Policy Resourch | WORKING PAPERS Soclo-Economic Data International Economics Department The World Bank July 1993 WPS 1157 Intertemporal and Interspatial Comparisons of Income The Meaning of Relative Prices Sultan Ahmad Harmonizing methods for comparing prices is essential for improving comparisons of outputs over time and space. And a synthesis of methods would reduce the cost of collecting and disseminating relevant information. Policy Research WorkingPapers disseminate the findings of work in progress and encourage the exchange of ideas among Bank staff and allothers interested in developmenissu Thespspera.disuibutedbytheResearchAdvisory Staffcarrythenames oftheauthors, reflect only theirviews, andshouldbeused and cited accordingly. Thefindings, interpretations, and conclusions antheauthos'own.lhey should not be attributed to the World Bank, its Board of Directors, its management, or any of its member countries. Policy Research| Soclo-Eoonomlc Date WPS 11.57 Thispaper-aproductofthe Socio-Economic Data Division, IntemationalEconomics Department-was presented at the meetings of the Allied Social Sciences Associations in Anaheim, Califomia, January 5- 7, 1993. Copies of this paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Elfrida O'Reilly-Campbeil, room S7-136, extension 33707 (July 1993,20 pages). The conceptual issues confronting compilers of particularly those encompassing economically price indices have not changed much over the heterogeneous areas. years. They include the intractability of basic index-number problems, the practical difficulties The global ICP exercise has provided useful of sampling and matching prices, and the uncer- insights into the issues involved. At the same tainties about the appropriate weighting scheme time, intemational comparisons of the type ICP for comparing events in specific locales over aims to facilitate are now seen as being more time and across locales. sensitive than expected to changes in relative prices. ICP has given little attention to this issue, Ahmad considers inconsistencies in some but there is a rich literature on the subject with measures of time-to-time and place-to-place respect to CPIs. The common ground for the two comparisons of income. He argues for a method logics is essentially national accounts, broadly that harmonizes price work across generally defined. recognized national price compilations, such as consumer price indices (CPIs), the International Through conceptual and practical work done Comparison Programme (ICP), and national by the World Bank on the topic, Ahmad suggests accounting. that harmonizing the various methods is essential to a proper interpretation of the market signals Modem economies tend to be more open, so that prices send to economic agents. He also relative prices should be more similar, but it is explains how a better synthesis reduces the increasingly apparent that price levels and trends overall cost of collecting relevant information can differ considerably even within a nation- and disseminating it to users. ThePolicy Research Working Paper Series disseminates the findings of work under way in the Bank. Anobjectiveof the series is to get these findings out quickly, even if presentations are less than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center Inter-Temporal and Inter-Spatial Comparisons of Income: The Meaning of Relative Prices By Sultan Ahmad Socio-Economic Data Division International Economics Department World Bank Paper presented at Allied Social Sciences Assodiations Meetings Anaheim, California, January 5-7, 1993 Table of Contents I. Introduction ......................... 1 Il. Current status of ICP ................... 1 m. ICP methods in a nutshell. . ......................... 2 IV. The ICP and the World Bank ........................... 4 V. Inconsistency between national accounts and ICP benchmark eFtimates ..... 4 VI. Some evidence of felative prices in ICP ........................ 6 VII. Proposed actions for reducing inconsistencies. 6 A. Harmonization of Price Work ..10 . Representativeness .10 ui. Sampling techniques and Computation of national annual averages . 10 fi. Consistency with national accounts .12 iv. Comparability and quality correction .12 B. Improvement of GDP expenditure data ..13 C. Improving global linkage: the "core commodity" .approach . .13 D. Development of ICP software and dissemination of data . .14 VIII. Concluding remarks .14 Table 1: Discrepancies in extrapolations over actual actual estimates .5 Table 2: Discrepancies in extrapolations over actual estimates. 5 Chart 1 . Chart 2 .8 Chart 3 .9 Annex Table .15 Anne. Table 2 16 Annex Table 3 .17 Annex Table 4 .18 Anne. Table 5 .......................................... 19 Annex Table 6 ........... 20 Inter-Temporal and Inter-Spatial Comparisons of Income: The Meaning of Relative Prices1,2 By Sultan Ahmad Focio-Economic Data Division International Economics Department World Bank I. Introduction 1. The inter-spatial indices of prices and outputs generated by the UN International Comparison Programme (ICP) are not always consistent with the inter-temporal indices compiled by national statistical offices. This paper discusses the issues involved and describes what the World Bank is doing about them. Althowgh the inconsistencies are due to a combbiation of factors ranging from intractable index ixumber problems, problems associated with compiling inter-temporal indices and those related to measurement of inter-spatial indices, the paper concentrates on concepts and practices of ICP. The paper concludes that the inconsistencies would be reduced and the remaining discrepancies better understood if there is harmonization of statistical work related to national accounts, consumer price indices (CPI) and ICP. 2. After brief descriptions of the status of ICP, its methods and World Bank objectives, the paper highlights some inconsistencies, discusses contributing factors and outlines actions designed to deal with them. HI. Current status of ICP 3. The ICP is a UN sponsored program to develop a system of comparing GDP and its components across countries by converting national currency estimates into a common currency, say the US dollar, by purchasing power parity (PPP)3 rates rather than exchange rates. The program involves collecting prices of 400 to 1,000 comparable items, classified into 150 or more basic headings covering the entire I This paper is a revised version of "Improving Inter-Spatial and Inter-Temporal Comparability of National Accounts" presented at the Conference on Data Base of Development Analysis, Yale University, May 15-16, 1992. 2 The author would like to thank Mark R. Rosenzweig of the University of Pennsylvania who reviewed an earlier draft and made helpful comments. While the paper focuses on inconsistencies between inter-spatial and inter-temporal indices of national accounts estimates and ways to improve consistency, Rosenzweig would have liked the paper to include 'details on and empirical or practical consequences" of the inconsistencies. These are legitimate concerns of users of ICP data and should be dealt with, perhaps in another paper. I PPP is defined here as the number of units of a country's currency required to purchase the same amounts of goods and services in the country as one dollar would buy in the United States. 1 gross domestic product (GDP). The price relatives of the basic headings are then aggregated into average price relatives or PPPs for GDP and its various components using GDP expenditure weights. These PPPs are then used as conversion factors which, by equalizing prices in every country, improves international comparability of national accounts as against conversions by exchange rates which do not necessarily reflect relative prices. 4. ICP comparisons are made in two stages: first each region4 makes its own comparisons, and then the regional numbers are linked together to form the global comparison. The ICP surveys, which are conducted about every five years, have covered 90 countries in one phase or another over the six phases from 1970 to 1990 completed so far'. The 1990 survey was confined to OECD and other European countries with the then USSR participating for the first time. However, work is well underway for a global exercise in 1993 which will include comparisons in Africa, Asia, perhaps Latin America also, as well as OECD an! other European countries, with many new prospective participants. Thne most notable prospective newcomers are China and several republics of the Former Soviet Union (FSU). 5. T11e ICP is often viewed especially by some developing countries as an activity imposed upon countries by international organizations and not too relevant for country policy work. But ICP generates a vast number of inte.-nationally comparable observations of prices which span the entire economy, its different geographic regions, urban and rural areas, and formal and informal markets; this price information is potentially very useful for improving indices of prices and outputs, including national accounts, and formulating policies for structural adjustment and resource allocation. The ICP has already become a more or less regular program in the developed countries. The task ahead is to demonstrate to the developing countries the usefulness of the numbers and persuade them to adopt ICP as a part of regular national statistical work. MI. ICP methods in a nutshell 6. ICP moves from the traditional binary comparisons, which are not transitive, to the realm of multilateral comparisons where a premium is paid for transitivity, additivity and base country invariance of the index numbers. For reasons of convenience, ICP concentrates on the expenditure side of GDP rather than on the production side. It collects national annual average prices paid by the final purchaser, takes GDP expenditure data from national accounts and, relying on the identity that expenditure equals price times quantity, obtains estimates of quantity in-licitly at the most detailed level, called basic bh.adings, for which expenditure data are available. Using the Geary-Khamis6 (GK) formula for 4 A region may refer to a geographic region such as "Africa," or a group of countries such as the "OECD." 5 Country coverage of the benchmark sur"eys in the six phases completed so far is as follows: 10 in phase I for 1970, 16 in phase II for 1973, 34 in phase m for 1975, 60 in phase IV for 1980, 64 in phase V for 1985 and 30 in phase VI for 1990. However, none of the 16 mainland Latin American countries that took part in the 1980 survey was present in the 1985 exercise. 6 The Geary khamis aggregation consists of solving a set of (m+n-1) simultaneous equations for where there are m basic headings and n countries, one of which is the base. The system delivers m international prices and n-I PPPs. PPP of a country is the ratio between expenditure in national currency and expenditure in "international prices", while the "international price" of a basic heading is the quantity 2 multilateral aggregation., these implicit quantities of every country are then valued at a single set of average prices, called "international prices", thus establishing a parallel with constant price inter-temporal quant;.y indices. The PPPs are simply the ratios between expenditures in national currencies and the corresponding expenditures in international prices. The PPPs are in effect quantity-weighted average price relatives - or indices - expressed in numbers of units of a country's currency per unit of a base- country currency, say the US dollar. When divided by the prevailing exchange rates, they are indices measuring inter-spatial differences in price levels. 7. The Geary-Khamis is by no means the only aggregation procedure available in the literature; among the alternatives are the EKS, van-Yzeren, Walsh, Implicit Prices, etc.,' each with its strengths and weaknesses. Although the GK satisfies conditions of transitivity, additivity across countries and components of GDP, and base country invariance, it has several serious drawbacks. Because of the simultaneous nature of its solutions, the results between two countries are subject to change if countries are added or deleted from the set; ICP seeks to counter this by using super-country weights where the sum of the weights is arbitrarily made equal to world GDP. Also, +he GK is particularly vulnerable to the so-called (ierschenkron effect, which results in an upward evaluation of the quantities of countries whose price structures differ significantly from the structure represented by average "international" prices.! Since the quantity weighted average prices tend to be closer to the prices of the more affluent countries in the set, the quantity estimates of the less affluent will tend to be relatively high. Although the incidence of Gerschenkron effect has been reduced by regionalizing ICP, the European Communities (EC) prefers to use the EKS method which minimwzes this effect. OECD has also published its official 1990 tables using the EKS method, postponing the publication of the GK results for a later date. However, since the EKS is not additive (i.e., estimates of components would not necessarily add up to totals), the GK is still the preferred method in all other regions and at the United Nations Statistical Division (UNSTAT), the global coordinator of ICP. It may be noted that comparisons of countries within the same region with similar price structures are not affected much by changes in the aggregations method. 8. Regionalization of ICP, by segmenting the work, has improved comparisons with neighboring countries as well as made surveys more manageable. However, this has made comparing countries across regions more difficult as the method of linking regions via binary comparisons of "core countries" in each region has not worked too well in the past. A new method of linkage by "core commodities' will be tried in the 1993 round of comparisons. weighted average prices of countries when the prices have been made commensurate with each other by converting them to the base country currency by the PPP. Thus PPP and "international prices" need to be determined simultaneously. I See Kravis, Kenessey, Heston and Summers, A System of International Comparison of Gross Product and Purchasing Power (Baltimore.JHUP, 1975), chapter 5, for a discussion of various multilateral methods. See I"on -eports of expert group meetings on ICP methodology, Luxembourg: June 6-10, 1988 (by Eurostat), id Es June 14-16, 1989 (by OECD). 8 See Kravis, et. al., A system..., pp. 74, 184. 3 IV. The ICP and the World Bank 9. With respect to ICP, the objectives of World Bank are to: a. maximize country participation in ICP by integrating it with national statistical routines, thereby reducing the cost of the exercise and increasing the poiicy relevance of the numbers; b. refine methods of extending coverage to non-benchmark countries and years, and c. disseminate detailed data in machine readable form and promote research in evaluating the usefulness of the numbers for policy decisions. This paper deals with World Bank approaches related primarily to the first objective, and elaborates on the Larmonization of statistical work designed to improve consistency between national accounts and ICP. The paper also focusses on a number of technical issues that need to be addressed at both the global and codntry levels to enable the objective of integration beirg met. V. Inconsistency between national accounts and ICP benchnark estimates 10. Users of ICP data have long recognized the fact that growth rates implicit in two benchmark year estimates are not always consistent with those obtained from national currency constant price series. In the absence of annual benchmark estimates, the conventional method of extrapolating benchmark estimates (expressed in US dollars) to other years is to use growth rates computed from national currency constant price GDP data and adjust the estimates by US inflation rates to express the values in current dollars. Alternatively, one could use the relative rate of inflation between the country in question and US, as measured by implic.. deflators, to extrapolate the base year PPP to the new year and use the resulting PPP te convert current price national currency GDP data to obtain the estimate of the new year in current US dollars. TRhe results of the two methods of extrapolations would be the same, but extrapolations of ICP estimates of one benchmark year to another will not usually match the actual benchmark results. 11. Table 1 summarizes the extent of discrepancies in the results of extrapolating per capita GDP. For the ICP Phase HI countries, extrapolated from 1975 to 1980, the average discrepancy in the index of GDP (with US = 100) was 11 percent (when the signs of the discrepancies were ignored), ranging from a negative 13 percent to a positive 34 percent. For estimates from Phase IV to Phase V (1980 to 1985), the average discrepancy was lo percent, with a range of a negative 46 percent to a positive 35 percent. In both these cases, the estimates were on average higher than the benchmark data. The range of discrepancies in the case of 1985 data extrapolated to 1990 were lower ( 7 percent) and the range much narrower (-14 to +9 percent) but the countries included in this comparison were also much more homogeneous (only OECD countries) than in the others. 4 Tablt ' Discrepancies in extrapolations over actual estimates ICP GDP per capita, US = 100 1975 to 1980 1980 to 1985 1985 to 1990 Average absolute difference (%) 11 16 7 Range of difference (%) -13 to +34 -46 to +35 -14 to +9 Source: Annex Tables 1, 2 and 3. 12. It must be noted that this inconsistency is by no means confined to ICP data. As Table 2 shows, using the same sets of countries as in Table 1, the inter-country comparison of per capita income by the World Bank's Atlas m-hod suffers from similar inconsistencies: index of values of a given year extrapolated to a target year by real growth rates do not usually match the direct estimates of the target year. This paper, however, concentrates on problems related to ICP. Table 2 Discrepancies in extrapolations over actual estimates Atlas GNP per capita, US = 100 1975 to 1980 1980 to 1985 1985 to 1990 Average absolute difference (%) 18 47 30 Range of difference (%) -30 to +94 0 to +99 0 to -43 Source: Annex Tables 4, 5 and 6. 13. Both ICP and Atlas estimates reflect current year data, with its own brand of relative price indices (i.e., the conversion factors: PPPs in ICP and average exchange rates in Atlas) appropriate for the year; an attempt to impose a fixed price structure of a base year to other years will indeed cause discrepancies in cases where there have been significant changes in relative price structures. Similarly, in inter- country comparisons of the ICP type, since the estimates are based on the average price structure of the countries being compared, one should take a critical look at estimates of countries whose price structures are significantly different from the average. For analysts, ICP provides a vast wealth of price and expenditure data which would permit them to recognize significant differences in relative prices both within and acroFs countries and guard them against drawing unwarranted X nclusions. For the compilers of ICP data, an examination of the detailed data would point up areas of exceptional errors and inconsistency in matching and sampling of items and should trigger corrective action to minimize such errors. The following section presents some evidence of the kind of problems the analysts and compilers should be aware of. S VI. Some evidence of relative prices in ICP 14. Chart I provides two pictures which coirpare respectively the relative price structure of Turkey and Austria vis-a-vis USA for 1985. For each of the 152 basic headings (which sum up to GDP), the price relative was divided by the average overall price relative (PPP) for GDP as a whole. Thus points above or below the horizontal line at 1.0 signifies higher or lower average price of the basic heading than the overall price level in the country, compared wit: USA. The curve for Turkey shows a much wider range of variation than for Austria signifying that analysts should be more confident about comparative levels of per capita income for Austria and USA than for Turkey and USA as measured by these data. For compilers, there would be much more scope for correcting errors of mismatch of items in rurt-ey than in Austria. 15. Chart 2 presents a comparison of real quantities of USSR and Hungary in 1985 and 1990. TLe 1985 data were drawn from a study by the Council of Mutual Economic Assistance vCMEA) in Moscow while the 1990 data were taken from ICP surveys coordinated by Austria. For each of the major headings, the real quantity estimate of USSR is divided by the corresponding estimate for Hungary. This is done separately for 1985 and 1990 and the ratio of the two (1990/1985) (in percentage terms) is plotted in Chart 2. The chart shows that comparative structures of real quantities between Hungary and USSR were very different in 1985 and 1990. Most analysts would find it hard to believe that this indeed was the case. Again the compilers should look for possible areas of mismatch of items and errors in relative prices. 16. Chart 3 presents a worst case scenario. It plots the price relative (ruble to schilling) of individual items divided by the estimated overall PPP (ruble to schilling) for 1990. The chart shows that individual item prices in USSR are sometimes sixty times higher and sometimes 30 times lower than in Austria. The standard deviation of the array of price ratios is 184 percent! While some the extreme values may be true, the widespread nature of the variation point to the difficulty of determining what constitutes market prices in a command economy and how to compare items that are not available in each other's markets. Again this shows opportunities for compilers to improve the comparisons and for analysts to be on guard about making conclusions based on these data. 17. Work in progress at the World Bank shows that national annual average prices smooth out vast differences in prices in different regions of the country or in markets types, especially for countries with economically heterogenous areas. Detailed price data generated by ICP for rural or urban areas or for formal or informal markets are expected to provide analysts with opportunities to analyze differences within as well as across countries and help formulate structural adjustment policies. VII. Proposed actions for reducing inconsistencies 18. There is a certain amount of incomparability among the phases of ICP resulting from changes in country coverage, weighting pattern, selection and pricing of items, formulas for computing basic heading price relatives and aggregations, introduction of regionalization and the method of regional linkage. Impact of these issues is likely to be reduced as country coverage becomes more or less complete and ICP methods and practices, especially with respect to inter-regional linkage, become stabie. 19. Some of these discrepancies are inevitable, however. Conceptually, inter-temporal quantity indices keep prices constant and let quantities vary. In ICP, as both quantities and prices are allowed to 6 CHART 1 RelI'*Ive Turlksh prices for 1IS commodity groups Tutkeoy-l!SA I~~~~~~~~~~~~S4 ll f129II 3 4 I 1 9II 172~ ii29 33 37 41 424 3197 61 4.9 69 72 7 SI 83 899 9$9 191 lO 0 I l I 1234 1 3 9 4 4 4 It7 15 19 23 7 It 33 39 43 47 SI 33 39 43 67 7 73 29 3 87 91939991 1 47 I I 119 123 117 931 133 In 143 147 131 Relative Austrian prices for 152 commodity groups Austria-USA 10 0.1 3 9 I) 17 i, 2 29 ,, 3*741 4 49 .3 69 7 77 * .9 9 97 191 190 109 113 117 121 123 a29 153 137 141 143 169 3 7 9 II 19 I "232337.1 33 394471 ,3 7 9 . a 771 73 79 83 7 91 . 1939 I7 1 1 119 123 1 31 1313 I" 19 1417 IN -7- Ouantities 1 2 3 4 5 6 7 Hungary-Austria: difference between 1985&1990 Group]] Compauson USSR Hungary Hungary RatioU/H RatioU/H DifferenceDifference 11ungamy-AuAria relative quantities in 1985=100 ALsUia.100 1990 1990 1985 1990 19854toS 2to3 3 1 Bread&cereals 86 39.9 91.2 215.5 103 209.3 43.8 2 Meat 45.3 60.8 82.3 74.5 76.5 97.4 739 g Fish 200.8 20.8 26 965.4 434 222.4 80.0 4 Milk,Eceese&eggs 81.9 67.6 77 121.2 92.3 131.3 87.S s Oils&fats 41.5 41.4 59.2 100.2 74.4 134.7 69.9 a Frui,vegies&potaos 41.5 50.4 52.3 82.3 123.1 66.9 96.- 7 Other food 26 35.5 54.9 73.2 81.9 89.4 64.7 aNon4albdolicbevages 10.9 26.5 40.2 41.1 25.1 163.9 65.9 9 Alcdholic beveraes 53.4 67.9 54.1 78.6 79.5 98.9 125 ,, 0 Tobacco 51.2 81.4 62.9 62.9 68.6 91.7 129A .T4 7i 4 tl CltOting 40.5 17.5 29.3 231.4 115.8 199.9 59.7 s9o / I 12 Footwear 49.8 22.5 46.7 221.3 157.5 140.5 48.2 A [ 7 f0t 13 Reais 10.7 37 50 28.9 85.3 33.9 74.0 -* 14 hld&power 36.9 81.9 S0 45.1 59.2 76.1 12.4 15 Furniture 20.5 19.9 31.3 103.0 80.8 127.5 63.6 . ie Household textiles 31.7 39.9 44.1 79.4 33.8 235.1 90.5 17 Appliances 12.2 36.7 51.7 33.2 35 95.0 71.0 isOierhbsddgoods&serv 58.4 44.5 72.6 131.2 70.4 186.4 61 ' 19 Medied care 65.5 95.3 67.7 68.7 138 49.8 140.S 20 Trasport eq 2.9 13.4 16 21.6 59.4 36A 83.8 21 Tranport opetion 9.9 32.7 18.2 30.3 22.9 132.2 179.7 22 Purchased tansp seice 100.4 45.6 64.4 220.2 185.8 1185 70.8 s.. . . . ... ... . co 23 Conmunication 28.1 18.2 15.7 154.4 96.8 159.5 115.9 k G.FbO&O =% O, 3. .AM 24 Reauional eq 13A 19.2 37.8 69.8 49.9 139.9 S0.8 f _. 25 Recreaiionl&altural se 26.2 96.6 63.8 27.1 42.8 63.4 151.4 I 25 Printing mterals 87.8 90.4 63.1 97.1 182.7 332 143. USSR-Hungay: difference between 1985&1990990 27 Educatn 123.9 82.3 73.9 150.5 94 160.2 111.4 USSR/Hungay rative quantyes in 85= 100 ga3 28 Retaurantsels 25.4 19.3 36.9 131.6 132.8 99.1 523 i -__ 29 Other goodsservices 29.9 32.4 53.7 92.3 55.5 166.3 60.3 1 so Govertmd a_umptio 132.4 51.7 64.6 256.1 216.4 1183 80.0 3s Residmdal buildings 73.1 43.5 48.4 168.0 149.3 112.6 89.9 32 Non-residetial buldings 120.2 33.1 41.2 363.1 324.6 111.9 803 33 Othler COsuction 117.3 42 562 279.3 125.9 221.8 74.7 34 TranSport equipeut 42.9 4.1 26.1 1046.3 605.5 172.8 15.7 35 Nowaelecrical machney 30.6 13.1 29.4 233.6 193.7 120.6 44.6 2 27 36 Electiicl nmchinery 17.6 4.5 21.4 391.1 142.7 274.1 21.0 9 GrossDomestcProdud 46.6 40.7 47.2 114.5 121.9 93.9 86.2 ' 7 ' Mm vW4.kT.dA..-. I 5 C.0.-- T* .9. M_ V, I h F_~iW 4~ ~ of. OnqS _7 _a sou a- Relative Soviet prices of 903 commodities (Austria= 1.00) 100 10 0.1 0.01 I I I I I I I I I I L 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 Preliminary ICP results for Group II countries, 1990 Standard deviation 184. 1% vary from one benchmark to another, the vector of international prices differs from one benchmark to another. In this sense, as mentioned earlier, comparing two bench mark estimates is like comparing two current price values; naturally the growth rate computed from them is not expected to match the one computed from constant price values. We discuss below some of the other factors that may be responsible for the discrepancies, and propose actions to reduce them where possible - through harmonization of price work, improvement in GDP expenditure data, and better documentation, accessibility and dissemination of data. A. Harmonization of Price Work 20. The most important reason for the discrepancy is that the prices observed in ICP are often not consistent with those underlying national accounts, resulting in estimates of implicit quantities which are different from those underlying national accounts. Since prices are the crux of the ICP exercise, we discuss at some length the steps needed to harmonize the price work in national accounts, CPI and ICP. 21. As mentioned earlier, PPPs are weighted averages of price relatives expressed in different currencies. The requirements of price data for ICP are: a. The items priced and the prices collected must be representative of the expenditure pattern of the country concerned. b. The sampl of prices must be amenable to computation of correct estimates of national annual averages. C. The annual average prices should be consisten with national accounts, so that both SNA and ICP estimates represent the same physical quantities embodied in the national accounts aggregates. d. The items priced should be as comparable as possible across countries. 22. Since it is not possible to fully satisfy each condition, the art of ICP is choosing the right trade- offs among them. In the Bank's view, too little attention has been given to the first three requirements. It is possible that comparability, or correct alignment of item prices across countries, is the key element in well developed economies where market prices prevail. However, that is clearly not the case when one is dealing with economies where price controls are prevalent and there is only limited diversity of goods and services. It is this issue of emphasis, perhaps, that distinguishes the Bank's approach to ICP from others. i. Representativeness 23. If the items selected for ICP are not representative of the country concerned, the country will not be enthusiastic about collecting the prices which will have little policy relevance. Also, the cost of collecting representative prices will be less partly because existing price collection mechanisms can be used or readily extended but also because it is inherently harder to find unrepresentative items. 24. The prices should be representative of a country's expenditure pattern both in the selection of specifications and in the sampling of markets and outlets. The guiding principle in ICP is to price items that are volume sellers in quality and packaging, and choose markets and outlets that account for most 10 expenditures on the item. Most nationally compiled price indices, notably the CPI, follow this principle, and the first place that ICP should look for a list of representative items would be the lists of specifications included in these indices. However, the CPI list, which covers consumption, may refer to a particular location (capital city) or a target group of consumers Oow income families), and may require additions and deletions to make it representative of the entire GDP expenditure which is the focus of ICP. Similarly for the other indices such as the WPI (wholesale price index), construction price index, labor cost index, etc. - items selected for ICP should have a maximum representation of the country as a whole rather than specific location or group of people. 25. It is unlikely, however, that all the items needed for ICP can be found in the lists of specifications of CPI plus other existing index number programs. Since ICP seeks to measure place-to-place differences in prices, it is much more concerned about detailed identification of item characteristics than national price indices that are mainly used to measure time-to-time changes. For measuring time-to-time price changes, countries may use sampling techniques that allow price collectors to choose specific items for each surveyed outlet without much concern about item specifications beyond continuous reference to the same items/outlet mix overtime. This does not mean that the outcome is less representative nor that the items chosen are less comparable. The documentation necessary to answer such questions has been stripped from the prices. One step in the harmonization process will be to develop ways for a sample of price collectors to feedback fidler descriptions of all items priced in various places in order to create a pool of specifications from which ICP items could be selected. ii. Sampling techniques and Computation of national annual averages 26. In ICP, national annual average prices are applied to national annual expenditures to obtain estimates of quantities implicitly. It is difficult, if not impossible, to compute national averages if the CPI is based entirely on outlet specific items without any mechanism to put them on a comparable basis. One recourse to ensuring this is the quality correction factors discussed later. Another would be to require each outlet to price some nationally established specifications under each basic heading of consumption - a set of national "core commodities", so to say - for linking outlets and markets within the country. We propose to work with experts from the US Bureau of Labor Statistics (BLS) to delve deeper into the sampling issue. 27. Since the sample of prices is likely to be different in different countries, and expenditure weights at the level of individual items are not usually available, mismatching of weights to items is quite possible; in which case comparisons based on national averages may be misleading. As a minimum, there seems to be a strong case for analyses of how much such higher-level averages are influenced by the imputations needed to link available price and expenditure details. This suggests that it may be better to compute price relatives by types of outlets and markets first and then average tie price relatives, rather than average the prices first and then compute the price relatives. A second step in the harmonization process will be for countries to retain as much of the details of prices by geographic regions, fornal or informal markets, rural or urban areas, etc. in order to make it possible to experiment with various options. 28. As ICP and CPI indices both use expenditure weights, a third step would be for countries to adopt uniform expenditure classifications of the type proposed by ICP and stratify their CPl price samples accordingly so as to make the components directly comparable across countries. 11 iii. Consistency with national accounts 29. As mentioned earlier, ICP observes prices and estimates quantities implicitly using the basic accounting identity: Expenditure = Quantity x Price. The expenditures are those estimated under the SNA (with the specific and fairly mechanical reclassification of collective expenditures on health and education). If the average prices are not consistent with the prices used in national accounting, the quantity comparisons will be inaccurate. Thus, for instance, if national accounts are based on controlled prices rather than market prices, the ICP prices should also be the controlled prices. This is because although the quantity comparisons are made via the PPP, it is the comparison of real quantities which is the ultimate goal of ICP. 30. In many countries the CPI is often used for linking current and constant price estimates of various components of consumption. A fourth step in the harmonization proces.. would be to establish mechanisms of collecting prices which would serve the needs for CPI, national accounts as well as ICP. The Bank has worked in Nigeria to incorporate as many of ICP items as possible into the country's CPI, and in Bangladesh to help set up a price collection mechanism based on ICP work to develop constant price deflators. 31. Since countries usually price a different sub-set of items included in a basic heading, the average price relative for the basic heading can only represent what is priced. For instance, for rice, one country may price mostly long grain rice sold loose in a predominantly informal market, and another may price mostly "basmati" rice in 1-kg bags sold in city supermarkets. The implicit quantities derived from these prices would thus embody a potentially different quality mix, even though each country's prices may be consistent with their national accounts. This kind of inconsistency can be reduced by improving comparability which is discussed below. Iv. Comparability and quality correction 32. For the purposes of ICP, items must be comparable across countries; truly unique products ? id services are not usable in ICP, because by definition there will be nothing with which to compare them. And the items priced should be comparable in unit, quality, packaging and conditions of sale. For example, the price of a kilogram of peanuts sold in one-ounce packages in stadiums in one country would not be comparable to the price of a kilogram of peanuts sold in kilogram-cans in super markets in another. 33. But a robust PPP requires a larger number of matching items. The more rigid the determination of "uniqueness," the less robust the PPP. Hence, ICP experts often make simple conversions, say taking half the price of two ounce-packages of peanuts in one country as comparators for one-ounce packages elsewhere. 34. However, easily compared items, such as American brand-named processed cheese, are simply not representative for many countries. It is necessary to strike an adequate balance between the conflicting requirements of representativeness and comparability. Preliminary investigation shows that it is difficult to find exact cross-country matches of items in CPI baskets; the items are usually described very broadly, leaving out differences in essential elements of quality which bear on prices. At the same time, the comparable items that are implicitly the bridges between representative items, with traditional ICP methods, cannot carry much weight when they refer to specialty goods (perhaps for countries on each side of the bridge) that may be subject to different national pricing regimes (e.g. not controlled like 12 representative items). A further step In the harmonization process will be to develop quality correction factors to ensure a closer match with regularly collected prices and minimize the need for specifying endrely new Items requiring special price collection. 35. The East European countries, participating in ICP through binary comparisons with Austria, were the first to recognize the need for quality correction and to develop a mechanism for implementing it. The Austrian Central Statistical Office (ACSO) has assumed a leadership role in this area. Although a potato is a potato at some level of abstraction, it may differ from market to market in packaging, return facility, associated services, and the like, all of which would add to price but which would be viewed as an element of quantity rather than price if corrections for these quality aspects are not made. The task is to minimize the element of arbitrariness in the decisions and to agree in advance how the corrections will be made. We believe some quality correction mechanism should be established for all regions. At the same time, we are beginning to explore (with colleagues at the US Bureau of Labor Statistics as well as the ACSO) better ways to catalog the "characteristicity" of items to be priced with the immediate goal being better identification of the inference process underlying the country- product dummy (CPD) and similar gap-filling methods. 36. The harmonization process cannot be completed overnight and will require the concerted effort of international organizations as well as national governments. We propose to make case studies of the harmonization process with some selected countries from Africa, Asia and Latin America for the purpose of developing guidelines for implementation. B. Improvement of GDP expenditure data 37. In most countries consumption, which is the largest component of GDP expenditure, is usually estimated as a residual and ICP details at the basic heading level is derived on an ad hoc basis. Understandably, countries concentrate on the production side to develop their basic national accounts; but time has come for countries to pay more attention to the expenditure side of national accounts which are needed in dealing with poverty alleviation and structural adjustment policy issues. 38. One action the World Bank has taken in this respect is to make desk estimates of )P expenditure at the same level of detail as required by ICP for every country (regardless of ICP participation) and pass them on to countries to improve upon them and maintain them on a regular basis. It is hoped that this will trigger an interest in countries to make the estimates of expenditure directly from price and quantity observations, improving the quality of national accounts and at the same time making it easier for them to supply the same prices to ICP as they use in expenditure estimates. This World Bank exercise is expected to be completed in 1992. C. Improving global linkage: the "core commodity" approach 39. Since each regional comparison is based on its own average price structure, they need to be linked together globally to allow comparisons across the regions. In the last two phases, the linking was done by the method of "core countries", in which several countries from each region, designated as "core countries", collected additional prices to match those in other regions. Binary comparisons between countries in different regions provided linking factors with which regional results were adjusted. The process did not work too well because of the lack of an adequate number of countries volunteering to take up the additional burden of work. Also the choice of a different core country would give a different link as it would generate different detailed bi-lateral PPPs, raising questions about advisability of using a 13 "core" country because it is willing to be one rather than that it is optimal. We are now supporting the core commodity approach in which a list of "core" items will be priced in every country. These items will provide the links between regions. The process will distribute the burden of work to all participants and would not be viewed as a separate exercise. The details of this proposal are still being discussed. D. Development of ICP software and dissemination of data 40. The ICP generates average prices of about 400 to 800 items per country. Each of these national averages is based on sometimes hundreds of observations in various markets and outlets all over the country - urban or rural, formal or informal. All this detail needs to be captured and be available to users. Two hurdles have to be overcome: countries should have the computer hardware and software to process these data themselves, and the shroud of confidentiality should be lifted from as much of this information as possible. The World Bank is committed to developing a PC-based portable software, and working with the countries and various international organizations to increase the amount of data that should be in the public domain. The Bank is collaborating with EUROCOST in Luxembourg to develop ICP-software which is expected to be ready for the next round of surveys in 1993. We have also begun making the published data available to users in diskettes using our *STARS* software. Also, to promote discussion of technical issues and minimize international travel, we are developing an ICP electronic mail network, elements of which are already working. 41. In order to facilitate speedy processing of data and bring them closer to country policy makers, there should be regional processing centers for each region. Eurostat has been doing the work for Europe, Africa and some countries in the Caribbean. The ESCAP seems to be ready to do the Asian comparison. ECLAC supported by IDB is expected to take on the responsibility for Latin America. Austria has been coordinating the work in Eastern European countries and the OECD has been linking its member countries with those of Europe. The UNSTAT will coordinate the global comparison. Participating countries would be expected to work closely with the regional processing centers as well as the UNSTAT. The regional processing centers should be fully operational by 1993 and the first set of results should begin to flow in 1994. VY. Concluding remarks 42. The ICP's objective of developing a system of comparing national accounts internationally on a consistent basis has been more or less adopted by the developed countries (OECD, EC); the work outlined here is expected to lead the developing countries in that direction. We believe it is feasible as well as desirable to harmonize the price work for national accounting, CPI and ICP, and to improve expenditure details which will have the added benefit of increasing consistency between various inter- temporal measures and the inter-spatial measures provided by ICP. 43. The World Bank is working in support of UNSTAT and various national and international organizations to integrate ICP with national statistical routines and develop the necessary infra-structure (computing facilities, communication network) to facilitate its implementation. The base year for the next phase of ICP will be 1993 and the various organizations, working closely together, should be able to bring out the first estimates in 1994. We are particularly interested in the work of Austria with the East European countries because how these countries cope with the transition is of immense interest to policy makers and provide an empirical basis to the development of economic thinking. 14 ICP 1975/1980 ANNEX TABLE: 1 Cw1parison of Extrapolated and Actual Per Capita GDP in InternationaL Dollars Based on PPP Conversion, 1980 Actual Actual Extrapolated Extrapolated Actual Actual Difference SI US=100 SI US=100 St US100 t%) Col.4 1975 1975 1980 1980 1980 1980 over col.6 1 2 3 4 5 6 7 1 Austria 4,995 69.6 8,520 73.8 8,625 75.3 -2 2 Belgiun 5,574 77.7 9,320 80.7 9,436 82.4 -2 3 Brazil 1,811 25.2 3,220 27.9 3,349 29.3 -5 4 CotwTbia 1,609 22.4 2,710 23.5 2,838 24.8 -5 5 Denmark 5,911 82.4 9,560 82.8 9,831 85.9 -4 6 France 5,877 81.9 9,700 84.0 9,780 85.4 -2 7 Germany 5,953 83.0 10,160 88.0 10,200 89.1 -1 8 Hungary 3,559 49.6 5,970 51.7 4,632 40.5 28 9 India 470 6.5 710 6.1 570 5.0 23 10 Ireland 3,049 42.5 5,150 44.6 5,480 47.9 -7 11 Italy 3,861 53.8 6,920 59.9 7,788 68.0 -12 12 Japan 4,907 68.4 8,500 73.6 8,414 73.5 0 13 Kenya 470 6.5 760 6.6 637 5.6 18 14 Korea, Rep. 1,484 20.7 2,890 25.0 2,583 22.6 11 15 Luxefbourg 5,883 82.0 9,410 81.5 10,626 92.8 -12 16 Mslawi 352 4.9 560 4.8 415 3.6 34 17 Netherlards 5,397 75.2 8,580 74.3 9,316 81.4 -9 18 Pakistan 590 8.2 990 8.6 1,097 9.6 -11 19 Philippines 946 13.2 1,640 14.2 1,740 15.2 -7 20 Spain 4,010 55.9 6,010 52.0 6,353 55.5 -6 21 Sri Lanka 668 9.3 1,140 9.9 1,226 10.7 -8 22 United Kingdem 4,588 63.9 7,260 62.9 8,253 72.1 -13 23 United States 7,176 100.0 11,550 100.0 11,447 100.0 0 24 Urusuay 2,844 39.6 4,970 43.0 4,259 37.2 16 25 Yugoslavia 2,591 36.1 4,830 41.8 4,042 35.3 18 26 Zsmbia 738 10.3 930 8.1 730 6.4 26 Average 3 Abs. average 11 Rarne -13 to +34 Source: 1975 and 1980 actuals are fran ICP Phase III and Phase IV reports respectively. 1980 estimated values are derived by extrapoLating 1975 actuals by per capita GDP growth rates corpited fran constant price national accounts data and scalirg up the results by US inflation measured by implicit GDP deflator. - 15 - ICP 1980/1985 ANNEX TABLE: 2 Comparison of Extrapolated and Actual Per Capita GDP in International Dollars Based on PPP Conversion, 1985 .... .... ............................................. ........................................................................... ............................ Actual Actual Extrapolated Extrapolated Actual Actual Difference Si 1U=100 ST US=100 SI US=100 (X) Col.4 1980 1980 1985 1985 1985 1985 over col.6 1 2 3 4 5 6 7 1 Austria 8,630 75 11,780 73 10,900 66 11 2 Belgiun 9,440 82 12,550 78 10,670 65 21 3 .otswana 1,590 14 2,790 17 2,670 16 7 4Cameroon 910 8 1,420 9 2,720 16 -46 5 Canada 11,620 101 16,400 102 15,260 93 11 6 Cote d'Ivoire 1,370 12 1,350 8 1,710 10 -19 7 Denmark 9,830 86 14,350 90 12,230 74 21 8 Ethiopia 280 2 320 2 300 2 10 9 Finland 8,640 75 12,450 78 I ;6er 69 12 10 France 9,780 85 13,200 82 11,430 69 19 11 Genmynv 10,200 89 13,910 87 12,170 74 18 12 Greece 5,100 45 6,760 42 5,860 36 19 13 Hong Kong ',170 63 11,140 69 10,200 62 12 14 Hungary 4,630 40 6,520 41 5,140 31 30 15 India 570 5 850 5 750 5 17 16 Ireland 5,480 48 7,650 48 6,740 41 17 17 Italy 7,790 68 10,560 66 10,830 66 0 18 Japan 8,410 73 12,520 78 11,800 72 9 19 Kerya 640 6 770 5 990 6 -20 20 Korea, Rep. 2,580 23 4,640 29 3,980 24 20 21 Luxembourg 10,630 93 15,330 96 13,420 81 18 Z2 Madagascar 570 5 590 4 630 4 -4 23 Malawi 420 4 500 3 630 4 -18 24 MaLi 340 3 390 2 360 2 11 25 Morocco 1,200 10 1,590 10 2,370 14 -31 26 NetherLands 9,320 81 12,250 76 11,260 68 12 27 Nigeria 890 8 860 5 980 6 -10 28 Norway 11,330 99 16,840 105 13,910 84 25 29 Pakistan 1,100 10 1,680 10 1,340 8 29 30 Philippines 1,740 15 1,850 12 1,790 11 6 31 Poland 4,320 38 5,320 33 4,040 24 35 32 Portugal 3,830 33 4,990 31 5,570 34 -8 33 Senegal 690 6 880 5 1,330 8 -32 34 Spain 6,350 55 8,460 53 7,590 46 15 35 Sri Lanka 1,230 11 1,890 12 1,850 11 5 36 Tanzania 360 3 410 3 420 3 0 37 Tunisia 1,990 17 2,760 17 3,450 21 -18 38 United Kingduon 8,250 72 11,580 72 10,900 66 9 39 United States 11,450 100 16,030 100 16,490 100 0 40 Yugoslavia 4,040 35 5,090 32 4,810 29 9 41 Zambia 730 6 800 5 940 6 -12 42 Zimfbebwe 890 8 1,170 7 1,680 10 -28 Average 4 Abs. average 16 Rarge -46 to +35 Source: 1980 and 1985 actuals are fron ICP Phase IV and Phase V (global) reports respectively. 1985 estimated values are derived by extrapolatirg 1980 actuaLs by per capita GDP growth rates ccWputed from constant price national accounts data and scaling up the results by US inflation measured by implicit GDP deflator. - 16 - ICP 1985/1990 ANNEX TABLE: 3 Ccmperison of Extrapolated ard Actual Per Capita GDP in Intermationat Dollars Based on PPP Conversion, 1990 .. .............. ............................ ..... ........ ......... ................. ...... .......... . ................. Actual Actual Extrapolated Extrapolated ActuaL Actual Difference Sl US=100 SI US=100 Sl US100 (%) Col.4 1985 1985 1990 1990 1990 1990 over col.6 . 1 2 3 4 5 6 7 ~~.. ... .. ... .. ... . .. . . . . . .. . . . . . .. . . . . . . . .. . . . .. . . .. . . .. .. . . . . . . . 1 AustraLia 11,720 71 14,430 67 16,000 75 -10 2 Austria 10.900 66 14,610 68 16,520 77 -12 3 BeLgium 10,670 65 14,520 68 16,350 76 -11 4 Canada 15,260 93 19,700 92 19,190 89 2 5 Deruerk 12,230 74 15,420 72 16,790 78 -8 6 Finlard 11,460 69 15,670 73 16,490 77 -5 7 France 11,430 69 15,340 71 17,380 81 -12 8 Germany 12,170 74 16,240 75 18,220 85 -11 9 Greece 5,860 36 7,350 34 7,330 34 -0 10 IreLand 6,740 41 9,910 46 10,620 50 -7 11 Italy 10,830 66 14,640 68 15,950 74 -8 12 Japan 11,800 72 16,950 79 17,660 82 -4 13 Luxembourg 13,420 81 18,930 88 19,290 90 -2 14 Netherlands 11,260 68 14,650 68 15,710 73 -7 15 New ZeaLand 10,050 61 11,670 54 13,560 63 -14 16 Norway 13,910 84 16,800 78 16,050 75 4 17 Portugal 5,570 34 7,990 37 8,750 41 -9 18 Spain 7,590 46 10,970 51 11,740 55 -7 19 Sweden 12,680 77 16,090 75 17,040 79 *6 20 Turkey 3,600 22 5,060 24 4,630 22 9 21 United Kingdom 10,900 66 14,790 69 15,880 74 -7 22 United States 16,490 100 21,510 100 21,450 100 0 Average -6 Abs. average 7 Rarne -14 to +9 . .............. ............................ . ........ ............ .......... .................... Source: 1985 and 1990 actuats are fron ICP Phase V (GlobaL) and Phase VI (OEM) reports respectiveLy. 1990 estimated vaLues are derived by extrapoLating 1985 actuaLs by per capita GDP growth rates conputed fram constant price nationaL accounts data and scalire up the resuLts by US inflation measured by inpLicit GDP deflator. - 17 - ATLAS 1975/1980 ANNEX TABLE: 4 Caoparison of Extrapolated ard Actual Per Capita GNP in US Dol Lars Based on Atlas Conversion, 1980 ............................................................................................................... Actual ActuaL ExtrapoLated Extrapolated ActuaL Actual Difference USS US=100 USS USC100 USS US=100 (%) CoL.4 1975 1975 1980 1980 1980 1980 over col.6 1 2 3 4 S 6 7 1 Austria 4,730 63.9 8,070 67.8 10,000 83.2 -19 2 Belgiun 5,930 80.1 9,910 83.2 12,160 101.2 -18 3 Brazil 1,07D 14.5 1,900 16.0 2,070 17.2 -7 4 Colombia 550 7.4 930 7.8 1,200 10.0 -22 5 Dennark 6,900 93.2 11,160 93.7 13,150 109.4 -14 6 France 5,990 80.9 9,890 83.0 11,890 98.9 -16 7 Germany 6,666 90.1 11,370 95.5 13,270 110.4 -14 8 Hungary NA NA NA NA NA NA NA 9 India 160 2.2 240 2.0 240 2.0 1 10 Ireland 2,640 35.7 4,460 37.4 5,060 42.1 -11 11 Italy 3,690 49.9 6,610 55.5 7,500 62.4 -11 12 Japan 4,520 61.1 7,830 65.7 9,840 81.9 -20 13 Kenya 230 3.1 370 3.1 420 3.5 -11 14 Korea, Rep. 580 7.8 1,130 9.5 1,620 13.5 -30 15 Luxwatourg 7,460 100.8 11,940 100.3 14,940 124.3 -19 16 Malawi 120 1.6 190 1.6 180 1.5 7 17 Netherlands 6,000 81.1 9,540 80.1 12,030 100.1 -20 18 Pakistan 130 1.8 220 1.8 290 2.4 -23 19 Philippines 340 4.6 590 5.0 650 5.4 -8 20 Spain 2,770 37.4 4,150 34.8 5,370 44.7 -22 21 Sri Lanka 290 3.9 500 4.2 260 2.2 94 22 United Kingdomn 3,900 52.7 6,170 51.8 7,980 66.4 -22 23 United States 7,400 100.0 11,910 100.0 12,020 100.0 0 24 Uruguay 1,330 18.0 2,330 19.6 2,720 22.6 -14 25 Yugoslavia 1,380 18.6 2,570 21.6 3,250 27.0 -20 26 Zambia 550 7.4 690 5.8 600 5.0 16 ......... ...................................................................------....................----------------------------------------..-...----- .... ............ . Average -9 Abs. Average 18 Range -30 + 94 ........................................ .................. ......... .. Source: Actual data for 1975 and 1980 are fron current World Bank fiLes. 1975 Atlas vaLues are extrapolated to 1980 by per capita GDP growth rates cowputed from constant price national accounts and scaling tp by US inflation. 1 - 18 - Atlas 1980/1985 ANNEX TABLE: 5 Camparison of Extrapolated and Actual Per Capita GNP In LS DoLLars Based on Atlas Conversion, 1985 ............................................................................................................... Actual ActuaL Extrapolated Extrapolated Actual ActuaL Difference USS USL100 USS USaloG USS US=100 (X) Col.4 1980 1980 1985 1985 1985 1985 over coL.6 1 2 3 4 5 6 7 ..................................... ................................................... ...................... 1 Austria 10,000 83 13,660 81 9,100 54 51 2 BeLgiuu 12,160 101 16,170 96 8,280 49 96 3 Botswana 870 7 1,530 9 1,050 6 46 4Cameroon 760 6 1,190 7 820 5 46 5 Canada 10,610 88 14,980 89 14,230 84 6 6 Cote d'lvoire 1,160 10 1,150 7 660 4 75 7 Dennmrk 13,150 109 19,200 114 11,380 67 69 8 Ethiopia 120 1 130 1 110 1 19 9 Finland 10,130 84 14,590 87 11,040 65 33 10 France 11,890 99 16,050 95 9,810 58 64 11 Gernny 13,270 110 18,090 107 10,916 65 66 12 Greece 4,380 36 5,810 35 3,700 22 57 13 Horg Kong 5,220 43 8,110 48 6,120 36 33 14 Hureary 1,930 16 2,720 16 1,940 11 41 15 India 240 2 360 2 290 2 25 16 IreLand 5,060 42 7,060 42 4,940 29 43 17 ItaLy 7,500 62 10,170 60 7,750 46 32 18 Japan 9,840 82 14,640 87 11,430 63 28 19 Kenya 420 3 500 3 310 2 62 20 Korea, Rep. 1,620 13 2,910 17 2,340 14 25 21 Luxebourg 14,940 124 21,550 128 14,080 83 54 22 Madagascar 430 4 440 3 310 2 42 23 Malawi 180 1 220 1 170 1 30 24 Mali 240 2 280 2 160 1 76 25 Morocco 930 8 1,230 7 620 4 99 26 Netherlands, 12,030 100 15,820 94 9,420 56 68 27 Nigeria 920 8 880 5 850 5 4 28 Norway 12,900 107 19,190 114 14,560 86 32 29 Pakistan 290 2 440 3 370 2 19 30 PhiLippines 650 5 690 4 540 3 28 31 PoLand NA NA NA NA NA NA NA 32 PortugaL 2,380 20 3,100 18 1,980 12 57 33 Senegal 510 4 650 4 380 2 72 34 Spain 5,370 45 7,150 42 4,360 26 64 35 Sri Lanca 260 2 400 2 390 2 3 36 Tanzania NA NA NA NA NA NA NA 37 Tunisia 1,280 11 1,770 11 1,180 7 50 38 United Kirn9dn 7,980 66 11,200 67 8,500 50 32 39 United States 12,020 100 16,830 100 16,880 100 0 40 Yugoslavia 3,250 27 4,090 24 2,060 12 99 41 Zaffbia 600 5 650 4 370 2 76 42 Zintabwe 710 6 930 6 640 4 46 Average 47 Abs. average 47 Rarne 0 to +99 . . .... .... . .. ... . . ..... ........ .... ...... .............................. ..... Source: 1980 and 1985 actuaLs are fron latest WoroLd Bank files. 1985 estimated values are derived by extrapolatirg 1980 actuals by per capita GDP growth rates ccopzted fron constant price national accounts data and scalirg up the results by US inflation nwasured by inplicit GDP deflator. - 19 - Atlas 1985/1990 ANNEX TABLE: 6 Cwaparison of Extrapolated and Actual Per Capita GNP in I'S Dollars Based on Atlas CCrwersion, 1990 .. ........ ............ .. ........... ................... ........... ..................... ... ... ................... Actual ActuaL ExtrapoLated Extrapolated ActuaL A.Wual Difference USS ULS100 US LSO0 US$ US 100 (X) Col.4 1985 1985 1990 1990 1990 1990 over col.6 * 1 2 3 4 5 6 7 1 Australia 11,760 70 14,480 66 16,560 76 -13 2 Austria 9,100 54 12,190 55 19,C0 87 -36 3 BeLgiun 8,280 49 11,260 51 17,580 81 -37 4 Canada 14,230 84 18,370 83 20,380 93 -11 5 Oernerk 11,380 67 14,350 65 22,440 103 -37 6 Finland 11,040 65 15,100 69 24,540 113 -39 7 France 9,810 58 13,170 60 19,590 90 -33 8 Genrany 10,920 65 14,570 66 22,363 103 -35 9 Greece 3,700 22 4,640 21 6,010 28 -24 10 Ireland 4,940 29 7,270 33 10,370 48 -31 11 Italy 7,750 46 10,470 48 16,880 77 -39 12 Japan 11,430 68 16,420 75 25,840 118 -37 13 Luxenbourg 14,080 83 19,860 90 29,010 133 -32 14 Netherlands 9,420 56 12,260 56 17,570 81 -31 15 Now ZeaLand 6,910 41 8,020 36 12,570 58 -37 16 Norway 14,560 86 17,590 80 22,830 105 -24 17 Portal 1,980 12 2,840 13 4,950 23 -43 18 Spain 4,360 26 6,300 29 11,010 50 -43 19 Sweden 12,020 71 15,260 69 23,780 109 -36 20 Turkey 1,080 6 1,520 7 1,640 8 -8 21 United Kingdan 8,500 50 11,530 52 16,080 74 -29 22 LUnited States 16,880 100 22,020 100 21,810 100 0 Average -30 Abs. average 30 Rarne 0 to -43 ........................................…... Source: 1985 and 1990 actuals are from latest World Bank files. 1990 estimated values are derived by extrapolatirg 1985 actuats by per capita GDP growth rates com.puted fran constant price nationaL accouts data and scaling up the results by US inflation measured by inplicit GDP deflator. - 20 - Policy Research Working Paper Series Contact Tlit Author Date for paper WPS1 137 Railway Reform In the Central and Philip W. Blackshaw May 1993 TWUTD Eastern European Economies Louis S. Thompson 31005 WPS1 138 The Economic Impact of Military Daniel Landau May 1993 C. Jones Expenditures 37699 WPS1139 Should Sub-Saharan Africa Expand Jonathan R. Coleman May 1993 G. liogon Cotton Exports? M. Elton Thigpen 33732 WPSI 140 How Retail Food Markets Responded Bruce Gardner May 1993 C. Spooner to Price Liberalization in Russia Karen M. Brooks 32116 after January 1992 WPS1141 Foreign Direct Investment in a MaxwellJ. Fry May 1993 R. Vo Macroeconomic Framework: Finance, 31047 Efficiency, Incentives, and Distortions WPS1 142 Pent-Seeking Trade Policy: A Time- Martin Rama May 1993 D. Ballantyne Series Approach 37947 WPS1 143 Tariff Rates, Tariff Revenue, aaid Lant Pritchett May 1993 M. Fernandez Tariff Reform: Some New Facts Geeta Sethi 33766 WPS1 144 The Foreign Trade Dimension of the Bartlomiej Kaminski June 1993 P. Kokila Market Transition in Poland: The 33716 Surprising Export Performance and Its Sustainability WPS1145 The Simple(r) Algebra of Pension Dimitri Vittas June 1993 P. Infainte Plans 37664 WPS1146 Is Poverty Increasing in the Shaohua Chen June 1993 P. Cook Developing World? Gaurav Datt 33902 Martin Ravallion WPS1 147 Interest Rates, Growth, and External Stijn Claessens June 1993 R. Vo Debt: The Macroeconomic Impact of Daniel Oks 33722 Maxico's Brady Deal Sweder van Wijnbergen WPS1 148 Economic Instability and Aggregate Robert S. Pindyck June 1993 S. Moussa Investment Andres Solimano 39019 WPS1 149 How Labor Markets and imperfect Martin Rama June 1993 D. Ballantyne Competition Affect Tariff Policy 37947 WPS1150 Wealthier is Healthier Lant Pritchett June 1993 WDR Lawrence H. Summers 31393 Policy Research Working Paphr Series Contact Title Author Date for paper WPS1151 Is Growth Bad for thQ Environment? Charles van Marrewlik July 1993 J. Verbeek Pollution, Abatement, and Federick van der Ploeg 33935 Endogenous Growth Jos Verbeek WPS1152 Population, Health, aid Nutrhion: Denise Vaillancourt July 1993 0. Nadora Annual Operational Review for Fiscal Stacye Brown 31091 1992 and Others WPS1 153 North American Free Trade Alberto Musalem July 1993 P. nf ante Agreement: Issues on Trade in Dimitri Vitas 37664 Financial Services for Mexico Asli DemirgOl-Kunt WPS1 154 Options for Pension Reform in Tunisia Dimitri Vittas July 1993 P. Infante 37664 WPS1 155 The Regulation and Structure of Martin F. Grace July 1993 P. Infante Nonlife Insurance in the United M; nael M. Barth 37664 States WPS1 156 Tropical Timber Trade Policies: What Panayotis N. Varangis July 1993 D. Gustafson Impact Will Eco-Labeling Have? Carlos A. Primo Braga 33714 Kenji Takeuchi WPSI 157 Intertemporal and Interspatial Sultan Ahmad July 1993 E. O-Reilly- Comparisons of Income: The Meaning Campbell of Relative Prices 33707