98846 ESTIMATING PER CAPITA INCOME FOR OPERATIONAL PuRPOSES 1 Introduction ...................................................................................................................................... 2 2 Problems Measuring Income in National Currencies ....................................................................... 3 2.2 Differences in International Standards .................................................................................... 4 2.3 Differences in Coverage and Quality of National Estimates ................................................... 4 3 Choosing a Conversion Factor .......................................................................................................... 5 3.1 Using Exchange Rates as Conversion Factors ........................................................................ 5 3.2 Using Purchasing Power Parities as Conversion Factors ........................................................ 6 4 The World Bank Approach to Country Comparisons for Operational Purposes .............................. 8 4.1 Estimating GNI in National Currency ..................................................................................... 8 4.2 Estimating Population ............................................................................................................. 8 4.3 Using the Atlas Conversion Factor ......................................................................................... 9 4.4 Measuring International Inflation .......................................................................................... 10 5 Recommendations and Conclusions ............................................................................................... 11 AnnexA Current Atlas Methodology .............................................................................................. 13 AnnexB Synthetic Atlas Conversion Factor ................................................................................... 15 AnnexC Modified Atlas Methodology based on U.S. Inflation Rates ............................................ 16 AnnexO Effects on Country Classification of Using Alternative Atlas Methodology ................... 17 AnnexE Effects on Country Classification of Using PPP-Based GNI Estimates ........................... 21 AnnexF Participants in ICP and Related Exercises, 1975-96 ........................................................ 26 This report provides an overview of the methodological and data issues involved in estimating per capita income for operational purposes. The main conclusion is that the Bank should retain its current Atlas methodology for estimating internationally comparable per capita income for operational purposes. The procedure of updating income thresholds based on international inflation rates should also be continued. Because international comparability of income per capita is affected by differences in quality of national statistics, priority should be placed on enhancing statistical capacity in developing countries. Once national statistics improve, the use of PPP-based GNI estimates should be reconsidered. 1 Introduction The purpose of World Bank OP 3.10, "Loan Charges, Currencies, and Payment Terms of ffiRD Loans and IDA Credits," is to ensure that the Bank sets its lending terms in a clear and transparent manner and provides the neediest borrowers with the most attractive loan terms. Under OP 3.10, an indicator of economic wellbeing is used to rank countries, and thresholds are established to determine the conditions applied to borrowing member countries. Since 1984 the methodology used to establish these rankings has been reviewed every five years. This report represents the 2000 assessment of the Bank's methodology. Historically, countries have been ranked by per capita gross national income (GNI) (previously called per capita GNP) 1, converted into U.S. dollars using the World Bank Atlas conversion facto~. These rankings are important because members classified as low-income countries receive more preferential lending terms (longer grace and repayment periods, lower fees, and lower interest rates) than those countries classified as middle-income countries. The Bank has identified the need to reduce the number of poor people in the world as its most important global objective. The methods used to achieve this goal are being scrutinized in a broader context in order to identify policies that will prove most efficient in a rapidly changing environment. While overall prosperity and human progress depend on long-term growth, the dimensions of development extend well beyond narrow notions of average per capita income and its rate of accumulation. Per capita income serves as a proxy for individuals' potential command over resources that enhance their well-being, but it does not indicate how well income is shared within the community. This raises an 1 In 1993 the Commission of the European Communities, the International Monetary Fund, The Organization for Economic Cooperation and Development. the United Nations, and the World Bank revised the System of National Accounts. One of the changes from A System of National Accounts 1968 to its successor System of National Accounts 1993, was the replacement of the term GNP (Gross National Product) with GNI. 2 See annex A. 2 important question: Is the use of average per capita income as a major operational criterion for determining the Bank's lending strategy appropriate? Or should it be complimented by the use of social indicators, such as those selected for the international development goals? This report recognizes the need for a broader discussion of how the World Bank can best allocate its resources to fulfill its mission, but limits it scope to an analysis of the best way to measure internationally comparable per capita income. In compiling internationally comparable income numbers, three conceptual, methodological, and practical issues arise, namely: what is the best measure (in national currency) of a country's income, how can that measure best be converted into a common currency, and is the methodology used to generate the data simple and transparent? Each of these issues is examined in this report. 2 Problems Measuring Income in National Currencies The use of GNI as a basis for estimating comparable per capita income measures is accepted by all international organizations, and has been used by the Bank for many years. GNI measures the income generated by a nation's residents from international and domestic activity. It is preferred to Gross Domestic Product (GDP), which measures income generated from domestic activity by residents of the economy as well as by non- residents3. A common and valid argument against ranking countries using per capita GNI is that it does not take into account factors that affect a country's ability to sustain or increase its current income level. These factors include demographic characteristics, such as differences in age or gender composition; whether income is used for consumption or capital formation; and the evolution of a country's natural resources. Using per capita income also fails to take account of the distribution of income. In view of the Bank's focus on targeting its activities to help the poor directly, a better measure might be the number of people living below a given poverty level, such as the widely quoted "dollar a day" criterion. Regrettably, the income distribution and PPP data required to compile such an estimate with confidence and in a timely manner are not available at a sufficiently high quality to justify its use in determining operational lending criteria. In general, the more factors taken into consideration in determining lending criteria, the less transparent the methodology will be. Moreover, the likelihood of creating internationally comparable and reliable data diminishes. For these reasons, in practice, GNI represents the best measure for estimating internationally comparable income. This conclusion seems to have been reached by several multilateral organizations, including 3 GNI equals GOP plus net compensation of employees and property income from abroad (GOP plus net factor income from abroad in 1968 SNA terminology). 3 the United Nations and the European Union, which use per capita GNI as the major criterion for operational purposes such as calculating membership contributions. International guidelines for compiling national accounts are designed to yield internationally comparable income measures by including production of all goods and services (with a few exceptions) as income-generating activities, whether or not they are produced for the market, for own use, or provided to others free of charge. In particular, informal, illegal, and subsistence activities should all be included. In practice, however, international comparability is hampered by the use of different vintages of international standards and by differences in compilation practices, in coverage, and in quality of the estimates. The latter is the most severe problem, and may result in systematic biases in the income estimates; it is generally thought that income in developing countries is underestimated compared to that of developed countries. 2.2 Differences in International Standards Many countries are currently in the process of updating their national accounts compilation systems to reflect new international guidelines, resulting in the existing estimates not being entirely comparable. Several countries have already implemented the System of National Accounts 1993 (1993 SNA) while some still base their estimates on the principles laid out in the 1968 guidelines. There are some minor differences between the definition of GDP and GNI in the 1968 SNA and in the 1993 SNA. However, analyses carried out for some of the countries that have adopted 1993 SNA indicate that the change in levels of GDP and GNI has been rather small (a one to two percentage point increase). The long-standing problem of international comparability due to differences between the Material Product System and the SNA has disappeared, as all transition economies are now following the 1993 SNA. 2.3 Differences in Coverage and Quality of National Estimates The lack of adequate source data and estimation methods for measuring informal, illegal, and subsistence activities generates severe problems. Several countries try to include explicit estimates, or adjustments, for informal and subsistence activities; some have also tried to obtain estimates of illegal production. Due to poor or unavailable source data, these estimates are generally very approximate, however, and the procedures for estimating these parts of the economy vary substantially from country to country. For some developing countries, information on subsistence activities is available from household or agricultural surveys conducted every 5-10 years. These estimates provide a base for establishing a benchmark level for these activities, but information on how to update them is typically lacking. These data problems introduce systematic biases in income figures. Because monetization and market orientation rise with the level of development, subsistence and informal activities constitute a larger share of the total economy in developing countries than they 4 do in industrial economies. As a result, income levels tend to be underestimated less in industrial countries than in developing countries. Capturing the nonmonetized and hidden economy has allowed several countries to revise their income estimates upward. Zimbabwe, for example, increased its estimates of GDP (and GNI) by some 20 percent in 1998 after it improved its estimates for subsistence farming and the hidden economy in general. Ethiopia is in the process of revising its GDP upward by about 50 percent to reflect new estimates of the nonmonetized economy based on a recent household survey. In several low-income countries with particularly weak statistical systems, even coverage of formal activities is problematic, because of difficulties in establishing and maintaining comprehensive business registers and low response rates on surveys without appropriate imputation for nonrespondents. In some countries, in fact, undercoverage of formal sector activities may be as important as undercoverage of activities in the informal economy. Coverage of formal activities is generally not a major problem in countries using advanced statistical methods. Transition economies have experienced their own set of problems. In the beginning of the transition period most of the data reporting systems in the transition countries did not operate effectively, and most of the countries in transition have developed growing informal and partly illegal sectors, which are difficult to measure. The transition economies are in the process of building statistical data collection systems that are better tailored to the needs of a market economy. But the process is slow and costly. Although data coverage is improving, it appears that several countries have been unable to make retroactive adjustments. As a result, their measures of economic growth may be distorted by changes in coverage over time. Due to the weaknesses in their statistical system, GNI may well be underestimated for many transition economies. However, several transition countries incorporate adjustments for what they call the "shadow" economy. Furthermore, substantial problems in adjusting the measure of output for holding gains on changes in inventories can result in an overestimation for some transition countries. So the issue is far from clear-cut. 3 Choosing a Conversion Factor To make cross-country comparisons of income levels possible, income must be converted to a common currency, such as the U.S. dollar. Exchange rates and purchasing power parities (PPPs) of the International Comparison Programme (ICP) offer two alternatives for conversion. PPPs have typically been used for analytical purposes, while exchange rates have remained the basis for comparing income levels for operational purposes. 3.1 Using Exchange Rates as Conversion Factors Because they are observable and generally available for most countries, exchange rates are commonly used as conversion factors. Conversion using exchange rates can be 5 misleading, however. The use of exchange rates as conversion factors assumes that relative prices on goods and services are about the same for all countries. If free trade, stable exchange rates, low transportation costs, full capital mobility, and no market imperfections existed, the use of exchange rates to convert income measures from a national to a common currency would generate comparable estimates across countries. Market imperfections exist, however, and many services are not traded across national boundaries. According to the purchasing power theory, changes in the exchange rates are caused by changes in relative price levels between countries. In practice, however, exchange rates are affected by capital flows, speculation, and interventions by governments and central banks. Thus what one U.S. dollar buys in the United States does not necessarily correspond to the amount of goods and services that one U.S. dollar converted into another country's currency buys in that country. A potential problem in exchange rate conversion is the effect of abrupt changes in the exchange rate. By applying a moving average exchange rate rather than a single year's exchange rate, the effect of sudden changes can be dampened. The Atlas conversion factor is an average of the past three years' exchange rates, with the earlier years' rates adjusted to reflect the ratio of domestic to international rates of inflation. 3.2 Using Purchasing Power Parities as Conversion Factors PPPs measure the relative purchasing power of different currencies over equivalent goods and services. PPP is the number of units of a national currency required to purchase the same amount of goods and services as a numeraire currency unit (for example, one U.S. dollar) would buy in the numeraire country. By correcting for differences in relative prices, PPPs allow meaningful comparisons to be made across countries. Ideally, the use of PPPs eliminates the inconsistencies in valuation inherent in exchange rate conversions, though concerns remain about methodology, quality, timeliness, and geographic coverage. PPPs have been estimated for different groups of economies under the ICP. The basic ICP approach is to collect detailed data on prices and GDP by type of expenditure from all participating economies for individual goods and services. The implicit volumes obtained from the price and expenditure data are then explicitly or implicitly revalued at a uniform set of average international prices denominated in a numeraire currency. The resulting values based on international prices can be considered internationally comparable. The ratio of the original national currency value to the corresponding value in international prices is a PPP. PPPs can be derived for various national accounting categories at different levels of aggregation, up to the level of GDP. Equivalently, the volume revalued at international prices can be obtained by converting the national currency value at PPP. PPPs tend to be more stable than exchange rates, which, as noted, are more sensitive to international capital movements, currency speculation, and direct government intervention. PPP-based data thus provide a more robust basis for comparing prices and quantities for economies at comparable levels of development. Furthermore, PPPs and PPP-based estimates at disaggregated levels provide relevant information for analysis and 6 policy work, particularly on subsectors of the economy or subgroups of the population. PPPs for basic products, for example, which are consumed mostly by the poor, can be used for poverty analysis, including comparison of real incomes of the poor across countries. The Bank uses PPP-based data extensively in poverty analysis. Such data have not been used for operational purposes, however, because of concerns about conceptual and methodological issues, comparability of price data across countries, gaps in geographical coverage, and timeliness. Conceptually, the valuation of different countries' national accounts at uniform international prices eliminates the systemic differences in price structures of economies at different stages of development. However, estimates of PPP conversion factors are sensitive to the choice of aggregation procedure. PPP-converted income differences between poorer and richer economies are much wider under one aggregation method than under another, and the ranking of economies by income level may also change. In collecting price data, it is difficult to match ICP specifications (including specifications for quality) for individual products in different countries. This problem is especially serious for products that are not internationally traded. Typically, the prices of personal, medical, educational, and other services are much lower relative to those of other products in low-income countries. These prices are used in the ICP without adjustment or with subjective adjustments for substantial differences in service quality. Ideally, PPPs should be computed based on prices of products that are representative for the country concerned and comparable across countries. But representative products for individual countries, particularly low-income countries, may not be internationally comparable, and internationally comparable products may not be representative for individual countries. Some of these problems can be mitigated by estimation techniques that make regional comparisons first and then align them through "bridge" countries. Another problem involves lack of coverage. Country coverage of the ICP survey is almost complete for high-income countries, but it remains incomplete and irregular for low- and middle-income economies, even though the number of participating economies has generally increased over time (34 countries participated in 1975, 60 in 1980, 64 in 1985, 118 in 1993/96 (annex F)). For countries that did not conduct a survey in a given year and for the years in which the ICP survey was not conducted, current estimates are extrapolated from the most recent survey available. For countries that have not participated in an ICP survey in the last 15 years, estimates are based on regressions. Using PPP-converted income for ranking and classifying countries would also require estimations of PPP-based income thresholds. One option would be to set income thresholds so that the smallest number of countries were reclassified when switching from the Atlas methodology. The thresholds could then be annually updated applying U.S. inflation rates4 • As the classification structure would change every five years or so (when 4 U.S. inflation rates would be used because PPP-based income for the years between benchmark years are extrapolated with real growth adjusted for U.S. inflation. 7 new benchmark data became available), income thresholds would be re-estimated following the same simple procedure. To compare the two alternatives, the results of Operational Guidelines FYOl were compared with the same GNI data converted into U.S. dollars using available PPP estimates. Differences in ranking and classification of countries were significant (annex E). However, concerns about timeliness, coverage, and methodology suggest that using PPP-based income estimates for operational purposes is not recommended at this stage. 4 The World Bank Approach to Country Comparisons for Operational Purposes The World Bank uses Atlas per capita GNI estimates as a transparent and readily available indicator for operational purposes, such as determining country eligibility for preferential credit. It updates income thresholds annually based on international inflation. Measures other than income are also used to make operational decisions. In deciding which countries are eligible for IDA lending terms, for example, relative poverty, lack of creditworthiness, and other special factors are taken into consideration. Exceptions are also made for countries such as small island economies with relatively high per capita incomes but without access to international funding because of their small size. 4.1 Estimating GNI in National Cu"ency Estimates of GNI are provided to the Development Data Group (DECDG) by the Bank's country economists or analysts, either on the basis of preliminary estimates from the national statistical offices or, when official data are not available, as best estimates by the country economists. The quality of the income data in national currencies underpins the use of per capita income in U.S. dollars for operational purposes. Intensive checking for consistency and validity of sources is therefore conducted in order to obtain the best estimates possible. For seven out of 143 countries in the FYOl Operational Guidelines (a decrease over previous years), there was no basis on which to derive reliable estimates of GNI. It has been the Bank's practice to not show the per capita GNI for these countries. 4.2 Estimating Population Population estimates are compiled jointly by the Human Development Network and DECDG, in consultation with operational staff. These estimates are usually based on censuses, but the frequency and quality of the censuses vary by country. Most countries conduct a complete enumeration no more than once a decade. Population estimates are thus typically extrapolations based on demographic models. 8 4.3 Using the Atlas Conversion Factor The Atlas conversion factor, based on official or market exchange rates, is used to convert income estimates from national currency to U.S. dollars (annex A). Special methods are used for countries where multiple exchange rate systems exist, the real exchange rate changes substantially, or market exchange rates are lacking. These methods are described below. The Atlas conversion factor is the simple arithmetic average of the current exchange rate and the exchange rate in the previous two years adjusted for the ratio of domestic to international inflation. The SDR deflator is used to represent international inflation, compiled from inflation measures in the G5 countries (Germany, France, Japan, the United Kingdom, and the United States). The change in the GDP deflator will, in the future, be used as the measure of domestic inflation. This represents a slight change, as the GNP deflator, in theory, was applied in the past. This change is made for two reasons. First, what is needed in estimating the Atlas conversion factor is a broad measure of inflation that captures the changes in prices that, according to purchasing power theories, lead to change in the nominal exchange rate. The strong correlation among the GDP deflator, the GNP deflator, and the GNI deflator means that they have roughly the same ability to predict exchange rate movements5 • Second, per capita GNI figures are needed in U.S. dollars less than four months after the end of the year, and the GDP deflator is available sooner than the GNP deflator. The Atlas method lessens the effect of fluctuations and abrupt changes in the exchange rate, which can be heavily affected by capital flows. Income measures converted using the Atlas conversion factor tend to be more stable over time, and changes in income rankings are more likely to reflect changes in relative economic performance than exchange rate fluctuations. 4.3.1 Estimating Conversion Factors in Countries with Multiple Exchange Rate Systems For countries with multiple exchange rate systems, the Bank's practice has been to apply a weighted average of the exchange rates, with shares of foreign exchange transactions subject to the different exchange rates used as weights. The weighted-average exchange rate is then used in the Atlas formula. 4.3.2 Estimating Conversion Factors in Countries with Exchange Rate Distortions To identify countries with exchange rate distortions, the Bank reviews real exchange rate movements. The objective is to identify distortions resulting from speculation, severe 5 The implicit GNP deflator is a World Bank created measure, and not a SNA concept. The difference between the GNP and the GNI deflators is that the GNI deflator takes into account the effect of changes in the terms of trade. 9 trade or foreign exchange restrictions, domestic price controls or policy interventions. The review is carried out by comparing the change in the nominal exchange rate with the change in the relative rate of inflation (the ratio of domestic to international inflation). If real appreciation or depreciation exceeds 30 percent over three years, an alternative conversion factor is applied. Typically, an alternative conversion factor is estimated by extrapolating a "normal" period's exchange rate with the change in relative inflation. Each case must be judged on its merits, and there is always an unavoidable element of professional judgment. 4.3.3 Estimating Conversion Factors in Countries without Market-Based Exchange Rates Much of the international trade of transition economies has been carried out at exchange rates that differ markedly from official rates. In these cases, official exchange rates cannot be used to convert GNI from local currency to U.S. dollars. The alternative approach used for market economies of selecting a past exchange rate as a basis for forward extrapolations is not feasible, given the history of managed price and exchange rate regimes in these countries. In view of these problems, an alternative method, known as the synthetic Atlas-type conversion factor, has been developed (annex B)6. In theory the synthetic Atlas-type conversion factor and the official exchange rates should converge as the transition economies become more open and market oriented. Lack of such convergence may reflect the fact that some transition economies have not yet become sufficiently market oriented. It may also indicate that the synthetic Atlas-type conversion factors were based on unreliable data. The synthetic Atlas-type conversion factors were used in operational guidelines for the first time in FY92, when they were used to convert per capita income for 16 transition economies. In FY98 official exchange rates replaced the synthetic Atlas-type conversion factor for 11 of these countries (Armenia, Azerbaijan, Estonia, Georgia, Kazakhstan, the Kyrgyz Republic, Latvia, Lithuania, Moldova, the Russian Federation, and Ukraine). Two more countries, Macedonia and Turkmenistan, moved to official exchange rates in FYO 1. The synthetic Atlas-type conversion factor continues to be used in the remaining three countries, Belarus, Tajikistan, and Uzbekistan. 4.4 Measuring International Inflation A measure of international inflation is needed to apply the Atlas methodology and update the income thresholds. The change in the SDR deflator expressed in U.S. dollar terms is chosen for both purposes. The SDR deflator is calculated as a weighted-average of the GOP deflators of the G5- countries in SDR terms, where the weights are determined by the amount of each 6 See: SecM93-589: ''Estimating dollar per capita income for the states of the Former Soviet Union," June 15, 1993. 10 currency included in one SDR unit and the exchange rate from local currencies to SDRs7• The SDR deflator is then converted into U.S. dollars using a three-year inflation-adjusted average exchange rate. Using U.S. rather than international inflation rates in the Atlas methodology and in updating income thresholds would improve transparency. But a recent review suggests that international inflation rates generate slightly better results. In most circumstances using U.S. inflation rates in the Atlas methodology would not change country classifications, as long as the income thresholds are updated to reflect U.S. inflation. To compare the two alternatives, the results of Operational Guidelines FYOl were compared to the same GNI data converted into U.S. dollars using the Atlas methodology with U.S. inflation rates (annexes C and D). No changes in rankings or classification were evident. If, however, the dollar were volatile, the SDR-deflator might predict the coming year's exchange rate more accurately than using U.S. inflation. Replacing international inflation rates with U.S. rates would result in a simpler, more transparent method, but possibly at the cost of some accuracy. Retaining the Bank's current practice of using international inflation in the Atlas methodology thus continues to make sense. 5 Recommendations and Conclusions For operational purposes the Bank should retain the Atlas methodology to derive per capita GNI estimates. Priority should be placed on enhancing statistical capacity in developing countries as the most effective means of improving the quality and comparability of national income statistics. Because current PPP-based per capita income estimates are not available for all countries and methods of extrapolations over time and space continue to suffer from the lack of recent survey data, a strong case cannot be made for using PPPs rather than the Atlas conversion factor. Furthermore, because the current Atlas methodology tends to yield more accurate estimates than the modified Atlas method based on the U.S. GOP deflator (annex C), the current Atlas method should be retained. Divergent national accounting practices, differing statistical capacity, and resource constraints across countries are probably the greatest limitation to intercountry comparisons of per capita GNI. Despite extended technical assistance to improve the available statistics in low-income countries, the quality of the data, especially in Africa, appears to be declining. The 1993 SNA specifies that all subsistence, informal, and illegal activities should be included in the measure of production, but lack of source statistics in many countries makes it difficult to do so. There are positive signs, however. Although 1 The composition of the SDR deflator changed slightly in 1999, as a consequence of the introduction of the euro. 11 some transition economies still have far to go before their statistical systems are adequate, many have come a long way, and both the coverage and the quality of the statistics are improving. PPP-based statistics, such as PPP-converted income estimates, income distribution data, and the "dollar a day" measure of extreme poverty, are of particular importance to the Bank. But such statistics are meaningful only if the underlying price data are timely and reliable. Efforts are needed to strengthen price data collection. Toward that end, the Bank has supported the ICP since the program's inception in 1968. In addition to funding the program, the Bank has supported research on ICP methodology, as well as the collection and processing of basic data. In recent years, the Bank has focused on integrating ICP surveys into national statistical work as part of its efforts to strengthen national statistical capacity. For the future, priority should be given to statistical capacity building, and to the comparison work carried out by the international organizations. Such work is needed before PPP and income distribution statistics can be used extensively in operational work. 12 AnnexA Current Atlas Methodology The Atlas conversion factor in year t is given by: ATLAS ei,t ={ ei,t + ei,t-1 [ I Pi.t Pi.t-1 ) SDR($)1 SDR($) + ei,t-2 [ Pi.t Pi.t-2 SDR($)1 SDR($) I )~ I J Pt Pr-J Pt Pr-2 where e(jTLAS is the Atlas conversion factor for country i in year t, eu is the exchange rate from national currency to U.S. dollar for country i in year t , Pi.t is the GOP deflator for country i in year t , and pfDR($).is the international deflator (SOR deflator) in U.S. dollar-terms in year t. In SOR- terms, the SDR deflator in year tis given by: 17 p~fR ]~ 112 SDR Pt ={( 7 p~f!z [ P j,t SDR } ) ( j,t-1 • I P j,t-1 [ SDR } j,t where pfDR is the international deflator (SDR deflator) in SOR-terms in year t. The GS country j 's GOP deflator in SOR terms is given by: and the weights are calculated as: where p~fR is the GDP deflator in SDR-terms for country j in year t, ef~RILC is the exchange rate from local currency to SOR for country j in year t, p j,t is the national GOP deflator in local currency for country j in year t , wj,t is the weight of country j in year t, and ca j,t is the currency amount for country j included in one unit of SOR in year t . In US-dollar terms, the SDR-deflator is given by: SDR($) Pt ={ SDR )• {e$1_SDR) 1/'t ~ ad.J-t where e~~~R is the conversion factor from SDR to U.S. dollars in year t. The conversion factor is calculated as: $/SDR_{ SISDR eadj-t - e, SISDR[PUSA,t/PUSA,t-1] + et-1 SDRj SDR + et-2 SISDR[PUSA,t/PUSA,t-2]~ SDRj SDR IJ Pt Pt-1 Pt Pt-2 13 where e~ 1 SDR is the exchange rate from SDR to U.S. dollar in year t, and PuSA.t is the U.S. GDP deflator in U.S. dollar terms in year t. Internationally comparable per capita GNI estimates can then be derived by dividing local currency per capita GNI estimates by the Atlas conversion factor. 14 AnnexB Synthetic Atlas Conversion Factor For economies in transition, where market exchange rates were lacking, a synthetic Atlas conversion factor was introduced in 1992. The synthetic exchange rate was estimated in the following manner: U.S. dollar GNI estimates in PPP terms were derived for all countries in transition and for comparator countries in the low- to middle-income group. For the comparator countries linear regression techniques were used to determine the relation between the ratio of the official exchange rate to the PPP and the PPP-converted per capita GNI as a ratio to the U.S. GNI per capita. This relation was then used to deduce a synthetic Atlas conversion factor for each country in transition, given the estimated PPP for the particular economy. A hypothetical exchange rate deviation index (ERDI), the ratio of the nominal exchange rate to PPP, was estimated for each transition economy. The ERDI was derived from an estimated inverse relationship in low- to middle-income market economies between the ERDI as the dependent variable and PPP-converted per capita income as the independent variable: 1 =(PPP) =a,+ p,{[GN/i.t • - 1 ]{GNIUSA,t ]~ ERDii,t ER i,t Ni,t PPP;,t NUSA,t If where GNI;,t is gross national income in local currency for country i in year t , a 1 ,p1 are regression coefficients for year t , estimated on the basis of data for m comparator countries in the low- and middle- income group in year t, and Nf is mid-year population for country i in year t. The synthetic Atlas conversion factor for each year and economy in transition was then derived as the product of PPP and ERDI: SACFi,t = PPlj.t j ,l PPP) ER i.t =PPPi,t ,A- -) = l ERDI 1 i,t PPP;,, • ERD/i,t. The Atlas conversion factor was then defined as a three-year average: ATLAS( SACF J _ { ei,t - SACF;,t + SACFi,t-1 ( Pi,t I Pi,t-1 SDR($) Pt I Pr-J l ( SDR($) + SACFi,t-2 SDR($) Pt Pi,t I Pi,t-2 SDR($) I Pr-2 )~ I 3 . 15 AnnexC Modified Atlas Methodology based on U.S. Inflation Rates A simpler, more transparent Atlas conversion factor, based on U.S. rather than international inflation rates, is given by: ATLAS* - eu -1 eu + ei.t-1 ( I Pi.t Pu-1 I PusA.r PuSA,t-1 ]+ ( ei.t-2 I Pi.t Pi.t-2 I PuSA,t PuSA.r-2 J~ 13 . Internationally comparable per capita GNI figures can then be derived by dividing local currency per capita GNI estimates by the alternative Atlas conversion factor. 16 AnnexD Effects on Country Classification of Using Alternative Atlas Methodology Income thresholds are currently updated with international inflation. If substituting international with U.S. inflation rates in the Atlas conversion factor, the corresponding income thresholds should be developed with U.S. inflation. The choice of inflation rate affects the Bank's operational categories (table 0.1). Table D.l Operational Categories Based on Current and Alternative Atlas Methodologies, FY2001 (in U.S. dollars) Operational category Current Atlas per capita Alternative Atlas per capita income thresholds income thresholds 1: Civil works preference Less than 756 Less than 776 II: IDA eligibility, or Less than 1,446 Less than 1,486 20-year IBRD terms ill: 17-year IBRD terms 1,446 - 2,995 1,486- 3,075 IV: 15-Year IBRD terms More than 2,995 More than 3,075 V: IBRD graduation More than 5,225 More than 5,355 Source: DECDG, World Bank. The choice of inflation rate had no effect on the ranking or classification of countries in FY2001. If real exchange rate movements between the U.S. dollar and the SDR are sufficiently large, however, rankings could be affected. Table D.2 Classification of Economies Based on Current and Alternative Atlas Methodologies, FY2001 (in U.S. dollars) Per capita GNJ based on Per capita GNJ based on Category current Atlas methodology alternative Atlas methodology Category V More than 5,225 More than 5,355 Slovenia 9,890 10,200 Korea, Rep. 8,490 8,740 Argentina 7,600 7,830 Seychelles 6,540 6,740 St. Kitts and Nevis 6,420 6,630 Uruguay 5,900 6,090 Antigua and Barbuda n.a. n.a. CategoryN More than 2,995 More than 3,075 Czech Republic 5,060 5,220 Chile 4,740 4,880 Hungary 4,650 4,790 Croatia 4,540 4,680 Brazil 4,420 4,550 Mexico 4,400 4,540 Trinidad and Tobago 4,390 4,520 Poland 3,960 4,080 St. Lucia 3,770 3,890 Lebanon 3,700 3,820 Venezuela 3,670 3,790 Mauritius 3,590 3,700 17 Per capita GNI based on Per capita GNI based on CateRorv current Atlas methodolo_gy alternative Atlas methodology Slovak Republic 3,590 3,700 Estonia 3,480 3,600 Grenada 3,450 3,560 Malaysia 3,400 3,500 Botswana 3,380 3,480 Gabon 3,350 3,440 Dominica 3,170 3,270 South Africa 3,160 3,260 Panama 3,070 3,170 Palau n.a. n.a. Category III More than 1,446 More than 1,486 Turkey 2,900 2,990 Costa Rica 2,740 2,830 Belize 2,730 2,810 St. Vincent and the Grenadines 2,700 2,780 Belarus 2,630 2,710 Lithuania 2,620 2,700 Latvia 2,470 2,550 Peru 2,390 2,460 Jamaica 2,330 2,400 Russian Federation 2,270 2,340 Colombia 2,250 2,320 Fiji 2,210 2,270 Tunisia 2,100 2,170 Thailand 1,960 2,020 Dominican Republic 1,910 1,970 El Salvador 1,900 1,960 Namibia 1,890 1,940 Micronesia, Fed. Sts. 1,810 1,870 Iran, Islamic Rep. 1,760 1,810 Tonga 1,720 1,780 Macedonia, FYR 1,690 1,740 Guatemala 1,660 1,710 Paraguay 1,580 1,630 Marshall Islands 1,560 1,610 Algeria 1,550 1,600 Romania 1,520 1,570 Jordan 1,500 1,570 Suriname n.a. n.a. Category II Less than 1,446 II -less than 1,486 Egypt 1,400 1,450 Bulgaria 1,380 1,430 Swaziland 1,360 1,400 Cape Verde 1,330 1,370 Ecuador 1,310 1,350 Kazakhstan 1,230 1,260 Morocco 1,200 1,230 18 Per capita GNI based on Per capita GNI based on Category current Atlas methodology alternative Atlas methodology Maldives 1,200 1,230 Equatorial Guinea 1,170 1,200 Vanuatu 1,170 1,200 Samoa 1,060 1,100 Philippines 1,020 1,050 Syrian Arab Republic 970 1,000 Kiribati 910 940 Albania 870 900 Sri Lanka 820 850 Papua New Guinea 800 820 Djibouti 790 810 China 780 810 Honduras 760 790 Guyana 760 790 Bosnia and Herzegovina n.a. n.a. Category I Less than 756 Less than 776 Ukraine 750 770 Solomon Islands 750 770 Uzbekistan 720 740 Cote d'Ivoire 710 730 Congo, Rep. 670 690 Turkmenistan 660 690 Georgia 620 640 Cameroon 580 600 Indonesia 580 590 Lesotho 550 570 Azerbaijan 550 560 Zimbabwe 520 530 Guinea 510 530 Senegal 510 530 Bhutan 510 530 Armenia 490 510 Pakistan 470 490 Haiti 460 480 India 440 460 Nicaragua 430 440 Ghana 390 400 Mauritania 380 400 Benin 380 390 Moldova 370 380 Vietnam 370 380 Bangladesh 370 380 Kenya 360 370 Mongolia 350 360 Comoros 350 360 Yemen, Rep. 350 360 Gambia, The 340 350 Sudan 330 340 19 Per capita GNI based on Per capita GNI based on CateRory current Atlas methodology_ alternative Atlas methodolo_gy Zambia 320 330 Togo 320 330 Uganda 320 330 Nigeria 310 320 Kyrgyz Republic 300 310 Central African Republic 290 300 Tajikistan 290 300 LaoPDR 280 290 Sao Tome and Principe 270 280 Cambodia 260 270 Rwanda 250 260 Madagascar 250 250 Tanzania 240 250 Mali 240 250 Burkina Faso 240 250 Mozambique 230 230 Angola 220 230 Nepal 220 220 Chad 200 210 Eritrea 200 210 Niger 190 200 Malawi 190 190 Guinea-Bissau 160 170 Sierra Leone 130 140 Burundi 120 130 Ethiopia 100 110 Mghanistan n.a. n.a. Congo, Dem. Rep. n.a. n.a. Liberia n.a. n.a. Myanmar n.a. n.a. Somalia n.a. n.a. n.a. Not available. Source: DECDG, World Bank. 20 AnnexE Effects on Country Classification of Using PPP-Based GNI Estimates The first step in using PPP-converted GNI estimates for Operational Guidelines is to define the appropriate cut-off levels for various terms of lending. Because country ranking may differ depending on the conversion method used, some countries will change to a different income category. The strategy adopted was to minimize such changes. As the classification structure will change every five years or so (when new benchmark data become available), all values will be recalculated and the cut-off levels reset in a way that minimizes changes. Within these five-year intervals, levels should be extrapolated using U.S. rates of inflation. Table E.l Operational Categories Based on Current Atlas Methodology and PPPs, FY2001 (in U.S. dollars) Operational Category Atlas per capita PPP-based per capita income thresholds income thresholds I: Civil works preference Less than 756 Less than 2,701 II: IDA eligibility, and Less than 1,446 Less than 4,251 20-year ffiRD terms ill: 17-year ffiRD terms 1,446 - 2,995 4,251 - 6,650 IV: 15-year ffiRD terms More than 2,995 More than 6,650 V: IBRD graduation More than 5,225 More than 10,750 Source: DECDG, World Bank. Based on the cut-offs in table E.l, 23 of 143 countries would have been reclassified into different income categories in FY2001 (table E.2). TableE.2 Reclassification of Economies as a Result of Using PPP-Based Income Estimates Rather than the Atlas Method, FY2001 Countries moving up one category Countries moving down one category (less favorable lending terms) (more favorable lending terms) Category change Economy Category change Economy From I to II Turkmenistan From II to I Bolivia Ukraine Honduras Papua New Guinea From II to III Bulgaria From III to II Guatemala Kazakhstan Jordan Swaziland Jamaica FromilltoiV Belarus From IV to III Dominica Russian Federation Gabon Grenada Lebanon Panama St. Lucia Venezuela FromiVtoV Czech Republic FromVtoiV St. Kitts and Nevis Uru ua 15 countries would have moved down a category (qualified for more favorable lending terms); eight countries would have moved up a category (qualifying for less favorable lending terms). Three of the 23 countries changed category in Operational Guidelines FYO 1 (Ukraine from II to I, Honduras from I to II, 21 8 and Dominica from ill to IV), and another four countries are less than 5 percent from a threshold. Thus, it may be that the use of PPP conversion would push some countries to change category a year earlier, while others would change a bit later. Table E.3 Distribution of Economies by Income Category Based on Atlas Methodology and PPPs, FY2001 Operational Category Atlas Guidelines PPP Guidelines 1: Civil works preference 63 64 II: IDA eligibility, and 23 22 20-year IDRD terms ill: 17-year IBRD terms 28 33 IV: 15-year IDRD terms 22 17 V: IDRD graduation 7 7 Source: DECDG, World Bank. The ranking and classification of countries differ significantly depending on conversion method used (table E.4). Table E.4 Per capita GN/ Based on Atlas Methodology and PPPs, FY2001 (in U.S. dollars) Per capita GNI based on Per capita GNI based on Country_ Atlas methodology PPP conversion Per capita GNI Category Per capita GNI Category Slovenia 9,890 v 16,080 v Korea, Rep. 8,490 v 15,630 v Czech Republic 5,060 IV 13,120 v Argentina 7,600 v 12,090 v Hungary 4,650 IV 11,190 v Seychelles 6,540 v 11,080 v Antigua and Barbuda n.a. v n.a. v Slovak Republic 3,590 IV 10,470 IV St. Kitts and Nevis 6,420 v 10,460 IV Mauritius 3,590 IV 9,240 IV Chile 4,740 IV 8,930 IV South Africa 3,160 IV 8,920 IV Uruguay 5,900 v 8,840 IV Malaysia 3,400 IV 8,500 IV Poland 3,960 IV 8,430 IV Estonia 3,480 IV 8,350 IV Mexico 4,400 IV 8,240 IV Trinidad and Tobago 4,390 IV 7,770 IV Croatia 4,540 IV 7,330 IV Belarus 2,630 m 6,960 IV Botswana 3,380 IV 6,770 IV 8 In the FY01 Operational Guidelines, per capita GNI estimates for nine countries are available in ranges only. This number increases to 14 when PPP rates are used. 22 Per capita GNI based on Per capita GNI based on Country Atlas methodology PPP conversion Per capita GNI CateROTY Per capita GNI Catel(ory Russian Federation 2,270 m 6,770 IV Brazil 4,420 IV 6,740 IV Palau n.a. IV n.a. IV Turkey 2,900 m 6,550 m Lithuania 2,620 m 6,500 m Latvia 2,470 m 6,310 m Grenada 3,450 IV 6,240 m Costa Rica 2,740 m 6,160 m Colombia 2,250 m 6,090 m Romania 1,520 m 6,030 m Thailand 1,960 m 5,990 m Tunisia 2,100 m 5,850 m Namibia 1,890 m 5,730 m Iran, Islamic Rep. 1,760 m 5,730 m Gabon 3,350 IV 5,680 m Venezuela 3,670 IV 5,620 m St. Lucia 3,770 IV 5,360 m Panama 3,070 IV 5,350 m Bulgaria 1,380 II 5,250 m Dominica 3,170 IV 5,150 m Algeria 1,550 m 5,090 m St. Vincent and the Grenadines 2,700 m 4,980 m Dominican Republic 1,910 m 4,970 m Fiji 2,210 m 4,840 m Belize 2,730 m 4,790 m Kazakhstan 1,230 II 4,710 m Peru 2,390 m 4,680 m Macedonia, FYR 1,690 III 4,630 m Tonga 1,720 m 4,570 m Swaziland 1,360 II 4,480 III Paraguay 1,580 m 4,450 m Lebanon 3,700 IV 4,410 m El Salvador 1,900 III 4,320 m Marshall Islands 1,560 III n.a. m Micronesia, Fed. Sts. 1,810 m n.a. m Suriname n.a. III n.a. m Samoa 1,060 II 4,180 II Philippines 1,020 II 4,070 II Maldives 1,200 II 3,870 II Cape Verde 1,330 II 3,790 II Jordan 1,520 m 3,790 II Guatemala 1,660 III 3,770 II Egypt 1,400 II 3,530 II 23 Per capita GNI based on Per capita GNI based on Country_ Atlas methodology PPP conversion Per capita GNI Category_ Per capita GNI Category China 780 II 3,510 II Jamaica 2,330 m 3,500 II Guyana 760 II 3,460 II Kiribati 910 II 3,420 II Morocco 1,200 II 3,410 II Ukraine 750 I 3,350 II Turkmenistan 660 I 3,310 II Sri Lanka 820 II 3,260 II Albania 870 II 3,090 II Vanuatu 1,170 II 2,960 II Syrian Arab Republic 970 II 2,950 II Ecuador 1,310 II 2,780 II Equatorial Guinea 1,170 II n.a. II Bosnia and Herzegovina 990 II n.a. II Djibouti 790 II n.a. II Zimbabwe 520 I 2,640 I Indonesia 580 I 2,600 I Georgia 620 I 2,530 I Moldova 370 I 2,520 I Azerbaijan 550 I 2,480 I Papua New Guinea 800 II 2,420 I Honduras 760 II 2,410 I Kyrgyz Republic 300 I 2,370 I Armenia 490 I 2,360 I Bolivia 1,010 II 2,340 I Lesotho 550 I 2,270 I Nicaragua 430 I 2,260 I India 440 I 2,250 I Uzbekistan 720 I 2,230 I Solomon Islands 750 I 2,000 I Guinea 510 I 1,880 I Pakistan 470 I 1,880 I Ghana 390 I 1,870 I Vietnam 370 I 1,870 I LaoPDR 280 I 1,840 I Mauritania 380 I 1,660 I Cote d'Ivoire 710 I 1,650 I Bhutan 510 I 1,600 I Mongolia 350 I 1,600 I Gambia, The 340 I 1,590 I Bangladesh 370 I 1,570 I Cameroon 580 I 1,540 I Haiti 460 I 1,500 I Comoros 350 I 1,450 I Togo 320 I 1,440 I 24 Per capita GNI based on Per capita GNI based on Country Atlas methodology PPP conversion Per capita GNI Category Per capita GNI Category Senegal 510 I 1,430 I Sao Tome and Principe 270 I 1,430 I Sudan 330 I 1,390 I Cambodia 260 I 1,370 I Nepal 220 I 1,300 I Uganda 320 I 1,210 I Central African Republic 290 I 1,210 I Eritrea 200 I 1,080 I Tajikistan 290 I 1,050 I Kenya 360 I 1,040 I Burkina Faso 240 I 960 I Congo, Rep. 670 I 960 I Benin 380 I 950 I Chad 200 I 860 I Mozambique 230 I 850 I Madagascar 250 I 820 I Nigeria 310 I 800 I Yemen, Rep. 350 I 790 I Niger 190 I 780 I Zambia 320 I 730 I Mali 240 I 730 I Angola 220 I 670 I Ethiopia 100 I 640 I Guinea-Bissau 160 I 640 I Malawi 190 I 620 I Burundi 120 I 590 I Tanzania 240 I 510 I Sierra Leone 130 I 440 I Rwanda 250 I n.a. I Afghanistan n.a. I n.a. I Liberia n.a. I n.a. I Myanmar n.a. I n.a. I Somalia n.a. I n.a. I Congo, Dem. Rep. n.a. I n.a. I n.a. Not available. Source: DECDG, World Bank. 25 AnnexF Participants in ICP and Related Exercises, 1975-96 PPP data have been estimated for all but seven countries in which the Bank has operations (table F.l). In most cases PPP data are based on price information gathered through participation in the ICP exercises. For some countries the PPP data are based on more limited price surveys or no price statistics at all. Where no price statistics were provided, special regression techniques were applied to deduce implicit knowledge about price levels. Table F.1 Participating economies in /CP and Related Exercises Economy 1975 1980 1985 1990 1993 1996 Albania ICP Algeria• Angola• Antigua and Barbuda RIA Argentina ICP RIA Armenia ICP ICP Australia ICP ICP ICP ICP Austria ICP ICP ICP ICP ICP ICP Azerbaijan ICP ICP Bahamas, The ICP RIA Bahrain RIA Bangladesh ICP ICP Barbados ICP RIA Belarus ICP ICP Belgium ICP ICP ICP ICP ICP ICP Belize RIA Benin ICP ICP Bermuda RIA Bolivia ICP RIA Botswana ICP ICP ICP Brazil ICP ICP RIA Bulgaria ICP ICP Burkina Faso• Burundia Cambodia• Cameroon ICP ICP ICP Canada ICP ICP ICP ICP ICP Cape Verde• Central African Republic• Chad a Chile ICP RIA China ICP (limited) Colombia ICP ICP Comoros• Congo, Dem. Rep.• Congo, Rep. ICP ICP Costa Rica ICP 26 Economy 1975 1980 1985 1990 1993 1996 Cote d'Ivoire ICP ICP ICP Croatia ICP ICP Czech Republic ICP ICP Denmark ICP ICP ICP ICP ICP ICP Dominica RIA Dominican Republic ICP Ecuador ICP RIA Egypt ICP ICP El Salvador ICP Estonia ICP ICP Ethiopia ICP ICP Fiji ICP Finland ICP ICP ICP ICP ICP France ICP ICP ICP ICP ICP ICP Gabon ICP Gambia, The" Georgia ICP ICP Germany ICP ICP ICP ICP ICP ICP Ghana• Greece ICP ICP ICP ICP ICP Grenada ICP RIA Guatemala ICP Guinea ICP Guyana ICP Haiti a Honduras ICP Hong Kong (China) ICP ICP ICP Hungary ICP ICP ICP ICP ICP ICP Iceland ICP ICP ICP India ICP ICP ICP Indonesia ICP ICP Iran, Islamic Rep. ICP ICP ICP Ireland ICP ICP ICP ICP ICP ICP Israel ICP ICP Italy ICP ICP ICP ICP ICP ICP Jamaica ICP ICP RIA Japan ICP ICP ICP ICP ICP ICP Jordan RIA Kazakhstan ICP ICP Kenya ICP ICP ICP ICP Korea ICP ICP ICP ICP Kuwait" Kyrgyz Republic ICP ICP LaoPDR RIA Latvia ICP ICP Lebanon RIA Lesotho" Lithuania ICP ICP 27 Economy 1975 1980 1985 1990 1993 1996 Luxembourg ICP ICP ICP ICP ICP ICP Macedonia, FYR ICP Madagascar ICP ICP ICP Malawi ICP ICP ICP ICP Malaysia ICP RIA Maldives• Mali ICP ICP ICP Malta• Mauritania• Mauritius ICP ICP Mexico ICP RIA ICP Moldova ICP ICP Mongolia ICP Morocco ICP ICP ICP Mozambique• Namibia• Nepal ICP ICP Netherlands ICP ICP ICP ICP ICP ICP New Zealand ICP ICP ICP ICP Nicaragua• Niger- Nigeria ICP ICP ICP Norway ICP ICP ICP ICP ICP Oman RIA Pakistan ICP ICP ICP ICP West Bank and Gaza RIA Panama ICP RIA Paraguay ICP Peru ICP RIA Philippines ICP ICP ICP ICP Poland ICP ICP ICP ICP ICP ICP Portugal ICP ICP ICP ICP ICP Qatar RIA Romania ICP ICP ICP ICP Russian Federation ICP ICP Rwanda ICP Samoa• Saudi Arabia RIA Senegal ICP ICP ICP Sierra Leone ICP ICP Singapore ICP Slovak Republic ICP ICP Slovenia ICP ICP Solomon Islands• South Africa• Spain ICP ICP ICP ICP ICP ICP Sri Lanka ICP ICP ICP ICP St. Kitts and Nevis RIA 28 Economy 1975 1980 1985 1990 1993 1996 St. Lucia ICP RIA St. Vincent and Grenadines RIA Sudan• Suriname ICP Swaziland ICP ICP Sweden ICP ICP ICP ICP Switzerland ICP ICP ICP Syrian Arabic Republic ICP RIA Tajikistan ICP ICP Tanzania ICP ICP ICP Thailand ICP ICP ICP Togo• Trinidad and Tobago ICP RIA Tunisia ICP ICP ICP ICP Turkey ICP ICP ICP Turkmenistan ICP ICP Uganda• Ukraine ICP ICP United Arab Emirates RIA United Kingdom ICP ICP ICP ICP ICP ICP United States ICP ICP ICP ·ICP ICP ICP Uruguay ICP ICP RIA Uzbekistan ICP ICP Vanuatu• Venezuela ICP RIA Vietnam ICP Yemen, Rep. RIA Zambia ICP ICP ICP ICP Zimbabwe ICP ICP ICP Former Czechoslovakia ICP Former Soviet Union ICP Former Yugoslavia ICP ICP ICP ICP TOTAL 34 60 64 30 118 52 a Regression estimates. ICP: International Comparison Programme (full scale price survey). RIA: Reduced information approach (limited price survey). Source: DECDG, World Bank. 29