sense of security to countries cies. In order to further these objectives, it was proposed involved in the production of exhaustible natural resour- that the Bank should strengthen cooperation with national ces. Likewise, the discovery of new natural resources may and regional agencies and should participate in the Inter- have to be reflected as a positive improvement in a secretariat Group on the Revision of the SNA. country's economic status and potential. 19. In recent years, Bank staff efforts to obtain im- Purposes proved and more comparable GNP data have focussed on 16. Apart from its general analytical significance, per the basic statistics themselves, and on achieving much capita GNP is used to gauge country eligibility for intema- closer cooperation and coordination of activities with other tional preferences. The Bank uses per capita GNP as a international agencies, particularly the IMF and UN guideline for granting preferences for domestic civil works regional offices. While substantial, these efforts have been contractors and for various borrowing terms (see Box limited by the paucity of the resources that could be A5.1.). The United Nations Development Program devoted to such activities. The Bank does not maintain a (UNDP) relates eligibility for assistance to per capita GNP large, central statistical staff; there are two dozen staff levels and assesses contributions on this basis. Certain members in EC's Socio-Economic Data Division (in con- industrial countries refer to per capita GNP to determine trast to about a hundred in the Fund's Bureau of Statistics, eligibility for preferential trade treatment, or for the dis- for example) and their responsibilities include a variety of tribution of aid programs. Per capita GNP is also a broad analytical and operational support functions beyond those criterion for distinguishing rich and poor countries. of purely statistical units. Limitations 20. In terms of the basic national statistics, there is 17. Beyond the inherent limitations of national ac- reason to believe that the modest increase in Bank staff counting concepts, two main issues arise in intercountry efforts to review and improve them has not even been comparisons of GNP: sufficient to counteract the widespread deterioration of basic statistical systems in developing countries. The staff i. the accounting concepts and practices used by effort itself has essentially taken the form of strengthened countries for compiling national accounts may diverge and systematic cooperation between IEC and country more or less from each other (and from SNA rules). economists. In addition, limited technical assistance to This is obvious for countries that use the Material national statistical authorities through missions of IEC ProductSystem (MPS) rather than the SNA, to compile staff have helped to deal with specific statistical issues. national accounts; but more or less important diver- Annex 1 contains further information on technical assis- gences from standard practice exist in most other tance and adjustments to national accounts. countries. 21. Given that data are weak in many countries, ii. National accounts of countries are compiled in a conclusions based on them should be formed carefully, variety of currency units. To compare or aggregate particularly when they are highly sensitive to minor varia- them, they must be converted into a single unit of tions. Country economists and other operational staff account, and this presents certain theoretical and should give adequate priority to statistical improvements, practical difficulties. and view data with healthy scepticism. However, lively consciousness of the weaknesses of data should not be 3 ap- imperfections of the domestic market. As for the prices allowed to degenerate into nihilism. The procedures of nontradeable goods and services, they would bear and oc- plied by Bank staff, including in-house reviews little relationship to foreign prices. In brief, the volume casional field missions and technical assistance, constitute of goods and services that can be purchased for a dollar a useful complement to national efforts, and at least indi- in one country does not necessarily bear a close and cate the order of magnitude of the major aggregates, of relationship to the volume of goods and services that the direction of changes, yielding a broadly usable set can be purchased for a dollar in another country, even data for most countries. Only in a small number of cases, if there are no tariffs, quantitative restrictions nor when the data are deemed to be particularly difficult to transport costs. interpret or unreliable, has it been decided not to include them in Bank documents. "Such ideal conditions never actually prevail. One rela- tively minor issue concerns fluctuations. While ex- 22. In view of the weakness in countries' GNP data, change rate fluctuations have attracted most attention it is proposed that the Bank should provide additional in the 1970s, since the devaluation of the US dollar and technical assistance within the constraints of budgetary the official end of the so-called fixed parity system, resources. Among the most urgently needed activities is an there was much instability even before then. Real ex- evaluation of the impact of certain accounting procedures change rate relationships are the relationship of nominal on growth estimates, particularly in economies where price exchange rates, deflated by the relationship of domestic controls are or have been dominant. There is reason to prices. Precisely because they were fixed, nominal ex- believe that in such economies national accounts may change rates did not move in exact parallel with domes- overstate growth rates in constant prices by understating tic price movements. Thus, even if one had known that price increases. the relationship of income "volumes" indicated by a particular period's exchange rates was "right" in some Findinga Common Num6raire sense, or at least constituted a useful reference, that relationship would no longer necessarily prevail in the 23. In practice, both in the Bank and elsewhere, most following year, if the exchange rate remained constant international comparisons are based on indicators con- and price movements were not exactly identical. Then, verted at official exchange rates. However convenient, from time to time, these relationship might return such conversions are known to suffer from serious limita- towards their original parity, or beyond, through the tions. As stated in the 1983 study: reverse movement as a devaluation abruptly lowered one country's prices calculated in foreign currency at "The Bank's past practice has been to convert national the official exchange rate. With the advent of floating data into a common numeraire (almost always the US exchange rates, which some had expected to stabilize dollar) by applying to them a conversion factor which real price parity relationships, in fact even more fre- generally has been the official exchange rate or some quent and wide, although less abrupt, fluctuations oc- multi-year average or other variant of it. The use of this curred. particular conversion factor has not seemed to require justification. It was there, and it has long been used to "Fluctuations apart, even for traded goods domestic convert one country's prices into another's. Neverthe- prices are not equal to their export or import prices, less, the official exchange rate has usually been viewed, converted at the official exchange rate. This is because at least implicitly, as a convenient approximation to transport costs, tariffs and subsidies intervene; quantita- some other conversion factor, and there have always tive restrictions affect their trade and their final prices. been cases where it was sensed that its use was, in some In some cases, extreme but not necessarily very rare, sense, wrong... the official exchange rate can bear very little relation- ship to the rate at which transactions actually take place. "In fact, there are strong reasons for considering that A uniform set of taxes and subsidies may be imposed converting national currency GNP data at the official on practically all foreign transactions, as in the early exchange rates assures only imperfect comparability. 1970s in Sri Lanka: exports may be subsidized by the This would be so even if this rate was uniformly applied granting of "import entitlements," and imports subject to all foreign transactions (exports and imports). The to their surrender, as in the 1960s in Pakistan; or prac- domestic prices of goods and services actually traded tically all trade and service transactions may take place internationally would then be exactly equal to their at prices quite unrelated to the domestic price level, and foreign prices multiplied by the exchangerate (abstract- apparently with little if any reference to it, as in the ing from transport costs). However, the prices of similar Soviet Union." tradeable goods could be very different because of 4 24. The 1983 study had argued that, ideally, volumes InternationalComparison Project of goods and services, aggregated at a common set of 26. prices or purchasing power parities (PPP) are the Although ICP information has become available in a somewhat more timely manner and progress has been legitimate basis of intercountry comparisons. The Interna- made on some of the above points, developments related tional Comparison Project (ICP) and related research, to to the ICP itself reduced or, at best, delayed the possibility which the Bank had contributed over $2 million during of introducing ICP-based methods. They relate to coverage 1975-1985 (ICP was by far the biggest single user of Bank of ICP and to a variety of methodological issues. The most research funds), is the only source of such estimates, important of these relate to consistency over time of ICP provided through five-yearly benchmark exercises. How- data and dependency of ICP-based rankings on ever, in 1983 the latest ICP benchmark was for 1975 and methodological choices, including the choice of the base covered only a handful of the Bank's developing member year and the process of aggregation. countries. Furthermore, ICP still presented some un- resolved methodological issues. The 1983 study proposed 27. Phase III of the ICP, for 1975, covered 34 use of ICP when Phase IV (1980) of the ICP was com- countries, of which half were developing countries. In pleted, provided that: Phase IV, for 1980, coverage grew to 60, including 39 developing countries (15 in Africa, 8 in Asia, and 16 in * important methodological issues have been satisfac- Latin America). Phase V (1985), now nearing completion, torily resolved; has seen major changes. Coverage increased in Africa and * countries grouped by income levels can be compared the Caribbean. The United States resumed full participa- tion, joining the majority of OECD countries including all in terms of price weights characteristic of them, and members of the European Communities. However, Latin then each group can be appropriately linked to America is now entirely absent. Lack of ICP data for a others; major region, whose relations with the World Bank have *a set of guidelines appropriate to the scale of GNPs important operational implications, renders the present PPPscanbe repred,andbe om- coerg coverage ofIPiaeut 'ootoeaoa n converted by PPPs can be prepared, and be com.. of ICP inadequate for most operational and parable in its operational impact to the guidelines broadly intemational comparative uses. It does not, how- ever, detract from its usefulness for intra-regional and corresponding to exchange-rate converted GNPs, limited inter-regional comparisons. which will have to continue to be applied simul- taneously to the countries for which PPP information 28. Staff have studied methods for operationaly link- is not yet available. ing ICP data for a limited number of countries to exchange- . PPP converted GNP numbers can be updated an- rate converted data available for others, into a single set of nually;andasatisfactorysolutionhasbeenfoundfor ordinally ranked per capita GNPs. Research into the application of the guidelines when the PPP-con- econometric formulas relating purchasing power parities verted information and the exchange-rate converted to per capita income levels through existing exchange rates information give different results. (so-called "shortcut estimates") are based on a presumed constant relationship between these variables. As succes- One additional implicit consideration was that the country sive rounds of ICP showed changes in this relationship and coverage of ICP should continue to expand after Phase IV. suggested less close determination, research into shortcut estimates has lost much of its earlier appeal. 25. During the 1984 Board seminar, it became clear that the majority of Executive Directors encouraged fur- 29. Research continued to be pursued by Bank staff ther examination of the theoretical issues related to the use into reduced information estimates. These are methods of the ICP-based methods. However, they also insisted on allowing the estimation of the full range of GNP com- the need forfurther Board discussions and clearance before ponents in terms of ICP methodology, but on the basis of ICP-based methods were introduced, even partially, in the a much more limited, and therefore much cheaper, sample criteria for operational decisions; no Executive Director of price data. Unfortunately, budgetary constraints and objected to this emphasis on the need for a cautious ap- inability to obtain additional financing, notably from the proach. Pending resolution of these issues, staff were to Bank's extemal research support funds, also brought this continue using conversion methods based on the exchange work to a halt, at least for the time being. rate for all countries. Systematic efforts were to be made by staff to reduce or eliminate certain anomalies associated 30. Furthermore, ICP work in Phases IV and V high- with the use of exchange-rate-based conversion methods. lighted methodological issues, discussed more extensively in Annex 3. These are not unique to ICP. They, or their 5 is also ag- The evolution of the effective real exchange rate often variants, are common to all comparisons of complex that systematically examined. A sharp increase in this rate gregates; they are inherently index number problems: by compare indicates an apparent overvaluation compensated is, problems that arise when an attempt is made to severe import controls (and perhaps export subsidies). In brief, con- complex aggregates using a single number. have These create a gap between the official exchange rate and round to the next ceptual and other changes from one the rate effectively paid and received in international trade. income caused substantial instability in the per capita relationships established by successive ICPs. Furthermore, com- 35. Estimating Alternative ConversionFactors. The even within a given round of ICP, direct bilateral Bank rarely uses such alternative conversion factors. for This cannot be made to yield the same parisons of two countries is partly because they are difficult to estimate, and results result as (and sometimes may yield quite different operational purposes it has been thought best to minimize from) indirectcomparison of the ratio of theirrelationships the opportunities for arbitrary decisions. It has been Bank broadly, to a third country or to a country group. More practice to tolerate departures from free trade conditions bearing on the value, methodological choices have a strong over a wide range. This conforms to the letter and spiit of and even the rank order, of intercountry ICP comparisons. alternative con- the 1983 study, which indicated that such only in "egregious" cases, version factors would be used ICP findings shed valuable light on the compara- by an 31. i.e., in cases that depart from common practice tive structures of economies, including both the relation-of exceptionally large measure. ship of prices in various economies and the volumes goods and services allocated to certain uses (e.g. personal or 36. Only for about a half dozen countries were es- and government consumption, business investment...) timated conversion factors (as opposed to official and comprised in certain categories (food, medical expenses, multiple authorized exchange rates including authorized housing, etc.). Staff will pursue the study of these dimen- exchange rates), used for GNP per capita calculations in Bank sions of ICP in order to ensure familiarity within the the operational guidelines and the Atlas in each of the past of the results. sub- five years (see Box A4.1). These countries had been which suggested that trade and ject to thorough review, 32. The dependency of ICP results on methodologi- exchange restrictions were particularly severe, and the to its cal choices is inherently related to its advantages, i.e. difference between official exchange rate and effective But goal of comparing volumes of national expenditures. transactions rate was exceptionally large during the since international comparisons of GNP must necessarily also signalled used to specified period. Usually, this problem was use a common num6raire, some method has to be into by a sharp appreciation in real terms of the officiall ex- convert values (typically in different currencies) fac- change rate. volumes. At present, exchange-rate-based conversion tors are most practical. 37. While an effective transactions rate cannot be the help measured precisely, it may be approximated with Official Exchange Rates of information on trade restrictions and effective protec- 33. IEC has expanded its collection and analysis of tion rates. When such information is inadequate for form- various exchange rates. It has improved conversion ing an estimate, a less satisfactory alternative procedure is factors methods, notably by creating average conversion where used, but one that is nevertheless preferable to a highly from officially recognized multiple exchange rates ex- distorted official exchange rate. A past year is identified appropriate. It has also clarified procedures to flag where "normal" conditions were thought to prevail in the and the greatly from rates effec- change rates that seem to differ exchange rate of that year adjusted for the changes transac- tively used in exports, imports, and other foreign country's domestic absorption deflator (relative to that of rates). EC now collects tions (the effective transactions the US) during the period. The resulting figure yields a parallel (black or free) market rates. the but does not publish conversion factor equal to the real exchange rate of These do not provide a direct indication of the appropriate other chosen base period. Even this, though, is not an unequivo- exchange rate, but in conjunction with ICP and and cal concept. Other reference countries and other price indicators they shed useful light on conversion issues indices might be equally reasonable. set outer limits on any possible adjustment of conversion factors. 38. There are limitations to this method. The choice is dif- of the base period is necessarily judgmental, as it that 34. Review andAdjustment. Exchange rates that dif- ficult to assess the impact of existing trade restrictions ransac- fer markedly from rates actually used in foreign and are not well documented. Moreover, even under free trade tions are identified through Fund reports on trade sometimes of country economists. conditions real exchange rates can change, exchange restrictions, and findings 6 substantially. To take one example, a fall in the price major export crop would both lower the domestic of a 43. In cases when an alternative conversion factor is deflator and tend to depress the exchange rate price estimated, this, by defnition, diverges markedly (since the from the value of exports has fallen). Increased debt service, official exchange rate. To obtain reasonably coming meaningful after years of heavy resource inflows, may have values for per capita GNP where trade-related components effects. In some countries, where black markets similar reflect the official exchange rate, it is necessary to return develop, to the original national accounts, and adjust the appreciation of the real exchange rate may be them ap- under- propriately. stated by the official price indices, while the increase nominal value of GNP may be similarly understated. in the 44. When an alternative conversion factor, not based 39. It is important to take such circumstances on the official exchange rate, is used for converting account when choosing the base year or adjusting into capita GNP, a related single-year conversion factor theper is derived conversion factor for them. The judgments the used for converting overall GNP and GDPand their also com- volved are difficult and their correctness cannot in- ponents for various analytical purposes. Niz.urally, the verified. However, as has happened in several cases, be types of adjustments to the national accounts that were subsequently the countries concerned have undertaken when described above then also apply. major adjustment programs involving trade liberalization and devaluation, post-devaluation exchange rates 45. While these procedures are necessarily often been within a reasonable range of the have what oversimplified, for analytical purposes the some- substitute resulting conversion factors estimated earlier, thus providing aggregates are more reasonable than those obtained some use of highly distorted official exchange rates. with ex-post support to staff judgment. In ticular, one avoids overstating the weight of countries par- with 40. Following sharp devaluations accompanied grossly overvalued exchange rates in respect to global relaxation of trade restrictions, the resulting exchange by a comparisons. Moreover, certain internal relationships to rate GDP like those of trade and resource inflows, could be extrapolated backward (in real terrns) debt, the to derive current account balance, and, if foreign aid contributes historical conversion factors. However, the real prevails immediately after the nominal devaluationrate that significantly to the budget, sometimes even the lower than the average rate of any representative may be deficit tend to be similarly underestimated in the fiscal period, if national the nominal devaluation also aimed atpartly compensating accounts of countries with grossly overvalued exchange for anticipated subsequent inflation. Consequently, rates. Even simple corrections in these cases yield in- exchange rate chosen as the base must either be the the real dicators that are more useful analytically. average for a relatively long period, or the rate for a given period 46. Smoothing Fluctuations in Conversion Factors corrected for effects of estimated or projected future tion. It is largely for this reason that staff rarely use infla- with SDR Rates. The variability of exchange rates has exchange rates to re-estimate past conversion factors. recent been particularly great in recent years. Conceptually, variability can be separated into two components: 41. As mentioned before, estimates of effective variability of one economy's "real" exchange rate transactions rates are difficult, hazardous, and to those of its trading partners, or to some otherrelative to global extent arbitrary. All that can be said for them is some composite; and variability of the num6raire, (which, that they are preferable to the use of grossly unrepresentative international comparisons, is usually the US dollar). for offi- If one cial exchange rates. Hence, staff are committed conveniently defined the latter as the variability to making of the US and using such estimates only when distortions dollar/SDR exchange rate, it could easily be attenuated of the adopting the SDR as num6raire. Indeed, to shield by official exchange rate are truly egregious. opera- tional decisions from fluctuations in the dollar num6raire, 42. When GNP or GDP is computed by a country's operational guidelines are now based on an SDR deflator, national accountants, the price actually paid merely translated into dollars at the prevailing for exchange goods determines the values attributed to specific traded rate. As the dollar-SDR exchange rate has fluctuated over sectors. the years, these movements have been paralleled The price thus influences the values of major by those aggregates of the dollar values of the operational guidelines, like consumption, investment, as well as imports and of ports. Its value in domestic currency partly dependsand ex- per capita GNPs of developing countries. Hence, the exchange rate. Converting GNP back into on the relationship of GNP per capita to the operational dollars at guidelines has not been affected, although the entire another exchange rate would therefore give somewhat struc- misleading results. ture has appeared quite variable. 7 for United 50. On balance, three-year averaging seems, 47. Exchange Rate Averaging. Like the operationalpurposes,areasonablecompromiseof various real value their of about by longer States, individual countries allow the considerations. Greater stability brought value of the curren- by itself. currencies to move relative to the real nominal ex- periods of averaging may not be desirable for cies of their trading partners. The official to day; some- Moreover, if, through the use of a very long period day toper change rate sometimes fluctuates from the real exchange only applies averaging, the moving average, which the Bank (in the as times it is subject to major realignment relativeprice capita GNP numbers, as published by rate gets gradually out of line; and sometimes divergence were to be by a gradual Atlas, etc.) and used in operational guidelines, rate, problems changes are partially compensated actually used for very different from the annual exchange would arise. between the official rate and the rate growing use of 38-40 similar to those discussed in paragraphs present practice the foreign transactions, brought about by subsidies. To at- It is therefore proposed to continue the issue needs to be or taxes and quantitative restrictions uses a three-year of three-year averaging. However, this tenuate these fluctuations, the Bank for converting per kept under review. moving average of real exchange rates dollars. capita GNPs from national currencies into 51. In summary, it is proposed to keep using as the last available averaging reference for operational guidelines 48. Arguments can be made in favor of years annual per capita GNP normally converted at the past or shorter than the three rate, adjusted rates over periods longer Changes in three year's average official exchangedeflator relative operational guidelines. used for the Atlas and new realities, (e.g. annually by the movements of the GNP estimate altema- real exchange rates sometimes reflect in the face of to that of the US. Staff will continue to devalue their currencies the rate actually applied to when countries after 1985, when tive conversion factors when sharply deteriorating terms of trade; or the official rate averaging in foreign transactions diverges from large diver- the dollar declined sharply). Any multi-year recognition egregiously. Such cases of truly exceptionally the face of such movements merely delays the a dozen cases in wish to reflect gence are unlikely to much exceed half of a lasting reality. Ideally, one would mentioned in para- to smooth out cycli- any year. For analytical purposes, as immediately permanent changes, and to convert national distinguish a priori graphs 40 and 41, staff will continue cal fluctuations. But it is difficult to account aggregates at the applicable annual conversion rate changes. between lasting and temporary exchange factor. the Bank 49. The seven-year moving average usedby Atlas 52. This discussion highlights some of the consider- World Bank of to calculate per capita GNPs for the able technical, practical, and theoretical difficulties smoothing. Indeed, on the conversion of before 1974 resulted in considerable to a three-year comparisons and aggregations based a seven a single num6raire. the Bank changed its method from averaging was national accounting aggregates into seven year methodological moving average because the rate move- Staff will continue to make pragmatic theoretical pos- attenuate exchange thought in the early 1970s to improvements, and also to investigate the that the movements and aggregation ments excessively. It was then thought prices constituted sibilities for devising better comparison continued close increase in oil include following the 1973-1974 rates, whose intro- methods. These investigations will a permanent readjustment of exchange Comparison Project, not be unduly involvement with the International duction into the GNP calculations should but will not be limited to it. delayed. COUNTRY CLASSICATION granted to domestic eligibility for (a) special preferences Adjusting OperationalGuidelines contractors, (b) IDA credits, (c) IBRD loans civil works IBRD loans for of per with maturities longer than 17 years, (d) The original 53. The Bank has established "thresholds" 15 years, and (e) graduation from IBRD. countries' capita GNP categories to determine borrowing They are: thresholds for IDA eligibility and IBRD graduation were eligibilities for various Bank loan programs. 8 Box 1: SDR Deflator in US Dollar Terms - Present and Proposed Formulas MS BASIC DATA FOR COMPUTATIONS 1980 1981 1982 1983 1984 1985 1986 1987 (A) SDR deflator in SDR terms (1970-100) 209.4 229.4 244.4 253.7 263.3 271.4 286.5 294.1 (B) SDR deflator In SDRs, 3-yr average 1/ .. . 227.7 242.5 253.8 262.8 273.7 284.0 (C) US GNP deflator(1980-1), annual 1.000 1.079 1.148 i.185 1.232 1.272 1.306 1.345 (D) US GNP deflator, 3-yr weighted average 2/ 1 . . 1.076 1.137 1.190 1.231 1.271 1.308 (E) S/SDR rate 1.3015 1.1792 1.1040 1.0690 1.0250 1.0153 1.1732 1.2931 (F) S/SDR, 3-yr average i/ .. .. 1.1949 1.1174 1.0660 1.0364 1.0712 1.1605 (G) Relative Inflation adjusted average S/SDR rate 3/ .. .. 1.187 1.113 1.065 1.038 1.053 1.154 FORMULAS COMPUTATION RESULTS 1980 1981 1982 1983 1984 1985 1986 1987 (8xFxC 40) SOR deflator in US S (Present method) 4/ .. .. 290.3 282.3 280.2 281.4 301.3 338.8 (AxG) SDR deflator in US S (Proposed method) 5/ . .. 290.1 282.3 280.5 281.8 301.7 339.3 NOTE: 1/ Simple arithmetic average for the target year and two preceding years. 2/ Aritheetic average weighted by real US GNPs for the target year and two preceding years; computed as ratio of the sum of nominal US GNPs to the sum of real US GNPs for the three years. 3/ Simple arithmetic average of the target year's S/SDR exchange rate and the two preceding years' exchange rates that are adjusted for differences between US inflation and the average inflation for the SDR basket countries. More specifically, this avorage S/SDR rate for year t is computed as: [E(t-2) { P ( P$(t 2) Psdr(t-) -Ps-dr(t)-7 + E(t_-1) P tt) t Psdr(t)I + E(t)] * 3, rPS(t) Psd r(t) 11 where ECt) * S/SDR exchange rate for year t, PS(t) * US GNP deflator for t and Psdr(t)=SDR deflator in SDRs for t. 4/ The computation procedure is: the 3-year averagp SDR deflator is first converted to US dollar terms via the 3-year average S/SDR exchange rate and then inflated to the target year price level via US Inflation rate; designed to be consistent with the previous Atlas method for computing GNP per capita. For example, the SDR deflator in USSterms for 1982 is: 290.3 * 227.7 x 1.1949 x 1.148 - 1.076. 5/ The computation procedure is: the SDR deflator in SDR terms for the target year is directly converted to US dollar terms via the relative inflation adjusted, average S/SDR exchange rate for the target year; designed to be methodologically consistent with the present Atlas method for computing GNP per capita. For example, the SDR deflator in USS terms for 1982 is: 290.1 * 244.4 x 1.187. 9 set in 1970 prices. The threshold for civil works preference The Need for Reasonably Stable Categories was set in 1971 prices, and that for "longer-term" IBRD loans, in 1972 prices. 58. Nevertheless, the simple application of the in- come criterion has occasionally moved countries from one 54. The Bank regularly updates the original threshold category to the other, and then back again. Since the FY83 of per capita GNP categories to reflect the applicable operational guidelines, 44 countries have changed lending international inflation rate. Until 1984, the thresholds were terms categories. Of these, 15 have subsequently moved updated by the US inflation rate only, i.e. kept constant in back into their previous category at least once. (See Box terms of base year US price levels. In 1984, the Board A5.2 for country specific changes in lending categories for approved the use of an SDR deflator for this purpose. The FY89, Operational Guidelines.) It would completely original thresholds are now updated via an SDR deflator defeat the purpose of any operational benchmarking if which is based on the average inflation rates of the five countries could not move from one category to another, countries whose currencies make up the SDR, weighted by and it is an unfortunate fact of the 1980s that many their currency compositions in the SDR basket and con- developing countries have suffered a reversal of previous verted into US dollar terms. It is proposed to keep this income gains. Nevertheless, it would be desirable to min- method, although with a slight technical change to be imize disorderly movements back and forth between consistent with the presentAtlas method for computing per categories, if reasonably objective means could be devised capita GNP (see Box 1). for doing so. 55. As noted earlier, despite its limitations, per capita 59. Such movements may occur for a number of GNP is the best single indicator of countries' eligibility for reasons. The most undesirable probably relates to fluctua- various financial terms from the World Bank Group. How- tions in real exchange rates. If a country pegs its nominal ever, it is rarely used alone. Economic performance and exchange rate, and devalues it from time to time, its real creditworthiness are key considerations in determining exchange rate would crawl up and then fall abruptly at each IDA-eligibility; the IBRD graduation benchmark is only a devaluation. If this is done within a trade and exchange threshold for starting a process, and the actual graduation system ofjust normal restrictiveness, the official exchange is determined by other considerations. Even graduation rate would be used for the Atlas-style conversions. If the from IDA is not truly automatic, as the benchmark does devaluation cycle is markedly longer than the three years not constitute an entitlement; many countries below the used for the moving average, the moving average real IDA graduation benchmark do not now receive IDA exchange rate could also be fluctuating sharply. A country credits, and several others borrow both from the Bank and would be pushed down to a lower category as its last IDA. Thus graduation from IDA on the grounds of rising devaluation gets fully incorporated into the three-year per capita income tends to be a gradual process. moving average, and then move back up into the higher category as its exchange rate gradually appreciates in real 56. Nevertheless, present Bank policy is to use per terms under the impact of its domestic inflation. A similar capita income as the sole determinant of eligibility for cycle has also characterized countries whose currencies granting preferences to domestic suppliers of civil works are linked either to the U.S. dollar or to European curren- financed by Bank group lending. The same benchmark is cies: their real exchange rates fluctuate relative to the used, by design, as the dividing line between low-income "real" SDR, used to calculate the guidelines. and middle-income countries. See Box A5. 1.Eligibility for softer IBRD terms is also determined by per capita GNP 60. Another cause might relate to temporary terms of alone. trade changes. If a country's major product and dominant export is a primary commodity, say coffee, a rise in the 57. Income categories also constitute analytically intemational coffee price will raise the country's GNP: the useful distinctions. The World Development Report states price level of domestic production will inkrease, and as (page xi of the 1988 edition) that, "The overall classifica- export earnings rise, there will be upward pressure on the tion uses GNP per capita as the main criterion." The value exchange rate. The same phenomenon in reverse would of such categories is reduced if the dividing line does not tend to reduce the GNP when coffee prices fall. Fluctua- follow logical transparent criteria, or if the categories are tions may also be due to a temporary change in national unduly unstable. As discussed earlier, the three-year production, due (for instance) to an exceptionally high (or averaging of the exchange rates used for calculating per low) harvest, or sometimes even to construction on a major capita GNP aims at reducing one cause of instability. foreign-financed investment project. 10 61. In principle, one could separate the issues of country classifications from those regarding operational point of view of the main criteria. Including among the decisions. For instance, it may be decided that certain upper middle-income countries a few economies with operational decisions take effect only after a country has much higher income levels than most other members of the belonged to a specific category for a certain length of time. group, and even exceeding some members of the higher However, the analytical uses of country classifications are income group, reduces the value of group indicators. More also important, and would also be disturbed by frequent broadly, given the Bank's general practice of grouping back and forth changes. countries into categories based on per capita incomes, it makes little sense to maintain a category that has a floor but no ceiling, and that includes the widest-and potential- High-Income Countries:A Special Case ly unbounded-array of per capita income levels. Israel, 62. The World Development Indicators (WDI) tradi- Hong Kong, and Singapore account for two percent of tionally lists countries in order of increasing per capita population, six percent of GNP, and about a quarter of the trade of their present group. Each has a well-developed income (see Box A5.3 for a list based on information currently available).The countries are grouped according financial system and an international investment position to income level: low-income, middle-income, and high-in- very different from those of most middle-income come. However, there has long been an overlap between countries. By their social indicators, too, they are com- the per capita incomes of the highest, so-called upper parable to other high-income countries. middle-income economies, and the lowest, so-called in- dustrial market economies. Furthermore, certain oil ex- 67. Whether a country is classified as "high-income," porting countries have been separated on the grounds or "middle income" has no operational implications for the of Bank itself. Bank graduation starts well before the high-in- their high-income levels, but also listed separately from the industrial country group, rather than in order of ascending come category denomination issue arises. Other organiza- incomes. tions follow a variety of practices in this field, usually on the basis of historical or other non-economic criteria (see 63. The only country ever reclassified by the Bank Aamex 5). The question is not without practical implica- tions, however. For example, GATT rules allow preferen- from the "developing" to the "industrial" category is Spain, in 1982. This issue was discursed in the Board at the time, tial treatment to be granted to developing countries. While in the context of operational criteria. Staff justified the the rules do not explicitly refer to any classification proposed reclassification by noting that Spain's per capita scheme, there is a widespread feeling that some industrial countries granting preferences may take Bank classifica- GNP had by then for many years equalled or exceeded the per capita GNPof Ireland, a country traditionally classified tion into account, although they also can, and occasionally as belonging to the "industrial" country group. Sub- do, terminate preferences for economies still classified as sequently, though Ireland's per capita GNP has risen mar- "developing" by the Bank. A less closely operational, but ginally above Spain's, this has not prompted any new nevertheless politically charged issue may also arise in proposal to reclassify Spain as a developing country. granting and defining "development assistance." 64. 68. All arguments bearing on use of per capita GNP For the past few years, three other economies classified in the middle income group have had higher per as opposed to other criteria for operational Bank decisions, capita GNP than Spain (1987, $6,010) and Ireland (1987, are equally valid in terms of the dividing line between $6,030); Hong Kong since 1976(1987,$8,260), Singapore middle and higher income countries. If "maturity" in some since 1976 (1987, $7,940), and Israel since 1983 (1987, sense is the true criterion, social systems, social and $6,810). demographic indicators, and even political circumstances might be relevant. Some argue that even within the purely 65. Another group of countries once classified as economic field, considerations relating to GNP should be "developing" has been separately listed in recent years, supplemented with others, like the role of foreign invest- that of the "high-income oil exporting countries." These ment, dependency on foreign markets, the role of nonfactor have had income levels markedly higher than those service payments, or the definition of "residents." of Spain and Ireland for some years. They were separated from both developing and industrial countries in analytical 69. However, proposals to supplement or replace the presentations. income criterion by some composite measure of social progress that would incorporate such indicators as health, 66. To be useful for analytical purposes, country education, nutrition and female advancement suffer from categories should be reasonably homogeneous from the several basic problems. Data in this field are weaker than economic measures. Consensus is also unlikely on what 11 data published by the Hong Kong and Singapore govern- measures assess social development, and on the relative ments, which also follow SNA rules, we have also calcu- weights attached to them. There is likely to be broad lated the incomes adjusted for these governments' favored agreement that low infant mortality and high life expectan- definitions; the adjustment is not large enough to affect the cy are signs of progress; and that so is education. But what classification issue. is the tradeoff between them? What is the tradeoff between primary and secondary education? Is income distribution 74. While the Bank does not directly use ICP for to be included, and what type characterizes "advanced" classification purposes, ICP is a valuable comple- country societies? What about criminality? And should a country ment to other information and is always carefully that devotes a large share of income to social improvement analyzed. In the last fully published ICP, Phase IV for be graduated by the Bank sooner than one that does not? 1980, the per capita GDPs of Hong Kong and Israel (Sin- The more one reflects on these points, the more it appears gapore is not a participant) are markedly higher than those that social indicators as much reflect the choices made by of Ireland and Spain. It is noteworthy that among countries a society on uses of its income, as its economic advance- now classified by the Bank in the middle-income group, ment. these are the only ICP participants for which this is the Many of the other indicators proposed rely on case. Venezuela, Greece, Hungary and Poland (in descend- 70. ing order) all had ICP-determined 1980 per capita GDP criteria on which it would be difficult to reach universal volumes lower than Spain and Ireland. agreement. For instance, some countries have relatively high life expectancies and education levels; this would A related country classification issue concerns 75. hardly warrant reclassifying them into categories which the economies now classified as "high-income oil ex- receive fewer operational benefits. Indeed, a composite of economic and social indicators might run counter to the porters." These are economies heavily dependent on hydrocarbon exports, with high per capita income levels. Bank's avowed concem about social progress, by ac- It has been said that the GNPs of these countries, because celerating graduation of the countries that respond most of present national accounting methods, do not properly effectivly in this dimension. reflect their dependency on exhaustible resources. As It would be even more difficult to agree on the noted above, this cannot, in any case affect gross national 71. importance to be attached to criteria like the share of product, which does not provide for any depreciation or foreign capital in the domestic economy (or a specific wear and tear on capital, man-made or natural. Were one considering net national product, it is by no means certain sector of it, e.g. manufacturing). Inasmuch as the foreign that such provisions for these producers of natural resour- investors are actually eaming incomes, the use of the national(as opposed to domestic) product concept already ces with very large reserves would be larger than the makes allowance for that. Less tangible aspects of such provisions for depreciation of other items, say, environ- investment ("domination" of the domestic economy by mental degradation, for other countries. High-income oil producing countries have reserves sufficient for at least a foreigners, sensitivity to outside decisions, etc.) have im- century, at the end of which the economic value of the plications that one cannot weigh objectively. For instance, ability to attract foreign investment in manufacturing, or a remaining reserves cannot now be known. Making high export to GNP ratio, can be interpreted as a weakness, provision for such a remote event, and deducting it from current income, does not have compelling appeal. In any because they show dependency on the rest of the world; or a strength, i.e. ability to compete for capital and for exter- case, if oil reserves were treated as an asset, new dis- coveries of proven reserves might be treated as income; the nal markets. resulting adjustment might be upwards, not downwards. 72. In any case, the immediate reclassification The ProposedCountry Categories decisions are not dependent on the way these arguments are settled. Hong Kong, Singapore, and Israel not only As agreed by Executive Directors in 1984, stated 76. have high incomes; they are also advanced in terms of most social indicators. prominently in WDR, and reaffirmed in this paper, per capita income constitutes the Bank's main criterion for The governments of the British Crown Colony of classifying countries. We now propose to elaborate this 73. Hong Kong and of Singapore have argued that residency as follows: definitions different from those of the SNA should be used The Bank classifies economies for certain operation- for calculating their per capita GNPs for operational pur- al and analytical purposes according to their per poses. While the Bank has continued to adhere to the capita GNP. The high-income, middle-income (itself generally accepted SNA definitions, and to use the official 12 divided into upper and lower) and low-income groups Ireland, along with Israel, Singapore, and Hong Kong, will constitute the main categories. In general discussion, the set of middle-and low-income economies may some- therefore be classified as "high-income economies," as tines be referred to as "developing countries." The use will also Kuwait, Qatar, Saudi Arabia (subject to confirma- of this term is convenient; it is not intended to imply tion of preliminary estimates) and the United Arab either that all economies in this group are actually Emirates. Also included in this group, although they are developing or that economies excluded from it have all not actually reported in the main WDI tables because their reached some final stage of development. Moreover, the populations are less than a million will be: Bahrain, Green- term "country" does not imply any judgment by the land, U.S. Virgin Islands, Bahamas, Faeroe Islands, World Bank on the legal or other status of any territory. Brunei, Luxembourg, Iceland, and Bermuda. Subject to confirmation of preliminary estimates, Libya would be classified as a middle-income developing country. For 77. operational purposes, the middle-income group would The dividing line between "low-income" and be "middle-income" countries will continue to be $200 in further subdivided by income levels to distinguish countries that are eligible for IDA loans, 17-year IBRD 1971 prices. The dividing line between "lower middle-" and "upper middle-income" economies will continue to be loans, or graduation. In addition, criteria other than income $850 in 1972 prices, both adjusted by the Bank's "SDR" would be used for forming, from time to time, appropriate regional and analytical subdivisions (Sub-Saharan Africa, deflator (to $480 and $1,940, respectively, for 1987). It is proposed that $6,000, in 1987 prices, become the highly indebted countries, etc.). These categories will also benchmark for separating "middle-income" from be used in the annually published World Development "high-income" economies. The two lowest income Indicators which contain the most complete presentation of statistics for countries grouped into categories. countries traditionally classified as "industrial," Spain and 13 Annex I NATIONAL ACCOUNTS ISSUES and ALTERNATIVE MEASURES OF INCOME paral- prices, and index number problems may obscure the Concepts andDefinitions lelism. 1. A nation's GNP is the sum of the value of goods 3. GNP is gross. No deduction is made for the wear and services produced in its territory, plus the factor in- No factor and tear and reduction of capital, man-made or natural. and come of nationals temporarily living abroad, the theless territory. account is taken of the deterioration of highways income of foreigners temporarily living in natural to be machinery, of pollution, or of the exhaustion of This definition needs to be rendered more precisedefini- resources. Anet national product (NNP) concept would be usable, and indeed whole volumes are taken up by One would then United more meaningful for many purposes. tional and similar issues in the description of the The issues deduct from the acquisition of new capital goods oron ac- gross Nations' System of National Accounts (SNA). for the costs incurred Bank analyses are indicated in Box Al .1. investment some allowance capital mostrelevant for (e.g. count of aging and intensity of use of the existingare not passing Issues about prices are noted there only in considered are stock. Unfortunately, estimates of depreciation between factor costs and market prices) but context of always available for statistically advanced countries, and more generally in Annexes 3 and 4, in the con- are even scarcer (and less reliable) in developing countries. choosing conversion factors. Hence, Box A1.1 for This iswhy most discussion of economic progressthan focusses centrates more on coverage issues such as imputations of gross, rather net, and con- on changes and comparisons certain products and services (the food produced(domestic product. (See paragraphs 53-55 in this annex.) sumed by the same farming family), exclusions services rendered by family members, and racketeering is- 4. Under current SNA conventions, even in NNP, no transactions, e.g., drugs and prostitution), and similar allowance is made for environmental degradation or deple- sues. tion of natural resources. In other words, the derricks and other equipment on an oil-well are depreciated, and the 2. The upper section of Box Al.1 summarizes the reduction in their value is deducted from GNP to arrve at on three basic accounting idendties of SNA, which proceed In the SNA NNP; but no such adjustment is made for the reduction in tracks: production, income, and expenditure. irrigation equipment would be such as GDP and national income oil reserves. Similarly, adjust- itself, major aggregates in the depreciated, but, even under the NNP concept, no (NNP) are explicitly defined. GNP is not included GDP ment is made for the fact that increasing use is made of SNA framework but the item of difference between of SNA limited amounts of available water. and GNP (net factor income) is; and the revision Also on the envisaged for 1990 reinstates GNP in SNA. 25-28) is the 5. The disregard of the impact of human activity on borderline of SNA (as discussed in paragraphs national income natural resources clearly limits the overall significanceand of Bank's constant price measure of gross run, it may be desirable, into two NNP numbers. In the long (GNY). The lower section of the box is divided that are in perhaps even feasible, to develop more comprehensive iUustrates economic activities parts. The first surveys and the concepts. The difficulty ofso doing is, however, enormous. SNA but difficult to measure (even through activities that For instance, the definition of "natural resources" varies. like) and the second shows some economic forests are a natural iden- In today's industrial countries are not conceptually included in the SNA.The basic price resource, but not very long ago natural resources, away that is tities and issues arise equally in current and constant increased by clearing the cultivable area, were national accounts except that GNY exists only in constant 14 Annex 1 Box A1.1: SNA concepts: measurement and coverage SNA CONCEPTS Production = Income = Expenditure Value added in: Agriculture Compensation of employees Private consumption + Mining + General government consumption + Manufacturing + Operating surplus of enterprises + Construction + Utilities + Depreciation + Investment + Trade and transport + Exports of goods and + Other private services nonfactor services + Govemment services - Imports of goods and nonfactor services = Gdp at factor cost + Indirect taxes less subsidies + Indirect taxes less subsidies Gdp at market prices Gdp at market prices Gdp at market prices + Factor income payments less receipts + Factor income payments less receipts - Depreciation = Gnp at market prices .National income (NNP) Additionally, in constant prices, the Bank measures: + Terms of trade adjustment = Gross national income (GNY) Included in SNA, but difficult to measure (shown where measurement may occur): Production Income Expenditure * Small-scale activities *Wages and salaries from informal * Own consumption (subsistence, etc.) services * Major repairsand maintenance *Informal, moonlighting *Profit and income from activities self-employment * Contraband trade * Imputed rents and depreciation for *Reforestation owner-occupied dwellings Not included in SNA (but shown where measurement could occur): . Household services of *Allowance for depletion *Routine repairs and maintenance family members of natural resources * Discovered resources *Racketeering 15 forests. During the nineteenth century, coal became an employment of the residents of a country who work abroad increasingly valuable natural resource, before losing much for less than one year) and interest (including due but not of its economic value in recent decades. Moreover, if one paid, i.e., in arrears). For highly indebted countries, such is to treat the use of some natural resources as the sale of as Chile and Jamaica, interest payments on external debt assets, one must first find means to reflect the state of such are high (and far outweigh factor income inflows). Conse- assets and also to account for valuation changes in them. quently, GNP is substantially lower than GDP. For some To incorporate natural resources in national accounts, other countries, such as Yugoslavia and the Yemen Arab ways would have to be found to include their contribution, Republic, the addition of labor income from abroad and not just to record their depletion; before deducting the renders GNP noticeably larger than GDP. Thus resources reduction in mineral reserves, one must first account for available to the residents of an economy can be substan- their discovery. tially reduced or increased as a result of external factor income. However, the compilation of these data often 6. The United Nations Statistical Office (UNSO), in causes problems. coordination with an Intersecretariat Group that includes the Statistical Office of the European Communities 10. Measuring labor income paid to or received from (SOEC), the OECD, the World Bank, and the International abroad is difficult. "Residency" is defined in SNA as stay- Monetary Fund, is at present revising the SNA. The issue ing for over one year in a country. However, compilers of the depletion of natural resources has been taken up by often do not have information on the length of time the Intersecretariat Group, and it is expected that the new workers stay abroad. The income of those temporarily (less SNA (to be completed by 1990) will recommend that than one year) working abroad should be treated as a factor satellite accounts should be built to measure certain natural income receipt by the home country but may not be distin- phenomena, without trying to incorporate them into a guishable in practice from rela.ed transfer receipts (remit- single number like NNP. tances of workers who stay in the country for a year or more). National practices for distinguishing between labor 7. Many alternative measures of income can be income and workers' remittances differ substantially. developed from the SNA; still more can be constructed by Since the 1983 study, which alluded to the problem, Bank recognizing variants that have been found useful by some staff have made concerted efforts to improve the intema- countries for their own analytical purposes but have not tional comparability of these practices. swayed the international community. Indeed, GNP itself has a checkered career in the SNA, having been exiled by 11. The treatment of population-the denominator of the last (1968) revision but undergoing rehabilitation in the per capita GDP and GNP-should in principle, be consis- forthcoming (1990) revision. The first decision, then, is to tenL For GDP it should include foreigners temporarily decide how many, and what type of alternative measures living in the country (though not tourists) and exclude of income should be studied for the Bank's purposes. nationals temporarily living abroad; and conversely, for calculating per capita GNP. In fact, such adjustments to Domestic andNationalProduct - GDP vs GNP population are almost never done, largely for lack of data and because demographers use different concepts (de jure 8. The distinction between domestic and national versus de facto population). For a given GDP, a change in measures of income is important. Conceptually, GDP, as residency status should affect per capita GNP only by the defined in the SNA, measures the value of the total final amount by which the per capita income of the migrant output of goods and services produced in an economy. (workers and their families) differs from that of the rest of GNP measures the domestic andforeignoutput claimed by the population. Migrant workers will tend to have higher residents of an economy. The two measures differ by the earnings and per capita incomes than prevail in their factor incomes received from and paid abroad, i.e. income countries of origin. They will often have lower earnings received by residents as compensation for factor services than prevail in the host country (though expatriate rendered overseas, less payments abroad for factor ser- managers are often an exception). Whether or not they also vices provided by nonresidents. These factor incomes have lower per capita incomes will partly depend on the comprise investment and property income, including inter- location of their families: "guestworkers" may have low est on debt, and labor income, i.e., the income that in- wages but relatively high per capita incomes if they leave dividuals earn in an economy other than the one in which their families behind. Even if they send money home, this they reside. should count as a transfer if they themselves are considered residents of the host country. 9. For most Bank borrowers significant differences between GDP and GNP relate to labor income (from 16 12. It should be noted that migrant worker residents of the host country may later migrate back country. They are likely to transfer their to their home Box A1.2: Cameroon: Bank staff adjustments to savings home official estimates periodically while resident in the host country, sum, at the end of their stay. Periodic transfers, or in a lump Cameroon is a case in which operationally classified in the SNA as current transfers, which are nificant deficiencies in the national accounts sig- reduce the na- led to tional disposable income of the host country a mission and subsequent adjustments in and increase GNP. The the national disposable income of the major issue was reporting of petroleum production home country.I and exports. The mission, in 1988, reviewed Lump sum transfers are classified as capital the transfers in the entire national accounts. Adjustments for SNA. They do not affect the national petroleum disposable income of as well as subsistence agriculture, rural construction, either the host or the home country. Thus, on the basis of commerce, banking, and investment were the SNA methodology, host and home countries' made. The disposable income would be affected by national revised estimates better reflect the actual migrant workers' structure of choice between periodic and lump sum the economy; it is understood that the new transfers. estimates tice, few countries are able to distinguish between In prac- are now being adopted by the national compilers. remittances and migrants' transfers. Even workers' the distinction national accounts, and sometimes to persuade between labor income and transfers is difficult, and assist somewhat arbitrary. and often national authorities make substantial methodological provements to their national accounts. 2 im- (See Box A1.2.) 13. In principle, though, countries whose receipts Occasionally, such missions helped provide of benchmark periodic labor remittances are larger than estimates of per capita GNP (i.e. Pacific payments, have Islands) and in disposable incomes larger than GNP. Similarly, some cases, missions have confirmed earlier countries the official national account series, without doubts about whose payments are larger than receipts obtaining have disposable enough information to replace them. incomes smaller than GNP. Improvements to Data: Bank Actions 16. Effective technical assistance has been possible only in rare cases. One precondition is operational staff and the government give that both Bank 14. Bank staff have made systematic though strong priority efforts to improve the quality of data. The modest to resolving specific statistical issues. economic in- Central staff dicator tables attached to the Country resource allocation must balance compilation biannually to the Board now constitute a Briefs circulated nation of statistics, a major priority against and dissemi- key technical assis- cooperation between central staff (IEC) instrument of tance to improve the data. Operational staff economists for that purpose. These tables and regional ly direct their statistical efforts towards quite reasonab- evaluating and statistics deemed to be operationally most embody the monitoring specific Bank projects and programs, but this which therefore deserve the closest attention.significant and does not in itself promote continuity and balance in the among available series are systematicallyDivergences basic data work of national statistical systems. Other in- country economists are requested to obtain analyzed; stitutions of the United Nations family additional in- have primary formation in the course of their visits to the responsibility for statistics, and some have adjusted series are prepared within the Bank countries; and efforts markedly in response to budgetary reduced their when there is constraints. The convincing evidence about specific deficiencies Bank too has reduced, albeit marginally, the staff and other official national statistics and there are reasonable of the resources allocated to statistical and data work. to correct them. Given the scarcity of resources means also recognize that, given the many priorities One must that can be ber governments, good-quality, timely facing mem- devoted to such quality checks and improvements, data have not al- regular dialogue has been found to be a ways been accorded high priority. the most effective means for improving statistics in the majority countries. of the 17. Technical assistance by itself cannot improve 15. In special cases, missions of IEC national data if this is not given adequate priority by countries have also helped evaluate the staff to specific national authorities. Such 'riority is rarely given. In some weaknesses of 1. While disposable income is formally an SNA concept, defined as GDP plus and was omitted from BoxAl. 1 for net transfers from abroad, it is rarely the sake of simplicity. used in practice 2. Brazil, Cameroon, Chile, Congo, Egypt, Jamaica, Liberia, Mali, Niger, Nigeria, Senegal, Sierra Leone, Sri Lanka, and Uganda. 17 crop produc- influence plex for informal sector activities other than systematic cases, national governments may purposely that is tion, from which there is usually no regular and statistical organizations to report information true for Upward biases in growth data collection program. This is especially favorable to the govemment. wholesale and retail trade and services. generally, rates have been identified in several cases. More time to focus on the organiza- governments rarely have the or to devote resources to an 21. Most developing countries include in their value tional needs of good statistics, the informal to come and added estimates some contribution made by area where improvements are necessarily slow sector, but it is unknown what proportion is actually in- This is even whose benefits are not always fully perceived. quan- cluded. Country estimates of informal sector activities detailed the case for well-trained economists, whose "The vary in coverage, valuation, and estimation methodology, titative discussions often begin with the statement: depend " within a given country, because the estimates data are known to be weak, but ...... heavily on informed guesswork, changes in staff can over time in the importance of produce spurious variations A Major Source of Difficulties:Informal Economic such activity within a given country. Activities of 22. There are clearly major methodological differen- 18. One particular source of weakness and lack These do not ces in the treatment of informal activities. its sectoral in the treat- comparability in data derives from differences merely affect the absolute size of production, or usually par- ment of informal activities, whose share is distribution. Similar difficulties and differences affect es- Their scope ticularly important in developing countries. but also timates of uses of production, i.e., the size and shares of varies not only with the structure of economies and savings. It has been alleged political, and economic his- consumption, investment rates and with their legal, institutional, is probably that significant differences in national savings of such activities tory. The relative importance where on capital-output ratios are sometimes due to methodological greatest in some Sub-Saharan African countries differences. The same may be true for income levels andto reflect infor- average around one-half of total output may may be as some extent even for growth rates. mal activities. In agriculture, subsistence output activities high as four-fifths of total output. Such informal 23. Obviously, the costs of statistical systems must but they are must, in principle, be fully included in GDP, essentially be borne by national governments. However, national accounts. In by no means uniformly recorded in given the major analytical use made of the data by the little about estimation many cases, the Bank staff know international community, and the resources devoted to a procedures used by compilers. analytical exercises variety of cross-sectional and other data, the level of and statistical studies which rely on the 19. Informal sector activities are mainly performed the international effort directed at improving the data individually by households producing goods and services seems paradoxically low. Resources permitting, we their own consumption, but sometimes and perhaps or collectively for are diverse, at times difficult propose to focus attention on this issue, agencies, These activities also for sale. difficult to develop, in cooperation with other international to locate for reporting purposes, and otherwise in open a work program better to understand, and uliimately to measure. Some producers, such as vendors of regular reduce divergences in this field. markets, have no fixed address for mailing more reliance on expert statistical surveys, which leads to Growth and Growth Rates:Analytical uses and (but nonetheless subjective) estimates. relationshipto "benchmark"-basedcomparisons 20. The quality of estimates of informal sector output 24. Most comparative analytical examination of na- of sub- depends on the data collection system. Estimates virtually all tional products is less concerned with levels than with will be reliable only if in the index sistence sector output and all growth rates. For a variety of reasons, inherent established producers are covered (albeit through samples) surveys all number problem, even if the basic data were growth rates production is captured. In many cases, however, that are precisely and in a fully comparable manner, influenced and even censuses are based on sampling frames and growth rate comparisons would be greatlychoices are commodities outdated and heavily geared towards major in place. by methodological choices. Some of these where other reporting mechanisms are generally who by administrative records, and espe- made by the national accountants themselves, indexes, Even when supported may require subjective weighing generally (if implicitly) use Paasche-type price to cially for crops, estimates evaluating each year's GDP at prices corresponding results can of partial and possibly conflicting reports. The of the sub- those of a "base year," itself moved up from time to time. only be as good as the estimator's knowledge For an intemational agency like the Bank, the two main more com- ject-matter. The measurement problem is even 18 issues (beyond all those concerning comparisons for a affected a country, then GDY may be a more relevant given year) relate to the choice of a base year and the consideration, or indeed - because it also reflects the im- related issue of partial rebasing; and to the choice between pact of changes in interest and other factor payments, income and product at constant prices. Gross national income (GNY). 25. The second one of these issues is intuitively more 29. The issue of the relationship between income and obvious. Atany one time, as shown in Box Al.1, inaclosed product at constant prices arises even if one considers a economy an identity prevails between production (agricul- single country, and no matter what base year one uses. ture, industry, etc.) incomes (of labor, property owners When the Bank considers, for comparison or aggregation, etc.) and expenditure (consumption and investment). In an many countries, a separate issue arises even for GDP. open economy, i.e. one including economic contacts with National accounts at constant prices are compiled in a the rest of the world, the identity still prevails provided variety of base years by national authorities. For compar- appropriate account is taken of foreign transactions. Thus ing levels and aggregating growth rates, they must be domestic absorption may exceed domestic production by "rescaled" to a sihgle "base year." The base year currently the same amount as imports exceed exports; this will also used by the Bank is 1980. These base years have been show up as a reduction of net foreign assets, i.e. negative changed every five years in the past, but, because 1985 was foreign investment. a particularly distorted year in exchange markets, most international organizations decided to await more normal 26. Over time, however, the equality does not hold. exhange rate relationships before rebasing. Suppose domestic production is constant in volume, i.e. GDP in constant prices does not move. Suppose, further- 30. In rescaling to 1980, for each major sector (in- more, that the price of exports rises sharply relative to that dustry, agriculture, and services) one starts with the value of imports. Then, exports stable in constant prices can buy of production in 1980 currentprices. The value of produc- greater quantities of imports, which in turn allows con- tion in previous and succeeding years is derived by multi- sumption and investment to rise. plying that 1980 value with the ratio of the sector's produc- tion in the target year to its production in 1980, in the 27. To analyze such phenomena, the Bank developed country's own constant price national accounts. the concept of gross domestic income in constant prices, abbreviated as GDY. GDY is derived from GDPat constant 31. For each sector, this constitutes a mere rescaling; prices by explicitly adjusting for changes in the terms of the growth rate of the sector is derived from the country's trade. The terms of trade adjustment is the difference own national accounts, and is not changed by the rescaling between exports deflated by the implicit deflators for ex- process. However, because relative price changes differ ports (the normal procedure), and exports deflated by the between sectors, the sum of the rescaled sectoral products implicit deflator for imports. As in all indices, this one is will differ from the rescaled GDP. Hence, a choice must be also sensitive to the choice of the base year, whose in- made between using the original constant price GDP (res- fluence on the terms of trade adjustment can be quite caled to 1980 prices) and its growth rate, and record a significant. Indeed, in certain cases, particularly in periods "rescaling adjustment" between the value of GDP and the remote from the base year, the terms of trade adjustment sum of the values of its rescaled components; and using relative to the previous year can actually be positive even the sum of the rescaled sectoral components as the proxy if the terms of trade continue to deteriorate (if the volume for GDP (partialrebasing). of exports is also declining). 32. The choice is not trivial: a research study under- 28. When to refer to the growth of GDY rather than taken for the Bank 3 has found that, out of 60 countries that of GDP depends mostly on the purposes of the studied, in about one-fifth of them, the average growth rate analysis. By and large, if growth "performance" is con- of the partially rebased (1980) series over the 1970-1981 sidered, the ability of governments to obtain satisfactory period differed from that of the original series growth rate growth rates, GDP would seem to be the more satisfactory by more than 1 percentage point per annum in the extreme measure. On the other hand, if the preoccupation is with cases. ultimate results, the way in which the combined impacts of their own efforts and of global circumstances have 3. MethodologicalProblemsin Cross-CountryAnalyses of Economic Growth, June 1988, by Jean-Paul Azamn and Sylviane and Patrick Guillaumont. PPR Discussion Paper. 19 33. After careful consideration, the Bank has decided 37. Finding a year with minimal misalignnents to use the partially rebased GDP in all comparative work. would be a matter of judgement. Only with great difficulty This shift has finally been completed in 1988, and is fully can one imagine that those for whom the choice has reflected in the latest WDR and WDI. It is recognized that operational implications should all agree on the same year. thismethodyieldsconstantpriceGDPgrowthrateswhich, Even then, the growth rate itself will depend on for certain countries and periods, can be significantly methodological choices. The necessity to make such different from those of the countries' own data, from which choices is inherent in the index number problem. Analysts they are derived. This is obviously a disadvantage, but it understand this, even if they do not often acknowledge it was felt that it is outweighedby two factors. One, relatively by appropriate testing of conclusions, i.e., verifying the trivial, is that partial rebasing does away with the "rescal- robustness of results in the face of different methodologi- ing adjustment" that would otherwise have to be added to cal decisions. rescaled sectoral values to bring their total up to the value of rescaled GDP. The more important consideration is that, 38. The relationship to each other and to various in cases of major relative price changes when partial guidelines of per capita GNP or income levels derived from rebasing makes a substantial difference, the growth rate of benchmark estimates and growth rates would also depend partially rebased GDP is in some sense closer to the "true" on choices between legitimate methods. Countries opera- growth rate than the original constant price GDP with an tionally disadvantaged by the choice ofaparticular method outdated base year, as the "reference year" is normally would unavoidably argue for another, equally legitimate, changed more often, and tends to be more recent, than the method that yields results more favorable to them. true "base year." However, because of the exceptional decision not to change the reference year to 1985, a few 39. This does not mean that one must necessarily give countries have now acquired true base years that are more up all hope of ever deriving indicators of development recent than the reference year; the growth rates of the from data that are more detailed but less frequently avail- partially rebased GDP's of these countries are, presumably, able than the routine, annual, national accounts. Indeed, less "true" than the original constant price growth rates. the Bank intends to pursue research into possible better ways of comparing economic capacities and development Benchmark Estimates levels. However, even if such research is successful, deriv- ing annual comparisons calculated by linking constant 34. This discussion has obvious implications for the price growth rates to some benchmark is unlikely to be- analysis of growth rates. Not only must one choose be- come an acceptable procedurefor operationalpurposes. tween the product and income approaches for the particular analytical purposes one has in mind; with either approach, Some Illustrationsof GNP and GNY the outcome is dependent on methodological choices whose impact and complexity are great, particularly when 40. Despite the above-mentioned reservation, the intemational consistency is also demanded. process of linking occasional, detailed studies with more frequent but less rigorous estimates is worth noting. The 35. The particular choice bears importantly on the objective, annual estimates of per capita GNP in a common result. This is not always fully recognized by analysts who numeraire, can be reached with detailed "benchmark" make great use of growth rates, particularly for cross-sec- work on a less periodic basis plus extrapolation of the tion studies. benchmark estimates with a growth rate of real income obtained by conventional national accounting methods. 36. The dependency of results on purely The main advantage of such a procedure is that it does not methodological choices weighs heavily against operation- return each year to conversion via exchange rates, which al use of GNP estimates derived from "benchmark" levels is unarguably flawed when exchange rate policy moves multiplied by the appropriate index number. A priori, the against market forces. The main disadvantage is that the benchmark idea is temptingly simple; instead of compar- flaws tend to accumulate and reappear in benchmark years ing per capita GNP numbers converted at the prevailing as debates about measurement of real income. exchange rate, which necessarily fluctuates and often reflects various misalignments, one might seek out a 41. Real Income (GNY) represents the volume of benchmark year when misalignments were minimal, and goods and services that is (or could be) purchased with the multiply its per capita GNP levels, converted into the total national product. common numeraire, by the appropriate index number of GNP at constant prices. 42. Table Al.1 shows how per capita GNP estimates can be derived from benchmark data and growth rates. The 20 first section of the table provides three GNP estimates for 1983, based on the different per capita IV (1980), and V (1985) have been used as benchmarks for growth of per capita estimating per capita GNP in 1983. ONY. Column 2 extrapolates 1975 (See also Annex 4). then-current US dollars to 1983 withper capita GNP in growth rate expressed in 1975 prices a per capita GNY 47. The variance among ICP and ges (GNP deflator) from 1975 to 1983. U.S. price chan- country is affected by changes in ICP Phases for a given 3 and 4 are calculated by extrapolating Similarly, columns probably also reflects the impact of methodology, but it a capita GNP in current US dollars by 1980 and 1985 per exercise inherent in the ICP method. "partial rebasing" the The classic index growth rate ex'ressed in 1980 and 1985 per capita GNY number problem becomes a major concern when estimat- prices, respective- ing per capita GNP for operational and ly, and adjusting each for US price analytical purposes benchmabk and target iears. Hence, changes betgeen the if a year of turmoil in international currency in each case the figures 1985, is used as a base year. These markets, like allow fob real growth and are inflated considerations US infladion rate for the corresponding(or deflated) bi the explain issues involved in finding a substitute help period, i.e. 1975- method. for the Atlas 83, 1980-8#, and 1985-83. 43. As shown in the table, the per capita 48. TheAtlas method relies on conventional GNPderived accounts and, with rare modifications, nadional by applying the per capita GNY growth official exchange the choice of base yearin some countries rate is sensitive to rates. This very broad methodological (e.g. Bangladesh, subsidiary choice of one index number choice precludes the Burma, and China) but not in others, over all others as (e.g. Algeria, Bolivia, all other methods necessarily must and Colombia). The pattern is do. When comparing not consistent across current values, by definition, the index countries; i.e., no one benchmark year higher (or lower) values. For 28 of always generates not arise. Nor is it likely that a country number issue does the 102 countries, the undervalue its exchange rate or undercount would deliberately 1975 base year generates the highest per capita income; for income in order to obtain more favorable its national another 61 it is the 1980 base year. Only in 13 cases does treatment. World Bank the 1985 benchmark provide the highest quate data did not exist to carry out esdimate. Ade- a similar comparison 49. for a further 35 countries. Neither All of the above does not preclude the was there any marked of "benchmark" estimates and other usefulness uniformity in the pattern of relationships references to the past industrial market economies and the between the as potentially useful checks on the low income develop- validity ing countries in the choice of base year. estimates. In particular, staff will continue of current attention to sharp movements of real to pay close effective exchange 44. General issues also arise rates as possible signals of the need benchmark esdimates extrapolated when considering closely the continued usability of the to examine more by the GNP growth official exchange rate rates. A crucial one is to decide whether as a conversion factor. In some cases, the benchmark and despite all the refers to an "equilibrium" year, i.e., reservations formulated above, it may major shocks in the economy have a year in which no to resort to references to the past even be necessary and devaluation, changes in relative prices, occurred, such as methods to derive estimates of present "benchmarking" per capita GNP changes. Various deflators may be or major structural levels; but such cases will be kept to a minimum, unless used extrapolated benchmark GNP per capita, to inflate the better and more objective methods are developed. basket or US GNP deflators or including SDR- the manufacturing unit Other Concepts, Measures and Variants value index, etc. 45. The second trio of benchmark results 50. NationalIncome estimated for distribution in Table population for consumption (or saving) to the A1.1 shows per capita GNP estimates is represented for 1983 using much the net national product (NNP) at by the same technique, except that the factor cost. This takes capita GNP, rather than of per capita growth rate of per account of the need to allocate some of difference between this and the first GNY, is used. The from production first to the maintenancethe current income set a rough gauge of how the gains and of results provides (to preserve the future flow of incomeof existing capital losses from intema- down resources) before distributing without running tional trade affect the comparison. it in the form of goods and services to the population. 46. Benchmarking exercises can also be using ICP estimates of per capita performed 51. In practice, even the capacity to GDP in "international depreciation of man-made capital estimate the dollars." For example, results from is severely ICP Phases III (1975), most countries, it is much more roughly limited. In often at much greater intervals, than estimated, and the gross product 21 relatively spealdng, analyses, both of the objective was to ascertain whether,at similar stages of itself. Hence the vast majority ofover time and of their when compared with countries of small island evolution of national economies focus on gross development, the per capita GNP by factors relationship with other national economies, economies was biased or otherwise distorted product. flows of national or domestic that influence their economic development. Discus- 52. That is certainly the practice at the Bank.the wear 56. remoteness The study arose from a belief that to exploit to refer to sions and comparisons sometimes seek combined with small size (i.e., an inability but statistical presentations especially affect the and tear on physical capital, mainly because potential economies of scale) might GNP. It generally and comparisons focus on gross product, on net product are comparability of estimates of per capita do indeed have to for most countries reasonable data national) product concluded that, although small islands features of open- (or simply not available. Gross domestic structures are contend with several disadvantageous problems were not is the focus of growth discussions; economic ness, remoteness, and small size, these agriculture, industry, handicaps were often analyzed in terms of the shares ofinvestment, consump- unique and that such development etc.; in terms of the shares of gross countries. Whilst having to payments shared by many other developing not considered tion, exports; and in terms of their relationship an influence on GNP, these features were distor- balance or foreign debt. to produce any major to be of such significance as for small islands. tions in the measure of per capita GNP by GNP plus 53. DisposableIncome is represented of grants in the form develop its net transfers received from abroad in cash or 57. IEC continues to maintain and economies transfers on island and other private or public unrequited economic and social data series It measures the total income available to a country missions and closer contact with the main kind. (In practice, the range through its own area, such as the services. this for acquiring goods and goods and services may international agencies working in Commission, of these Pacific and sovereignty of choice Asian Development Bank, the South Commissions and Social be restricted). the UN Regional Economic Secretariat, Income or Real Expendi- (especially ESCAP), the Eastern Caribbean the Common- Purchasing Power of University, and 54. when analyzing the the Australian National twe. For comparative purposes, wealth Secretariat. it is meaningful to economic performance of countries, output and national determine the value of a country's net prices in order "Indigenous"Production expenditure at a uniform set of international in international cur- productive Some countries wish to subdivide the transactors to ascertain how much it would cost and services. Pur- 58. of rency to purchase a given set of goods between those cost in national curren- activities within their territory chasing power parities reflect the who are implicitly concemed, the equivalent legally-defined as nationals and others, rather directly cies of acquiring in the country For individuals this tends to return that a US dollar foreign. adoption of analytic amrrount (quantity of goods and services) income (GNP) to the issue of residency, and SNA's national of intemational buys in the USA. Thus, when the into dollar values rather than legal criteria in the interest the of its components are converted the issues are fundamentally or any for individual items of comparability. While are less using purchasing power parities single overall official ex- same for enterprises, the lines of argumentation naturally in at a expenditure, rather than with other countries direct because retained eamings accumulate owned. change rate, the resulting comparison or domestically real differences in enterprises, whether foreign added by foreign- in "intemational" dollars represents the value as valued on the same Hence, even in the long run, exceed quantities (volumes) of net output are discussed more owned or controlled corporations will normally factor in- in net common basis. The accounting issues repatriated earnings (which are included to GNP). This GDP fully in Annexes 3 and 4. come and thus in the adjustment from foreign ownership of Islands. A November 1985 leads some countries with sizeable on SNA overstate 55. The Issue of Small 4 reviewed the quality, consistency, and inter- enterprise to argue that income measures Board Paper World Bank Atlas their income levels. national comparability of the published economies. The island per capita GNP estimates for small to Board Paper R85-335, annex, 'The Measurement of GNP in Small Island Economies"; 4. Specifically, the statistical from IDA." to Small Island Economies Graduating IDA/R85-134, 'Terms of Lending 22 Table A1.1: Per capita GNP for 1983: Atlas methodology and selected alternatives Atlas Gross National Income (GNY) Actual Real Gross National Product 75 Base 80 Base 85 ICP Results (by Phase) Co-tmty 1983 75 Base 80 Base 85 GNY gr. GNY gr. GNY gp. II IV V GNP gr. GNP gr. GNP gr. Lao, PDR 90 GNP gr. GNP gr. GNP g. Bhutan 110 . . Ethiopia 120 .. . Chad 150 140 130 . 130 150 140 130 180 200 100 . 3 40 180" Bangladesh 150 180 200 100 260 170 130 Mozambique 250 170 130 160 Nepal .. 170 160 170 Burkina Faso 170 .. 180 160 130 190 230 150 Bumna 180 200 230 150 Mali 270 230 160 180 250 240 160 240 280 170 Guinea-Bissau 240 280 170 190 260 400 230 190 170 230 Malawi 190 180 170 180 200 190 Maldives 230 210 200 170 530 Burnmdi .. 470 390 240 220 170 190 2006 230 220' Rwantda 260 200 240 220 210 290 260 Haiti 200 280 260 270 270 260 Togo 260 300 270 270 260 Benini 430 380 250 280 480 390 250 290 370 250 India 280 300 390 250 320 300 250 Central African Rep. 290 280 300 250 270 340 250 950 760 860 Guirnea 270 340 250 290 410 340 Zaire 260 440 330 290 350 270 Madagascair 430 280 410 300 280 460 290 China 330 290 360 310 470 330 290 550 Tanzania 420 230 380 340 310 250 420 230 300 300 270 Comoros 300 290 370 320 240 Niger 310 340 430 320 300 Sri LAnka 470 500 350 330 470 450 350 540 340 300 Ghana 340 450 330 290 350 360 290 1,660 1,610 Kenya 340 460 370 300 380 430 320 Sbo TomOe & Principe 430 460 320 340 790 670 480 740 380 360 Gambia, The 1,060 420 360 350 350 410 Pakistan 410 360 430 350 290 360 410 Sudan 300 270 380 380 470 530 310 1,260 1,370 Sierra Leone 450 450 520 390 340 350 460 420 350 380 Cape Verde 430 400 830 510 370 Senegal 830 540 380 430 560 620 Mauritania 420 530 610 460 420 530 430 .. 870 710 Bolivia 430 530 510 490 500 470 470 Liberia 490 580 470 500 460 540 510 .. 1,510 480 660 560 Yemnen, PDR 500 500 . . .. Lesotho 470 540 510 510 520 Guyana 550 .. 500 520 490 570 570 550 Solomnon Islands 570 990 650 550 Zambia .. 480 510 570 570 670 620 560 570 510 830 Indonesia 660 580 850 690 580 570 620 430 Egypt, Arab Rep. 520 460 610 630 520 .. 1,430 850 650 600 Yenen Arab Rep. 640 720 670 580 390 610 560 Honduras 670 320 610 560 600 610 640 Morocco 710 700 670 640 930 1,020 660 .. 1,160 Philippines 1,000 1,030 660 750 720 .. 1,390 1,200 810 840 740 C&e dlvoire 760 820 840 1,890 2,070 El Salvador 830 1,140 720 760 920 1,200 760 570 670 690 .. 1,340 970 Papua New Guintea 760 710 730 690 800 790 720 .. 1,270 Tonga 770.. 870 880 710 . ... Thailand 810 .. . 780 840 750 Saint Vincent 760 860 730 2,050 830 820 .. 2,880 Zimbabwe .. 670 690 750 850 870 710 Cameroon' . 790 890 870 870 830 1,030 790 . 1,120 .2 Botswana 700 760 1,050 920 870 820 730 .. 1,240 1,220 700 940 1,080 Swaziand 740 .. 1,640 1,260 940 570 680 Nigeria'1 .. 920 1,000 840 970 660 1.030 960 750 980 950 .. 900 520 Not available 23 Real Gross National Product ICP Results (by Phase) Atlas Gross National Incomne (GNY) Base 80 Base 85 II NV V Base 80 Base 85 75 Actual 75 gr. GNP gr. GNP gr. GNP gr. GNP g. GNP gr. GNP 1983 GNY gr. GNY gp. GNY gr. Country 700 960 880 .. Domiinica 1,000 -- 960 900 -- -- . 930 .. 3,050 Grenada 1,030 910 1,810 1,930 1,060 1,540 1,880 Costa Rica 1,210 980 . .1,760 Belize 1,070 .. 990 910 1,110 .. 1,020 2,380 SL Lucia 1,080 .. 1,140 1,210 1,180 .. 1.060 1,200 1,190 Dominican Rep. 1,130 1,340 1,470 970 980 1,430 960 2,410 Mauritius 1,140 1,640 980 1,040 .. 1,430 950 1,060 Peru 1,140 1,170 .. 2,340 1,180 1,130 1,150 1,170 950 1,130 Guatemnala 770 780 1,300 710 700 1,300 3,330 Nicaragua 1,170 1,440 1,740 1,040 ... 1,430 1,700 1,020 Turkey 1,180 930 1,370 1,170 1,300 1,390 1,130 2,450 1,870 Congo, People's Rep. 1,210 1,450 1,550 1,100 .. 1,520 1,580 1,140 Tunisia 1,250 1,210 .. 2,700 1,250 1,080 1,360 1,310 1,050 1,320 2,000 Ecuador 1,600 1,440 1,540 1,520 1,350 1,450 Jamaica 1,370 1,540 1,320 1,310 1,320 1,300 3,260 3,230 Saint Kitts and Nevis 1,380 .. 1,140 1,400 1,320 1,130 1,390 1,320 3,680 Colomnbia 1,440 1,700 1,780 1,650 1,790 1,780 1,650 Syrian Arab Rep. 1,640 1,570 .. 2,340 1,610 1,370 1,480 1 1,660 1,330 1,480 Pamaguay 1,440 1,830 1,620 1,490 1,840 1,600 Jordan 1,680 .2,130 1,850 . . Namibia 1,690 .. 1,880 1,640 1,400 1,700 .. 1,820 . 1,590 Antigua and Barbuda 1,590 1,960 1,840 1,700 1,590 1,740 3,370 Fiji 2,140 1,540 3,190 2,100 1,540 2,260 1,820 1,900 .. 4,540 3,640 Brazil .. 1,560 .. 1,560 . 4,250 Poland 1,830 . 1,870 2,170 1,790 2,280 2,060 1,770 3,410 Malaysia 1,900 1,970 2,030 1,710 .. 1,580 1,960 1,800 3,680 Chile 1,920 3,150 1,900 1,890 .. 2,540 1,870 1,910 Argentina 1,950 1,850 .. 3,730 1,790 2,040 2,040 1,980 1,830 2,000 3,990 3,670 3,600 Panamna 1,300 2,310 1,700 1,560 2,310 1,690 Korea, Rep. 2,020 * . . . 7,180 5,680" 5,200 Macao 2,140 . 2,430 1,990 2,360 2,010 . Hungary 2,160 .. ... 2,100 . Romania 2,180 . .. 2,080 .. 4,390 5,630 2,080 3,030 2,750 2,230 2,970 2,680 4,650 Portugal 2,770 2,560 2,050 2,600 2,460 2,070 Mexico 2,290 2,130 2,020 . .1,870 Seychelles 2,380 .. 1,780 2,490 2,220 2,230 2,570 2,230 Algeria 2,430 2,470 2,290 2,270 2,180 2,130 2,250 South Africa 2,430 2,320 4,730 4,070 2,380 2,850 2,720 2,460 2,250 2,680 5,470 4,660 4,750 Uruguay 2,990 3,740 2,440 2,910 3,740 2,500 Yugoslavia 2,640 2,650 2,690 . .2,430 Suriname 2,680 .. 3,990 3,960 3,140 3,940 3,890 3,110 Malta 3,480 3,210 4,020 3,880 . 4,620 5,770 Gabon 3,950 . 3,580 .. 5,910 3,620 4,730 5,030 3,970 4,620 5,060 4,640 Greece 3,580 3,400 3,970 .. 3,040 2,970 3,820 Venezuela 3,980 3,220 3,330 3,610 . . 6,930 7,340 7,270 Barbados 4,010 .. 4,880 6,280 4,200 4,720 0,130 4,160 5,570 6,350 6,620 Spain 4,670 5,010 5,690 4,640 4,700 5,860 4,660 Ireland 5,030 9,700 9,590 5,750 7,160 5,430 .. 6,880 7,090 5,230 Hong Kong 6,230 5,510 7,430 4,880 7,340 4,420 8,500 6,450 4,500 6,430 6,100 .. Trinidad and Tobago 7,420 6,650 6,230 7,100 Israel 6,480 .5,860 6,800 5,950 . Singapore 6,930 .. 1,760 6,380 5,440 6,250 5,700 Omnan 7,090 .. 5,870 5,630 7,790 6,450 7,170 .. 7,060 7,730 9,090 10,350 Bahamas 7,840 8,700 6,660 7,480 7,680 8,750 6,660 7,630 .. .. 9,880 htaly 7,680 7,680 8,510 10,290 8,390 8,140 8,460 8,22 10,360 10,860 New Zealand 8,170 11,590 14,300 10,360 9,110 10,910 13,940 8,110 9,060 10,170 Belgium 8,130 7,250 9,720 9,180 7,780 9,730 10,580 10,370 United Kingdomn 12,220 8,250 10,270 12,250 8,260 9,580 9,31f0 9,750 9,860 10,940 10,830 Austria 11,270 13,970 9,080 10,950 14,100 9,040 10,570 10,750 10,820 Netherlands, The 10,080 9,370 12,630 8,850 9,590 12,600 8,830 11,770 11,770 11,200 Japan 10320 11,940 14,130 9,770 12,090 14,270 9,710 France 10,630 Not available 24 Atlas Gross National Incone (GNY) Real Gross National Product ICP Results (by Phase) Actual 75 Base 80 Base 85 75 Base 80 Base 85 m IV v Country 1983 GNY gr. GNY gr. GNY gr. GNP gr. GNP gr. GNP gr. GNP gr. GNP gr. GNP g. Finland 10.900 10,630 12,580 9,640 10,650 12,670 9,690 10,730 10,860 Demnark 11,390 12,870 15,980 9,900 12,990 16,130 9,940 11,070 11,970 11,420 Germany, Fed. Rep 11,450 13,300 15,850 10,050 13,520 15,990 10,020 11,900 Iceland 12,110 11,470 12,240 13,400 14,980 10,910 14,210 14,730 10,890 Australia 12,480 12,730 12,200 11,210 12,800 12,210 10,990 Sweden .. .. 11,340 12,540 13,940 16,730 10,940 14,740 17,040 11,050 .. .. 11,870 Canada 12,670 14,260 12,470 11,090 14,080 12,500 10,980 .. 13,660 14,370 Saudi Arabia 12,960 14,400 15,790 .. 9,890 12,820 8,680 Norway 14,150 14,850 16,160 11,940 13,730 16,000 11,870 United States .. 14,210 12,610 14,560 14,320 14,610 12,610 14,230 14,490 12,630 13,830 13,920 15,330 Luxembourg 14,780 17,700 21,350 12,920 15,370 21,500 12,900 13,860 14,960 12,600 Switzerland 16,440 15,420 21,300 14,530 14,970 20,760 14,460 Kuwait 18,540 22,880 19,520 14,630 13,530 21,080 14,610 Qatar 19,170 .. .. .. 12,620 20,160 17,780 Brunei 20,980 .. .. .. 13,370 23,960 15,190 United Arab Em. 24,540 24,050 26,470 .. 26,610 20,370 20,340 = Not available Note: The first section of the table provides three different per capita GNP estimates for 1983 based on GNY. Column 2 extrapolates 1975 per capita GNP in then-current US dollars to 1983 with a per capita GNY growth rate expressed in 1975 prices and US price changes (GNP deflator) from 1975 to 1983. Simnilarly, columnm. 3 and 4 are calculated by extrapolating 1980 and 1985 per capita GNP in current US dollars by the per capita GNY growth rate expressed in 1980 and 1985 prices, respectively, and adjusting each for US price changes between the benchmark and target years. Hence, in each case the figures allow for real growth and are inflated (or deflated) by the US inflation rate for the corresponding period, i.e. 1975-83, 1980-83, and 1985-83. T}he second trio of benchmark results (columns 5-7), shows per capita GNP estimates for 1983 using much the same technique, except that the growth rate of per capita GNP, rather than of per capita GNY, is used. Based on current data and methodology; actual FY84 classification was one class lower. Source: World Bank, Socio-Economic Data Division. 25 Annex 2 ENVIRONMENT AND RESOURCE ACCOUNTING 1. Complex issues and inter-relationships associated directly or indirectly with the conventional system of na- impinge on the broader, longer-terrn viability of an tional accounts are sometimes treated in separate "satel- economy. The overall surpluses and distributable financial lite" accounts linked with the SNA. Satellite accounts are rewards determined by corporate accounting methods, which are officially incorporated into the subsequent a technique, developed in France, to portray important and more detailed aspects of sectoral behavior and inter-sec- statistical assessment of output, value added, and income toral linkages, usually between economic and social ac- in the national accounting context, may impart a false tivities. This technique has been adopted, for example, to sense of security to countries involved in the production of analyze the education and health sectors. It helps illustrate exhaustible natural resources. Likewise, the discovery of what services are provided publicly and privately, and who new natural resources may have to be reflected as a posi- uses them, usually by linking administrative data with tive improvement in a country's economic status and household survey information. potential. 2. 5. Prudent economic management requires that The Expert Group for the Revision of the SNA has recommended that complex issues, such as the assessment governments (and households) distinguish sharply their of environmental degradation and the depletion (or dis- income from resources obtained through the reduction of covery) of natural resources, with all their varying their net wealth by sale, borrowing, or reductions in its economic and social ramifications, be treated in a satellite volume and usability. Many feel it is important therefore system of data presentation linked to the SNA, at least for that national income should as closely as possible reflect the time being. A more detailed investigation of environ- truly sustainable income. Satellite accounting systems may mental concerns, including resource depletion, can also be used to reflect the economic impact of deple- be tion/degradation of natural resources and public "defen- carried out within the proposed UN framework for en- vironmental statistics. We agree with these proposals. sive expenditures." While in practice the distinction may often be difficult or blurred, conceptually it is clear: raising 3. this year's income does not, by itself, reduce prospects for These satellite approaches, in telling a clearer and more expanded story, would, to a large extent, take care of future income; while raising the resources obtained important issues like so-called "defensive expenditures." through the reduction of finite net wealth clearly affects These are costs incurred to protect the environment and the amounts that can be so obtained in future. to combat, either privately or at a collective, public level, those undesirable aspects of economic activity that lead to 6. Such issues arise partly as a result of taking a environmental decay, resource degradation and depletion, long-term perspective of the production process and a view and pollution. At present, when defensive expenditures are beyond national geographic boundaries. But there is also incurred privately, they are treated as other costs and they an evident asymmetry in the SNA treatment of man-made are not normally part of GNP. The contribution to GNP of, assets and natural resources. Man-made assets, buildings say, a steel plant is equal to the value of the steel produced and equipment, for example, are valued as productive minus all costs incurred, whether to buy iron ore or to abate assets and are written off against the value of production pollution. However, when pollution abatement or environ- as they depreciate and thus deducted from net national mental protection is incurred publicly, it is deemed to product. Natural resources, which are also assets, are not contribute to GNP without any corresponding deduction so valued or at least, not adequately accounted for in most having been made necessarily elsewhere. instances. This practice may sometimes confuse the sale of assets with the generation of income. 4. Satellite accounts can also be developed to reveal more clearly how the exploitation, production, and sale of 7. Accounting for the discovery and depletion of scarce subsoil and other non-renewable natural resources non-reproducible, exhaustible natural resources in the derivation of annual GDP or income is a highly compli- 27 Annex 2 10. Mostproduction and consumption activities have cated question, even in theory. The issue of sustainability time some impact on the physical environment. As they have economic arises, at least partially, from the more limited In growth and population expansion have occurred, income. perspective conventionally taken in measuring increasingly put pressure on the environment small, and the that the SNA the case of natural resources, it is arguable created natural resource base. When the pressure was still treats the sale of an asset as if it were value added years ago, there may have been some justification to make Thus, the calculated by the original factors of production. derived no reference to the contribution made by the environment value added of these resource producing sectors, to economic activities, both as a resource base and as a the sale of the final output and the and as a residual between the "waste sink," receiving the residues of the production reflect cost of the intermediate or basic inputs, will also costs for consumption process. But there is littlejustification a sig- this as the scarcity value of the produced resource as well form. The now that output has expanded worldwide to such of bringing it to the point of sale in a marketable nificant level that it causes "externalities" that often lead natural economic rent derived from the production of the and to unexpected additional costs that are not bome by the asset concemed can thus vary according to the actual to him. the asset, quite independent of the producer and are not easily attributable perceived scarcity of by the factors of production. contributions made 11. An understanding is gradually gaining ground has done that, in many instances, the production process Those 8. The justification for including resource values in and is still doing enormous harm to the environment. GDP assessments rests, historically, on conventional and "external" costs will eventually have to be borne by sys- some- standard accounting practices followed by enterprises one. The costs are definitely "internal" to the global that, once the widely accepted commercial assumption are tem. At least at the global level, is it important to account discovered, goods "in the ground," like oil and copper, effect, is properly for income and costs, and to distinguish clearly "free gifts of nature." The implication of this, in drawing down of and between true income generation and the that in total such stocks of natural assets are unlimited or degradation. explora- capital assets through resource depletion that new stocks can always be tappedby incurring As a matter of practical management of these costs, it expenses. When these latter are to internalize tion and development they are would also be desirable as much as possible incurred by the producing entities themselves, country over time, them, i.e., deduct them from the incomes of thedo so, in deducted from NNP when they are amortized It is and producer responsible for them. However, to and often even from GDP when they are "expensed". some cases related benefits would also have to be internal- true that the new resources discovered through exploration ized first. have often exceeded the amounts used up worldwide. Known or assured reserves of most mineral resources have 12. GDP, the most commonly used variant of ag- value of tended actually to grow over time. The economic gregate income, is valuable mostly for indicating short- to decline certain natural resources, say, oil or coal, may well and in medium-term changes in the level of economic activity, the resource itself runs out, worldwide long before that this and is widely used for demand management an stabiliza- a given country. However, one cannot be assured country may be tion policies. However, as calculated at present, it part- is always the case, particularly at the individual growth, specific less useful for gauging long-term sustainable level. Natural resources have become exhausted in ly because there can be undesirable side effects of produc- in Kiribati, before their countries, e.g., phosphate tion that are not taken into account. has disappeared. Conse- economic value, worldwide, quently, GDP in these countries has fallen dramatically. 13. On the grounds of simplicity and continuity, the likely to Expert Group for the Revision of the SNA seems 9. Two main conceptual approaches have been recommend the continuation of the conventional treatment proposed to deal with the depletion/degradation of natural presently recognized internationally (but with a reminder resources: the depreciation and the user cost approaches. could to users to be more conscious and explicit about the The principle of depreciation of man-made capital shortcomings). However, as indicated above, the construc- to the consumption of renew- the main be applied straightforwardly not be tion of separaie satellite tables will be proposed as able and non-renewable resources, but this may lies the method for deriving a measure that more closely ap- entirely satisfactory. Behind the user cost approach such a can be proximates sustainable income. To move toward notion that the net sales of a depletable resource satellite system, the UN has drafted a separate Framework split into a capital element, or user cost, and a value-added under element for Environmental Accounting, which is currently these element, representing true income. The capital review (the Bank, UNEP, and UNSO are involved in and so an equivalent amount represent real asset erosion, discussions) and this should provide a more satisfactory from total income should be actually or hypothetically basis for accounting for environmental issues. reinvested in other assets so that it continues to generatebeen the same level of income after the resource has exhausted. 28 Annex 2 Annex 3 THE INTERNATIONAL COMPARISON PROJECT 1. As in the case of adjusting value series for price Methodological Issues changes over time to determine real GNP or the underlying real physical quantity movements, the rationale for defin- 5. With the completion of new rounds of the ICP, new ing a similar standard basis for making intemational com- theoretical and practical problems appeared and some old parisons of quantity aggregates is compelling. ones seemed no nearer solution. Many of these problems are highly technical, in particular those that relate to 2. The theory, as it has been developed for the Inter- choices between different types of aggregation procedures national Comparison Project (ICP), focuses on the deriva- and index numbers. There are continuing debates as to tion of appropriate purchasing power parity (PPP) which ICP methodology is most consistent with the nation- measures to convert national accounts aggregates al accounts. The aggregation procedures used and the level measured in national currencies into a common unit of at which aggregation occurs influence the measured real account to facilitate comparisons on a uniform basis. A economic relationships between two (or more) countries, brief history of ICP is given in Box A3.1. In providing a i.e., the proportionality is not stable. All methods may tend consistent relative price statement as a reference standard, to generate similar results if the same set ofpricesis used. the conversion of GNP aggregates (and their various sub- However, the choice of the set of prices and linking proce- components) by calculated PPPs permits a better under- dures to be used (bilateral, linked multilateral, or extended standing of the income levels of countries, according to multilateral averages) bears strongly on the result. their relative command over actual goods and services. By the same token, the approach provides a more rational basis 6. The basic approach of the ICP is to calculate "real" for weighting national growth rates for regional com- quantities (Q3 on the basis of estimated expenditure values parisons. PPPs, being the rates of exchange that equalize (V) divided by the relevant reported national prices (P) and the purchasing powers of different currencies, also serve to revalue all such derived quantities of goods and services as a basis for analyzing differences in internal relative at a common set of "average international prices" (IP) prices between countries. based on the countries involved in the particular com- parison. The derived measure (Q x IP) is deemed to repre- 3. The use of PPPs to adjust GNP data leads to figures sent an internationally comparable real economic value or of the volume of goods and services available in an "volume." However, how "true" the "Q" obtained from economy. For analysis of real differences in the composi- "V/P" is depends on several considerations, in particular, don of national expenditures, particularly for comparisons on the ability to overcome differences in the types of goods between countries within a region, it is preferable to use present in each country and on the ability to compensate PPPs for some purposes. for differences in quality, including style (fashion), which affect "P" and basic price collection issues. 4. However, ICP methodology is such that different PPPs obtain depending on how one responds to various 7. Some products are not present in every country. methodological issues. In practice, country rankings will Therefore, one must decide how to compare the prices of vary with the PPP methodology adopted. This appendix bundles of goods containing different products: e.g., the considers the main conceptual issues as well as some more price of a fruit basket containing (apples, pears, and oran- prgmatic aspects of the exercise that may have to gel ges) with that of other baskets containing, respectively, before a consensus can develop that some unique set of (apples, pears, and plums), (plums, pears, and melons), and PPP calculations is not only valid but also applicable in the (plums, melons, and mangoes). Moreover, while price analytical and operational context of Bank work. comparisons must be based on precisely defined types of products (e.g., French bread of a certain weight and cotton shirts of a certain fineness and size), national accounts data 29 Annex 3 correspond only to much broader categories, and therefore on all available observations; the EKS (Elteto-Koves- do not provide adequate guidance in determining the Szulc) method does the same thing through a series of weights to be attributed to specific products. A related weighted binary comparisons. When all data are available, problem is that of differences in the representativeness of the two methods generate identical results. Data imperfec- given products. Packaged, sliced bread is available in tions will affect ihe two techniques differently but in France, but a relatively scarce luxury, while the same is practice the results appear similar at very high levels of true of freshly baked French baguettes in the U.S. To aggregation whichever technique is used, according to overcome these problems, the 1980 and 1985 ICP studies available data. use the respective country "characteristic product" price comparison approach. This handles differences in the 9. Other less technical but no less complex issues availability of goods in various countries by means of an relate the comparison of qualities. Within a country, if to implied weighting procedure. the quality and the value of some products or services increase jointly over time, this is reflected in the national 8. There are two main methods for calculating tran- accounts as an increase in the volume, i.e., the quantity of sitive purchasing power parities at the group heading level production. For goods and certain services, ICP strives, in (e.g., "fruit") when some of the item prices are missing. principle, to make similarallowances in intemational com- The CPD ("country-product dummy") method uses a parisons. As far as goods proper are concerned, ICP prac- regression technique to estimate the group average based tice seems to have evolved towards increasingly taking Box A3.1: The International Comparison Project: an historical perspective some funding from the US In 1968, Professor Irving B. Kravis of the University of Pennsylvania secured The purpose of the project Project (ICP). National Science Foundation to start the International Comparison countries by converting them was to develop a system of comparing national accounts aggregates of different in domestic markets rather to a common numeraire currency on the basis of purchasing powers of currencies the then OEEC with Milton than exchange rates. Building on his experience in similar work in the fifties at Robert Summers and Alan Heston sought to develop a Gilbert, Kravis in collaboration with his colleagues, economic system (socialism system which could be applied to all countries of the world irrespective of their in several phases, initially or capitalism) or geographical location. The proposal was to develop the system to an increasing number of concentrating on methods and gradually refining them as they were applied research phase was over, it moved countries. The work began at the University of Pennsylvania and, after the the ICP right from and funded to the United Nations for implementadon. The World Bank liked the proposal a stamp of approval and but it also gave it the start. Not only did the World Bank provide own research money secured funding for the project from other countries and organizations. methods. It worked with data The ICP has now more or less completed five phases. Phase I concentrated on results that were transitive with for ten countries and adopted a system that was multilateral rather than binary, Phases II and III extended the coverage to sixteen and thirty-four and independent of the base country chosen. III, which was completed in countries respectively with some refinements in comparing services. After Phase 1982, the project moved to the United Nations. stage system was adopted in Phase IV introduced a radical change in methods. In response to demand, a two linked to form a global and then the "regions" which countries were first compared within a regional setting own groups. From Phase of the countries in their comparison in a manner that retained the relative positions providing additional technical IV,the EEC became the main organizer and financier of ICP, with the OECD process which is expected to support. Phase V introduced some further refinements to this regionalization continue in Phase VI for 1990. an interest in the project and Major World Bank funding for the project ceased in 1982, but the Bank retains funding in special areas. It has continued to provide technical advice and support as well as supplementary has also provided research funds for ICP related studies. Annex 3 30 into account quality differences. Nevertheless, problems 14. The same procedure can, of course, be performed remain, particularly for products where brand names and forwards or backwards. One can derive from ICP IV data "fashions" are important. (Is a sturdy pair of blue jeans for 1980, and from national accounts information on made in India closer to a sturdy Levi or to an equally sturdy values and implicit price indices relating to other years, the but much more expensive Calvin Klein? How to compare prices and volumes of GDP components and of total GDP the two qualities of mutton, "with bones" and "boneless" for both 1983 and 1979, for example. Similarly, one can sold in rural African markets, to the many grades of chops, derive separate estimates of, say, 1983 GDPat intemation- legs of lamb, and shoulders, sold in high-income alpricesfromPhaseIIIICPdatafor 1975 andPhaseVdata countries?) The answers to such questions can mean dif- for 1985. ferences of several hundred percent in the volumes im- puted to certain products and for products whose qualities 15. Because of index number problems relating to are truly very different, growth rates, the estimates for 1983 based on Phase III, IV and V benchmark values cannot be expected to be identi- 10. Even more difficult quality comparison issues cal. However, they should be reasonably close. In fact, arise for services. How do the annual services rendered by however, for certain countries they have proven to be an Indian and an American physician or teacher compare? substantially different. In certain cases the ratios between What about general administrators? Answers to such in- volumes of per capita GDP, and even their ranking, are herently difficult questions determine overall PPPs. substantially different depending on the choice of base reference study. 11. For certain purposes, an average intemational price needs to be calculated. Hitherto, this was weighted 16. Some of these differences can be explained in by the "real" (i.e. PPP-adjusted) incomes of the participat- light of the methodological evolution of ICP; in particular, ing countries. The difficulty is that, to calculate the incom- the improved treatment of quality differences. Other es, the PPPs, i.e. the international price, needs to be known. causes, however, are not well understood. Clearly, a Price and incomes have been calculated simultaneously by methodology that is supposed to establish the relationships the method developed by Geary and Khamis. Recently, this of the volumes of the GDPs of various countries, i.e., to method has been criticized on technical grounds, and there correct for price differences, should yield, for any given is some question of replacing it. year, similar results independent of the choice of the par- ticular benchmark year used for the calculation. This is 12. Beyond these issues, two problems dominate and certainly necessary if ICP results are to be used for opera- form, together with the inadequate coverage of developing tional purposes in preference to using exchange rate con- countries, the principal obstacles to the operational use of version methods. But even for many analytical purposes ICP-generated data by the Bank, and even to their the data are credible and usable only if they are reasonably widespread analytical use in international comparisons. independent of the choice of the benchmark year. ICP has One is the lack of consistency of findings between different yet to meet this test. phases of ICP. The other problem relates to intercountry relationships within each phase, bound up in the issues of 17. The basic index number problem ensures that the international transitivity and fixity. estimates based on different base years will never com- pletely converge. However, as ICP methods improve, and The Consistency of QuinquennialBenchmark Findings theoretical choices stabilize, this problem should gradually be eased. Without ever becoming identical, the annual data 13. Rounds of ICP (called Phases) take place every obtained by extrapolating forwarn ast benchmark results five years; they constitute benchmarks. Annual price data can be expected gradually to conv( with those obtained can, in principle, be derived from these benchmarks if the by extrapolating backwards suosequent benchmark intenational price of each category of GDP is multiplied results. For this progress to occur, it is of course necessary by the price index denoting its evolution over time. In this to make careful and consistent choices based on a stable fashion, the international price index corresponding to methodology, but these conditions can be expected to be each component of GDP can be estimated for each of the established. years intervening between ICP benchmark years. These indices can then be multiplied by the value of each com- 18. Variations from one benchmark phase to the next ponent of GDP in those years, and yield an estimate of have deeper implications too. The negative relationship these components (and of their sum, total GDP), at inter- established by early ICPphases between income levels and national prices. Other techniques can be adopted to achieve price levels (ratio between PPP and exchange rate) was the same objective with more or less similar results. deemed to be due to differences in the productivity be- 31 Annex 3 Box A3.2: Ratio of PPP to exchange rate related to per capita GDP ICP Phase III, 1975 [CP Phase IV, 1980 Per capita GDP (ICP) (US =100) Per capita GDP (I:P) (US =1.00) ca Actual + Estitnated aActual + Estimated 4) ICP Pbase V. 1985 Rreg sion lJnes 1975, 1980 & 1985 Per capita GDP (ICP) (US =100) Per capita GDP (ICP) (US = 100) a Actual + Estimated 3 32 tween tradeables and non-tradeables. Earlier shortcut countries only within the OECD comparison; the global methods used this relationship to estimate data for comparison, therefore, maintains the fixity of the intra- countries not covered by ICP. However, a shift in the nature OECD ranking. However, the separate publication of the of this relationship from ICP III to ICP V (see Box A3.2) European comparison retains the relative positions within and the reduced strength of the relationship within each the Central European group, and these are significantly new phase has made earlier shortcut methods obsolete and different, Finland being somewhat above Austria at OECD have led researchers to look for alternatives. prices, but below Austria in European prices. Transitivity andFixity: The Relationship between 23. Methodology and Country Rankings In the published results for other regional group- ings and overall international comparisons, procedures were adopted to maintain, within the international com- 19. The joint problems of transitivity and fixity are parisons, the same rank orderings and real GDP relation- inherent to the index number problem. Transitivity is a ships that had been observed in the separate regional desirable property of data relationships. It requires that studies. Thus, in 1980, the relationship of Tanzania's GDP comparisons between two countries should yield the same to that of Zimbabwe was 0.40, whether the UN Intemation- result, whether made directly, or indirectly through joint al (60 country) or the European Communities Regional comparison with one or several other countries. If the GDP African (17 country) estimates are taken. This is because of country A is twice as high as that of country B, and that the regional 'fixity' was observed in producing the more of country B is 20 percent higher then that of country C, generalized world results. Had an average intemational then the GDP of Country C should be about 41 percent of price system been adopted, the results would have been country A's [if A=2B, B=1.2C, then A=2.4 C]. different. 20. In early rounds of ICP, transitivity was easily 24. obtained by the use of a single set of international prices Thus, the published relationship of overall and per capita GDPs is that prevailing within certain country based on a weighted average of all participating countries. groups on the basis of the weighted average of prices However, the relationship between any two countries prevailing within these country groups. Full adherence to could change with the introduction of more countries in the this approach, however, would prevent many interesting set. In Phase III, the relationships between per capita comparisons between groups. Therefore, a different proce- GDP's of ten member countries of the European Com- dure was adopted for calculating detailed category-wise munities (EC) were found to be significantly different, in expenditures; to evaluate these, a single set of weighted a separate study of EC countries, from those obtained for international prices (calculated by the Geary-Khamis the same countries from a global comparison of 34 method) has been used. This has the advantage that countries. Not only were the ratios different; the country category-wise expenditures (e.g., food expenditures) are with the highest income within the EC comparison was not valued at the same set of prices in all countries; if it were the highest income European country within the global not so, the data would not allow meaningful comparisons comparison. of, say, food consumption, to be made between countries of different regions. However, this means that the sum of 21. Subsequent rounds of the ICP have moved to a expenditures for a given country, valued at global intema- two-stage comparison focussing first on intra-regional tional prices, will not equal its GDP, which is valued at comparisons (EC. 10 and 12, Central Europe, other OECD, regional prices. Asia, Latin America, Africa) which are then linked to each other via bilateral comparisons of "core countries" (e.g., 25. The solutions adopted no doubt constitute a Austria links Central Europe to the EC via Germany, and reasonable compromise between conflicting goals: Japan links Asian countries to the OECD) in a manner that retains relationships from the regional comparisons even * allowing bilateral rankings of GDP to be made on in global comparisons. the basis of bilateral data only, uninfluenced bv 22. Some countries belong to several "regional" prices and consumption patterns prevailing in third groups. For instance, Austria and Finland both belong to countries (bilateral fixity); the Central European group as well as to the overall European and OECD groups. It has been decided that the * allowing the same thing within groups of countries maintaining special relationships (the European publications giving the results of the global comparison Communities OECD, the Asian region, etc.); and those giving the results of the OECD-wide com- (groupwise fixity); parison, should show the relative positions of these two 33 dependent on choices between different methods. The problems of the ICP approach are inherent to basic index * allowing a comparison of overall GDPs between number problems, and they should not preclude increased groups; use of the method for a variety of analytical processes. . allowing category-wise comparisons between and However, the highly technical nature of and the choices, their within groups; and lack of transparency to non-specialists, above all the possibility that different but equally valid choices might . allowing GDP to equal the sum total of its com- yield significantly different results, all this is likely to ponents (additivity). of the ICP 26 Asinlcompromises, some of these goals had constitute a serious obstacle to the adoption methodology for rateoperational decisions. Comparisons 26. As in all based on exchange conversion avoid the index number to be sacrificed. Bilateral fixity was sacrificed to group- problem though at the cost of not correcting for price wise fixity. Additivity was sacrificed to the goals of fixity differences, nor therefore for the quantity differences and category-wise comparability. which may lie behind identical values. 27. These problems, and the solutions given to them, 29. ICP work is costly, and absorbs resources also are sketched out here in their broadest outlines only. What needed in other areas of statistical enquiry, particularly in they illustrate is that even when the coverage of ICP developing countres which see little relevance of ICP to improves, complex methodological choices will continue domestic policy decisions. Preferably, ICP research should to influence the relationships and rankings of GDPs (and on be more closely integrated into regular statistical work other comparative information) indicated by the data. ntiona acounting and domestcar indices This Bilateral comparisons provide the most acceptable would both reduce its costs and increase the priority likely relationship between two countries. They could be linked to be attached to it by developing country governments direct comparison of two crityou ld not yield the facing acute budget constraints.The Bank should continue saersl steidirect comparison of twoconriswol noryeldatinhep to support ICP related research in a modest way. It should same result as the idirect comparison of their relationship also continue to advise membercountries thatparticipation to a third. beneficial over the long run. Such ICP work is levels, in ICP is dif- 28. When counitries are far apart in income investiga , already shedding some valuable light on intercountry t8 heyanberanked withountrs careful statistimmcome ferences between price structure and consumption theycanberankedwithoutcarefulstatsticalivestigation. problems. Once it gets into more general use, it should help Such careful evaluations are needed only when one wishes mostly to explore the relationship between related expen- to compare with each other (and perhaps to pre-determined ditures and concrete results: for instance, the relationship benchmark levels), countries whose positions are so close between the "volume" of education expenditures and to each other that they cannot be distinguished readily. Yet, achievements, of health expenditures and indicators like the ICP approach has not assured us of a way to do this infant mortality. fully objectively, as the ICP rankings of such countries are 34 Annex 4 THE SEARCH FOR BETTER CONVERSION FACTORS 1. Economic analyses are often based on imperfect per capita terms. It also properly represents the relationship concepts, approximately applied and inadequately of the domestic economy to its international linkages: the measured. The search for an all-encompassing concept or burden of foreign debt or the share of foreign trade relative the perfect measure is without end. When domestic ag- to GNP, for instance. Many domestic, intemal relation- gregation is at all meaningful, some of the domestic ships, e.g., the taxation rate, or the share of govemment, or economies' international links will always have to be ex- the overall savings rate can also be well understood at such amined in light of a rate reflecting the relative scarcity of market prices. internationally traded goods. The official foreign exchange rate generally provides an acceptable approximation to 5. As soon as one departs from these ideal conditions, this. none of these relationships and aggregations is perfectly maintained. Indirect taxes cause the marginal transforma- 2. If free trade, competitive conditions, and stable tion rates of producers to differ from the marginal substitu- exchange rates prevailed, and transport costs were relative- tion rates of consumers; quantitative restraints on trade ly low, comparing or aggregating national accounts inter- enhance such differences. At the limit, in a command nationally, by converting them at prevailing exchange economy, where willingness to pay a price does not suffice rates, would not be conceptually different from the com- for a transaction, the very notions of "income" and "price" pilation of national accounts within a country, or their are less than meaningful. Nevertheless, one has come to comparison between regions. Under these ideal condi- accept the market imperfections of most economies, and tions, the market value of each product corresponds to its to use national accounting concepts to measure activities, marginal cost, to the marginal cost of the factors embodied even when these concepts do not fully capture the extent in it, and to the marginal utility it brings to its users. of economic activities. Throughout the economy, a dollar's worth of any product can, at the margin, be transformed into a dollar's worth of 6. Imperfections are much greater in international any other product, and substitute for it, both from the point markets. There is no present altemative to exchange rate of view of producers and of users. Market values cor- conversion for operational comparisons. Nevertheless, for respond to marginal transfomiation rates and to marginal lack of a better method, official exchange rates are general- substitution rates. This is what justifies their use for ag- ly close enough to the rate at which foreign transactions gregation and comparisons. actually take place to constitute acceptable proxies for most purposes. For instance, the ratio of foreign debt or 3. Under such ideal conditions an economy would interest payments to GDP, converted at official exchange need to give up one dollar's worth of its products in order rates are usually considered to be meaningful measures of to acquire one dollar's worth of the other economy's the burden of debt or the effort required to service it. products. Its overall consumer satisfaction would then stay constant if in order to import a dollar's worth of cloth it 7. Yet as trade becomes less free, as the official had to export a dollar's worth of, say, wheat; and to produce exchange rate is less and less closely related to the rate at the additional wheat, it might have to switch resources which intemational transactions actually take place, e.g., from another sector and thus reduce its production and use because of tariffs and quotas, the official exchange rate of some nontraded goods by one dollar. moves farther away from being a reasonable approxima- tion to the transformation rate of domestic values into 4. Under such conditions (and with no transportation foreign values. The point comes when the difference be- costs), converting national currency data at official ex- tween the official exchange rate and the rates effectively change rates reflects the degree of command of a given applied to foreign transactions becomes so large as to economy over world resources, both in aggregate and in render quite meaningless any conversion based directly on 35 Annex 4 8. For analysis, full use must be made of all measure- the official rate. In these egregious cases, an alternative ments and methods of comparison now available, and conversion factor can sometimes be meaningfully es- others must be developed through further research. ICP timated (during FY85-89, such conversion factors have sheds valuable light on the comparative structures of been estimated for about six countries per year, see domestic absorption in a complementary manner. One BoxA4.1.) Such alternatives also permit more meaningful should continue to pursue research into potential uses for calculations, inter alia, of the burden of foreign debt and it, as well as research into methods that would allow to the share of foreign trade, as well as of intercountry com- reduce its costs and to integrate it more into other statistical parisons. In the case of full command economies and, in processes. general, when imports are highly but unevenly restricted, no simple conversion factor may be fully meaningful. by IECSE for Atlas per capita GNP Box A4.1: Countries with conversion factors estimated from the effective transactions rate, and I. Countries for which the official exchange rate differed "egregiously" the split conversion method was used. FY87 FY88 FY89 FY85 FY86 Bolivia Bolivia Somalia Argentina Bolivia Ghana Ghana Ghana Bolivia Somalia Ghana Somalia Somalia Guinea-Bissau Uganda Somalia Uganda real exchange rates have evolved in recent As trade and exchange systems have been liberalized, and have differed "egregiously"from the effective years,fewer andfewer countries' official exchange rates transactionsrate. were estimated from officially recognized multiple II Countries for which trade weighted conversion factors exchange rates. FY87 FY88 FY89 FY85 FY86 Dominican Rep. Dominican Rep. Ecuador Ecuador Dominican Rep El Salvador Ecuador Ecuador Ecuador El Salvador El Salvador Egypt Egypt El Salvador El Salvador Egypt Egypt Guatemala Jamaica Egypt Nicaragua Jamaica Guatemala Guatemala Nicaragua Nicaragua Paraguay Paraguay Nicaragua Jamaica Nicaragua Paraguay Syria Syria Paraguay Syria Paraguay Syria Syria 36 Annex 4 Annex S COUNTRY CLASSIFICATION: INTERNATIONAL ORGANIZATION PRACTICES 1. Country classification is normally by listings, situation, and regions of special Bank development focus especially when institutional membership or geography is (Sub-Saharan Africa). The composition of groups based on involved; or typology, when general analytical objectives per capita income is reviewed annuaUy, and the composi- are pursued. Some international organizations, such as the tion of other analytic groups is reviewed every few years UN, emphasize listings or enumerative classifications. and revised according to evolving structural changes. The Others, notably the Bank and the Fund, combine enumera- present Bank country classification scheme, and its evolu- tion and typology in their countries classificaaion schemes. tion, is discussed more fully in IEC's Statistical Manual. The distinction between enumerative and typological groupings is usually self-evident, but becomes blurred for 5. The Fund's analytic classification distinguishes very broadly defined categories. This point should be kept between countries grouped by: (1) predominant exports, in mind when comparing what appear to be roughly (2) financial criteria, (3) other criteria, and (4) former equivalent country groupings reported by different inter- classification criteria. Classification in the Fund's IFS is national organizations (as in the accompanying table). It is regional; that used in the World Economic Outlook (WEO) also important to an understanding of the staff's proposal is reviewed frequently and revised as countries' economies to deemphasize use of terms that appear to be typological, and financial situations change. Bank and Fund staff con- such as developing or developed countries, but in practice tinue to develop ways to harmonize country classification tend to be defined by enumeration. schemes; it is understood that Fund staff are currently reconsidering their country groups. 2. Most international organizations make some dis- tinction between developed or developing and industrial 6. The United Nations Statistical Office (UNSO) and market economies. However, no organization attempts to the UN Conference on Trade and Development (UN- support the distinction with objective criteria, and CTAD) classification schemes are virtually identical. coverage tends to be defined historically. The distinction UNIDO (not shown in the attached table) follows the same was once broadly correlated with levels of per capita basic classification scheme but further subdivides income. developing economies into three income groups on the basis of their 1978 per capita GDP. Developing countries 3. The country classifications used by major intema- in the high income group range from Chile to Kuwait. The tional organizations have basic similarities. There are some classification of the UNSO and UNCTAD has remained precise country groupings, notably those based on or- quite stable over the years since only partial weight is given ganizational membership; some with relatively clear to income changes. The Least-Developed group, as clas- geographic or analytical objectives, which are usually well sified by the UN General Assembly for example, entails documented, and a few that seem to remain for historical criteria such as proportion of population in the subsistence reasons but are rarely explained. sector, population growth, and agricultural productivity. The General Agreement on Trade and Tariffs (GATT) 4. The Bank's analytic classification is based classification is also similar to that of UNSO. primarily on income. Economies are also grouped accord- ing to majorexports (oil, manufactures), their extemal debt 37 Annex 5 Table AS.1: Country classirications of World Bank and other international organizations UN Statistidcal Office UNCTAD GATT World Bank' IMF Developed Market Economies Developed Market Economies Developed Countries Industrial Market Economies Industrial Countries North America North America North America OECD I (excluding Greece, North America Canada Canada Canada Portugal, and Turkey) Canada USA USA USA USA Europe Europe Europe Europe 2 Communities 2 European Communities European Communites' European European Communites' 2 2 2 EFTA EFTA EFTA (excluding Greece and Gibraltar Portugal) Malta EFTA l Other Europe Other Europe Faeroe sId. Faeroe Isi. Turkey Gibraltar Gibraltar Yugoslavia Malta Africa Africa Africa South Africa South Africa South Africa Asia Asia Asia Asia Israel Israel Japan Japan Japan Japan Oceania Oceania Oceania Oceania Australia Australia Australia AusNZalia New Zealand New Zealand New Zealand New Zcaland Developing Economies Developing Countries Developing Market Economies Developing Market Economies Developing Areas America America Latin America Latin America and Caribbean Western Hemisphere 2 2 LAIA LAIA 2 2 CACM CACM 2 Other (including Cuba) CARICOM Other Europe Europe Europe Europe Yugoslavia Yugoslavia (including Cyprus, Malta Yugoslavia, Greece, Portugal, Turkey, Malta, Hlungary, Poland, and Romania) Africa Africa Africa Middle East and North Africa Africa (including South North North (excluding South Africa) Africa) Other 2 Sub-Saharan Africa Other 2 CEUCA 2 CEUCA 2 South Asia ECOWAS ECOWAS East Asia Rest of Africa (excd. South CEPGL' Africa) Other (excl. South Africa) Asia Asia Middle East and Asia Asia (excluding "Middle Asia Middle East West (excluding Japan, China, and East" but including Oceania) Other Asia South and South-East other Asian CPEs) Oceania Oceania 2 Middle East (including OPEC Major petroleum exporters' High-lincome Oil Exporters "high-income oil exporters") Centrally Planned Economies Socialist Counrries Eastern Trading Area Nonreporting Nonmembers USSR and Other Asia (including China) Asia China and Other Asian CPEs Nonmembers ni .e. Europe and USSR (including Eastem Europe (including Eastem Europe and USSR Hungary, Poland, and Hungary, Poland, and (including Hungary, Romania) Romania) Poland, and Romania) Other Analytical Groups Developing Countries Developing Countries Developing Countries Developing Economies Small low-income countries6 Least developed countries' Least developed countries' Low-income China and India Major exporters of Other low-income manufactures' Middle-income Lower middle-income Income groups Upper middle-income 7 Oil exporting countries Oil exporters' Exporers ot manufactures' Exporters of manufactures' FiCteen heavily indebted Highly indebted countries' countries'" Sub-Saharan Africa' Sub-Saharan Africa" 7. High income developing oil exporters, and Ubys, excluding Bahrain, Brunei, I. See World Development Report 1988, page xi, for detsils. Cameroon, Congo, Ecuador, Egypt, Gaboun, Mexico, Syria, and Trinidad and 2. See the Fund's Directory of Regional Econofmic Organizations and Intergovemmnental Corn- Tobago. modity and Development Orgsnizations for details. 8. Exporters of manufactures and Turkey, excluding Brazil and Prnugal 3. ltigh-income and devetoping oil exporters excluding Cameroon. 9. Exporters of manufactures and Argentina, excluding China, Hungary, India, 4. Other low income exctuding Burma, Ghana, Kenya, Madagascar, Mozambique, Pakistan, Tuvalu, Isrel, Poland, Porugal, and Ronania. Senegal, Sri Lanka, Viet Nam, Zaire, and Zambia, and including Botswana, Kiribati, 10. Highly indebted countries, excluding Cons Rica and Jamaica. Vanuatu, Western Samoa, Yetnen Arab Republic, and Yemen PDR. 11. Sub-Saharan Africa excluding NigeriaL 5. 1980 per capita GDP: above S1500, between $500 and $1,500, below $500. did not 6. Fund member countries whose per capita GDP. as estimated by the World Hank, exceed the equivalent of $4 10 in 1980 (excluding China and India). 38 Annex 5 Box A5.1: Operational guidelines and country classification Operationalguidelines Threshold level 1988 WDR/WDI FY88 1986 per capita GNP (US$) Per capita GNP , 1986 (partiallisting) * Civil works preference less than $425 Low-income * IDA eligibility and Middle-income 20-year IBRD terms $426 -$835 2 Lower middle-income * 17-year IBRD terms $836 -$1,725 3 * IBRD graduation more than $3,010 4Upper middle-income Israel 6,2105 Singapore 7,410 5 Hong Kong 6,910 High-income oil exporters Libya 5,500 6 Saudi Arabia 6,950 6 Bahrain 8,510 Qatar 3,200 Kuwait 13,890 UAE 14,680 Brunei 15,400 Industrial economies Spain 4,860 Ireland 5,070 Threshold level 1987 per capita GNP (US$) FY89 operationalguidelines are asfollows: 1. Less than $480. 2. $481-$940. 3. $941-$1,940. 4. More than $3,385. 5. 1987 per capita GNP: Israel -$6,810; Singapore -$7,940; Hong Kong -$6,910. 6. 1987 per capita GNP: Libya -$5,500; Saudi Arabia- $6,200 (est.). Note: Per capita GNP is the main classification criterion for the operational guidelines and the WDR/WDI. The four operational guidelines lending categories correspond to the WDR/WDI classification for low-income, lower middle-income, and upper middle-income (shown above). High-income countries are not Bank borrowers, so there is no corresponding guideline threshold level. In future, a single high-income category will be presented in the WDR. 39 Box A5.2: Changes in lending categories: FY89 operational guidelines (SecM88-1028) In FY89, fifteen countries moved from one lending category to another: (two moved to harder term categories and thirteen moved to softer term categories) based on 1987 per capita GNP'. These changes, and the percentage above or below the new threshold level are shown below (ranked by percentage above or below the threshold). Lending Lending 1987 per Threshhold Percentage Country category category capita GNP FY89 abovelbelow FY88 FY89 (US$) (US$) threshold Uruguay III IV 2,160 1,941 11 Senegal I II 510 481 6 Guatemala III II 940 940 0 Poland IV III 1,920 1,940 -1 Mexico IV III 1,820 1,940 -6 Indonesia II I 450 480 -6 Congo P.R. III II 880 940 -6 Malaysia IV III 1,790 1,940 -8 Liberia II I 440 480 -8 Solomon Is II I 420 480 -13 Yemen PDR II I 420 480 -13 Gabon V IV 2,750 3,385 -19 Guyana II I 380 480 -21 Fiji IV III 1,510 1,940 -22 Nigeria II I 370 480 -23 1. Lending categoriesare: I - Civil Works Preference; II - IDA Eligibility and 20-Year IBRD Terms; III - 17-Year IBRD Eligibility; IV - 15-Year IBRD Terms; V - IBRD Graduation. 40 Box AS.3: GNP per capita, 1987 Adts methodoogy GNP per capita GNP per capita GNP per capita GNP per capiat Country or territory (US) 1987 Contryor territory (US) 1987 Cosatry or ierritory (US) 1987 Cotryor territory (US) 1987 Low-Income Middle-Income High-Income Data Not Available Ethiopia 120 Cape Verde 500 Spain 6,010 Low-Incom Bhutan 150 Senegal 510 Ireland 6,030 Chad 150 Westem Samoa 560 Saudi Arabia 6,200 Afghanistan Mozambique 150 Bolivia 570 Israel 6,810 Burma Zaire 150 Philippines 590 Singapore 7,940 Guaina Bangladesh 160 Yemen Arab Republic 590 New Zealand 8,230 Dem Guinea-Bissau 160 Zimbabwe 590 Hong Kong' 8,260 Kampuchea, Dem. 2 Lao, PDR 160 Morocco 620 Bahrain 8,510 Vanuatu Nepal 160 Swaziland 700 Greenland' 8,780 Viet Nam Malawi 180 Egypt, Arab Republic of 710 Virgin Islands (U.S.) 1,2 9,760 Burkina Faso 200 Tonga 720 Bahamas 10,320 Middle-Income Madagascar 200 Dominican Republic 730 Italy 10,420 Albania Mali 210 Papua New Guinea 730 United Kingdom 10,430 American Samoa Gambia, The 220 C6te dIvoire 750 Australia 10,900 Angola Tanzania 220 Honduras 810 Belgium 11,360 Bulgaria Burundi 250 Nicaragua 830 Netherlands 11,860 Cuba Zambia 250 Thailand 840 Faeroe Islands' 11,930 Czechoslovakia Niger 260 El Salvador 850 Austria 11,970 Djibouti Uganda 260 Congo, People's Republic 870 Qatar 12,360 Fed. States of Micronesia Sao Tome and Principe 280 Guatemala 950 France 12,860 French Guiana China 290 Cameroon 960 Finland 14,370 Gibraltar Sownalia 290 Jamaica 960 Germany, Federal Republic 14,460 Guadeloupe Togo 290 Paraguay 990 Kuwait 14,870 Guam Benin 300 Botswana 1,030 Denmark 15,010 Iran, Islamic Republic India 300 Ecuador 1,040 Canada 15,080 Iraq Maldives 300 St. Vincent 1,070 Brunei '2' 15,390 Korea, Democratic Republic Rwanda 300 Turkey 1,200 Sweden 15,690 Lebanon Sierra Leone 310 Tunisia 1,210 Japan 15,770 Macao Central African Republic 330 Colombia 1,230 United Arab Emirates 15,770 Marshall Islands Kenya 330 Belize 1,250 Luxembourg 15,860 Maftinique Sudan 330 Chile 1,310 Iceland 16,670 Mongolia Pakistan 350 Grenada 1,340 Norway 17,110 Namibia Haiti 360 St Lucia 1,400 United States 2 18,430 Netherlands Antilles Comoros 370 Peru 1,430 Bermuda 20,410 New Caledonia Lesotho 370 Dominica 1,440 Switzerland 21,250 Reunion Nigeria 370 Mauritius 1,460 Rornania Guyana 380 Jordan 1,540 USSR Ghana 390 Costa Rica 1,550 Sri Lanka 400 Fiji 1,580 High-lncome Solomon Islands 420 SL Kitts and Nevis 1,700 Ahuba Yemen, PDR 420 Malaysia 1,810 Aruba Mauritania 440 Mexico 1,820 Charnel Islands Indonesia 450 Syrian Arab Republic 1,820 French Polynesia Liberia 450 South Africa 1,890 Ieoman Kiribati 480 Poland 1,920 Isle of Man Brazil 2,020 Uruguay 2,160 Hungary 2,240 Panama 2,240 Argentina 2,370 Suriname 2,370 Yugoslavia 2,480 Antigua and Barbuda 2,570 Algeria 2,680 Korea, Republic of 2,690 Gabon 2,760 Portugal 2,810 Seychelles 2,990 Venezuela 3,230 Malta 4,020 Trinidad and Tobago 4,220 Greece 4,350 Cyprus 5,210 Barbados 5,350 Refers to GDP per capita Libya 5,500 Puerto Rico 5,520 Data refer to 1986 Oman 5,830 Annex 5 41 PER CAPITA INCOME PAPER Errata 1. Some curious typographical garbles appear in the last six lines of paragraph 42 of Annex 1 (page2 1). These lines should read: growth rate expressed in 1980 and 1985 prices, respective- ly, and adjusting each for US price changes between the benchmark and target years. Hence, in each case the figures allow for real growth and are inflated (or deflated) by the US inflation rate for the corresponding period, i.e. 1975-83, 1980-83, and 1985-83. 2. On page 8, paragraph 50 should refer to paragraphs 42-44, and paragraph 51 should refer to paragraphs 44-45. 3. On page 41, the footnote to United States should be attributed to Bermuda. January 1989