## 1. The World Bank's Operational Income Thresholds ## Source details **Canonical URL:** [1. The World Bank's Operational Income Thresholds](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1131.pdf) ## Other formats - [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1131.pdf.md) - [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1131.pdf.json) --- ### I. Introduction: classification problem and practical observations - Classification systems (taxonomies) are central to scientific paradigms and data comparability (example: IMF resident/non-resident definition using a one year benchmark). - No generally accepted objective criterion for classifying countries by level of development; multiple competing terminologies exist (poor/rich, backward/advanced, underdeveloped/developed, North/South, Third World/First World). - Illustrative development differentials (2009): - Burkina Faso average income: US$510. - Japan average income: US$37,870. - Burkina Faso adult literacy: 29 percent. - Burkina Faso life expectancy at birth: 53 years. - Japan adult literacy: virtually all adults literate. - Japan life expectancy at birth: 83 years. - OECD membership used heuristically for “developed” status: - OECD has 34 members up from 20 members at its establishment in 1961. - Using OECD membership heuristically yields about 80–85 percent of the world’s countries designated as developing and about 15–20 percent as developed. - Definitions used in the paper: - “Development taxonomy” = system ordering countries by level of development. - “Development threshold” = criterion separating developing from developed. - Paper focus: taxonomies given a definition of development; compares UNDP, World Bank, and IMF taxonomies and proposes an alternative taxonomy. ### II. Development, development taxonomies, and thresholds (conceptual points) - Historical background: - Classical economics: emphasis on sustained increases in per capita real income. - Post-World War II development economics: broadened to social outcomes, culture, production structures. - UN actions: 1952 UN General Assembly resolution; 1954 Report distinguished standards of living (normative) and levels of living (positive). - Measurement advances: improvements in national accounts and PPPs increased feasibility of inter-country measures. - Poverty-based taxonomy considerations: - World Bank WDR 2000/2001 estimate: 1.3 billion people in 1993 below PPP US$1.08 per day (1993 dollars). - Chen and Ravallion (2010) revised estimate: at 2005 PPP US$1.45 per day (equivalent in real terms to 1993 PPP US$1.08 per day), 2.2 billion poor people in 1993. - At PPP US$1–2 per day, poverty in richer countries is insignificant; developed countries could be defined as those with negligible poverty at such a line. - Drawbacks: internationally comparable poverty data are imprecise, subject to large revisions, and not drawn directly from official country sources. - Alternative simple taxonomies: per capita income; single social indicator (e.g., life expectancy at birth); composite indices. - Threshold types: - Absolute threshold: fixed over time (e.g., many Millennium Development Goals). - Relative threshold: based on contemporaneous outcomes (captures changing expectations). - Example: absolute life expectancy threshold set in 1950 would classify Sri Lanka as developed today. - Threshold selection approaches: - Percentile-based: countries above x percentile designated developed. - Distance-based: countries within y percent of the most advanced country designated developed. - Any threshold requires rationale and invites scrutiny. ### III. International organizations’ country classification systems — UNDP, World Bank, IMF - UNDP (Human Development Index, HDR history and practice) - HDI composition (HDR 2010 definitions): - Composite of three indices: longevity, education, income. - Income measure: Gross National Income per capita (GNI/n) converted using PPP. - Longevity: life expectancy at birth. - Education: combined measures of actual and expected years of schooling. - Sub-index construction: - X = (X_actual – X_min)/(X_max – X_min) - Maximum values set to actual observed maximum values over the 1980-2010 period. - Minimum values: education = 0; longevity = 20 years; income = US$163 (income level in Zimbabwe in 2008). - Aggregate HDI: geometric average of the three sub-indices, bounded between zero and one. - HDI categories and thresholds over time: - HDR 1990: low-, medium-, and high-human development thresholds at 0.5 and 0.8. - HDR 2009: introduced “very high human development” threshold at 0.9. - HDR 1990 designated “industrial” vs “developing”; industrial largely subgroup of high human development (exception: Albania with HDI 0.79). - HDR 2007/08 alternative grouping introduced OECD membership and Central/Eastern Europe/CIS categories; overlaps and inconsistencies noted. - HDR 2009: “developed countries” = countries with very high human development; others designated developing (characterized as “somewhat arbitrary”). - HDR 2010: moved to relative thresholds — developed = top quartile in the HDI-distribution (UNDP uses equal country weights). In that distribution 15 percent of the world’s population lives in designated developed countries. - HDR 2010 did not explain shift to relative thresholds nor the choice of top quartile or equal country weights. - World Bank (operational and analytical classifications; thresholds and evolution) - Operational context: - IBRD statutory obligation: lend only to credit-worthy member countries that cannot otherwise obtain external financing on reasonable terms (Article 1 (ii) and Article III, Section 4 (ii) of the IBRD Articles of Agreement). - IBRD designated a subset of members as eligible borrowers; early judgments shifted toward a rule-based GNI/n criterion in early 1980s. - An informal threshold existed since 1973 (set at US$1,000 in 1970-prices). - Under GNI/n system, countries that borrow and exceed a certain income threshold move to non-borrowing status (graduation). - IDA (established 1960) partition: - Part 1 countries expected to contribute financially to IDA. - Part 2 countries were other countries of which only a subset would draw on concessional resources. - Historical partitioning noted as “a political exercise” with per capita income criterion “with a few exceptions” (Spain, Japan mentioned). - Historical operational thresholds and selected figures: - In 1964 an income threshold for IDA eligibility was set at an annual per capita income level of US$250; during the 1960s countries with income levels of up to US$300 accessed IDA resources. - In the 1970s operational guidelines used GNI/n thresholds (Atlas method); thresholds adjusted annually with inflation; GNI/n considered “the best single indicator of economic capacity and progress.” - Multiple operational thresholds (selected figures for FY 2011 and earlier years): - Civil works preference threshold: FY 1978 = 265; FY 1984 = 410; FY 1989 = 480; FY 2011 = 995. - IDA eligibility (operational): FY 1978 = 520; FY 1984 = 805; FY 1989 = 580; FY 2011 = 1,165. - IDA eligibility (historical): FY 1978 = 520; FY 1984 = 805; FY 1989 = 940; FY 2011 = 1,905. - IBRD term (threshold between softer and harder IBRD-borrowing terms; abolished FY 2009): FY 1978 = 1,075; FY 1984 = 1,670; FY 1989 = 1,940. - IBRD graduation: FY 1984 = 2,910; FY 1989 = 3,385; FY 2011 = 6,885. - World per capita GNI (Atlas method): 1978 = 1,562; 1984 = 2,486; 1989 = 3,179; 2011 = 8,741. - Observed pattern: percent-change and percent-of-world-per-capita-GNI lines show secular decreases in thresholds relative to average world income because thresholds are absolute, inflation-adjusted but not adjusted for trend growth in global real income. - Analytical classifications and evolution: - 1978 WDI classification: three categories — (1) developing countries, (2) industrialized countries, (3) capital-surplus oil-exporting countries. Developing split into low-income (GNI/n ≤ US$250) and middle-income (GNI/n > US$250). - 1989 WDI reforms: - High-income category for GNI/n above US$6,000 (no rationale provided). - Middle-income group subdivided into lower and upper middle using the IBRD borrowing-terms cut-off. - “Developing” label dropped though “low- and middle-income countries” sometimes referred to as developing. - Between 1987 and 2009: - Low-income threshold fell from 16 to 11 percent of average world income. - High-income threshold fell from 189 to 140 percent of average world income. - Table 2 selected entries: - Low income: 1976 = 250; 1982 = 410; 1987 = 480; 2009 = 995. - Lower middle income: 1987 = 1,670; 2009 = 1,940; 2009 (alternate) = 3,945. - High income: 2009 (and earlier) = 6,000; 12,195. - World per capita GNI (Atlas method): 1976 = 1,562; 1982 = 2,486; 1987 = 3,179; 2009 = 8,741. - IMF operational and analytical classification practices - IMF Articles of Agreement do not distinguish membership by development level; for the first three decades operational policies did not discriminate by development. - Post-1973 oil shock responses introduced distinctions: - Oil facilities (1974, 1975) financed by commercial borrowing; higher charges applied. - 1975 Subsidy Account subsidized charges for developing members designated “most seriously affected” (UN list of 41 countries). - Trust Fund proposals and later gold-sale proceeds financed concessional operations; initial eligibility limited to 61 members with per capita income ≤ SDR 300 in 1973 for loans; distributions “for the benefit of developing countries.” - Executive Board in 1977 designated 103 members as developing countries (after debate and adjustments including Singapore). - 1978 second amendment to Articles recognized that “balance of payments assistance may be made available on special terms to developing members… and that for this purpose the Fund shall take into account the level of per capita income.” - Concessional facilities evolved into the Poverty Reduction and Growth Trust (PRGT); new PRGT eligibility framework agreed in early 2010 determined eligibility based on per capita income, market access, and vulnerability. - Under the new framework the number of PRGT-eligible countries was reduced from 77 to 71. - These countries are recognized by the Fund to be “low-income developing countries.” - IMF analytical classifications evolved through multiple schemes; since May 1980 the WEO used the IFS classification. The industrial country group was renamed advanced in 1997; additions to advanced group since 1997 include Israel, Korea, Singapore, Cyprus (2001), Slovenia (2007), Malta (2008), Czech Republic (2009), Slovak Republic (2009). WEO reports generally do not provide rationale for reclassifications. - Cross-organization comparison (high-level) - Despite methodological differences, UNDP, World Bank, and IMF produce similar high-level splits: around 20–25 percent of countries designated as developed. - Differences: - UNDP uses a relative threshold (top quartile in HDI distribution). - World Bank uses an absolute threshold (e.g., US$6,000 GNI per capita in 1987-prices). - IMF threshold is not explicit; IMF classification appears most likely absolute. - Selected shares of countries designated 'developed': - Share in 1990: IMF = 13 percent; UNDP = 25 percent; World Bank = 16 percent. - Share in 2010: IMF = 17 percent; UNDP = 25 percent; World Bank = 26 percent. - Table 4 (in source) provides country-by-country listings reflecting each organization’s classification (datasets as of October/November 2010). ### IV. Alternative development taxonomy methodology (proposed, formal approach) - Goals and principles: - Data-driven thresholds determined by observed distribution of development outcomes. - Preserve both mean and ranking of the actual distribution when constructing synthetic distributions. - Use Lorenz curve framework: L(x) is cumulative outcome of lowest x percentile; preserve mean (anchoring) and ranking (no nonsensical re-orderings). - Error metric and optimization: - Define E as the area between the line of equality and synthetic distribution segments; minimize E to find thresholds. - Dichotomous taxonomy: - Error-minimizing threshold between developing and developed countries is the mean development outcome (the population-average). - Trichotomous taxonomy: - First-order conditions for optimal thresholds x0 and x1: thresholds must satisfy that each threshold country’s development outcome equals the average development outcome of a specified complementary set (e.g., threshold between LDC and MDC set at average of non-HDC). - General case for K categories: - For K categories each threshold xi is chosen such that the development outcome at xi equals the average development outcome of countries in the interval between adjacent thresholds. - Advantages of the method: - Thresholds are data-determined (relative thresholds) and remain relevant over time. - Flexible: implementable with any number of categories and any ordinal development proxy. ### V. Examples, empirical illustrations, and weighting choices - Development proxies used for empirical taxonomies: - Income: GNI/n, Atlas method, using market exchange rates. - Longevity: life expectancy at birth. - Lifetime income: income × longevity (expected lifetime income of a newborn at 0 percent discount rate, assuming no future income growth). - Weighting variants: - Population-weighted distribution. - Equally weighted distribution (equal country weights). - Weighting materially affects taxonomy; population-weighting may better reflect global implications due to concentration of population in large countries. - Empirical results (as reported in the source): - Trichotomous taxonomies constructed (Tables 5 and 6 in source) list countries ranked highest to lowest for each proxy and weighting scheme; Middle Development Countries highlighted in bold in source tables. - Dichotomous taxonomies corresponding to these trichotomous partitions identify marginal developed and marginal developing countries. - Observations: - The alternative dichotomous threshold (mean development outcome) lies well below existing thresholds used by international organizations. - A trichotomous taxonomy provides a better fit for observed diversity and aligns the Higher Development Countries group broadly with existing 'developed' groupings while creating two lower groups that provide more granular distinctions among developing countries. ### VI. Key findings, implications, and policy-relevant points - Existing international classification systems: - Employ different bases (absolute vs. relative thresholds) and often lack transparent rationales for specific cut-offs. - Use absolute thresholds that, while inflation-adjusted, are not adjusted for trend growth in global real income, causing thresholds to decline relative to average world income over time. - Proposed alternative taxonomy: - Provides a principled, data-driven method to set thresholds that minimize classification error given an observed distribution of development outcomes. - Flexible: can be implemented with any number of categories and any development proxy (including multivariate indices like HDI or lifetime income). - Produces dichotomous threshold at mean outcome (minimizes error) but typically recommends trichotomous taxonomy in practice because it better captures heterogeneity. - Practical considerations: - Choice of development proxy and weighting scheme materially affects classification outcomes. - The paper does not resolve the optimal number of categories; it specifies how thresholds should be set once number of categories and proxy are chosen. ### VII. Concluding remarks and open questions - UNDP, World Bank, and IMF reach broadly similar conclusions about which countries are developed, but their thresholds and methods differ and lack full transparency. - The proposed methodology delivers transparent, data-driven thresholds and can be implemented with any development proxy and any chosen number of categories. - Open questions: - What constitutes the optimal number of categories in a development taxonomy? - What is the most appropriate choice of development proxy or multivariate index for classification purposes? *Source: _wp1131 - 1. The World Bank's Operational Income Thresholds (PDF chapter content provided).* ### 1. The World Bank's Operational Income Thresholds .............................................................12 ### 1. The World Bank's Operational Income Thresholds ### Major sections - 1. The World Bank's Operational Income Thresholds .............................................................12 - 2. The World Bank's Analytical Income Thresholds ...............................................................14 - 3. Country Classification Systems in Selected International Organizations ............................19 - 4. Country Classifications in Selected International Organizations ........................................20 - 5. Trichotomous Development Taxonomies (Population Weighted) ......................................33 - 6. Trichotomous Development Taxonomies (Equally Weighted) ...........................................37 ### Figures listed in the content unit - Figure 1. Dichotomous Development Taxonomy ...............................................................................27 - Figure 2. Trichotomous Development Taxonomy ..............................................................................28 - Figure 3. Development Taxonomies ...................................................................................................31 *Source: _wp1131 - 1. The World Bank's Operational Income Thresholds* ### References ............................................................................................................. ### _wp1131 - References ### I. INTRODUCTION - Classification systems (taxonomies) are central to scientific paradigms and practical data comparability (example: IMF resident/non-resident definition using a one year benchmark). - There is no generally accepted objective criterion for classifying countries by level of development; multiple competing terminologies exist (examples: poor/rich, backward/advanced, underdeveloped/developed, North/South, Third World/First World). - Practical observations illustrating development differentials (2009): - Burkina Faso average income: US$510. - Japan average income: US$37,870. - Burkina Faso adult literacy: 29 percent. - Burkina Faso life expectancy at birth: 53 years. - Japan adult literacy: virtually all adults literate. - Japan life expectancy at birth: 83 years. - OECD heuristic for “developed” status: - OECD has 34 members up from 20 members at its establishment in 1961. - Using OECD membership heuristically results in designation of about 80–85 percent of the world’s countries as developing and about 15–20 percent as developed. - Definitions: - “Development taxonomy” = system ordering countries by level of development. - “Development threshold” = criterion separating developing from developed. - Paper focus: development taxonomies given a definition of development (not on conceptual definition of development per se). Comparison of UNDP, World Bank, and IMF taxonomies and proposal of an alternative taxonomy. ### II. DEVELOPMENT, DEVELOPMENT TAXONOMIES, AND DEVELOPMENT THRESHOLDS - Historical and conceptual background: - Classical economics emphasized sustained increases in per capita real income. - Post-World War II development economics broadened focus to social outcomes, culture, production structures. - UN actions: 1952 UN General Assembly resolution; Report on International Definition and Measurement of Standards and Levels of Living (UN, 1954) distinguished standards of living (normative) and levels of living (positive). - Measurement advances: - Improvements in national accounts and purchasing power parities (PPP) have increased the feasibility of inter-country measures. - Poverty-based taxonomy considerations: - World Bank World Development Report 2000/2001 estimate: 1.3 billion people in 1993 below PPP US$1.08 per day (1993 dollars). - Chen and Ravallion (2010) revised estimate: at 2005 PPP US$1.45 per day (equivalent in real terms to 1993 PPP US$1.08 per day), 2.2 billion poor people in 1993. - Implication: number of poor people in early 1990s larger than originally believed. - At global poverty line PPP US$1–2 per day, poverty in richer countries is insignificant; thus developed countries could be defined as countries with negligible poverty at such a line. - Drawbacks of poverty-based taxonomy: - Internationally comparable poverty data are imprecise and subject to large revisions. - Required data are not drawn directly from official country sources, reducing tractability and acceptance. - Alternative simple taxonomies: - Per capita income. - Single social indicator (e.g., life expectancy at birth). - Composite indices combining economic and social indicators. - Threshold types: - Absolute threshold: fixed over time (e.g., Millennium Development Goals mostly absolute). - Relative threshold: based on contemporaneous outcomes (captures changing expectations). - Example: an absolute life expectancy threshold set in 1950 would classify Sri Lanka as developed today; illustrates normative issues. - Threshold selection approaches: - Countries above x percentile designated developed. - Countries within y percent of the most advanced country designated developed. - Any threshold requires rationale and will invite scrutiny. ### III. INTERNATIONAL ORGANIZATIONS’ COUNTRY CLASSIFICATION SYSTEMS - The UN General Assembly has created specific groups (example: Least Developed Countries in 1971; current list includes 49 countries) but has not established a taxonomy for full membership. - This section compares taxonomies used by UNDP, World Bank, and IMF. A. United Nations Development Programme’s (UNDP) Country Classification System - Built around the Human Development Index (HDI), launched with the Human Development Report (HDR) in 1990. - HDI composition (HDR 2010 definitions): - Composite of three indices measuring longevity, education, and income. - Income measure: Gross National Income per capita (GNI/n) converted using PPP. - Longevity: life expectancy at birth. - Education: combined measures of actual and expected years of schooling. - Sub-index construction (basic building block): - X = (X_actual – X_min)/(X_max – X_min) - Maximum values set to actual observed maximum values over the 1980-2010 period. - Minimum values: education = 0; longevity = 20 years; income = US$163 (income level in Zimbabwe in 2008, minimum observed income level). - Sub-indices bounded between zero and one; aggregate HDI is geometric average of the three sub-indices and bounded between zero and one. - HDI categories and thresholds over time: - HDR 1990: low-, medium-, and high-human development using thresholds 0.5 and 0.8. - HDR 2009: introduced “very high human development” with threshold 0.9 (no explanations for thresholds provided). - HDR 1990 also designated countries as industrial or developing; industrial grouping largely a subgroup of high human development (exception: Albania with HDI 0.79). - HDR 2007/08 replaced industrial grouping with: (1) member countries of the OECD and (2) countries in Central or Eastern Europe or members of the Commonwealth of Independent States; overlap issues noted (e.g., Mexico and Turkey OECD members also designated developing; Czech Republic, Hungary, Poland, and Slovakia OECD members and Central/Eastern Europe). - HDR 2009 resolved overlaps by introducing “developed countries” = countries that have achieved very high human development; others designated developing. Distinction recognized as “somewhat arbitrary.” - HDR 2010: moved from absolute to relative thresholds. Developed countries = countries in the top quartile in the HDI-distribution; bottom three quartiles = developing countries. - UNDP uses equal country weights to construct HDI distribution; in this distribution, 15 percent of the world’s population lives in designated developed countries. - HDR 2010 did not explain shift to relative thresholds nor why top quartile chosen; report did not discuss choice of equal country weights. B. The World Bank’s Country Classification Systems - World Bank classifications used for operational and analytical purposes; the operational system preceded the analytical system. - Operational classifications: - IBRD (International Bank for Reconstruction and Development) statutory obligation: lend only to credit-worthy member countries that cannot otherwise obtain external financing on reasonable terms (Article 1 (ii) and Article III, Section 4 (ii) of the IBRD Articles of Agreement). - IBRD designated a subset of membership as eligible borrowers; early determinations were judgmental, but in early 1980s moved toward a rule-based system using a GNI/n criterion. - An informal threshold existed since 1973 (set at US$1,000 in 1970-prices). - Under the GNI/n system, countries that borrow from the IBRD and exceed a certain income threshold engage in a process to move to non-borrowing status (graduation). - IDA (International Development Association, established 1960) identified two lists of IDA member countries: - Part 1 countries expected to contribute financially to IDA. - Part 2 countries were other countries of which only a subset could be expected to draw on concessional resources. - The source text begins to discuss the basis for assigning countries to Part 1 or Part 2 but stops at this point. *Source: _wp1131 - References (excerpts as provided)* ### Part 2? ### _wp1131 - Part 2? ### Partitioning process and decision-making - The partitioning of countries into categories was "a political exercise: a civilized understanding among sovereign countries about how to label each other." - The Bank provided "a large number of economic criteria," including: - "the amount of capital exported by the country" - "the gross national product of the country" - "various other things of that sort" - The Board of Directors reviewed these criteria, but "in the ultimate analysis, the management of the Bank was invited to present a list of those countries which, in their opinion, and based on the background [work] of the World Bank, should be in category I and those which should be in category II." - "The management presented this list, and the various executive directors who were negotiating the charter discussed it and agreed that this was an adequate list." 11 ### Role of per capita income and noted exceptions - The partition "followed a per capita income criterion with a few exceptions." - Explicitly mentioned exceptions: - Spain: "stated that it was flattered to be asked to be in Part 1 but did not consider it belonged there" - Japan: "a capital exporter" that "was placed in Part 1 despite its relatively low per capita income level (Mason and Asher, 1973)." - While "an income criterion was not used to demarcate" (sentence fragment as in source). *Source: _wp1131 - Part 2?* ### Part 1 and Part 2 countries, it was decided in 1964—at the time of the first IDA ### _wp1131 - Part 1 and Part 2 countries, it was decided in 1964—at the time of the first IDA ### Historical use of income thresholds in World Bank operations - In 1964 an income threshold was established as a test for eligibility to access IDA resources. - The threshold was initially set at an annual per capita income level of US$250; during the 1960s countries with income levels of up to US$300 accessed IDA resources. - In the 1970s operational guidelines used GNI/n thresholds, based on Bank research showing a relationship between poverty incidence and infant mortality and per capita GNI (Atlas method). - Thresholds were adjusted annually in line with inflation; the Bank considered GNI/n to be “the best single indicator of economic capacity and progress.” - Multiple operational thresholds employed by the Bank include: - Civil works preference threshold (ceiling). - IDA eligibility (operational) threshold (ceiling). - IDA eligibility (historical) threshold (ceiling). - IBRD term threshold (threshold between softer and harder IBRD-borrowing terms; abolished in FY 2009). - IBRD graduation threshold (floor). ### World Bank operational thresholds (selected figures and comparisons) - For the 2011 fiscal year (July-June), operational thresholds range from US$995 (the civil works preference threshold) to US$6,885 (the IBRD graduation threshold). - Table 1 (selected entries shown in source): - Civil works preference (FY 1978, FY 1984, FY 1989, FY 2011): 265, 410, 480, 995. - IDA eligibility (operational) (FY 1978, FY 1984, FY 1989, FY 2011): 520, 805, 580, 1,165. - IDA eligibility (historical) (FY 1978, FY 1984, FY 1989, FY 2011): 520, 805, 940, 1,905. - IBRD term (FY 1978, FY 1984, FY 1989): 1,075, 1,670, 1,940 (abolished FY 2009). - IBRD graduation (FY 1984, FY 1989, FY 2011): 2,910, 3,385, 6,885. - World per capita GNI (Atlas method) (1978, 1984, 1989, 2011): 1,562, 2,486, 3,179, 8,741. - Percent-change and percent-of-world-per-capita-GNI lines in Table 1 show secular decreases in thresholds relative to average world income because thresholds are absolute and adjusted for inflation but not for trend growth in global real income. ### World Bank analytical classifications and evolution - 1978 WDI classification: three categories — (1) developing countries, (2) industrialized countries, (3) capital-surplus oil-exporting countries. Developing countries split into low-income (GNI/n ≤ US$250) and middle-income (GNI/n > US$250). Industrialized membership largely aligned with OECD membership but with notable exceptions and unexplained derogations. - 1989 WDI reforms: - High-income country category established for GNI/n above US$6,000 (no rationale provided for cut-off). - Middle-income group subdivided into lower and upper middle using the IBRD borrowing-terms cut-off. - Developing category label dropped though “low- and middle-income countries” sometimes referred to as developing. - Between 1987 and 2009: - Low-income threshold fell from 16 to 11 percent of average world income. - High-income threshold fell from 189 to 140 percent of average world income. - Table 2 (selected entries from source): - Low income (1976, 1982, 1987, 2009): 250, 410, 480, 995. - Lower middle income (1987, 2009): 1,670, 1,940, 3,945. - High income (2009, and earlier): 6,000, 12,195. - World per capita GNI (Atlas method) (1976, 1982, 1987, 2009): 1,562, 2,486, 3,179, 8,741. ### IMF operational and analytical classification practices - The IMF’s Articles of Agreement do not contain distinctions among membership based on development, and for the first three decades operational policies did not discriminate by development level. - Responses to 1973 oil shock introduced distinctions: - Establishment of oil facilities (1974, 1975) with eligibility open to full membership but financed by commercial borrowing; higher charges applied. - 1975 Subsidy Account created to subsidize charges for developing members designated as “most seriously affected” (UN list of 41 countries). - Trust Fund (proposal in 1974) and later gold-sale proceeds financed concessional operations with eligibility initially limited to 61 members with per capita income ≤ SDR 300 in 1973 for loans; distributions “for the benefit of developing countries.” - Executive Board in 1977 designated 103 members as developing countries (after debate and adjustments including Singapore’s case). - 1978 second amendment to Articles recognized that “balance of payments assistance may be made available on special terms to developing members… and that for this purpose the Fund shall take into account the level of per capita income.” - Concessional facilities evolved into the Poverty Reduction and Growth Trust (PRGT); a new PRGT eligibility framework agreed in early 2010 determined eligibility based on per capita income, market access, and vulnerability. Under the new framework the number of PRGT-eligible countries was reduced from 77 to 71. These countries are recognized by the Fund to be “low-income developing countries.” - IMF analytical classifications evolved through multiple schemes; since May 1980 the WEO used the IFS classification. The industrial country group was renamed advanced in 1997; additions to the advanced group since 1997 include Israel, Korea, Singapore, Cyprus (2001), Slovenia (2007), Malta (2008), Czech Republic (2009), Slovak Republic (2009). WEO reports generally do not provide rationale for reclassifications. ### Comparison across international organizations - Despite methodological differences, the UNDP, World Bank, and IMF produce similar high-level splits: around 20–25 percent of countries designated as developed. - Differences include: - UNDP uses a relative threshold (75th percentile in HDI distribution). - World Bank uses an absolute threshold (US$6,000 GNI per capita in 1987-prices). - IMF threshold is not explicit; IMF classification appears most likely absolute. - Table 3 from the source summarizes key taxonomy differences and share of countries ‘developed’: - Share of countries 'developed' in 1990: IMF 13 percent, UNDP 25 percent, World Bank 16 percent. - Share in 2010: IMF 17 percent, UNDP 25 percent, World Bank 26 percent. - Table 4 in the source provides country-by-country listings reflecting each organization’s classification (as of October/November 2010 datasets used). ### Alternative development taxonomy methodology (proposed) - Goals and principles: - Data-driven thresholds determined by the observed distribution of development outcomes. - Preserve both mean and ranking of the actual distribution when constructing synthetic distributions. - Use Lorenz curve framework: L(x) is cumulative outcome of lowest x percentile; preserve mean (anchoring) and ranking (no nonsensical re-orderings). - Error metric and optimization: - Define E as the area between line of equality and synthetic distribution segments; minimize E to find thresholds. - Dichotomous taxonomy: show that the error-minimizing threshold between developing and developed countries is the mean development outcome (the population-average). - Trichotomous taxonomy: derive first order conditions for optimal thresholds x0 and x1; thresholds must satisfy that each threshold country’s development outcome equals the average development outcome of a specified complementary set (e.g., threshold between LDC and MDC set at average of non-HDC). - General case: for K categories each threshold xi is chosen such that the development outcome at xi equals the average development outcome of countries in the interval between adjacent thresholds. - Advantages: - Thresholds are determined by the data (relative thresholds) and remain relevant over time. - The methodology is flexible and can be implemented with any number of categories and any ordinal development proxy. ### Examples and empirical illustrations - Three development proxies used to construct taxonomies in the source: - Income (GNI/n, Atlas method, using market exchange rates). - Longevity (life expectancy at birth). - Lifetime income (income × longevity; expected lifetime income of a newborn at 0 percent discount rate, assuming no future income growth). - Weighting variants: - Population-weighted distribution and equally weighted (equal country weights) distribution were both constructed because weighting affects taxonomy (population weighting may better reflect global implications given concentration of population in large countries). - Empirical results (trichotomous taxonomies presented in Tables 5 and 6 of the source): - Tables list countries ranked from highest to lowest development outcome for each proxy and weighting scheme; Middle Development Countries are highlighted in bold in the source tables. - The dichotomous taxonomies corresponding to these trichotomous partitions identify marginal developed and marginal developing countries (noted in the source as marginal developed country and marginal developing country in a dichotomous taxonomy). - Observations from examples: - The alternative dichotomous threshold (mean development outcome) lies well below existing thresholds used by international organizations. - A trichotomous taxonomy provides a better fit for observed diversity and aligns the Higher Development Countries group broadly with existing 'developed' groupings while creating two lower groups that provide more granular distinctions among developing countries. ### Key findings, implications, and policy-relevant points - Existing international classification systems: - Use different bases (absolute vs. relative thresholds) and often lack transparent rationales for specific cut-offs. - Operate with absolute thresholds that, while inflation-adjusted, are not adjusted for trend growth in global real income, causing thresholds to decline relative to average world income. - Proposed alternative taxonomy: - Provides a principled, data-driven method to set thresholds that minimize classification error given an observed distribution of development outcomes. - Offers flexibility to choose number of categories and development proxies (including multivariate indices like HDI or lifetime income). - Produces dichotomous threshold at mean outcome (minimizes error) but recommends trichotomous taxonomy in most practical settings because it better captures heterogeneity. - Practical considerations: - Choice of development proxy and weighting scheme materially affects classification outcomes. - The paper does not resolve what is the optimal number of categories; it defines how thresholds should be set once the number of categories and proxy are chosen. ### Concluding remarks and open questions - The UNDP, World Bank, and IMF reach broadly similar conclusions about which countries are developed, but their thresholds and methods differ and lack full transparency. - The paper’s proposed methodology delivers transparent, data-driven thresholds and can be implemented with any development proxy and any chosen number of categories. - Remaining questions include: - What constitutes the optimal number of categories in a development taxonomy? - What is the most appropriate choice of development proxy or multivariate index for classification purposes? *Source: _wp1131 - Part 1 and Part 2 countries, it was decided in 1964—at the time of the first IDA (PDF chapter content provided).* ### References ### _wp1131 - References ### References: Bibliographic sources - Bell, Clive, 1989, The New Palgrave: Economic Development (New York: W.W. Norton & Company, Inc.) - Boughton, James M., 2001, Silent Revolution: The International Monetary Fund 1979–1989 (Washington: International Monetary Fund) - Chen, Shaohua and Martin Ravallion, 2010, “The Developing World is Poorer than we Thought, but no Less Successful in the Fight Against Poverty,” Quarterly Journal of Economics, Vol. 125 (4), 1577-1625 - Garritsen de Vries, Margaret, 1985, The International Monetary Fund 1972–78: Cooperation on Trial (Washington: International Monetary Fund) - Horsefield, J. Keith, 1969, The International Monetary Fund 1945–1965: Twenty Years of International Monetary Cooperation (Washington: International Monetary Fund) - International Monetary Fund, World Economic Outlook (Washington, various issues) - International Monetary Fund, 2009, Selected Decisions and Selected Documents of the International Monetary Fund, Thirty-fourth Issue (Washington) - International Monetary Fund, 2010, “Eligibility to Use the Fund’s Facilities for Concessional Financing,” January 11 (Washington) - Mason, Edward S. and Robert E. Asher, 1973, The World Bank Since Bretton Woods (Washington: The Brookings Institution) - Pearson, Lester B, et al, 1969, Partners in Development: Report of the Commission on International Development (New York: Praeger Publishers) - Sen, Amartya, 1999, Development As Freedom (New York: Random House) - United Nations, Report on International Definition and Measurement of Standards and Levels of Living, 1954 (New York) - United Nations Development Programme, 1990, Human Development Report 1990 (New York: Oxford University Press) - United Nations Development Programme, 2007, Human Development Report 2007/08 (New York: Oxford University Press) - United Nations Development Programme, 2009, Human Development Report 2009 (New York: Oxford University Press) - United Nations Development Programme, 2010, Human Development Report 2010 (New York: Oxford University Press) - World Bank, 1978, World Development Report 1978 (Washington) - World Bank, 1989, World Development Report 1989 (Washington) - World Bank, 2000, World Development Report 2000/2001 (Washington) - World Bank, 2010, World Development Indicators (Washington), available at http://data.worldbank.org/data-catalog/world-development-indicators/wdi-2010 ### Appendix: Taxonomy error-term derivation and area decomposition - Context: In a taxonomy with N categories, the error term associated with that taxonomy consists of the sum of N triangles and 1N rectangles less the area under the Lorenz curve. - Areas of triangles (as stated): - 1 T: 11 1 () 2 xLx - 2 T: 21212 2211 211 11111 ()(( ) ())( )()( )() 22222 xx LxLxxLxxLxxLxxLx  - 3 T: 323233322322 11111 ()(() ())()()()() 22222 xx LxLxxLxxLxxLxxLx - 1n T : 12121112212 2 11111 ()(()( ))()( )()( ) 22222 nnnnnnnnnnnn xxLxLxx Lxx Lxx Lxx Lx       - n T: 111111 11111 (1 )(1())()() 22222 nnnnnn xLxLxxx Lx     - Sum of areas of the triangles (as expressed): - 122 1111 1111 11111 ()()( )( ) 22222 nnnn iiiiiii nn iiii TxLxx LxxLxLxx   + +      - Areas of the 1N rectangles (as stated): - 1 R: 21 1 21 11 ()()()()xxLxxLx xLx - 2 R: 32 2 32 2 2 ()()()()xxLxxLx xLx - 3 R: 43 3 43 33 ()()()()xxLx xLx xLx - 2n R : 12 2 12 11 ()()()() nn n nn nn xxLx xLx xLx      - 1n R : 11111 (1) ()()() nnnnn xLxLxx Lx    - Sum of areas of the rectangles (as expressed): - 121 11 111 () ( )() nnn iii nii iii RxLx LxxLx  +     - General form for the error term for n categories (as presented across pages): - The document shows a lengthy expression culminating in: - 1 1 11 0 122 1111 111 1 21 11 11 0 22 111 11 () 11111 ()()( )( ) 22222 () ( )()() 1111 ()( )( ) 2222 nn ni i ii nn nnn iiiiiinn iii nn iinii ii nn ii iin ii ET R L x xL xx L xxL xL xx x LxLxxLxLx x LxxLxLx     + +     +    + +       + +          1 1 0 1 () 2 n xLx  +   - By defining 1 n x and recalling the definition of 1 E the document gives a more compact formulation: - 11 111 11 11 ()( ) 22 nn niiii ii EExLx xLx  + +    - First order conditions (as presented, showing structure and multiple terms): - The text presents first order conditions in block form including expressions such as: - 1 212 1 2 132123 2 2432343 3 2121 1 '( )( ) '( ) ()'( )'( ) () '( ) 1 ()'()'() () 0'( 2 ()'()'()1 n n n nnnn n n E x E xL xLx Lx x LxxL xxL xLx Lx E LxxL xxL xLxL x x LxL xx L x E x                                        - Additional structured lines include terms such as: - 2 2 31 31 42 42 1 2 2 ()() ()() () )'() 1( ) 1 n n n Lx x LxLx xx LxLx xx Lx Lx x                                    *Italic: Source: _wp1131 - References (PDF)_* --- _Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1131.pdf_