sEIU:U*I:E±U/E_ REG I ONA L AND SECT UR AL ST U DIES Income, Inequality, and Poverty during the Transition from Planned to Market Economy (71911 Feb. 19q8 RuR5ia's GDP during \>,~~tetransition US GDP,duin;g v th d. reat Depressin \ BRANKO MILANOVIC Income, Inequality, and Poverty during the Transition from Planned to Market Economy WORLD BANK REGIONAL AND SECTORAL STUDIES Income, Inequality, and Poverty during the Transition from Planned to I Market Economy BRANKO MILANOVIC The World Bank Washington, D.C. © 1998 The International Bank for Reconstruction andDevelopment/TheWorldBank 1818HStreet,N.W.,Washington,D.C.20433 All rights reserved Manufactured in the United States ofAmerica First printing February 1998 The World Bank Regional and Sectoral Studies series provides an outlet for work that is relatively focused in its subject matter or geographic coverage and that contributes to the intellectual foundations of development operations and policy formulation. Some sources cited in this publication may be informal documents that are not readily available. The findings, interpretations, and conclusions expressedin this publication are those of the author and should not be attributed in any manner to the World Bank, to its affliated organizations, or to the members ofits Board ofExecutive Directors or the countries they represent. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination ofits work andwill normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910,222 Rosewood Dr., Danvers, Massachusetts 01923, U.S.A. Branko Milanovic is principal economist in the World Bank's Development Research Group. Cover design by Sam Ferro and Sherry Holmber,. The graphic shows that the decline in Russia's gross domestic product (GDP) during the transition (1987 to 1995) was even greater than that of the United States during the Great Depres- sion (1927 to 1935). U.S. GDP in 1927 and Russia's GDP in 1987 = 100. Photograph: church in Tula, Russia, by Jan Pakulski. Library of Congress Cataloging-in-Publication Data Milanovic, Branko. 1957- Income, inequality, and poverty during the transition from planned to market economy / Branko Milanovic. p. cm. - (World Bank regional and sectoral studies) Includes bibliographical references. ISBN 0-8213-3994-X 1. Income distribution-Europe, Eastern. 2. Income distribution- Former Soviet republics. 3. Poverty-Europe, Eastern. 4. Poverty- Former Soviet republics. 5. Europe, Eastern-Economic conditions-1989- 6. Former Soviet republics-Economic conditions. I. Title. II. Series. HC244.Z91547 1997 339.2'0947-dc2l 96-32776 CIP Contents Preface xi 1. The Tectonic Changes 1 Political Developments: The New States 1 Social Costs 6 2. The Way It Was 12 Income Composition and Inequality 12 Ideological Underpinnings 19 3. Income 23 The Post-Communist Great Depression 23 The Effect on Population Incomes: How Level and Composition of Income Changed 30 4. Inequality 40 Income Inequality 40 Distribution of Income Sources: Wages, Social Transfers and Private Sector Income 47 Disparity among Social Groups 54 5. Poverty 60 What Happens to Poverty When Income Goes Down? 60 By How Much Has Poverty Increased? 64 How Much Is Needed to Cover the Poverty Deficit? 76 How to Explain Increases in Poverty? 85 Who Are the Poor? 92 The Incidence of Cash Social Transfers 108 v vi Contents 6. Selected Issues in Social Policy 115 Should OECD-Like Social Assistance Be Introduced in Transition Economies? 115 Transfer Payments to Different Groups of Recipients 120 Guaranteed Minimum Income and the Supply of Labor 122 Informal Sector and Pension Reform 126 7. A Look Ahead 132 Appendices 1. Description ot the Surveys Used and Data Problems 135 The Surveys Used 135 What Biases Are Inherent in the Data? 146 How Biases Affect Comparability between Pre-transition and Transition Years 152 2. Decile Shares of Total Income 156 3. Change in the Poverty Deficit Due to a Uniform Slide in Income 163 4. The Original Income Distribution Statistics 164 5. Poverty Headcount Calculations Based on the Original Income Distribution Statistics Given in Appendix 4 183 6. Distribution of International Funds Based on Minimization of Deprivation Function 192 7. Sources for Table 1.2 193 Country Data Sheets 195 Russia 196 Ukraine 202 Poland 208 Latvia 214 References 221 Indices 233 Contents vii Tables 1.1. Countries at War or under Economic Blockade, 1991-96 4 1.2. Speed of Reforms in Selected Countries, 1992-96 8 2.1. State Employment as a Proportion of the Labor Force, 1988 12 2.2. Composition of Income in Socialist Economies, 1988-89 13 2.3. Composition of Gross Income in Socialist, Market, and Developing Countries, 1980s 14 2.4. Family Allowance for Two Children as Percentage of Average Earnings, 1988 21 3.1. GDP Growth Rates in Eastern Europe and the Former Soviet Union, 1987-96 25 3.2. Russia's Output Decline after the Dissolution of the Czarist Empire and after the Dissolution of the U.S.S.R. 27 3.3. Ratio between Expenditures and Income Reported in Household Budget Surveys 33 3.4. Change in Real per Capita GDP and Real per Capita Population Income between 1988 and 1993 34 3.5. Population Income by Sources in 1987-88 and 1993-94 36 4.1. Changes in Inequality during the Transition 41 4.2. Decomposition of the Change in the Gini Coefficient between Pre-transition and 1993-96 48 4.3. Change in Real and Relative per Capita Income of Worker, Farmer, and Pensioner Households 57 5.1. Estimated Poverty Headcount and Poverty Deficit in 1987-88 and 1993-95 Using HBS Income 68 5.2. Estimated Poverty Headcount and Poverty Deficit in 1993-95 Using a Higher (Macro) Income instead of HBS Income 75 5.3. Estimated Poverty Headcount and Poverty Deficit in 1993-95 Using HBS Expenditures 77 5.4. Explaining Increase in Poverty Headcount between 1987-88 and 1993-94 90 5.5. Relative Poverty Rates for Different Types of Unemployed Households 95 5.6. Concentration Coefficients of Family Benefits before the Transition (1988-89) and in 1993-95 109 5.7. Concentration Coefficients of Unemployment Benefits and Social Assistance, 1993-95 110 5.8. Concentration Coefficients of Non-pension Cash Social Transfers before the Transition (1988-89) and in 1993-95 112 viii Contents 5.9. Percentage of Social Assistance, Unemployment Benefits, and Non-pension Cash Social Transfers Received by the Bottom Quintile Of Population 113 6.1. The Composition of the Poor, 1993-95 117 6.2. Relationship between Poverty Lines and Wages, 1992-94 125 Figures 1.1. Number of Independent States in Europe, 1860-1995 2 1.2. A Map of Europe and Central Asia 10 2.1. Distribution of Social Cash Transfers by Income Decile 17 2.2. Concentration Curve for Pensions in Hungary, Czechoslovakia, and the Russian Republic, 1988-89 18 3.1. Distribution of Countries' Growth Rates in Eastern Europe, 1987-96 24 3.2. Distribution of Countries' Growth Rates in the Former I Soviet Union, 1987-96 24 3.3. Real GDP in Poland and Russia (1987-95); and the United States, and Germany (1927-35) 26 3.4. Real Wages in Poland and Russia (1987-96); and in the United States, the United Kingdom, and Germany (1927-36) 29 3.5. Unemployment Rates in Hungary, Poland, and Russia (1987-95); and in the United States, the United Kingdom, and Germany (1927-35) 30 3.6. Real per Capita GDP and Real per Capita Population Income in 1993-95 35 4.1. Dispersal of Gini Coefficients in Transition Economies 42 4.2. Changes in Quintile Shares in Hungary, Slovakia, and Slovenia between 1987-88 and 1993-94: Little Change 43 4.3. Changes in Quintile Shares in Belarus, the Czech Republic, Latvia, Poland, and Romania between 1987-88 and 1993-94: Moderate Regressive Transfers 43 4.4. Changes in Quintile Shares in Bulgaria, Estonia, Lithuania, Moldova, Russia, and Ukraine between 1987-88 and 1993-94: Large Regressive Transfers 44 4.5. Relationship between Type of Adjustment and Increase in Gini Coefficient 46 4.6. Composition of Disposable Income in Bulgaria, 1987-95 50 4.7. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Bulgaria, 1989-95 50 Contents ix 4.8. Composition of Disposable Income in Poland, 1987-95 51 4.9. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Poland, 1987-95 51 4.10. Composition of Disposable Income in Slovenia, 1987-95 52 4.11. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Slovenia, 1987-95 52 4.12. Composition of Disposable Income in Hungary, 1987-93 53 4.13. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Hungary, 1987-93 53 4.14. Composition of Disposable Income in Russia, 1989-94 55 4.15. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Russia, 1989-94 55 4.16. Composition of Disposable Income in Latvia, 1989-96 56 4.17. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Latvia, 1989-96 56 5.1. Income Density Function and the Poverty Line 63 5.2. Average Monthly per Capita $PPP Incomes, 1993-95 70 5.3. Income Distributions in Moldova and Slovakia, 1992 72 5.4. Estimated Number of Poor before the Transition and in 1993-95 76 5.5. Estimated Annual Costs of Poverty Elimination Assuming Perfect Targeting, 1993-95 78 5.6. Estimated Monthly Spending Required to Lift an Average Poor Person out of Poverty, 1993-95 78 5.7. Optimal Allocation of Funds by Countries if the Objective Function Is to Minimize Deprivation of the Poor 83 5.8. Distribution of Income in the Czech Republic, 1988 and 1993 86 5.9. Decomposing the Change in the Poverty Rate: Parts Due to Change in Income and Change in Distribution 87 5.10. Breakdown of the Increase in Poverty Headcount between Growth and Distribution Effects in Poland, 1990-94 89 5.11. Breakdown of the Increase in Poverty Headcount between Growth and Distribution Effects in Russia, 1990-94 89 5.12. Social Group and Relative Poverty Rates, 1992-95 93 5.13. Poverty Headcount by Socioeconomic Group in Poland, 1987-94 94 5.14. Average Wage, Pension, and per Capita Income as a Ratio of the Poverty Line in Russia, 1987-96 97 5.15. Average Wage, Pension, and per Capita Income as a Ratio of the Poverty Line in Ukraine, 1987-95 97 5.16. Cumulative Wage Distribution in Ukraine (1992) and Poland (1993) 98 x Contents 5.17. Average Wage, Pension, and per Capita Income as a Ratio of the Poverty Line in Poland, 1987-96 100 5.18. Household Size and Relative Poverty Rates, 1993-95 101 5.19. Relative Poverty Rates for Single-parent Households, 1993-95 102 5.20. Age and Relative Poverty Rates, 1992-95 103 5.21. Age and Relative Poverty Rates in Eastern Germany, 1990 and 1992 104 5.22. Education of Household Head and Relative Poverty Rates, 1993-95 105 5.23. Size of Locality and Relative Poverty Rates, 1993-95 106 5.24. Sex, Age, and Relative Poverty Rates in Bulgaria and Russia, 1992-93 107 5.25. Poverty Rates for Male- and Female-headed Households in Poland, 1993 108 6.1. The Effect of Taxes on Supply of Labor of the Working Population 121 6.2. Labor Supply of the Poor (with and without Transfers) 122 6.3. Gross and Net Income if There Is a Minimum Income Guarantee 123 6.4. Marginal Tax Rate in the Presence of a Minimum Income Guarantee 124 6.5. Relationship between the Expected Future Tax Rate and Current Value of Contributions 128 6.6. Shift of Formal Sector Labor Supply Curve as a Function of Current Discount of Contributions 129 7.1. The Year When the Countries Are Expected to Reach Poverty Headcount of 10 Percent 133 Preface This is a book about income, inequality, and poverty during the remarkable period of collapse of Communism and the "construction" of capitalism in eighteen formerly socialist countries. It covers a period of almost ten years, from the time of the early Gorbachevian reforms of 1987-88 to approximately 1996. The book was made possible by two almost simultaneous revolutions that took place in the late 1980s. The first, of course, was the collapse of Com- munism. The second was the opening up of information on income, inequal- ity, and poverty, even in countries (such as the former Soviet Union) where for the better part of the last seventy years such information had been treated as a state secret. This is indeed a period of great turmoil-comparable with the period that followed both World Wars. It is the period of dramatic declines in in- come, the reappearance of diseases long forgotten, growing poverty and un- employment, and great uncertainty. But it is also a period when great for- tunes are being made, consumer goods of incomparably better quality are becoming available for many, and people have the opportunity to control and radically alter their lives. Unlike during some previous episodes of tur- moil, we now have relatively reliable and up-to-date information that allows us to follow and analyze the developments. The goal of this book is precisely this: to describe what happened during the transition in eighteen countries -from the Czech Republic in the West to Kazakhstan and Russia in the East. Specifically, the book will examine what happened to the real incomes of the population, to the inequality with which incomes and expenditures are dis- tributed, and to poverty. It will also attempt to find out why these changes occurred. A word about the data used in this book. All data on incomes and expen- ditures come from household budget surveys. Most of these surveys are con- ducted regularly (on an annual or quarterly basis) by national statistical of- fices, while some are done by independent non-governmental organizations, often in an ad hoc fashion. The surveys are of uneven quality. Some-for ex- ample, the regular, official surveys conducted in most Central European coun- tries-are of a quality comparable with surveys conducted in advanced mar- ket economies. Because the quality of these surveys has remained unchanged xi xii Preface or has improved during the transition, they allow us to measure the impact of the transition relatively well. But in some other countries, the quality of surveys has not improved, and their reliability may have deteriorated as cir- cumstances changed. For example, while the surveys may have once offered a more or less accurate picture of households employed in the state sector, the decline in the size of that sector has significantly reduced the importance of that information. These are only some of the problems peculiar to transition economies. Other data problems, common to surveys everywhere-such as inadequate coverage of the very rich and very poor, differing concepts of income, and differences in survey periods-further complicate the analysis undertaken here. The interested reader can consult appendix 1, where issues related to limitations of the data are discussed at length. The book has its origin in the encouragement I received from Alan Gelb of the World Bank. The book would not have been possible without the large and varied data sources the World Bank has collected and developed during the past seven years of work on the transition economies. In particular, I benefited greatly from Jeni Klugman's and Jeanine Braithwaite's early work on poverty in Russia, recently published by the World Bank as Poverty in Russia: Public Policy and Private Responses. I am grateful to the many colleagues, in the World Bank and elsewhere, who helped me locate necessary data, or who gave me the benefit of their advice. In particular, I am grateful to Jeanine Braithwaite, Mark Foley, and Jeni Klugman, who generously allowed me to use their Russian data. I am also grateful to Anna Ivanova, Nanak Kakwani, Alberto Martini, and Paolo Roberti, who provided me with data for Belarus; Jifi Vecernik and Thesia Gamer (Czech and Slovak Republics); Christiaan Grootaert, Gi-Taik Oh, and Zsuzsa Ferge (Hungary); Helen Jensen and Edmunds Vaskis (Lithuania and Latvia); Carlos Cavalcanti (Estonia); Mansoora Rashid (Romania); Jan Rutkowski and Irena Topifnska (Poland); Tom Hopengartner (Ukraine); Milan Vodopivec and Irena Krizman (Slovenia); Neeta Sirur (Bulgaria); Michael Mills (Kyrgyz Republic); and Timothy Heleniak (countries of the former Soviet Union). The book also benefited from the comments of Martha de Melo, Clara Else, Zsuzsa Ferge, Alan Gelb, Carol Graham, Christiaan Grootaert, Emmanuel Jimenez, Jeni Klugman, Janos Kornai, Mark Kramer, Kathie Krumm, Robert Liebenthal, Costas Michalopoulos, Aleksandra Posarac, Martin Schrenk, and Maciej Zukowski. I am also thankful to the six anonymous reviewers who read the draft manuscript and contributed valuable comments and sugges- tions. Preface xiii Parts of this book were presented at seminars sponsored by various or- ganizations: the Economic Development Institute of the World Bank in Wash- ington, D.C., Harvard University, the Council of Europe in Strasbourg, the Institute of Labor and Social Policy in Warsaw, the European Center in Vienna, and the 1995 meetings of the American Economic Association in San Fran- cisco. I am also grateful for the fine research assistance provided by Yvonne Ying and Nadia Soboleva. Rachel Cantor did an excellent editing job. Virginia Hitchcock ensured the consistency of the style and editing. Vicky Hilliard and Kristi Stoner worked hard to put in shape the text, tables, and graphs: it was not an easy book to produce. To my mother, who has survived it all: monarchy, capitalism, and Nazism; Communism, socialism, and nationalism. 1 The Tectonic Changes June 4,1989, was an eventful day. In Tehran, Imam Khomeini died. In Beijing, tanks rolled in to suppress student demonstrations. In Poland, the first freely contested election ever under a Communist regime took place.' Political Developments: The New States In June 1989, Communist regimes in Europe (including the U.S.S.R.) ruled more than 24 million square kilometers of territory, or 17 percent of the world's land area, and about 420 million people, or approximately 9 percent of the world's population. These people lived in the Soviet Union, six smaller Com- munist countries in Eastem Europe that were allied to the Soviet Union, and two independent Communist states (Yugoslavia and Albania).2 In no coun- try was organized opposition tolerated,' and in no country had genuine mul- tiparty and free elections been held after 1947.4 Eight years after 1989-at the time of this writing-the landscape has en- tirely changed. In all these countries, competitive elections have taken place, even if these elections have not always been fair and transparent. In all these countries, except a few (Armenia, Azerbaijan, Kazakhstan, Tajikistan, Turkmenistan, and Uzbekistan), all kinds of political parties exist.5 1. Although the election was freely contested, the number of seats for the lower house of Parliament was predetermined, giving Communists almost 50 percent of the seats regardless of the outcome of balloting. The first entirely free elections in Eastern Europe were held in March 1990 in Slovenia and Hungary. 2. Mongolia is not included in the analysis. 3. Technically, Bulgaria, the German Democratic Republic, and Poland were multi- party states. Elections were held with predetermined seat allocations, however, and the so-called opposition parties functioned as faithful allies of the ruling Communist party. Yet, in a historic twist, it was the defection of the two "faithful" allied parties after the June 1989 Polish elections that allowed non-Communists to form the government. 4. The last free election in any of these countries was the Hungarian election in August 1947. 5. Turkmenistan had only a single-candidate election for Parliament (in December 1994). The president's term was simply extended by a referendum. In Uzbekistan, Birlik and Elk opposition groups are banned and their leaders exiled. In Kazakhstan, parlia- mentary elections in December 1995 featured almost no opposition parties, but a collec- tion of government-sponsored parties, while the president's term was extended by refer- endum. In Azerbaijan, the president was elected in September 1993 with 99 percent of the vote and virtually no opponents. In February 1996, three major opposition parties were banned. In Armenia, the major opposition party Dashnak was suspended in December 1994. In Tajikistan, a collection of opposition groups and the government are at war. 1 2 Income, Inequality, and Poverty during the Transition The three Communist federations-Czechoslovakia, the Soviet Union, and Yugoslavia-have disintegrated, spawning in their stead a total of twenty- two countries. The number of countries in the world thus rose by approxi- nmately 15 percent. One Communist country (the German Democratic Re- public) disappeared. Only five of the former European Communist countries (Albania, Bulgaria, Hungary, Poland, and Romania) have remained within the same borders. Approximately 345 million are now citizens of countries that did not exist prior to 1992. Forty-three million people in the forner U.S.S.R. and 1.5 million in the former Yugoslavia now find themselves living outside their ethnic republics.6 As figure 1.1 shows, the number of countries in Europe is now greater than it has been in the last 140 years-even slightly greater than it was in 1860 before the Italian and German unifications. The difference now is that while in 1860 the West was segmented (Italy in three states and Germany in twenty- one), Eastern and Central Europe were largely unified under the Habsburg, Ottoman, and Russian Empires.7 Figure 1.1. Number of Independent States in Europe, 1860-1995 50 Breakup of Communist 40 federations Aftermath of World WarTI 30 20 German and l0 Italian Nazi unifications rule 0- I I I I 1. 1860 1880 1900 1920 1940 1960 1980 1995 Source: Calculated using Centennia software program. 6. The number for Yugoslavia does not include Bosnia and Herzegovina, the only republic of twenty-three in the U.S.S.R., Czechoslovakia, and Yugoslavia not to have had a "titular" nationality. 7. The Habsburg empire spun off six countries: Austria, Croatia, the Czech Repub- lic, Hungary, Slovakia, and Slovenia. Also, parts of the Empire were, at different times, taken over by Italy, Poland, Romania, and Yugoslavia. The Russian empire spun off twelve European countries: Armenia, Azerbaijan, Belarus, Estonia, Finland, Georgia, Moldova, Latvia, Lithuania, Poland, Russia, and Ukraine. The Ottoman empire spun off three countries: Albania, Bosnia, and Bulgaria. The Tectonic Changes 3 The current emergence of new countries implied huge needs for institu- tional build-up, since many republics (particularly in the former Soviet Union) had underdeveloped national institutions and little experience running their own affairs. They are all practically "new" countries in the sense that they either had almost no experience with self-government or were never inde- pendent.8 Of the seventeen countries born out of the ruins of the three Com- munist federations (not including Central Asian countries), five had not been independent states during the last two centuries: Bosnia, the Czech Repub- lic, Macedonia, Moldova, and Slovenia. Five (Armenia, Azerbaijan, Belarus, Georgia, and Ukraine) were independent for three years or less during the turmoil of the Russian civil war that followed the 1917 Revolution. Two coun- tries (Croatia and Slovakia) were independent for a brief period of four to five years during the Nazi rule over Europe. The three Baltic states had ap- proximately twenty years of independent existence between the two World Wars. Serbia was independent for about sixty years, from about 1860 until 1915, before it merged into Yugoslavia in 1918.9 And, finally, Russia, the big exception, went through practically all the stages: an Empire, a Soviet Re- public, and now, as the Russian Federation, an independent democratic coun- try. Excluding Russia, however, the average duration of independence for the other 'new" countries was less than nine years over a 200-year period from the French Revolution until 1990. In addition to the increase in the number of recognized countries, a num- ber of states not recognized internationally have appeared-a new phenom- enon in Europe. In virtually all countries with civil strife (see table 1.1), re- gions have in their turn declared independence. Often such regions control their own territory, a fact which in the past was sufficient to ensure them international recognition. Such self-declared independent states include the Transdnestrian Republic in Moldova, Chechnya in the Russian Federation, South Ossetia and Abkhazia in Georgia, Nagorno-Karabakh in Azerbaijan, and Kosovo in Yugoslavia. It is revealing that all regions, except the Transdnestrian Republic, enjoyed the status of autonomous republic in the U.S.S.R. or of province in the former Yugoslavia-a status one step below that of a republic.'" The fissure along republican boundaries that rendered asunder the three Communist federations is apparently continuing at the next level of administrative organization. 8. Independence and self-governance are not necessarily the same thing: Hungary, for example, was self-governing in the Austro-Hungarian Empire but was not indepen- dent. 9. Montenegro, interestingly enough, had more experience of independence than any other country (except, obviously, Russia). Both Montenegro and Serbia, as well as Bulgaria and Romania, were officially recognized as independent states at the 1878 Berlin congress. 10. Transdnestria was "added" to Moldova (then called Besarabia) when Besarabia was annexed by the Soviet Union in 1940. 4 Income, Inequality, and Poverty during the Transition Table 1.1. Countries at War or under Economic Blockade, 1991-96 Estimated Percentage War on its internally decline in territory displaced GDP (pre-war Estimated persons (DP) between population) dead and refugees 1987 and Economic Country in millions) (in thousands) (in millions) 1996 sanctions Armenia No (3.2) 0 DPs: 0.376 58 Azeri and Ref.: n.a. Turkish blockade Azerbaijan Yes (7.2) 15 DPs: 0.9 66 No Ref.: n.a. Bosnia Yes (4.4) 250 DPs: 2.7 70a No Refs.: 0.6 Croatia Yes (4.8) 20 DPs: 0.187 47 No Refs.: 0.4 Georgia' Yes (5.5) 11 DPs: 0.28 67 Yes Refs.: n.a. Macedonia No (2.1) 0 n.a. 47 Greek blockaded Moldova Yes (4.3) 1 DPs: 0.015 61 No Refs.: n.a. Russia Yes (148.2) 100' DPs: 1.4 38 No Refs.: 0.3 Tajikistan Yes (5.2) .50 DPs: n.a. 70 No Refs.: 0.1 Yugoslavia No (10.4) 0.2 DPs: 0.65 41 UN Refs.: 0.2 sanctionsb Total 7 countries "450 =8f 40 with wars; 48 million peoples n.a. = not available. a. Estimate. b. Imposed in May 1992; suspended in November 1995. c. Includes two conflicts: in Abkhazia and Southern Ossetia. d. Lifted in September 1995. e. Casualties in Chechnya. f. The numbers cannotbe added because of double-counting: out-refugees of one country are in-refugees of another. g. Excluding Russia. Source: Data on refugees in Bosnia and Croatia are from UN High Commissioner for Refugees, August 1995 (reported in Nasa Borba, August 20,1995). For Federal Republic of Yugoslavia, from the census of refugees in April-June 1996. For countries of the former Soviet Union, from International Migration Bulletin, No. 6, May 1995 (data at the end of 1994) and Heleniak (1997). Sources for casualties: Tajikistan: Akchurin (1995); Azerbaijan: The Economist, November 13,1993, p. 61; Croatia and Bosnia: Zimonjic-Peric (1995); Russia: RFE/RL Daily Service April 1994; Moldova, Georgia and Azerbaijan: Gurr (1994). The Tectonic Changes 5 War and civil strife have affected approximately 50 million people in the war-tom countries profiled in table 1.1. Almost half a million people have been killed and about 8 million (or approximately 15 percent of the countries' population excluding Russia) have become refugees or displaced persons, fleeing either persecution and war or leaving the newly independent states where they are minorities for the relative safety of their "mother countries." As shown in table 1.1, 4 million refugees and displaced people are from the territory of the former Yugoslavia (or almost 20 percent of the former Yugoslavia's population); 1.5 million are from the Transcaucasian countries; and 1.5 million are from the Russian Federation. Those in war-torn countries who have not been killed, maimed, or injured, or who have not become refu- gees, have experienced plummeting standard of living. Best available esti- mates suggest that nowhere has real per capita income decreased by less than one-third. Major economic and social dislocations, of a size unseen since the Second World War and its aftermath, are clearly underway. By the end of 1997, only the bloodiest of these conflicts, the Bosnian had been resolved. The intensity of all the conflicts, however, was less by the end of 1997 than it had been one or two years before."' Political instability has also been accompanied by coups, successful and unsuccessful popular uprisings, and assassination attempts. A bloody coup against Georgia's first democratically elected president brought Georgia's former Communist party boss back to power. Later, Shevardnadze himself was the target of a failed assassination attempt. Similarly, a creeping coup against Azerbaijan's first elected president brought Azerbaijan's former Com- munist leader, Haidar Aliyev, to power. Aliyev himself later faced an unsuc- cessful coup. In Russia, the president dissolved Parliament, and the Parlia- ment responded by attempting an armed rebellion. The president bombed the deputies out of the Parliament building. Macedonia's first president nar- rowly escaped an assassination attempt. The results of Armenia's presiden- tial elections were so contested that the re-elected president had to call in tanks and troops against his opponents. In Serbia, the ruling Socialists re- fused to accept electoral defeat until three months of peaceful demonstra- tions and international pressure obliged them to do so. In Albania, after the fraud-marked parliamentary elections and collapse of a score of government- sponsored pyramid schemes, the country plunged into anarchy. In Tajikistan, an assassination attempt on the president failed but left several people dead. The building of new democratic institutions is also hampered by the pau- city of democratic traditions in virtually all transition countries."2 If democ- 11. The Abkhaz-Georgian, Nagomo-Karabakh, Chechen, and Transdnestrian conflicts are in the cease-fire stage; war in Tajikistan is much less intense. 12. For a historical overview of political developments in Eastem Europe, see Polonsky (1975) and Walters (1988). For Hungary, see Sugar, Hanak, and Frank, eds. (1990). For Bulgaria, see McIntyre (1988). For the Baltic countries, see von Rauch (1974). For Transcaucasia, see Goldenberg (1994). Very useful is also a beautiful historical atlas with the text by Magocsi (1993). 6 Income, Inequality, and Poverty during the Transition racy is defined as (a) universal and secret (male) suffragel3 with competitive party elections, from which no important parties are banned; and (b) a gov- ernment responsible to a parliament or to a democratically elected president, then the longest pre-1990 democratic tradition is that of the Czech Republic and Slovakia, which have twenty-three years of experience (1918-38 and 1945- 48). Romania and Serbia each have about twenty years of (checkered) experi- ence with democracy,'4 followed by Latvia and Estonia each with about thir- teen years (1920-33). Several countries had only a few years of experience with democracy: Bulgaria during 1919-23 and 1926-35, Lithuania during 1920-26, Poland during 1922-26, and Armenia, Azerbaijan, and Georgia dur- ing their brief periods of independence. All other countries under discussion here have had no democratic experience at all-only a few democratic epi- sodes, such as the period from March to November 1917 when Russia was ruled by the provisional government, or Hungary between 1945 and 1947. During the entire period following World War I, when universal franchise became the norm in many Western countries, and until 1990, countries in Eastern Europe and the former Soviet Union had an average of nine years of democratic experience, and no country had lived in a democratic system for more than twenty-five years. Social Costs The total value of goods and services produced by the vast area covered by the transition economies, extending from the Baltic and the Adriatic in the West to the Northern Pacific in the East, has declined since the transition started by at least one-quarter in real terms. Expressed in dollars, the decline has been even steeper, as many local currencies have depreciated. In 1989, the value of this area's output, assessed at realistic (that is, not official) exchange rates, was about US$1.2 trillion, or virtually the same as the gross domestic product (GDP) of the Federal Republic of Germany, and three times that of China."5 The average an- nual per capita income was about $3,000. In 1996, transition economies (exclud- 13. Democracy is defined in this restricted way in order to identify discrimination based on social class and income, not gender. 14. Romania, from 1919 to 1938, and Serbia from 1903 to 1914, and then (as part of Yugoslavia) from 1919 to 1929. (Obviously, other successor states of the former Yugoslavia also experienced democracy between 1919 and 1929.) However, in both Romania and Yugoslavia, there were problems. The respective prerogatives of the kings and parlia- ments were not dearly spelled so that govemnments sometimes depended more on kings than on parliaments (for example, Romania under Carol II during 1930-38, and Yugosla- via under Alexander I during 1919-29). Not surprisingly, both kings (Alexander in 1929 and Carol in 1938) eventually suspended the constitution and assumed full power. Also, fraud was massive during several Romanian elections and two large parties were banned in Yugoslavia (the Communists after 1921 and the Croat Peasant Party briefly in 1925). 15. All data are from various years of the World Bank Atlas. All dollar amounts are current U.S. dollars. The Tectonic Changes 7 ing eastern Germany) produced goods and services valued at approximately $880 billion, or only 40 percent of the united Germany's GDP and just 20 percent more than China's GDP. The GDP of the Netherlands is equal to that of the Russian Federation; the GDP of Finland is equal to that of Poland; and the GDP of Hong Kong, China, is greater than the GDP of Ukraine. The region's average GDP per capita has declined to about $2,000. After the Great Depression of 1929- 33, this decline represents the largest peacetime contraction of world output. At the same time, poverty has increased substantially in the region. While it was estimated that, in 1989, the number of people living on less than $4 per day (at international prices) was 14 million (out of a population of approxi- mately 360 million), it is now estimated that more than 140 million people live below the same poverty line.16 (See chapter 5 for the calculations and the discussion.) Social transfers, and free health care and education, once taken for granted, are now rapidly shrinking. Mortality and morbidity, particu- larly in the countries of the former Soviet Union, have increased substan- tially in some cases, without peace-time precedent (Heleniak 1995). Unem- ployment, almost nonexistent (except in the former Yugoslavia) before the transition, affected more than 15 million people by 1996 and is still rising. On a more positive note, the economic landscape is rapidly changing for the better. In all the countries studied here, state ownership is being replaced by private ownership (see table 1.2). While in 1989, only Hungary, Poland, and Yugoslavia were members of the International Monetary Fund (IMF), and none of them had a convertible currency, by 1996, thirteen of eighteen countries shown in table 1.2 had convertible currencies. As an indicator of monetary stabilization, by 1996, triple-digit inflation was present in only two countries (Bulgaria and Turkmenistan), and in several countries (the Czech Republic, Slovakia, and Slovenia), inflation was reduced to single-digit levels. Such massive dislocations-the creation of new states, civil wars, and de- clines in GDP-have had huge social costs. Conceptually, we can divide these costs into three categories. First, costs associated with decreases in output due to systemic changes (that is, the transition to market economy) and to macroeconomic stabili- zation. These costs are expressed in lower incomes, higher inequality, and greater poverty. Second, job-loss costs associated with the transition. Job-loss is sometimes accompanied by poverty but not always. Unemployment is a distinct issue from poverty. Third, costs associated with civil strife. These are costs of lives lost, people becoming refugees, and destruction of property. In this book we shall deal only with the first type of costs. The second (unemployment) will be discussed only to the extent that it affects poverty 16. This calculation does not include the countries at war or with civil strife listed in table 1.1. (Russia is included though.) Table 1.2. Speed of Reforms in Selected Countries, 1992-96 Current account Estimated privatized Non-state sector Share of retail trade (exchange Explicit producer state assets share in total sales in private hands rate) subsidies CPI inflation Country (percent) employment (percent) (percent) convertibility (percent of GDP) in 1996 Belarus Slow 28 (96) 39 (96) No 9 (92) 40 13 (95) Bulgaria Slow No 310 Czech Republic Es Estonia Fs Hungary Medium 4 9)2 49 (end 94) Kazakhstan Medium 37 (95) 5 (93) 29 8 (92) Kyrgyz Medium T3 Republic 35 (93) ...... Latvia Eas Lithuania Fast Moldova Slow Current Non-state sector account Estimated privatized share in total Share of retail trade (exchange Explicit producer state assets employment sales in private rate) subsidies CPI inflation Country (percent) (percent) hands (percent) convertibility (percent of GDP) in 1996 Poland Medium 64(4 ). ... 35 (93). Romania Slow 9 No 6 (93) 57 8 (94) Russia Fat63(96 7(6)Ys7 (93) 2 70 (end 94) _ - _ Slovakia Medium 21 (92) 949)Ys5 (926 Slovenia Slow 28 (94) Turkmenistan Slw6 9)N 9)100 Ukraine Slow 33 (95) - 76 (96-),Ye 4(96) 40 1 (93) . . - -:B_.-.-.. ___ Uzbekistan Slow 63 (95) 74 (96) No 86 Note: Areas are shaded if a variable takes the value at least equal or 'better" than "fast" for privatized state assets, 40 percent for non-state share in total employment, 50 percent for share of private retail trade, 'yes" for current account convertibility, 'less than 5 percent" for producer subsidies, and "less than 20 percent" for inflation. All data in parentheses refer to the year. Source: See appendix 7. OL LZ6[ 13]YVAON ......... ... ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ - ~~~~~~ -4 ~ ~ ~ ~ ~ ~ ~ ~~~~~~~ a ~~~~~~~~~~~~~~~~~~~~~~~~~~ 1' 1 ~~~~~~~~~~~~~~~~ ~~~~~ Z 0 ... .. .. .. .........."~ ... .....s...1uJ unao~ 3f v~y ~~ The Tectonic Changes 11 and inequality. The third type is not, properly speaking, a cost of economic transition, but rather the cost of nation-building and the creation of a "new world order" Responsibility for the latter cost lies with politicians, not economists. In this book, the topics of income, inequality, and poverty shall be dis- cussed only in reference to the countries listed in table 1.2, that is, the eigh- teen countries unaffected by military conflict.17 They are shown in figure 1.2. The organization of the rest of the book is as follows. Chapter 2 will review the key features of income composition, income inequality, and social policy in the former system. Chapter 3 will look at what happened to the population's real income-its size and composition-during the transition. Chapter 4 will analyze the inequality with which income and expenditures are distributed and the change in the relative position of different social groups during the transition. Chapter 5 will assess what happened to poverty under the twin impact of declining income and grow- ing inequality. Chapter 6 discusses social policy issues related to the tran- sition. Chapter 7 presents a brief look at what might lie ahead. 17. The only exception is Russia, which had a limited military conflict on its terri- tory, but is obviously too important not to include. 2 The Way It Was Income Composition and Inequality The defining characteristic of socialist countries was state ownership of the means of production. Though this ownership took various forms-direct state ownership, "social" ownership, or some form of collective ownership-and though the management rights of the state varied-from direct allocative cen- tral planning to bureaucratic interference in decentralized decision-making- the role of the state was undeniably major. On average, 90 percent of the labor force was employed by the state as compared to members of the Organisation for Economic Co-operation and Development (OECD), where an average of 21 percent was so employed (see table 2.1). On the household level, the role of the state was reflected in the fact that most income was received through the mediation of the state: through wages paid to employees working in state- owned enterprises (SOEs) (including farms) or in government; pensions paid out of state-administered funds or directly out of the budget; and family al- lowances, scholarships, etc. Only in countries with private agriculture (Po- land, Yugoslavia, and, more recently, Bulgaria and Hungary) were private sources of income of some importance. In other countries, private income was Table 2.1. State Employment as a Proportion of the Labor Force, 1988 (percent) Country Share Socialist average 90.0 Czechoslovakia 98.8 U.S.S.R. 96.3 Romania 95.2 German Democratic Republic 94.7 Hungary 93.9 Bulgaria 91.5 Yugoslavia 78.9 Poland 70.4 OECD average 21.2 Note: All averages are unweighted. Czechoslovakia, Poland, and Romania, 1989; German Democratic Republic, 1987. The state sector includes the government, social services run by the state (health and education), and state-owned enterprises (SOEs), including agricultural cooperatives (kolkhozes in the U.S.S.R.). "Social sector" in Yugoslavia is treated as a state sector. OECD data from the late 1970s and mid-1980s. Source: Milanovic (1994a, appendix 1). 12 The Way It Was 13 minimal and limited to income from small-scale agricultural plots, self-con- sumption, gifts or remittances, moonlighting, and black market activities. Table 2.2 shows the composition of household income in socialist countries before the transition. Countries are arranged, from left to right, in the order of the increasing importance of private sources in total gross income.' At one end of the spectrum is Czechoslovakia, where an almost total "socialization" of agriculture and severe restrictions on non-agricultural private business meant that only a small fraction (5 percent) of total income was not received through state mediation. In the Soviet Union, approximately 14.4 percent of gross income was derived from private sources. In Bulgaria, Hungary, Yugo- slavia, and Poland, the private income share was greater than 20 percent. Labor income-virtually all of it derived from the state-ranged from 53 percent of total gross income (in Poland) to 72 percent (in the U.S.S.R.). Social transfers ranged between 13 percent (Yugoslavia and the U.S.S.R.) and 25 percent (Czechoslovakia) of gross income. Table 2.2. Composition of Income in Socialist Economies, 1988-89 (gross income=100) Czecho- Income source slovakia U.S.S.R. BulgariaHungary YugoslaviaPoland Primary income 72.9 78.8 71.2 71.7 83.1 78.2 Labor income 69.5 72.0 56.5 55.0 62.2 53.0 Self-employment income 3.4 6.8 14.7 14.0 20.9 25.2 Property income n.a. n.a. n.a. 2.7 n.a. n.a. Social transfers 25.4 13.6 21.2 22.4 13.3 20.7 Pensions 16.5 8.0 16.6 13.4 12.1 14.3 Child benefits 5.6 1.2 2.3 6.0 1.2 5.2 Other cash transfers 3.3 4.4 2.3 3.0 0.0 1.2 Other incomea 1.7 7.6 7.6 6.0 3.6 1.1 Gross income 100.0 100.0 100.0 100.0 100.0 100.0 Personal taxes 14.2 n.a. n.a. 16.5 1.2 1.6 Direct taxes 0.0 n.a. n.a. 10.7 1.2 1.6 Payroll tax (employee)b 14.2 0.0 0.0 5.8 0.0 0.0 Memo: Private incomec 5.1 14.4 22.3 22.7 24.5 26.3 n.a. = not available. a. Includes private transfers (gifts, alimony, remittances), insurance and lottery receipts, and rental income (if not included in property income). b. Except in Hungary and Czechoslovakia, the entire payroll tax was paid by enterprises. c. Equal to self-employment income, property income, and other income. Source: Czechoslovakia: Microcensus 1988. Poland, Bulgaria and Yugoslavia: 1989 Household Budget Surveys; Hungary: 1989 Income Survey; U.S.S.R.: 1988 Family Budget Survey. 1. The entire wage income is here allocated to the state sector. This is only a very slight exaggeration because the private sector (outside of self-employment) was minimal. 14 Income, Inequality, and Poverty during the Transition Table 2.3. Composition of Gross Income in Socialist, Market, and Developing Countries, 1980s (percent) Socialist Market Developing countries countries countries Primary income 77 85 90 Labor income 63 64 35 Self-employment income 13 14 48 Property income 1 5 6 a Occupational (private) pensions 0 2 0 Social transfers 19 14 3 Pensions 13 12 2 Child benefits 4 1 0 Other cash transfers 2 1 0 Other income b 6 1 7 Gross income 100 100 100 Total taxes 34 38 n.a. Direct taxes 3 20 n.a. Payroll tax (employee) 7 5 n.a. Payroll tax (employer) c 24 13 n.a. GDP in '000 $PPP (1988, per capita) 5.5 14.0 1.8 d Note and Source: All averages are unweighted. Socialist is the average of the countries listed in table 2.2. Market economies are the following eleven OECD countries: Australia, Canada, France, West Germany, Israel, New Zealand, Norway, Spain, Sweden, the United Kingdom, and the United States. All of the data are from 1979-81, with the exception of Australia and New Zealand (1985-86) and Spain (1988). The source for the first seven countries is LIS data reported in O'Higgins, Schmaus, and Stevenson (1989, p. 107); New Zealand and Australia from Saunders, Stott, and Hobbes (1991); France from Sologoub (1988, p.5); Spain calculated from INE (1989, pp. 366-67). Developing countries are represented by C6te d'Ivoire, Ghana, Jordan, Peru, Madagascar, and Vietnam with sources respectively Kozel (1991, p.15), Boateng and others (1992, p. 22), World Bank (1993, p. 33), World Bank (1994), World Bank (1996c, p.22), and Vietnam State Planning Comrmittee General Statistical Office (1994, p. 218). n.a. = not available. a. Includes inputed rent. b. Includes private transfers (gifts, alimony, remittances), insurance and lottery receipts. c. Estimate based on 40 percent enterprise-paid payroll in socialist countries and 20 percent in market economies. d. Excludes Vietnam. Features specific to socialism are more apparent when the average income composition for former socialist countries is contrasted with income composi- tion of developed market economies and of developing countries (see table 2.3). There are three such features. 1. The share of primary income, that is, income that results from eco- nomic activity (labor, capital, entrepreneurship), was smaller in so- cialist than in market economies or developing countries. This is a re- flection of three phenomena: (a) virtual absence of property incomes; (b) absence of occupational pensions (which are considered part of primary income); and (c) greater importance of income redistribution, The Way It Was 15 whether via the state (social transfers in socialist countries represented 19 percent of gross income as compared with 14 percent in market economies) or privately (6 percent in socialist countries as compared with 1 percent in market economies).2 2. Much lower direct taxation under socialism (personal income taxes plus employee-paid payroll tax): 10 percent of gross income as com- pared with 25 percent for market economies. Total payroll taxes were high, however (between 40 and 50 percent of net wages), and the over- all tax burden was not much different from that found in market econo- mies. 3. Child benefits were more important in socialist countries than in mar- ket economies (4 percent of gross income as compared with 1 percent). In addition, the following regularities ("stylized facts") regarding the dis- tribution of income in socialist countries are generally accepted. 4. Overall income distribution was more egalitarian than in most market economies even after allowing for (a) fringe benefits and various forms of implicit income received by the nomenklatura and (b) direct subsi- dies.3 The distribution of direct subsidies generally favored the poorer segments of the population. 4 Element (a) pulled income inequality up while element (b) pulled income inequality down. When both nomenklatura benefits and subsidies were added to measured income, overall inequality did not change much. The Gini coefficient was in the range of 23 to 265 or slightly above that of very egalitarian Nordic countries, but definitely below inequality in other OECD countries, or other countries at similar level of development. 6 2. Most of other income is private transfers. 3. There are numerous studies comparing income inequality in socialism and capitalism. I will cite only a few: Atkinson and Micklewright (1992), which present the most detailed study of inequality and poverty in Central Europe (Czechoslovakia, Hungary, and Poland) and Russia during the post-War period. Phelps-Brown (1988) and Morrisson (1984) contrast East and West European income distribution. Muller and oth- ers (1991) compare the German Democratic Republic and the Federal Republic of Ger- many. Bergson's (1984) excellent review of inequality in the Soviet Union, with its exten- sive bibliography, is a dassic in the field. Good comparative studies of several socialist countries include Asselain (1987), Kende (1987), Debroy (1986), and Flakierski (1986,1989, 1993). 4. See Milanovic (1994, pp. 193-95), where results of several studies of East Eu- ropean countries are discussed. For the German Democratic Republic, see Bird, Frick, and Wagner (1995, p. 6), who show that the simultaneous inclusion of consumer subsi- dies and benefits for the nomenklatura leaves the calculated Gini coefficient unchanged. For discussion of Czechoslovakia, see Hirsl, Rusnok, and Fassman (1995, p. 2). 5. Except in Yugoslavia, where it was greater because of huge regional differences in average income. For the exact values of the Gini before the transition, see table 4.1. 6. In the work on the Kuznets' curve, it was standard practice to introduce a dummy variable for socialist countries. The dummy variable was always found significant and negative. That is, socialism was found to reduce inequality below its expected level (see, for example, Ahluwalia (1976) and Kaelble and Thomas (1991)). 16 Income, Inequality, and Poverty during the Transition 5. Cash social transfers were distributed alnost equally per head in so- cialist countries in contrast to some market economies where such trans- fers were more focused on the poor.7As shown in figure 2.1, the abso- lute amounts of cash transfers in Czechoslovakia, Hungary, and Poland were approximately equal for the poor and the rich (a value of 1 indi- cates that the share of the decile in the total amount of transfers was 10 percent; a value of 2 indicates that the decile of recipients received 20 percent of transfers, and so on). The concentration coefficient of cash transfers was thus very close to zero in socialist economies.8"9 The approximate equality of cash transfers per capita regardless of the re- cipients' position in income distribution was the outcome of two separate distributions: pensions, and other cash transfers, among which family allow- ances were the most important. 10 Pensions tended to be slightly pro-rich in absolute terms. The positive cor- relation between recipients' positions in income distribution and their level of pensions was stronger in countries where pensions were relatively high (that is, where the average-pension-to-average-wage ratio was greater than 7. Market economies differ greatly in terms of concentration of cash transfers. Dif- ferences in results are driven by differences in (a) how pensions are administered (for example, whether the pension system is mostly state-financed, in which case pensions are a part of social transfers, or differently, occupational and other private pension schemes are important), and (b) the relative position of pensioners (whether they are relatively well off or not). Thus, while some countries (for example, Australia, Finland, and Swit- zerland) target social transfers on the poor (so that the concentration coefficients of social transfers is around -10), other countries (Belgium, Germany, and Sweden) have strongly positive concentration coefficients for transfers showing that there is a positive correla- tion between disposable income and the amount of received social transfers. For the defi- nition of the concentration coefficient, see footnote 9 below. (Data on market economies are derived from the LIS data base, and cover the late 1980s to the early 1990s; see Mlanovic 1995c, table 2.) 8. The discussion of the distributional incidence of transfers (stylized fact 5) and taxes (stylized fact 8 below) is based on Milanovic (1994, p. 178) and Milanovic (1995). Using an equivalent scale rather than a per capita measure (for example, assigning less weight to children than to adults) would tend to place families with children higher, and pensioners lower, in the income-distribution scale. Because pensions account for a bulk of cash social transfers, the distributional impact of cash transfers might then become more pro-poor. 9. The concentration coefficient shows the concentration (or cumulative percent- age) of an income source (for example, social transfers or wages), when recipients are ranked by amount of disposable income. The coefficient ranges from -1 (or -100 if ex- pressed in percentages as is done here), when the entire income source is received by the poorest (by income) recipient; through 0, when all recipients receive the same amount; to +100 when the entire income source is received by the richest recipients. A concentration coefficient's negative (or positive) value shows that a given source is negatively (or posi- tively) correlated with overall income. 10. Family allowances indude all family-related transfers: birth grants, maternity pay, regular monthly family allowances, and so on. The Way It Was 17 Figure 2.1. Distribution of Social Cash Transfers by Income Decile 2.0 - United Kingdom , 1.5 Ireland! 1.0 0.5 1 2 3 4 5 6 7 8 9 10 Income decile Note: Equal per capita distribution = 1. CSSR = Czechoslovak Socialist Republic. Source: Milanovic (1995, p. 495). 50 percent) as in Eastern Europe. In the late 1980s, pension concentration coefficients ranged between +5 and +10 in Eastern Europe (see the concen- tration curves for Czechoslovakia and Hungary in figure 2.2, which just barely depart-to the right-from the line of perfect equality). 11 By contrast, in the Soviet Union, where the pension-wage ratio was smaller, pensioners belonged to lower income groups, and the concentration coefficient was strongly nega- tive (see data for Russia in figure 2.2). Transfers other than pensions were mildly pro-poor with concentration coefficients ranging between -5 and -15. Most progressivity of non-pension transfers was due to family allowances and other family-related transfers (for example, child care and maternity al- lowances). Thus, the mildly pro-rich or neutral pensions and the mildly pro- poor family allowances combined to produce a basically flat distribution of 11. The concentration curve shows the cumulative percentage of income source (on the y axis) against the cumulative percentage of recipients (on the x axis). An income source is said to be pro-poor in absolute terms, if its absolute amount declines with in- creases in income (for example, social assistance). Such a source is associated with a nega- tive concentration coefficient and a concentration curve that lies above the 45-degree line (for example, see pensions for the Russian Republic in figure 2.2). The opposite is true for a source that is pro-rich in absolute terms. 18 Income, Inequality, and Poverty during the Transition Figure 2.2. Concentration Curvefor Pensions in Hungary, Czechoslovakia, and the Russian Republic, 1988-89 100 - .~80 CSR1988 0 4 Russia 1989 / 60- 40- CZ 40 0 20 40 60 80 Individuals ranked by household per capita income Note: CSSR Czechoslovak Socialist Republic. Source: Milanovic (1995, table 1). social transfers-social transfers were paid more or less equally across the entire income spectrum. 6. In countries where private sector incomes were low, the share of pri- vate sector incomes in gross income was broadly flat across income distribution. In countries with private agriculture, where private in- comes were more important, the share of private income in gross in- come exhibited an inverted U-shaped pattern-being high in relation to overall income at the low and high end of income distribution. This meant that most "private-oriented" households were either poor agri- cultural households (often living barely above the subsistence level) or very successful, even if not very numerous, private businesses (see Milanovic 1991, and Marnie and Micklewright 1993, figure 5). 7. There is some debate regarding inequality of wage distribution. Ac- cording to some (for example, Atkinson and Micklewright 1992, chap- The Way It Was 19 ter 4; Phelps-Brown 1988, pp.39-51; and Lydall 1968, pp. 152-62), wage distribution in socialist countries was more egalitarian than in market economies. 12 According to others (for example, Redor 1992), the dif- ferences between the two distributions were not systemic-that is, they could be explained by elements other than the difference in the eco- nomic system, such as the size of the country, enterprise concentra- tion, and participation rates (although the latter two can be regarded as system induced). Most authors, however, agree that the difference between average pay of non-manual and manual workers was less in socialist countries than it was in market economies and that returns to education were therefore less (see Redor 1992, pp. 60ff.; Rutkowski 1996; and Jackman and Rutkowski 1994).'3 8. Direct taxation was proportional to wages because most direct taxes were paid as proportional payroll taxes. Progressive direct taxes were minimal and no country had a personal income tax system similar to systems common in the West."4 Taxes had virtually no redistributive effect: if, on average, direct taxes were proportional to wages, which were not distributed much differently from overall income, it follows that tax distribution was close to the distribution of income-taxes, in other words, were proportional (or "flat," to use the current terminol- ogy) rather than progressive (Milanovic 1994, pp. 184-86). These eight regularities ("stylized facts") define the distribution of income in socialist countries relatively well. But what was the logic of the system that engendered such distribution? Ideological Underpinnings According to socialist ideology, most of the population was supposed to work in the state sector. To have people employed by the state was both a state- ment of the ideological objective and a means. The objective was to speed up realization of a "developed socialism" characterized by a predominance of the state sector. The means for achieving this objective was fast economic growth, which in turn meant fast growth of the state sector, because the state sector was (not unlike in Schumpeter's Capitalism, Socialism and Democracy) regarded as more efficient than the private sector. It is worth remembering that the time when Communists came to power was in many countries the epoch of the "big is beautiful." The state sector was also considered to be the 12. For example, Phelps-Brown (1988) notes the absence of the Paretian right-end tail (that is, highest earnings) in socialist wage distributions. 13. Redor (1992, p. 63) finds that the difference between average manual and non- manual wage explained between 17 and 30 percent of the overall wage inequality in the Federal Republic of Germany and France and practically none in Eastern Europe. 14. In 1988 Hungary became the first to introduce a real personal income tax system. 20 Income, Inequality, and Poverty during the Transition best vehicle for transferring the labor force from agriculture to industry and from rural to urban areas (particularly since socialism won mostly in agrar- ian countries). In consequence, the state sector was to be developed, while the private sector was to be allowed to wither away, or was, at best, to be tolerated. 15 High participation rates for both men and women were to be encouraged, as work in the state sector was also a way to be usefully integrated into society. High participation of women was needed both to speed up growth (by uti- lizing all labor resources) and as a proof of the equality of the sexes, an early Communist objective. High participation rates combined with a generally pro-children stance (as reflected in free health care and education) resulted also in a heavy emphasis on family allowances, the size of which, in relation to wages, could be several times greater than in market economies (see table 2.4). Family allowances also introduced an element of reward "according to needs"-a harbinger of distribution as it should be, according to Marx, in a developed Communist society. Ideological views on wage distribution were ambivalent. On the one hand, socialist ideology is against wide differences in income; on the other hand, people should be paid according to their contribution, at least in theory, which means that differences in abilities and effort should be recognized. Marx's dictum that under socialism, workers are paid according to their work (as opposed to "communism," where workers should be paid according to their needs), as well as the influence of Taylorism on the early Bolshevik practice, meant that wage differences were not only to be tolerated, but accepted. 16 Uravnilovka, or leveling, had, from the early days of Soviet Communism, ac- quired a negative ideological connotation. Manual labor was, in general, preferred to non-manual labor: small differ- ences in wage between manual and non-manual labor reflected this ideo- logical preference. The preference itself had roots in the crude (and actually mistaken) interpretation of Marx's concept of "productive" and "unproduc- tive" labor, and in the nomenklatura's belief that manual workers represented the "salt of the earth" for the Communist party and were politically more reliable than intellectuals."7 15. Incidentally, this view resulted in a desire to statistically demonstrate the progress of socialization by contrasting the shares of the state and private sectors. This explains why such statistics are more readily available in formerly socialist countries than they are in market economies. 16. Taylorism remained quite alive well into the 1970s. Redor (1992, p. 159) points out that in the late 1970s, piece-rate pay was much more common in Eastem Europe and the Soviet Union than it was in OECD economies. About 50 percent of industrial workers in socialist countries were paid according to the piece-rate system. Corresponding per- centages in the West ranged from 5 percent in Belgium, to 12 percent in France, to 22 percent in West Germany. 17. A high percentage of Communist leaders were workers or came from a working- class background. Often they were skilled (metal) workers rather than less-skilled (say, textile) workers. Examples include Brezhnev in the U.S.S.R., Tito in Yugoslavia, Kadar in Hungary, and Novotnr in Czechoslovakia. The Way It Was 21 Table 2.4. Family Allowancefor Two Children as Percentage of Average Earnings, 1988 Country Percentage of average earnings Hungary 24.9 Bulgaria 20.0 Czechoslovakia 19.6 Poland 17.0 Austria 16.9 Belgium 10.7 Netherlands 9.0 United Kingdom 8.2 France 6.5 Italy 5.4 United States, 0.0 Note: Does not include tax deductions available for families with children (important in the United States and France) and thus underestimates the amount of family benefits in the market economies. a. No universal child allowance. Source: Sipos (1994). Finally, individual accumulation of wealth was frowned upon for ideo- logical and pragmatic reasons. Ideologically, Communists were against big differences in wealth. Large wealth inevitably "overflows" into the pro- duction process: initially such wealth may be used for personal consump- tion (for example, for purchasing a house and a car), but eventually it will be used to obtain ownership of the "means of production" complete with hired labor, an outcome which is obviously antithetical to socialism. On pragmatic grounds, Communists were also against private wealth because wealth provides "an island of liberty" for those who have it. The rich can become independent from imposed political obedience. Thus, Communist authorities preferred collective consumption and income-in-kind to cash rewards. Collective consumption (for example, state-sponsored sanatoria and kindergartens, or free vacations) and income-in-kind (for example, plush houses for the nomenklatura) have the advantage that they appear more "collectivistic," they cannot be accumulated, and they can easily be withdrawn if one fails to toe the political line. From the above tenets, the shapes of "ideal" socialist income distribu- tion emerge. Most income is earned in the state sector (pensions received after retirement are an extension of state-sector employment); high partici- pation rates obtain; unemployment is non-existent; family allowances are relatively high; wages are compressed with low pay in intellectual profes- sions; collective consumption and income-in-kind are important; and wealth accumulation is minimal. Actual income distributions in socialist coun- 22 Income, Inequality, and Poverty during the Transition tries came reasonably close to this ideal. Actual income distributions, there- fore, did not evolve by accident: they were logical extensions of the ideo- logical premises of Communism, which, conveniently, often coincided with the political interests of Communist rulers. 3 Income Changes in income are the most decisive factor influencing poverty. This is especially true when income declines are as great as those found in Eastern Europe and the former Soviet Union. Changes in income distribution, as a factor influencing poverty, then take a second place. The Post-Communist Great Depression Incomes Table 3.1 shows the average (weighted) change in the officially recorded GDP during 1987-96.l In Eastern Europe as a whole, growth fell from about 2 percent in 1987 and 1988 to slightly above zero percent in 1989. In 1990, the average growth rate was strongly negative (-8.2 percent). The decline reached its maximum in 1991 (-14.7 percent). Eventually, in the 1994-96 period, East- em Europe grew at the rate of about 4 percent per year, and growth spread to almost all countries in the region.2 In 1996, overall GDP in Eastern Europe was approximately 80 percent of its 1987 level. The trend in the former Soviet Union was similar to that in Eastern Europe with about a year, at first, and then apparently longer time lag. The Soviet GDP continued to grow until 1989.3 It shrank by 2.5 percent in 1990 and by 6.5 percent in 1991, and in 1992 it fell precipitously by 16 percent. In the following two years, the region's GDP experienced a double-digit decline. The combined GDP of countries that composed the former U.S.S.R. was in 1996 approximately 60 percent of its 1987 level. 1. The officially recorded decline in GDP almost certainly exaggerates the actual decline (see the section, The effect on population incomes: how level and composition of income changed, below). It is nonetheless necessary to begin with official measures-which in the further text will be adjusted (to the extent possible). The year 1987 is used for comparison purposes because it was a year before serious reforms were initiated and because eco- nomic levels reached by socialist countries in 1987-while not likely to improve by much- could have been sustained for a prolonged period. 2. In 1994 and 1995, only the Former Yugoslav Republic of Macedonia experienced negative growth, as did Bulgaria in 1996. 3. According to the altemative calculations of Khanin (1992), however, the Soviet GDP had begun to decline already by the second half of 1988 (yielding zero growth for the year as a whole), and in 1989 it shrank by 2 percent. For a critique of Khanin's meth- odology, see Kudrov (1995,1996). For a general discussion of Russian post-Soviet macro- economic data, see Bloem, Cotterell, and Gigantes (1995) and Koen (1996). 23 24 Income, Inequality, and Poverty during the Transition Figure 3.1. Distribution of Countries' Growth Rates in Eastern Europe, 1987-96 14 12 ...D.LI .... 10 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 o Double-digit decline EL Single-digit decline Ol Growth Figure 3.2. Distribpution of Countries' Growth Rates in the Former Soviet Union, 1987-96 12 0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 o Double-digit decline El Single-digit decline O Growth Note: The Czech Republic, the Slovak Republic, and all republics of the former Yugoslavia and Soviet Union are shown here as separate countries. German Democratic Republic is not included. Bosnia and Herzegovina is not included after 1992. Income 25 Table 3.1. GDP Growth Rates in Eastern Europe and the Former Soviet Union, 1987-96 (percent per annum) Region 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Eastern Europe 1.9 1.6 0.5 -8.2 -14.7 -8.1 -1.9 4.1 5.5 3.6 Former Soviet Union 2.4 5.2 2.7 -2.5 -6.5 -16.1 -10.1 -14.0 -5.2 -4.7 Total 2.3 4.0 1.9 -3.7 -8.6 -14.3 -8.1 -9.2 -1.9 -2.0 Note: Eastem Europe indudes the German Democratic Republic until 1989. Source. Official goverment statistics. Data for the Soviet Union are official statistics except for Russia where the revised official GDP statistics published by the World Bank and Government of the Russian Federation (1995) are used also for the period before the Soviet Union's breakup. There are minimal differences between these statistics and PlanEcon statistics or the data published by World Bank Socio-economic Data Division (1992). How widespread, deep, and long-lasting was the post-Communist depres- sion? Depression is defined as involving at least two years of continuous GDP decline. According to this definition, all transition economies experi- enced a depression after 1990; no country experienced fewer than three years of consecutive GDP decline. Both the GDP decline and its subsequent recovery affected all the countries in the region (figures 3.1 and 3.2). In 1987, most socialist countries recorded growth rates of between 2 and 3 percent per year, and only Albania, Armenia, Georgia, Tajikistan, and the former Yugoslavia had negative growth.4 The situation did not change much in 1988 and 1989. By 1990, however, all East European coun- tries and all republics of the former Soviet Union (except two: Uzbekistan and Turkmenistan) were in a decline. For three years, during 1990-92, in a remark- able unison, the GDPs of almost all twenty-eight countries dropped, with the GDP of more than half the countries shrinking more than 10 percent annually. Only one country (Poland) posted positive, if minimal, growth (1.5 percent) in 1992. By 1993, half the East European countries (Albania, Poland, Romania, and Slovenia) were growing again, and by 1994 and 1995, almost all were. In the former Soviet Union, however, the decline was still universal in 1993. Three countries turned the corner in 1994 (Armenia, Latvia, and Lithuania), and by 1996, nine of fifteen countries were growing again, although the largest coun- tries (the Russian Federation and Ukraine) continued to post negative growth rates. The depth of the depression is best assessed by comparing it to the 1929-33 Great Depression. Figure 3.3 shows the GDPs of Russia and Poland, and the United States and Germahy. Data on Poland and Russia are presented be- cause of the size of these countries (combined, they produce almost one-half the output of all transition economies) and because of the radicalism of their reforms. For Poland and Russia, the base year is 1987; for Germany and the 4. The then-republics of Czechoslovakia, the U.S.S.R., and Yugoslavia are consid- ered separate countries here for the sake of comparison. 26 Income, Inequality, and Poverty during the Transition Figure 3.3. Real GDP in Poland and Russia (1987-95); and in the United States and Germany (1927-35) Germany, 100- ......... .4Poland 90 United StatesZ r| 80 " 70- Russia 60- 27-87 28-88 29-89 30-90 31-91 32-92 33-93 34-94 35-95 Years Note: 1927 = 100 for the United States and Gernany; 1987 = 100 for Poland and Russia. Source: The Great Depression data, unless noted otherwise, are from Fozouni, Gelb, and Schrenk (1992). United States, the most hard-hit countries during the Great Depression, the base year is 1927. For all countries, peak output is reached in the third year following the base year. If Russia is compared with Germany and the United States, it emerges that initially, the decline was steeper in Germany and, particu- larly, in the United States.5 By the third year of depression (1932 and 1992), GDPs in the United States and Germany were approximately 20 percent below pre- depression levels; in Russia, the GDP was 14 percent below. Then, the situation reversed. In 1933, Germany grew quickly, at a rate of approximately 10 percent per year, and the U.S. GDP declined, although by very little, while Russia's GDP continued its plunge, in both 1993 and 1994 by 9 and 13 percent, respectively. Thus, by 1935, Germany was above its base GDP level, the United States was approximately 15 percent below, while Russia in 1995 was 34 percent below its base level. In other words, the post-Communist depression in Russia is deeper than the Great Depression was in the United States and Germany. The depres- 5. Throughout, revised GDP statistics for the Russian Federation published jointly by the World Bank and Government of the Russian Federation (1995) are used. The revised GDP is approximately 6 to 7 percent higher for the period 1991-93 than that recorded by earlier official statistics. There is virtually no difference between the two sets of numbers for 1990 and 1994 (see World Bank and Government of the Russian Federation, 1995, p. 94). Income 27 sion in Poland, although deeper during the first two years (1990-91), was not as severe as the depressions in the other three countries. The Polish trough, reached in 1991, was approximately 15 percent below the base level. Moreover, as al- ready mentioned, Poland was the first transition economy to emerge from the depression: in 1995, its GDP had returned to its 1987 level. Russia had experienced another major GDP decline in 1917-21, during the Revolution and Civil War, when its GDP declined by approximately one-half (according to Block [1976] as quoted in Sokoloff [1993]). The decline of 1917-21 represented a fall of approximately 60 percent from the pre-War level (see table 3.2). The current post-Communist depression is about half as severe. During both crises, agricultural output, not surprisingly, contracted less than industrial out- put. The duration of the depression is also important. A longer depression prob- ably has a more deleterious effect on welfare than a shorter depression, even if the overall decline is the same. The same conclusions can be drawn from the permanent income hypothesis. A short, even if sharp, decline in income is treated as transitory because it does not affect long-term income or consumption. A drawn-out depression, on the other hand, reduces people's perception of their long-term income and consumption, possibly by even more than the actual de- cline, because they may expect decreases to continue. Thus, for example, the GDP in the United States dropped in 1938 by about 5 percent to quickly recover the following year. Understandably, no one talks of 1938 with nearly the same awe with which people still refer to the Great Depression. In terms of duration, the post-Communist depression appears as bleak as or worse than the 1929-33 De- pression. Most transition countries have experienced between three and four years of successive GDP declines, and some have experienced five (Romania, 1988-92; Estonia, 1990-94) or six years (Hungary, 1988-93; Slovenia, 1987-92;6 and Belarus, Table 3.2. Russia's Output Decline after the Dissolution of the Czarist Empire and after the Dissolution of the U.S.S.R. GDP Industrial output Agricultural output Year 1913=100 1987=100 1913=100 1987=100 1913=100 1987=100 1917 or 1991 75 101 77 97 100 97 1919 or 1992 54 86 26 79 88 87 1920 or 1993 45 79 18 66 62 84 1921 or 1994 38 69 n.a. 52 n.a. 77 1922 or 1995 n.a. 66 n.a. 51 n.a. 70 n.a. = not available. Source: 1913-22: GDP from Sokoloff (1993); industrial output (only large-scale) and agricultural output (only cereal production in Central Russia) from Kritsman (1926), quoted in Pipes (1990, p. 696). 1987-95: World Bank data. 6. Declines in Hungary and Slovenia were, however, small, at least by the stan- dards of other transition economies. War-affected countries are not included here; Azerbaijan, Croatia, and Georgia have each experienced six successive years of GDP de- cline. 28 Income, Inequality, and Poverty during the Transition 1990-95) or even seven years (Moldova, Russia, and Ukraine, 1990-96) . During the Great Depression, by contrast, GDP decreased for three successive years in Germany, four years in the United States, and six years in France, while the GDPs of the United Kingdom and Italy alternated between growth and decline. While the effects of the two depressions were similar for the populations con- cerned (and probably worse for Russians now than for Americans sixty years ago), the impact of the two depressions on the rest of the world was different. In the late 1980s, transition countries accounted for about 6 percent of the world's GDP (at current exchange rates); about 6 percent of world trade, only half of which was not among themselves; and approximately 9 percent of the world's population. The shrinking of their GDPs by a quarter thus reduced the world's GDP by approximately 1.5 percentage points. On the other hand, the major capi- talist countries where the Great Depression began (the United States, the United Kingdom, Germany, and France) accounted for more than half the world's out- put, 2; of world trade, and 70 percent of industrial production (Gazier 1983; Romer 1993, p. 20n). Their combined GDP decline of about 20 percent between 1929 and 1933 meant that the world's GDP shrank as much as one-tenth; world industrial output decreased by one-third (Bairoch 1993, p. 136), and the volume of world trade dropped by 25 percent.7 Since they were the "core" countries, their links with the rest of the world (via trade and capital flows), and the dependence of the rest of the world on them, was much greater than the dependence of the rest of the world on formerly socialist economies. The Great Depression led to export declines and a sharp deterioration in the terms of trade for African, Asian, and Latin American countries. Some authors hold that the Great Depression signaled the beginning of underdevelopment in the Third World. Wages and unemployment One major difference between the Great Depression of 1929-33 and the cur- rent post-Communist depression is the way in which wages and employment have adjusted (see figures 3.4 and 3.5). During the Great Depression, wages in all major countries (the United States, the United Kingdom, and Germany) remained stable in real terms. At the same time, unemployment grew from an initial rate of 5 to 10 percent of the labor force in the late 1920s to between 20 and 25 percent (and even more in Germany). The adjustment of the wage bill was thus entirely borne by a quantity adjustment: that is, unemployment went up. The wage bill was cut by between 15 and 20 percent in real terms, falling a few percentage points short of the decline in output. The labor share in GDP 7. GDPs in major market economies declined, between the peak and the trough, as follows: the United States by 30 percent (1929-33), Germany by 24 percent (1930-32), France by 17 percent (1930-36), the United Kingdom by 5 percent (1929-32), and Italy by 5 percent (1929-34). Japan's GDP over approximately the same period grew by 20 per- cent. Income 29 Figure 3.4. Real Wages in Poland and Russia (1987-96); and in the United States, the United Kingdom, and Germany (1927-36) 140- 130 United States 120 - ; _ - -.Germany -' ~United Kingdom 90 -- - -- 8 0 80 - -* -- Poland 70- 60- 50 - 40- 27-87 28-88 29-89 30-90 31-91 32-92 33-93 34-94 35-95 36-96 Years Note: 1927=100 for the United States, Germany, and the United Kingdom. 1987=100 for Poland and Russia. Source: For the United Kingdom and the United States: Bain and Elsheikh (1976, appendix E). For Germany, James (1986, p. 196) and Pierenkemper (1987). For Poland and Russia, World Bank data. thus expanded (Fozouni, Gelb, and Schrenk 1992; James 1986, p. 416). Labor adjustment during the post-Communist depression occurred differ- ently. In Russia and other countries of the former Soviet Union, registered, and even actual unemployment, is very small, while real wages have declined be- tween 40 and 60 percent.8 This type of wage-bill adjustment is thus exactly the opposite of the adjustment that took place during the Great Depression. The East European situation lies between these two extremes. Both real wages and employment have decreased. On average, real wages dropped by one-fourth between 1987-88 and 1994, while unemployment grew from zero percent to between 12 and 15 percent of the labor force (except in the Czech Republic, where unemployment is much lower). In no transition economy, except those affected by war, has the severity of unemployment reached Great Depression levels. The real wage bill was cut by approximately one-third in Eastem Europe 8. In 1995, of all former Soviet republics, only Latvia had a registered unemploy- ment rate of more than 5 percent of the labor force. Registered unemployment in Russia was 3.2 percent, and actual unemployment was estimated at 9 percent. 30 Income, Inequality, and Poverty during the Transition Figure 3.5. Unemployment Rates in Hungary, Poland, and Russia (1987-95); and in the United States, the United Kingdom, and Germany (1927-35) 50 - Germany 40 / - X~~~~~~~~~~ // \ 30 - / United \ . United -~~~~~~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~States 20 /- Tg9I. .. M 2 0 - / /^/~Poland =. -- ' 10 - `-~Hungary _ / <___~- -- Rssia 0*- 27-87 28-88 29-89 30-90 31-91 32-92 33-93 34-94 35-95 Years Note: Percent of labor force (annual average). Source: For the United Kingdom and the United States: Bain and Elsheikh (1976, appendix E). For Germany, Pierenkemper (1987). For transitions economies, World Bank data. and one-half in countries of the former Soviet Union; both cuts are larger than those experienced by labor in major countries during the Great Depres- sion. Further, in contrast with the Great Depression, the share of labor in- come in the GDP was reduced in countries of the former Soviet Union, and remained constant in countries of Eastern Europe. The Effect on Population Incomes: How Level and Composition of Income Changed Decline in real population income The decline in real population income provides a better indicator of hard- ship than an aggregate measure such as GDP. This is so because GDP ac- counting is faulty and because GDP movements do not always mirror changes in population income and welfare. Recorded GDP declines in former Communist countries overstate real de- clines because official accounts err at both ends: by overstating the level of GDP achieved prior to transition and by failing to record fully the growth of the most dynamic sector, that is, the private sector after the transition. Previ- ous GDPs were "padded up," principally for two reasons. First, all levels of the Communist hierarchy had incentive to report better than actual fulfill- Income 31 ment of planned targets because their promotions and income depended on target overfulfillment.9 Second, published GDPs underestimated inflation (and thus overestimated real output) and did not adequately account for the low quality of goods. On the other hand, the increase in private sector activities since the transi- tion is not fully reflected in macroeconomic statistics. The informal or "shadow" economy is estimated to account for between 10 and 15 percent of recorded GDP in more "orderly" transition economies (for example, the Czech Republic and Hungary) and up to 50 percent in countries affected by war and embar- goes (for example, Armenia and the Federal Republic of Yugoslavia). '0 It seems certain that in all transition countries, the share of the informal sector has in- creased, whether because the absolute size of the sector has expanded or be- cause it had shrunk less than the formal economy has. This increase in the share of the informal sector implies that GDP declines are overstated. The discrepancy between GDP and welfare in formerly socialist econo- mies existed because some goods produced and recorded as part of the GDP did not contribute much, or anything, to the welfare of the population (such as production of armaments and shoddy goods recorded at higher than mar- ket prices). llAlso, the transition-related declines in gross investments (which have no immediate impact on welfare and were to a large extent wasteful under socialism), in inventories (which tended to be excessively high), and 9. The most extreme example of this phenomenon was the hugely falsified cotton harvest results under Uzbekistan's Communist ruler Rashidov in the 1970s and early 1980s. 10. Arvay and Wrtes 1994 study of Hungary represents the most detailed study of the informal sector in a transition economy. They estimate that the informal sector not covered by statistics has grown from between 11 and 12 percent of the GDP before the transition to 17 percent in 1993. Kadera (1995) assesses the Czech shadow economy to be 8 to 10 percent of the recorded GDP. Slovak INFOSTAT agency estimates Slovak informal economy to be about 12 percent of the recorded GDP in 1995 (Narodna Banka Slovenska, 1996, p. 28). Sadowski and Herer (1996) estimate the Polish informal economy to be 25 percent of the recorded GDP. In Russia, the informal or "shadow" economy was esti- mated to be 20 percent of the GDP by the chairman of the Goskomstat (press conference on December 6, 1994). Croatia's informal economy was estimated to have increased (under very conservative assumptions) from 18 percent of recorded GDP in 1990 to 23 percent in 1995 (Madzarevic and Mikulic, 1995). The shadow economy in the Federal Republic of Yugoslavia (Serbia and Montenegro) was estimated to have risen from 32 percent of the official GDP in 1991 to 54 percent in 1993 because of hyperinflation and illegal activities connected with United Nations sanction-busting and the war in Bosnia (see Yugoslav Federal Government 1995). 11. Some argue that official GDPs in socialist countries were necessarily biased up- ward by the use of arbitrary prices, including higher-than-market prices for inferior goods. This view, however, fails to acknowledge that there were many other products (such as energy, industrial inputs), as well as dwelling rents, and land value that were unrecorded in GDP or assessed at less than market value. Also, the material concept of Net National Product used by all Communist countries yielded estimates that were approximately 20 to 25 percent below the United Nations' System of National Accounts (SNA) concept of gross national product (see Marer 1985, pp. 18-19). 32 Income, Inequality, and Poverty during the Transition in arms production (in the Soviet Union) had little to do with population welfare. 12 The problems with estimating population welfare either disappear or are muted if real population income or expenditure is used. Some problems, such as the decline in fixed investments or inventories, then disappear, by defini- tion. Other problems become less acute, because population incomes, and in particular (formal sector) wages and social transfers, are almost fully captured by statistics. Private sector activities, both formal and informal, are better captured by household-level statistics, for example, through household surveys of incomes or expenditures, than they are by GDP statistics. 13 Retail or consumer price indexes also reflect inflation-as it affects consumers-better than GDP deflators do. Estimates of real population income are derived from two sources: mac- roeconomic sources (national accounts) and household budget surveys (HBSs). Population income from macroeconomic sources is obtained by adding all money funds received by the population, including cash wages and, sometimes, in-kind fringe benefits; social transfers; entrepreneurial income; dividends; and other sources of income, such as remittances, in- come from the lease of assets, fees, and honoraria. Estimated home con- sumption is then added to this figure. HBS income is defined in the same way as it is in macroeconomic statistics, but it is obtained through regu- lar household interviews and then extrapolated to the level of the popu- lation. In some HBSs, an attempt is made to account for the differential response of various population groups and to correct for income underreporting. Almost all HBSs also include estimated consumption- in-kind. 14 Another way to assess what happened to population welfare is to look at real expenditures. Expenditures are also collected by most HBSs and are often considered a more reliable measure than income. The reason for this is that while people maybe reluctant to share information about their 12. According to Winiecki (1991), at least two components of the registered output decline do not matter to population welfare. One part of the decrease was purely a statis- tical artifact because of the earlier practice of padding output figures to show fulfillment of plan targets. The second part of the decline was due to behavioral changes among enterprises and, to some extent, individuals. For example, better availability of goods, higher interest rates, and harder budget constraint led enterprises to reduce inventory stocks, which in socialist economnies were inordinately high. The drawing-down of in- ventories produced a short-term output decline, but in reality it represented an adjust- ment to market conditions. Thus, the first component is fictitious, and the second, while involving a real decline in output, had no effect on population welfare. 13. There are numerous problems associated with household surveys in transition economies, however (see appendix 1). 14. Consumption-in-kind is an important source of income in transition economies. It includes not only own consumption, but also a portion of small-scale production that may be sold. The latter is not, strictly speaking, consumption-in-kind, but is often treated as such because households that sell a portion of their output may not do so regularly or may not be willing to report such sales as commercial activity. Income 33 income with interviewers whom they assume to be agents of the govern- ment, they may be less wary when asked to give information about ex- penditures. Results from eleven countries do, on average, confirm that HBS expenditures tend to be higher than HBS incomes: in six cases, ex- penditures are more than 20 percent greater than income; 15 in three cases, they are about the same; and in only two cases, income is higher than expenditures (see table 3.3). Table 3.4 and figure 3.6 show estimated changes during the transition in real per capita GDP and in the two measures (macro and HBS) of population real income. 16 Table 3.4 shows that (unweighted) income declines were the least sharp in Eastern Europe. According to GDP and HBS data, real per capita incomes were, on average, 21 percent and 25 percent smaller, respectively, in 1993-94 than they were in 1988; according to macroeconomic data, they were perhaps 10 percent less. In the Baltics, real incomes seem to have fallen by almost a half, according to the GDP data, and 40 percent according to HBS data. In Moldova and the Slavic republics of the former Soviet Union, the decrease in income was more than one-third, according to GDP data, one- Table 3.3. Ratio between Expenditures and Income Reported in Household Budget Surveys Country (year of survey) Mean expenditures:mean income Eastern Europe 1.03 Slovakia (1993) 0.83 Bulgaria (1993) 0.85 Hungary (1993) 1.00 Slovenia (1993) 1.02 Romania (1994) 1.19 Poland (1993) 1.30 Former Soviet Union 1.58 Estonia (1995) 1.02 Belarus (1995) 1.21 Russia (1993) 1.23 Ukraine (1995) 2.10 Kyrgyz Republic (1993) 2.32 Note: All regional averages are unweighted. Both income and expenditures are defined on a net (that is, after tax) basis. Source: For surveys, see appendix 4. See also information on surveys and caveats in appendix 1. 15. In two cases (the Kyrgyz Republic and Ukraine), expenditures are more than twice HBS income, suggesting that even HBS income must be hugely underestimated. 16. Success in transition should not be assessed from these figures because they refer to different dates. 34 Income, Inequality, and Poverty during the Transition Table 3.4. Change in Real per Capita GDP and Real per Capita Population Income between 1988 and 1993 (percent) Real per Real per Real per capita income capita income Country capita GDP (macro data) (HBS data) Eastern Europe -21 -7 a -25 Bulgaria (1989-93) -27 n.a. -45 Czech Republic -18 -7 -12 Hungary (1987-93) -15 +1 -26 Poland (1987-93) -12 -11 -26 Romania (1989-94) -26 -18 -43 Slovakia -29 -29 -29 Slovenia (1987-93) -21 +20 +8 Baltics -49 n.a. -41 Estonia (1988-94) -37 n.a. -37 Latvia (1988-95) -43 n.a. 45 Lithuania (1988-94) -66 -44 -42 Slavic republics -37 -49 -54 Belarus (1988-95) -34 -30 -44 Moldova -39 -67 -67 Russia -27 -33 42 Ukraine (1988-95) -49 -64 -62 Central Asia -28 -39 -54 Kazakhstan -26 -57 -61 Kyrgyz Republic -35 -58 -66 Turkmenistan -31 -40 -46 Uzbekistan -20 +1 43 n.a. = not available. Note: The end years are 1988 and 1993 unless different years are given between brackets. All regional means are unweighted. To insure comparability among the dates of different data sources, the actual years of comparison for GDP and macroeconomic data are the same as the years for which household surveys are available. Macroeconomic and HBS data for Lithuania do not include consumption-in-kind in 1994; HBS data for Hungary do not include consumption-in- kind in 1993. a. Does not include Bulgaria. Source: GDP and macro incomes: World Bank data. HBS incomes: calculated from surveys presented in appendix 4. See also caveats in appendix 1. half according to macroeconomic data, and more than one-half according to HBSs. Similarly, in Central Asia, GDP declines were smaller than income declines whether measured by macroeconomic data or by HBSs. It is remarkable that of the sixty-two observations in table 3.4 and figure 3.6 17 only five show an increase: minimal 1 percent macroeconomic income 17. Sixty-two observations represent three observations for each income concept for eighteen countries less three unavailable observations plus eleven expenditure observa- tions (shown only in figure 3.6). Income 35 Figure 3.6. Real per Capita GDP and Real per Capita Population Income in 1993-95 (pre-transition 100) 120 100 ..0 80 o 0 60 40 20 0 Po Hun Cz Uz Sn Rom Kaz Bul Rus SI Turk Bei Ky Est Mo La Uk Li - GDP Macro -HBS a Expenditures Note: Countries are arranged in ascending order of their GDP decline (Poland has the smallest decline, Lithuania the greatest). Source: Table 3.4. increases in Hungary and Uzbekistan, and macroeconomic and HBS income and expenditure increases in Slovenia. Of interest is the relationship between the two measures of population real income. For most countries where both sources of data are available, HBSs show larger income declines than macroeconomic data do. In Hungary and Romania, the difference amounts to more than one-quarter of initial (that is, pre-transition) income; in Poland, it is 15 percent; and in Russia, it is 9 per- cent."8 The correlation coefficient between changes in HBS and macroeconomic income is a high 0.88. HBS income change and macroeconomic income change are less strongly correlated with GDP change: the correlation coefficient is 0.68 for HBS income and GDP, and 0.66 for macroeconomic income and GDP. A stronger correlation between macroeconomic and HBS data is expected because they both measure the same thing: the real income of the population. GDP, of course, includes components other than population income. Perhaps more importantly, during the transition cost-of-living indices used to deflate the HBS and macroeconomic income have moved differently from (often in- creasing faster than) implicit GDP deflators. 18. Note, however, that the large difference between the two measures in Hungary and Lithuania may be explained by non-inclusion of consumption-in-kind in 1993-94. Table 3.5. Population Income by Sources in 1987-88 and 1993-94 (percent of GDP) Non-wage Social transfers Cash social private sector in kind (health Wages transfers income and education) Total Country 1987-88 1993-94 1987-88 1993-94 1987-88 1993-94 1987-88 1993-94 1987-88 1993-94 Eastern Europe 33 32 11 15 10 19 7 9 62 75 Bulgaria 27 25 11 14 9 21 7 11 55 71 Czech Republic 41 32 12 13 4 24 9 12 67 82 Hungary 32 37 13 19 10 14 7 12 63 83 Poland 27 32 9 20 22 25 7 9 65 85 Romaniaa 35 33 9 9 3 14 4 5 52 61 Slovakia 42 31 13 13 4 17 11 6 70 68 W Slovenia 28 34 11 16 10 20 7 7 57 77 Baltics 43 35 8 12 9 14 9 8 70 69 Estoniab 46 33 9 11 8 10 11 8 73 62 Latvia 38 33 8 14 8 12 8 9 62 69 Lithuaniab 46 40 8 10 12 18 9 8 76 76 Slavic republics 41 29 8 9 6 16 6 9 61 63 Belarus 40 37 6 8 7 17 7 12 59 74 Moldova 43 23 7 8 8 28 10 12 69 72 Russia 41 26 8 9 5 23 5 7 59 64 Ukraine 42 25 9 12 7 8 7 10 66 55 Note: All regional means are unweighted. Regional means as calculated avoid the rounding off errors. Definitions: Wages equal income from employment. Social cash transfers equal pensions, family and child allowances, sick leave payments, unemployment benefits, and social assistance. Non-wage private sector income equals income from sales of agricultural products, entrepreneurial income, interest and dividends, income from abroad, gifts, and income (or consumption) in kind. Health and education equals total government expenditures on health and education. a. 1992-93 instead of 1993-94. b. Does not include home consumption. Source: Income of the population as a whole (from macroeconomic sources): World Bank data. Income 37 Changing income composition The composition of population disposable income changed during the tran- sition. Population income can be divided into three categories: (a) wages; (b) cash social transfers; and (c) self-employment, home consumption, property income, private transfers, and other private sector income. The last category can be termed "non-wage private sector" income. It underestimates the true size of private sector income because it omits some informal sector income, as well as wages earned in the private sector. The latter are classified together with other wages. As the transition proceeds, this three-way classification becomes less of an indicator of transformation toward a private property- based economy than it was at the beginning of the transition process. This is because at the beginning of the transition, practically all income from private sector activities was included in the third category. It then served as a good proxy for the size of non-state sector. As privatization expands, however, more private sector income is eamed through private sector wages. The three- way classification is thus somewhat biased against faster reformers. In any case, it gives the lower-bound estimate for the size of the private sector. Table 3.5 illustrates the changing composition of population disposable income. The three types of income, as well as in-kind social transfers (health and education), which are not part of population disposable income, are shown as percentages of (current) market-price GDPs. Not surprisingly, the share of total population income (inclusive of health and education) in GDP increased in most countries between 1987-88 and 1993-94. The accounting explanation for this is that other types of income (for example, gross operat- ing surplus of enterprises, indirect taxes, and inventory build-up) contracted more severely than the GDP did. '9 A more meaningful explanation is that under conditions of rapidly declining GDPs, most governments tried to cush- ion the population as much as possible from the effects of the depression. Broad regularities revealed by data in table 3.5 are: ' The (unweighted) share of labor income in the GDP has remained con- stant in Eastern Europe and has declined in the Slavic republics of the former Soviet Union (by 12 points) and the Baltics (by 8 points).2' - The share of social cash transfers in the GDP has risen in all three re- gions (by between 1 and 4 percentage points of GDP). * The share of non-wage private sector income in the GDP has increased everywhere: by 9 percentage points of GDP in Eastern Europe, 10 points in the Slavic republics, and 5 points in the Baltics. 21 19. Because GDPs in table 3.5 are at market prices (and not at factor cost), popula- tion-income components cannot add up to 100 percent of GDP. 20. Data for the Central Asian republics are not presented because they are not reli- able. 21. Private sector data for Estonia and Lithuania do not include consumption-in- kind. Because consumption-in-kind has increased, in real terms probably, and certainly as a share of GDP, the growth of private sector income is underestimated in the Baltics. 38 Income, Inequality, and Poverty during the Transition * The share of health and education in the GDP has increased by about 2-3 GDP points in Eastern Europe and the Slavic republics of the former Soviet Union. In all East European countries (except Slovakia) the share of population in- come in GDP increased. For all these countries combined, the share went from 62 to 75 percent, because of a large increase in private sector income and a smaller increase in social transfers. In the Slavic and Baltic regions of the former Soviet Union, population income share in the GDP remained virtually unchanged. The situation is, of course, more differentiated at the country level. All tran- sition economies can be divided into three categories, depending on the type of change. The first category ("the non-compensators") is characterized by a declin- ing share of wages which is not "offset" in any meaningful way, by an in- creased share of cash social transfers. This particular configuration can be denoted as (- 0 +), where the negative sign, zero, and the positive sign denote a decreased, unchanged, or increased share, respectively, of wages, cash social trans- fers, and non-wage private sector income. "Non-compensators" include only countries where each $5 in lost wages is accompanied by $1 or less in greater social transfers: the Czech Republic, Estonia, Moldova, Romania, Russia, Slovakia, and Ukraine belong to this group. For example, in Russia, the share of wages decreased by 15 percentage points of GDP while social transfers increased by 1 GDP point.2 The second group ("the compensators") consists of countries with a (- + +) configuration, where compensation, in the form of social transfers for lost wages, is more generous. "Compensators" include Belarus, Bulgaria, Latvia, and Lithuania. The third group ("the populists") has a +++ +) configuration. In these coun- tries, all sources of population income increased in terms of GDP. "Populist" coun- tries clearly attempted to cushion the population, as much as possible, from the effect of real GDP declines. Only Central European countries (Hungary, Poland, and Slovenia) belong to this group. Growth of non-wage private sector income All configurations include a plus sign for non-wage private income, because non- wage private income has risen as a share of GDP in all transition economies. In Eastem Europe, its unweighted share increased from 10 percent of GDP before the transition to 19 percent of GDP in 1993-94; in the Baltics, it went from 9 to 14 percent; in the Slavic republics of the former Soviet Union, it grew from 6 to 16 percent. These figures make clear that Eastem Europe began its transition with a more sizable private sector than did the Soviet Union, particularly so the coun- tries where agriculture was private or semi-private or which exhibited a more 22. Interestingly, but perhaps not surprisingly, only countries affected by war and regional tensions (not shown in the table) display the configuration (- - +). Income 39 liberal attitude toward the small private sector (Hungary, Poland, and Slovenia). During the transition, however, the share of non-wage private sector income in GDP grew as quickly in the Slavic republics of the former Soviet Union as it did in Eastern Europe. In addition, it is important to recall that the informal sector (which, by definition, is not included in national statistics) is larger in the coun- tries of the former Soviet Union than it is in Eastern Europe. The share of private sector income in the countries of the former Soviet Union is thus more underes- timated than it is in Eastern Europe. The most important increase in the share of non-wage private sector income occurred in the Czech Republic and Moldova (an increase of 20 GDP points), Russia (18 points), Slovakia (13 points), and Bul- garia (12 points). The next chapter will investigate what happened to income distribution dur- ing the transition and how it was affected by changes in income composition. 4 Inequality Income Inequality Table 4.1 illustrates the change in inequality that has occurred since the be- ginning of the transition. 1 First, inequality increased in all countries except the Slovak Republic. The average Gini coefficient of disposable (or gross) 2 income rose from 24 to 33. While the first value is equal to the average value of low income-inequality OECD countries (such as the Benelux, the Federal Republic of Germany, and Scandinavia), the second value places transition economies at near the OECD mean, or at about the same level of inequality as Commonwealth countries (for example, Australia, Canada, and the United Kingdom) and Latin European countries (such as Italy and France).3 Second, the increase in the Gini coefficient was sharp: over a period of about six years, the average Gini rose by 9 points. This is, on average, 11. Gini points per year, a rise that is almost three times as fast as the rise recorded in those Western countries where inequality rose most rapidly in the 1980s: United Kingdom, the Netherlands, and United States (Atkinson, Rainwater, and Smeeding 1995, p. 25). Third, the dispersal of Ginis among transition economies increased. While before the transition, their Gini coefficients (with the exception of Central Asian republics, which were more unequal) lay within a very narrow range between 1. Results are based on HBSs presented in appendix 4. A discussion of problems associated with these surveys and biases in estimating inequality is presented in appen- dix 1. 2. Disposable and gross income in transition countries are not very different be- cause direct personal income taxes are still very small. 3. Atkinson, Rainwater, and Smeeding (1995, p. 16) divide OECD countries into four groups. Very low income-inequality countries, with Ginis between 20 and 22, in- clude Nordic countries (Finland, Sweden, Norway) and Belgium. Low income-inequal- ity countries, with Ginis between 24 and 26 include the Federal Republic of Germany and the Netherlands. Latin Europe (France and Italy) and Commonwealth countries (Aus- tralia, Canada, and the United Kingdom) have average levels of income-inequality, or Ginis of between 29 and 31. Finally, high income-inequality countries include Ireland, Switzerland, and the United States, with Ginis of between 33 and 35. Income concept is disposable income; distribution is per equivalent adult using the OECD equivalence scale. According to the empirical evidence presented in Coulter, Cowell, and Jenkins (1992), the use of per capita income (as in table 4.1) will lead to slightly higher estimates of the Gini coefficient than will the use of income per equivalent adult. For a discussion of Coulter, Cowell and Jenkins results, see Banks and Johnson (1994). 40 Inequality 41 Table 4.1. Changes in Inequality during the Transition Gini coefficient (annual)a Income Expenditures per capita per capita Country 1987-88 1993-95 1993-95 Balkans and Poland 24 30 Bulgaria 23b 34 Poland 26 28e 31e Romania 23b 29' 33C Central Europe 21 24 Czech Republic 19 27' Hungary 21 23 27 Slovakia 20 19 Slovenia 22 25 Baltics 23 34 Estonia 23 35d 31d Latvia 23 31d Lithuania 23 37 Slavic republics and Moldova 24 40 Belarus 23 28d 30d Moldova 24 36 Russia 24 48d 50e Ukraine 23 47c 44' Central Asia 26 39 Kazakhstan 26 33 Kyrgyz Republic 26 55d 43d Turkmenistan 26 36 Uzbekistan 28b 33 All transition 24 33 Note: For most countries income concept in 1993-95 is disposable income; in 1987-88, gross income. Personal income taxes are small, and so is the difference between disposable and gross income (see the exact definitions in appendix 4). Income indudes consumption-in-kind, except for Hungary and Lithuania in transition years. Regional averages are unweighted. a. Except when stated otherwise. b. 1989. c. Monthly. d. Quarterly. e. Semiannual. Source: Calculated from the countries' household budget surveys given in appendix 4. See also the discussion of HBSs and caveats in appendix 1. All expenditure data obtained from the same surveys as income data. 19 and 24, the current range goes from around 20 (Slovakia) all the way to high 40's (Ukraine and Russia), and even mid-50's (the Kyrgyz Republic) (see also figure 4.1). These results illustrate not only sharply increasing income differentiation among members of a single population, but also growing differences among countries. Central European countries have registered only moderate increases 42 Income, Inequality, and Poverty during the Transition Figure 4.1. Dispersal of Gini Coefficients in Transition Economies 58 - Kyrgyz Republic 48 Ukraine 0 * Russia : 38 - Lithuanias Estonae * *Moldova *Turkmen. :Bulgaria Kazakh.- Uzbek. i Romania- 28 - Czech Republic LatviaBelarus Poland0 OSlovenia No........ I No change Hungary. ................... 18 * S - -Siovak Republic _ I I I l 1, 1 1 1 18 19 20 21 22 23 24 25 26 27 28 Pre-transition Gini Source: Countries' household budget surveys (see appendix 4). in inequality: their Gini rose, on average, from 21 to 24. The increase was greater in the Balkans, even greater in the Baltics (where it rose from 23 to 34), and by far the greatest in Moldova and the Slavic republics of the former Soviet Union (where it rose from 24 to 40). Fourth, Ginis calculated on the basis of 1993-95 expenditures are not, on average, lower than Ginis calculated on the basis of 1993-95 income. In five cases, expenditure Ginis are higher (by 2 to 4 Gini points), and in three cases, they are lower. Higher expenditure than income Ginis suggest that surveys tend to underestimate both income levels (as discussed in chapter 3) and income inequality because we would normally expect to find, particularly during a depression, lower inequality among expenditures than among in- comes. While inequality rose everywhere, the exact shape of this change differed among countries. Figures 4.2,4.3, and 4.4 show the change between 1987-88 and 1993-94 in income shares received by the five quintiles. 4 4. Income distribution data, as they appear in the original sources, are presented in appendix 4. Decile data (from which the quintiles in figures 4.2-4.4 are calculated) are presented in appendix 2. Inequality 43 Figure 4.2. Changes in Quintile Shares in Hungary, Slovakia, and Slovenia between 1987-88 and 1993-94: Little Change 1.5 1 Slovak Republic 0. 0. ._.__._..._._----__ ...................... ---.-----'--'-----'--'-------'-----''--------------------- .... . .................................................... -------- Slovenia -0.5 J- 1 2 3 4 5 Quintile Figure 4.3. Changes in Quintile Shares in Belarus, the Czech Republic, Latvia, Poland, and Romania between 1987-88 and 1993-94: Moderate Regressive Transfers 8- 6 2 -2 - - - teh Republic 1 2 3 4 5 Quintile Source: Countries' household budget surveys (see appendices 2 and 4). 44 Income, Inequality, and Poverty during the Transition Figure 4.4. Changes in Quintile Shares in Bulgaria, Estonia, Lithuania, Moldova, Russia, and Ukraine between 1987-88 and 1993-94: Large Regressive Transfers 20 / 15- Rus,i, / E 10- Ukraine/ 0 Lithuania U Estonia ,.F*i / ~~~~~~~~Bulgaria Estonia~ 0 Moldova ;Bli -5- - - - - -10 I 1 2 3 4 5 Quintile Source: Countries' household budget surveys (see appendices 2 and 4). We can divide countries into three groups. 5 The first group consists of Hungary, Slovakia, and Slovenia (figure 4.2). Income shares in these coun- tries barely changed at all. No quintile gained or lost more than 1 percentage point of total income. 6 In the second group, maximum loss ranged between 1 and 2 percentage points of total income, and was sustained by the bottom three quintiles (fig- ure 4.3). The forth quintile either experienced a very small loss or retained its pre-transition share. In all cases, the top quintile alone was the "winner."7 5. Central Asian republics are not considered here because of the unreliable nature of the data. 6. A similar pattem of change is reported for eastern Gerrnany (see Speder 1995, table 1). Between 1990 and 1993, the bottom quintile lost 1.2 percentage points of income; the next three quintiles lost between 0.5 and 0.2 points, and the top quintile had gained 2.2 points. The Gini coefficient increased quite moderately from 19.5 in 1990 to 22.4 in 1993 (calculated from data provided by Speder 1995, table 2). 7. Note that this discussion relates to shares. Because overall income decreased in all countries, even an increased share received by the top quintile did not necessarily guarantee an increase in its real income. The real income decline of the bottom quintiles was that much more severe because both its share of the pie and the size of the pie itself decreased. Inequality 45 However, depending on the amount lost by the bottom three or four quintiles, the gain of the top quintile ranged from less than 2 percentage points in Po- land to about 6 points in Latvia and the Czech Republic. On average, the share of the bottom quintile was reduced from 10 to 11 percent of total in- come to 9 to 10 percent; the share of the top quintile rose from 32 to 35 per- cent of total income to 35 to 37 percent. By contrast, in 1992 the bottom quintile in the United Kingdom, whose inequality was about average among OECD countries, received about 7 percent of disposable income, while the top quintile received just over 40 percent. Income distribution in this second group of transition countries thus remained more equal than in the United Kingdom. This was not the case for the countries in the third group (figure 4.4). Their inequality is greater than the OECD average. The extent of income transfer from the bottom 80 percent of the population to the top 20 percent was much larger than among group 2 countries. Income loss by the bottom quintile var- ied between 4 and 5 percentage points of total income. Only a slightly smaller loss was sustained by the next two quintiles. In Russia and Ukraine, the sig- nificant income losses extended to the fourth quintile as well. In these coun- tries, sharp losses by 80 percent of the population translated into large gains for the top quintile. Thus, in Russia, Ukraine, as well as in Lithuania, the top quintile gained 20,14, and 11 percentage points of total income, respectively. In Russia, the bottom quintile's share was halved, declining from 10 percent of total income to less than 5 percent,8 while the share of the richest quintile rose from 34 to 54 percent of total income. These results show that: * In all cases (except the Slovak Republic), redistribution was regressive. e As regards poverty, the most unfavorable developments occurred in the countries of the third group (Russia, Estonia, Ukraine, Moldova, Bulgaria and Lithuania), where the poorest suffered greater absolute losses than did the middle or top income classes. Because in all these countries, real income decreased by between one-third and one-half, this translated into real income losses of up to two-thirds for the bot- tom quintile. • Inequality in Russia, Ukraine, and the Baltics (in that order) seems to be greater than the OECD average. In Eastern Europe, however, in- equality remains distinctly less than the OECD average. Is there a relationship between the type of adjustment identified in chapter 3 ("non-compensators, " "compensators, " and "populists") and increases in the Gini coefficient? The only relationship that could be detected is that be- 8. In other words, an average Russian in the bottom quintile had an income equal to half the mean before the transition; now his income is one-quarter of the mean. 46 Income, Inequality, and Poverty during the Transition Figure 4.5. Relationship between Type of Adjustment and Increase in Gini Coefficient 30 - 25 - Russia* Ukraine 20- w 15 - Lithuania 0 Estonia *. M ~~~Moldova --Bulgaria- 10 - Moldova--~~!~!aria a regression line --*Latvia Czech Republic 5 - Romama g Belarus* Slovenia Poland 2 0- Slovakia v Hungary v 0 Non-compensators Compensators Populists Type of adjustment Source: Increases in the Gini coefficient are calculated from table 4.1. tween the "populist" type of adjustment and low increases in inequality9 The average increase in Gini for the three "populist" adjustment coun- tries (Poland, Hungary, Slovenia) was less than 2 Gini points, and except for Slovakia they experienced smaller increases in inequality than any other country. The average increase among "compensators" and "non-compen- sators" was approximately 10 and 12 Gini points, respectively. Each move- ment toward "tougher adjustment" (i.e., from "populists" to "non-com- pensators") was associated, on average, with a Gini increase of about 4 points (see the regression line in figure 4.5). However, if Russia and Ukraine are excluded, the average inequality increase among the "non-compensa- tors" is less than among the "compensators:" the monotonic relationship between "softer" adjustment and lower increase in inequality no longer holds. 9. The partial correlation coefficient between the type of adjustment (where 1 = non-compensators, 2 = compensators, 3 = populist) and the Gini increase is -0.42. On the other hand, there is no statistically significant correlation (r = -0.04) between the type of adjustment and the change in an economic liberalization index (defined for transition economies by de Melo, Denizer, and Gelb 1996) that includes privatization, internal mar- ket liberalization, and external market liberalization. Similarly, there is only a weak rela- tionship between the change in the liberalization index and the change in the Gini (-0.11). The Original Income Distribution Statistics 47 Distribution of Income Sources: Wages, Social Transfers, and Private Sector Income How can the increased income inequality during the transition be accounted for? Disposable income can be defined as the sum of wages (w), cash social transfers (t), and non-wage private sector income (p). The Gini coefficient of disposable income, (G), is equal to the weighted average of the concentration coefficients of the three income sources (wages, transfers, and non-wage pri- vate sector income) Ci where weights are sources' shares (S) in total income:", 4.1 G =Y3S.C. = S C ,+ S,C,+ SC. i= zI t p p The change in the Gini between two dates (before and after the transition) can be written as: 4.2 AG =x AS,C, + ACQ,SW + AC,S + AC Sp +X3AS,ACi. pp i_ The first term on the right hand side shows the change in Gini due to chang- ing shares of different income sources; the next three terms show the change due to changing concentration coefficients of income sources; and the last term is an interaction term. Table 4.2 shows the decomposed change in the Gini for selected countries between a pre-transition year and 1993-96. The following conclusions can be made. First, the change in income composition has had little relation to increased inequality. In the only country where income composition did have a signifi- cant impact on inequality (Russia), it contributed to reduce inequality; that is, income composition in 1994 was more favorable to equality than it was in 1989. This is chiefly because social transfers, which were the most equally distributed income source in Russia before the transition, increased their share in overall income after the transition. In other countries, only about 1 Gini point was added to or subtracted from total inequality by changes in income composition. Second, higher concentration coefficients of wages drove the overall Gini up in all countries. It was the most important factor behind increases in in- 10. The concentration coefficient captures both the inherent inequality with which a given income source is distributed (source Gini coefficient) and the correlation of that source with overall income. Thus, an inherently unequal source, such as social assis- tance, with a high Gini coefficient will have a low or negative correlation with overall income (because most social assistance recipients are poor), and its concentration coeffi- cient will be low or negative. For a more detailed definition of concentration coefficient, see footnote 9 in chapter 2. 11. The analysis here is based on HBS data. The three income sources add exactly to disposable income. Table 4.2. Decomposition of the Change in the Gini Coefficient between Pre-transition and 1993-96 Due to: Change in concentration of: Change in Out of which: Non-wage Overall composition Social Non-pension private Interaction Gini Country (years) of income Wages transfers Pensions transfers sector term change Hungary (1989-93) -1.3 +5.9 -0.6 +1.4 -0.2 -0.6 -1.3 +2.2 Slovenia (1987-95) -0.2 +3.6 -0.6 -0.1 -0.4 +0.4 -3.8 +2.6 Poland (1987-95) -1.7 +3.4 +3.5 +3.2 -0.1 +0.8 +0.9 +7.0 Bulgaria (1989-95) +1.4 +7.8 +0.9 +0.4 +0.4 -0.4 +0.3 +10.0 Latvia (1989-96) -1.6 +15.0 -1.5 -2.0 +0.5 +1.4 -3.3 +10.0 Russia (1989-94) -3.4 +17.8 +5.1 +3.9 +0.4 +3.0 +1.2 +23.6 Note: All data in Gini points. The years in brackets show the dates between which the Gini change is calculated. The data sources and end-years for Bulgaria, Latvia, Poland and Slovenia are not the same as in table 4.1. This explains differences in the overall Gini change. Source: Calculated from the countries HBSs (see appendix 1). Inequality 49 equality. Increased wage concentration was responsible for between 3.5 and 8 Gini points of increase in Eastern Europe, and for 15 to 18 Gini point in- creases in Latvia and Russia. In the latter two countries, these huge increases were due not only to a greatly increased concentration coefficient of wages, but also to a high pre-transition share of wages in income. Thus, a very high weight attaches to a more unequal concentration of wages that occurred dur- ing the transition. An increase in the concentration of non-wage private sec- tor income was responsible for 3 Gini points of increase in Russia and about 1.5 points in Latvia, while its impact was negligible in Eastern Europe. Third, the effect of the changed concentration of transfers on inequality was not uniform across countries. In Bulgaria, Hungary, and Slovenia, for example, the concentration of transfers did not change. In Latvia, on the other hand, better targeting of transfers reduced inequality by 1.5 Gini points. In Poland and Russia, transfers contributed to an increase in inequality. This was due to a greater concentration coefficient of pensions. Non-pension trans- fers, because of their initially small size, did not anywhere have much of an impact on inequality. 12 Decomposition of the increase in inequality in table 4.2 is based on the change between the two data points (before the transition and 1993-1996). The two end-data points can mask changes in the distribution or shares of various income sources in the intervening years. For the six countries shown in figures 4.6-4.17, annual HBSs are available. They allow us to chart annual developments in the concentration and share of wages, cash social transfers and non-wage private sector. The developments in Bulgaria illustrated in figures 4.6 and 4.7 are straight- forward, and to some extent typical of the transition. The rising concentra- tion of wages (from around 20 to 35) contributed strongly to inequality. The concentration coefficient of non-wage private sector income, which was al- ready high before the transition, remained high while the share of non-wage private sector income increased. This also pushed up overall inequality. Pen- sions' concentration and share both remained unchanged, thus leaving in- equality unchanged. Finally, non-pension transfers were too small (less than 5 percent of total income) to make any difference in the overall Gini. Polish results illustrate a different story (see figures 4.8-4.9). Although wage concentration increased markedly, the most important development was in the area of pensions: their rising concentration and rising share in overall income. Pensions thus contributed strongly to increase inequality. In 1995, pensions had the same concentration coefficient as wages and non-wage pri- vate sector income. The fact that concentration coefficients of the three in- come sources converge means that income composition is almost equal across 12. To some extent, this conclusion differs from Cornia's observation that "the rela- tive importance of redistribution [via transfers] has grown... Targeting of these [social] transfers has generally improved or remained sufficiently progressive" (1994, p. 39). 50 Income, Inequality, and Poverty during the Transition Figure 4.6. Composition of Disposable Income in Bulgaria, 1987-95 (percent) 60 - 50- -40 - Private . ° 30 a,20-,, .. ' Pensions 10 Other transfers 1987 1988 1989 1990 1991 1992 1993 1994 1995 Figure 4.7. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Bulgaria, 1989-95 60 50 --- ---- -* Private -' 40- .. -'.^ 30 .8 ~Wages/ 2^ 20-- 20 tJ 10 Pensions - Other transfers , -10 1989 1990 1991 1992 1993 1994 1995 Note: The more unequally distributed the income source, the higher the coefficient of concentration. The concentration coefficient shows how much a given source "pushes" up the overall inequality. Source: 1989-95 Bulgaria Household Budget Surveys. Inequality 51 Figure 4.8. Composition of Disposable Income in Poland, 1987-95 (percent) 60 - 50 40 0 30 - Private ___ 0 o , -' ". 20 -- Q 0 Pensions ".,." 10 - Other transfers 1987 1988 1989 1990 1991 1992 1993 1994 1995 Figure 4.9. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Poland, 1987-95 5!0 _ Private,.- ,,, , 30 Wa 8 20 Pensions 0 10 -10 -Other transfers . -20 - 1987 1988 1989 1990 1991 1992 1993 1994 1995 Note: The more unequally distributed the income source, the higher the coefficient of concentration. The concentration coefficient shows how much a given source "pushes" up the overall inequality. Source: 1987-95 Poland Household Budget Surveys. 52 Income, Inequality, and Poverty during the Transition Figure 4.10. Composition of Disposable Income in Slovenia, 1987-95 (percent) 80 70- o 50- ° 40- 0 ii 30- 20 Private- . 10 - Pensions Other transfers O . .. . _,_,_.......___ 1987 1988 1989 1990 1991 1992 1993 1994 1995 Figure 4.11. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Slovenia, 1987-95 40 30 - Private, ^'* g1 10 - \Pjensions 200 o Other transfers U -20 -2 1987 1988 1989 1990 1991 1992 1993 1994 1995 Note: The more unequally distributed the income source, the higher the coefficient of concentration. The concentration coefficient shows how much a given source 'pushes" up the overall inequality. Source: 1987-95 Slovenia Household Budget Surveys. Inequality 53 Figure 4.12. Composition of Disposable Income in Hungary, 1987-93 (percent) 60 W g 50 - 40 40 - o 30 Private 20 -- , 20 - ~~Pensions - - -- - . -_ - --' 10 Other transfers 1987 1988 1989 1990 1991 1992 1993 Figure 4.13. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Hungary, 1987-93 40 - Private , -' - .g 3 Wages " u 20- A 10 - _Pensions -10 .............................................................................-- ' ' - ' - - - . . -10 - - Other transfers -20 - I I I I - 1987 1988 1989 1990 1991 1992 1993 Note: The more unequally distributed the income source, the higher the coefficient of concentration. The concentration coefficient shows how much a given source "pushes" up the overall inequality. "Private income" in 1993 does not include consumption-in-kind. Source: 1987,1989, 1993 Hungary Household Budget Surveys. 54 Income, Inequality, and Poverty during the Transition income brackets: in other words, a poor person will derive the same percent of income from wages, pensions, or non-wage private sector income as a rich person will. This is a rather unexpected outcome. In Slovenia and Hungary (figures 4.10-4.13) none of the concentration co- efficients showed a clear tendency to increase or decrease (the increased con- centration of wages in Hungary is the only exception). This explains very small increases in Ginis in both countries. Russia represents a unique case: all income sources' concentration coeffi- cients were higher in 1994 than before the transition and all have pushed overall inequality up (see figures 4.14 and 4.15). The only factor that has mod- erated the increase in inequality was a shift toward more equally distributed income sources: that is, towards transfers and non-wage private sector in- come, which prior to the transition had smaller concentration coefficients than did wages. Transfers and, in particular, pensions either left inequality unchanged (as in Bulgaria, Slovenia, and Hungary) or they contributed to an increase in inequality (as in Poland and Russia). The only exception noted here is Latvia, where an improved (that is, more pro-poor) concentration of pensions resulted from the introduction of almost flat pensions in 1992, a development which is reflected in the steep, downward-sloping line for pensions in figure 4.17.13 Disparity among Social Groups Changes in income distribution have also been accompanied by changes in the relative position of social groups. Under previous regime, care was taken to maintain some level of balance (or "parity") between the average income of workers and of farmers."4 Communist concern with the relative position of vari- ous social groups was also reflected in the design of HBSs, where survey repre- sentativeness was ensured at the level of the social group, but not necessarily at the level of the population (see appendix 1 for further discussion of HBSs). Table 4.3 shows what happened to the average per capita income of work- ers', farmers', and pensioners' households between 1987-88 and 1993-94. 15 As expected, average real incomes of all social groups have declined. How- ever, it is the difference in the decline between the various groups that is of interest here. When the difference in decline between two groups is less than 3 percentage points, it is assumed that no change in the groups' relative 13. The concentration coefficient of pensions decreased from 34 in 1989 to -4 in 1995. 14. Okrasa (1988, p. 637) argues that redistribution policies under Communism were mostly designed to insure vertical equality among social groups. In some aspects, how- ever (for example, in access to safe water and sanitation), rural populations, particularly in the less developed republics of the former U.S.S.R., were at a disadvantage. 15. Workers are, in principle, both those employed in the state and private sector. Many private sector workers are self-employed, however, and are thus included in the category "self-employed" or "other" which is not shown here. Inequality 55 Figure 4.14. Composition of Disposable Income in Russia, 1989-94 (percent) 80 - 70 c, 60 - 0 50 2 40 0 t 30 Private 20 - , -- '= - Pensions 10 ~ _ _ _ Other transfers 1989 1990 1991 1992 1993 1994 Figure 4.15. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Russia, 1989-94 70- 60 - Private 50- Wages S 40 -.,~Other transfers 0) U .-.. ~~~~~~~~~~~~Pensions 'IN 10 0 . ...........z /......... ....................................... ......................................... .. .. ......... -20_ * / | I I 1989 1990 1991 1992 1993 1994 Note: The more unequally distributed the income source, the higher the coefficient of concentration. The concentration coefficient shows how much a given source "pushes" up the overall inequality. Source: 1989: Family Budget Survey; 1992 and 1994: Russian Longitudinal Monitoring Survey, rounds 1 and 4 respectively. 56 Income, Inequality, and Poverty during the Transition Figure 4.16. Composition of Disposable Income in Latvia, 1989-96 (percent) 80 - 70 - 60- 0 i 50- U 40- 30 20 - Private , - 10 - ,, _,: _ _ ~ Pensions - * Other transfers 10 -l 1989 1990 1991 1992 1993 1994 1995 1996 Figure 4.17. Concentration Coefficients of Wages, Cash Social Transfers, and Non-wage Private Sector Income in Latvia, 1989-96 40 Private ' 30 Wage 0 h 20 10 u 10 ~~~~~~Pensions Other transfers -- --- - -10 - II 1989 1990 1991 1992 1993 1994 1995 1996 Note: The more unequally distributed the income source, the higher the coefficient of concentration. The concentration coefficient shows how much a given source "pushes" up the overall inequality. Source: 1989 and 1992-93 Family Budget Survey; 1995-96: New Household Budget Survey. Inequality 57 Table 4.3. Change in Real and Relative per Capita Income of Worker, Farmer, and Pensioner Households (workers' households real per capita income in 1987=100) Country 1987-88 1993-94 Change in relative position Belarus Workers 102 86 W=F Farmers 101 85 Bulgaria Workers 102 53 W>F Farmers 119 64 Czech Republic Workers 102 73 P>W=F Farmers 93 64 Pensioners 61 54 Hungary Workers 100 73 P>W Farmers 90 n.a. Pensioners 92 68 Latvia Workers 101 54 P>W>F Farmers 100 41 Pensioners 59 34a Lithuania Workers 102 43 P>W>F Farmers 100 33 Pensioners 58 46a Poland Workers 107 85 P>W>F Farmers 121 77 Pensioners 95 89 Romania Workers 100 76 F>P>W Farmers 74 59 Pensioners 88 68 Russia Workers 103 59 W>F Farmers 88 30 Slovak Republic Workers 102 71 P>W=F Farmers 95 63 Pensioners 64 52 Slovenia Workers 94 81 W>F Farmers 84 66 a. 1992. Source: Calculated from the countries' household budget surveys (see appendix 1). 58 Income, Inequality, and Poverty during the Transition positions has taken place. For instance, if workers' real income lost 15 points and farmers' real income lost 16 points (as in Belarus), their relative posi- tions are considered unchanged. However, if workers lost 49 points and farmers lost 55 points (as in Bulgaria), then workers are considered to have done better than farmers. Remarkably, the following rules apply for all coun- tries here: * Pensioners' position in relation to workers has improved: P>W. * Workers' position in relation to farmers has either improved (in six countries) or remained the same (in three countries): W2F. 16 * Pensioners' position in relation to farmers has improved in all coun- tries: P>F. Farmers' households real incomes have declined the most almost every- where. This was caused by a variety of factors: decreased agricultural pro- duction in all countries; 17 removal of input subsidies (for example, for fertil- izer and gasoline) combined with liberalization of food imports (in Poland, Czechoslovakia, Russia, and the Baltic countries); 18 chaos associated with land privatization and uncertainty of ownership (in Albania, Bulgaria, Esto- nia, Romania, and Russia); and the end of an explicit government policy of ensuring rural-urban income parity (Poland, Czechoslovakia). Workers have done slightly better than farmers, and pensioners have done better than both. How have the "new" private sector entrepreneurs, and the self-employed done? Anecdotal evidence and common sense suggest that they would be among those who benefited most from the transition. It is important, however, to keep in mind that this group is more heterogeneous than workers or farmers are. The private sector encompasses not only the self-employed professionals (for example, doctors, accountants, engineers, and computer specialists) but also coffee shop owners, shoemakers, hoteliers, and repairmen. It includes also small- scale employers and "capitalists" (that is, owners of larger plants or factories). The difference in income between a self-employed shoemaker and a large-scale capitalist can be so large that it becomes all but meaningless to include them in the same group. From the perspective of the transition, however, both are part of the private sector, because their income comes from private activities. Poland is the only country for which there is evidence about how the self- employed have done since the transition. The Polish HBS introduced the cat- egory of "self-employed" (outside agriculture) as a new social group in 1992.'9 The self-employed represent about 6 percent of the 1993 and 1994 sample. Their 16. Except in Romania, where farmers have done better than either pensioners or workers. 17. In 1993-94, agricultural output in Eastern Europe, Russia, and Ukraine was about 20 percent below its 1987-88 level. 18. In addition, wage arrears in the Russian agricultural sector have been particu- larly severe (see Braithwaite 1997, p. 65). 19. The first HBS results reflecting this new category were published in 1993. Inequality 59 average income and expenditures are the highest of any group: approximately 20 percent above the overall mean in 1993 (World Bank 1995a, p. 10) and 30 percent in 1994 (Polish Central Statistical Office 1995 table 7 and 9). Being self- employed reduces the probability of poverty by 11 percent, when other fac- tors, such as education, type and size of household, location, and so on, are controlled for.20 20. This result is obtained using the probit analysis. The coefficients on other ex- planatory variables are virtually the same whether or not the probit regression is esti- mated with the self-employment dummy variable. 5 Poverty What Happens to Poverty When Income Goes Down? Chapter 3 discussed the dramatic declines in output experienced in transi- tion countries. As incomes decreased, poverty went up. In this chapter, two simple measures of poverty will be used to study poverty: the headcount index and the poverty deficit. The headcount index gives the percentage of people who are poor because their income is below a certain threshold. The poverty deficit is the sum of all income shortfalls (difference between the threshold and one's income). In other words, it gives a weight to each poor individual equal to the amount by which that individual's income falls short of the poverty line. Poorer individuals thus "count" more, according to this measure. The poverty deficit shows the total amount of money needed to bring all those who are poor up to the poverty line. The poverty deficit is often expressed as a percentage of GDP in order to show the effort a country needs to make to "solve" its poverty problem.' When the poverty deficit is divided by the total number of the poor, the average income shortfall of the poor (sJ) is obtained. This shortfall can then be expressed as a percentage of the poverty line, and will be here denoted by s. For example, let the poverty line be 4, and let there be three individuals with incomes of 1, 2, and 10. The first two are considered poor. The headcount index is 2 out of 3 (that is, 66.6 percent). The total poverty deficit is equal to (4-1) + (4-2)=5. Five units of income will be needed, then, to bring the poor up to the poverty line. The average income shortfall of the poor in this ex- ample is equal to 5 divided by 2, or 2.5. The average shortfall as a percentage of the poverty line is 2.5 divided by 4, or 62.5 percent. In chapter 4 it was concluded that income distribution has become more unequal. Both the decline in overall income and the more unequal distribu- tion of income will increase the incidence of poverty. The poverty deficit should also then increase: if the poverty line is fixed and everyone's income declines, then the overall amount needed to "solve" the poverty problem must go up. The example provided in box 5.1 may help explain this. We may 1. The assumption behind the poverty deficit is that all money used for poverty alle- viation is paid only to the poor (with no leakage to the non-poor) and is paid in exact amount needed to bring the poor just up to, but not above, the poverty line (that is, there is no "spillover"). This is the assumption of so-called perfect targeting. In reality, no more than 50 to 60 percent of funds specifically destined for poverty relief are perfectly targeted. 60 Poverty 61 Box 5.1: How Poverty, According to Certain Measures, Changes as Income Declines Suppose there are 100 people with an average income of $10 (so that their total income is $1,000). Let the poverty line be $4 and let there be ten people with an income of less than $4. Let the average income of these ten people be $2. Their average income shortfall is thus $2, and the poverty deficit is $2 times 10 people, or $20. The shortfall represents 2 percent of overall income or GDP. Suppose, then, that income declines uniformly by 20 percent. Total income becomes $800. The ten people who were poor remain poor, and their average in- come drops to $1.60. The poverty deficit goes up to $2.40 times 10, or $24. In addition, however, let an additional ten people "slide" below the poverty line. Their incomes formerly ranged between $4 and $5 and are now between $3.20 and $4. Let their average income be $3.60. Their poverty deficit will be $(4-3.60) times 10, or $4. The new total poverty deficit will thus be equal to $28, or 3.5 percent of the new GDP. What does this example show? It shows, first, that a 20 percent uniform decline in income results in a 10 percentage-point increase in the headcount index. This gives an implicit elasticity of 0.5 (10 percentage points divided by 20 percent), which as will be seen later, is realistic. Second, doubling the poverty headcount from 10 to 20 led to a 40 percent increase in the poverty deficit (from 20 to 28). A smaller increase in the poverty deficit than in the headcount is due to the decline in the average income shortfall, which went down from $2 to $1.40. This is also very common as the "new" poor tend to fall slightly below the poverty line. Third, the poverty deficit as a pro- portion of GDP has risen by 75 percent, both because the poverty deficit has gone up and because the GDP has shrunk. In terms of percentage changes in the poverty measures, the following may be expected to apply: * The poverty headcount will increase the most. * The poverty deficit as a share of GDP will increase less. * The poverty deficit will increase even less. * The average income shortfall might fall. 62 Income, Inequality, and Poverty during the Transition also expect that the more income declines and the greater the increase in inequality, the greater the increase in the headcount index and the poverty deficit. Other aspects of changes in poverty measures in the presence of large in- come declines can be elucidated through simple algebra. For the sake of sim- plicity, it is assumed here that income distribution does not change and all incomes decrease by the same proportion. Such changes in income are con- sidered "uniform." The poverty deficit (PD) is obtained by multiplying the average income shortfall of the poor, s (in percent) by the poverty line (z) and the number of the poor (P). This amount can then be related to the country's GDP written as the product of its GDP per capita (y) and its population (N). After some simple manipulation, the poverty deficit/GDP ratio becomes: PD szP z P z 5.1 GDP = =s HC where HC is the headcount index. As income shrinks across the board, the ratio between the (fixed) poverty line and GDP per capita (z/y) increases by definition, as does the poverty headcount. The average shortfall s may be reduced if the distance between the new poor and the poverty line is small (as per the example in box 5.1). There is a further change in the poverty deficit that is not directly revealed by equation 5.1, however. To consider this, the poverty deficit is written as: 5.2 PD= f (z-y) f(y) dy. The poverty deficit is equal to the sum of all distances between the poverty line and actual income (z-y) multiplied by the number of people who have such income,f(y), wheref(y) is the density function. An across-the-board slide in income is formally equivalent to a corresponding increase in the poverty line z. Consider what happens to the poverty deficit when the poverty line increases infinitesimally. If equation 5.2 is differentiated with respect to z the following is obtained: 5.3 dPD_ fz f(y) dy + (Z-Z) f(z) z Jo f(y) dy = F(z). Equation 5.3 shows that the change in the poverty deficit following an infinitesimal uniform decline in income is equal to the area between 0 and z under the density functionf(y) as shown in figure 5.1.2 The clear implication 2. For a more intuitive explanation, see appendix 3. Poverty 63 Figure 5.1. Income Density Function and the Poverty Line f(z2) f(z,) ZI Z2 Mean Income is that the higher the poverty line, the greater the increase in the poverty deficit for a given absolute decline in income. If, for example, the poverty line is z2 then the poverty deficit will increase by more than it would if the pov- erty line were z1 (see figure 5.1). Because in the analysis in this chapter the same real poverty line is used for all countries ($120 per capita per month at international prices; see the section below, By how much has poverty increased?), the place of the poverty line within the income distribution curve will differ between the richer and poorer countries. For example, the purchasing power parity (PPP) $120 pov- erty line amounts to only 38 percent of Hungary's average income (approxi- mately equal to point z1 in figure 5.1), and 70 percent of Estonia's average income (approximately equal to Z2 in figure 5.1). The areaJ f(y) dy will be greater than the areaJ 2 f(y) dy. Thus it can be concluded that: Jo (A) A given absolute decrease in income will raise the poverty deficit more in the case of a poor country (Estonia) than it will in the case of a richer country (Hungary). In addition, a uniform decline in income (equivalent to an increase in the poverty line) will have a more than a proportional impact on the poverty deficit: if incomes shrink by, say, 10 percent, the poverty deficit will rise by more than 10 percent. This can be seen if equation 5.3 is multiplied by z/PD, 64 Income, Inequality, and Poverty during the Transition thus yielding the elasticity of the poverty deficit with respect to the poverty line (equation 5.4). The elasticity is equal to the reverse of the shortfall, and is always greater than 1. The percentage shortfall would normally increase with z and hence elasticity would decrease with z.3 This is equivalent to elasticity 5.4 increasing with income. dPD z Z fz f(y) dy dz PD 5.4 J ro-Z (z-y) f(y) dy z 1 = - >1 z-yp s where y- = average income of the poor. In other words, (B) If percentage declines in income are the same in poor and rich coun- tries, the poverty deficit will increase, proportionately more, in rich countries. Finally, with a uniform infinitesimal slide in incomes, the poverty headcount will increase by f(z), that is, 5 5 dHC dF(z) _ dz dz (C) Because the density function is "thicker" (higher) around z2 than it is around z, (see figure 5.1), the poverty headcount will, for a given absolute decrease in income, increase more for a poorer than for a richer country. Another way to express this is to say that the relationship between HC (with a given poverty line) and mean country income (or consumption) is convex: the higher the income, the smaller the effect of a given change in income on the headcount index (see Ravallion 1993, p. 6). The exception to this would be if a country were so poor that the poverty line was situated very much to the right of the mode where the density function is "thin." By How Much Has Poverty Increased? One of the central objectives, if not the central objective, of a comparative study of poverty is to compare poverty rates and poverty deficits both among 3. For example, for an extremely high poverty line equal to the highest income in the country (y), elasticity becomes y/(y- Y) and tends to 1. Poverty 65 countries and within a given country at different points in time. For this sort of comparison, however, a common poverty line is needed. The same poverty line across countries implicitly treats each individual equally, regardless of where he or she lives. To have the break-down of poor people among countries is important because it helps the development of an international anti-poverty policy. If the objective is to decide where, among a given group of countries, available resources should best be directed in or- der to help the poor, cross-country comparability is indispensable. It is rea- sonable to assume that international effort should be directed toward coun- tries in which, according to international standards, (1) many poor people live, or (2) poverty rates are high even if the absolute number of poor is small (because the country's population is small). For the first, it is necessary to use the same poverty yardstick in order to determine the distribution of the poor by country, and for the second, it is necessary to use the same yardstick to compare headcount indexes in different countries. Once the poverty line is fixed in time and among countries, a full comparison is possible. For ex- ample, we can compare Hungary's poverty headcount in 1993 not only with Hungary in 1989, but also with Russia in 1989. The first step in calculating such a poverty line is to establish a single bench- mark expressed in the same "currency" for all countries. Four international dollars per capita per day will be used here.4 These dollars have the same purchasing power over consumption goods in all countries. Then it is neces- sary to find the purchasing power parity PPP exchange rate for each country: how many units of local currency are needed to buy $1 worth of consump- tion goods at international prices? These data are obtained from the multilat- eral International Comparison Project (ICP), whose most recent round is for 1993.5 The PPP exchange rate for each country is then multiplied by $120 per month to obtain the monthly poverty line expressed in domestic currency. Of course, a poverty line that implies the same purchasing power over goods and services across countries requires different amounts of actual dollars in each country: the lower the country's price level compared with that of the world, the fewer actual dollars are needed to buy the same amount of goods. For example, in 1993, $51 per month was needed to reach the $PPP 120 pov- erty line in Poland; $57 in Hungary; $30 in Bulgaria; and only $21 in Ukraine. By contrast, it would require $165 per month to reach the same poverty line in Switzerland, $125 in Finland, and $59 in Turkey. Once the 1993 benchmark poverty line in domestic currency is established, poverty lines for other years can be obtained, if needed, simply by inflating 4. This is calculated at 1990 international prices. 5. Data supplied to the author by European Comparison Project. Hungary, Poland, and the former Yugoslavia have been included in the ICP since 1985. In 1990, Romania and the U.S.S.R. were added. In the 1993 ICP exercise, all transition economies partici- pated. Data for Central Asian countries were not published, however. For these countries the base price level used is the 1990 ICP result for the Soviet Union as published in United Nations Economic Commnission for Europe (1994). 66 Income, Inequality, and Poverty during the Transition the 1993 poverty line by the cost-of-living index. Full comparability of differ- ent points in time for a given country, as well as of different countries is thus possible. It is, however, worth pointing out again that the actual dollar amounts needed to reach the poverty line in each country, and in each year, will differ. In Poland, the poverty line in 1987 amounted to $27 per capita per month (at the official exchange rate); in 1993, because of the rapid real appre- ciation of the zloty, it was, as mentioned before, $51. On the other hand, in Moldova the poverty line in 1988 (as in the rest of the Soviet Union) was $25 per capita per month;6 but in 1994, one needed only $13 to reach the poverty line. More formally, the formula for calculating the poverty line can be written as follows: PLit = PL93 i, COLi,(t93) where PLk, = the poverty line in domestic currency for country i and year t; PL$93 = the common poverty line in international dollars at 1993 prices; ER*jg3 = the consumption purchasing power exchange rate of i country's currency in 1993; and COLi(tS93) = the change in the cost-of-living index between year t and 1993 for country i. The amount of $PPP 4 per capita per day is a relatively high poverty line. It is four times higher than the World Bank line of absolute poverty. But the level of income of East European and former Soviet countries, and the com- pression of their income distributions make $PPP 4 per day per capita a rea- sonable poverty line.7 This line is below the "accounting" social minimum lines for most East European countries, that is, the lines which are not used to define eligibility for social assistance but represent same vague "desirable" minimum. How- ever, they are widely published and often used to calculate the number of the poor. The "accounting" social minimum lines in Eastern Europe range from $PPP 170 to $PPP 300 per capita per month (see Milanovic 1996). For ex- ample, the Polish social minimum in 1987 was zloty 14,222 per capita per month, while the $PPP 120 poverty line for the same year is zloty 7,265. The gap is even greater in 1993. The Polish social minimum for June 1993 was zloty 2.1 million per capita per month; the poverty line as calculated here is zloty 880,000. The situation is somewhat different in republics of the former Soviet Union. At the beginning of the transition, like in Poland, social mini- mum lines8 were higher than the poverty line used here. The Soviet social 6. The 1988 ruble exchange rate is calculated according to a two-to-one "blend" between the official rate (Rs. 0.6) and the parallel market rate (Rs. 4.2). 7. In a recent World Bank study of income distribution and poverty in Latin America and the Caribbean, the poverty line used was $PPP 2 per day per capita (see Psacharopoulos and others 1992). 8. Actually, there was a single U.S.S.R.-wide line until 1991. Poverty 67 minimum in 1987-88 was Rs. 78,9 while the poverty line used here is only Rs. 54. With time, however, the social minimum in all these republics was scaled down in real terms, and by 1993-94, the new social minimums were very close to the $PPP 120 poverty line. For example, in 1992 Russia officially adopted a new subsistence minimum that was about two-thirds of the old social minimum.'0 In July 1993 (the date of the Russian survey used here), the official minimum was 21,206 rubles per month per capita, almost the same as the poverty line proposed here of Rs. 21,496.11 Because both lines are indexed by the cost-of-living index, their amounts continue to converge. The results: universal increase in poverty Table 5.1 shows the estimated poverty rates for the eighteen countries. For all countries, poverty rates are calculated from "uncorrected" HBS data from the two end-periods: 1987-88 and 1993-95. These "uncorrected" estimates are here referred to as INCOMEI estimates. The average per capita $PPP in- comes range from $PPP 75 per capita per month in the Kyrgyz Republic to $PPP 480 in Slovenia (see also figure 5.2). (For further discussion of data sources, biases, and main problems, see appendix 1.) The total estimated number of the poor in the eighteen countries has risen twelvefold from nearly 14 million before the transition or about 4 percent of the population, to 168 million in 1993-95, or approximately 45 percent of the population. Poverty increased in all eighteen countries. The headcount increase, how- ever, was very uneven. In the richer countries of Central Europe (the Czech Republic, Hungary, the Slovak Republic, and Slovenia) the percentage of the poor rose, on average, modestly from less than 1 percent to 2 percent. In Poland, it rose from 6 percent to 20 percent. It is more difficult to interpret results for the former Soviet republics because inadequate coverage of sur- veys before the transition tended to underestimate poverty, while during the transition, dramatic changes in the economy (such as, the expansion of the informal sector) were not reflected in household surveys, thus biasing cur- rent poverty estimates upwards. The two biases reinforce each other when the overall change in poverty is being considered. Despite these and other 9. Or Rs. 84 if goods were valued at free market prices (see Braithwaite, 1994, p.4). 10. The Russian subsistence minimum defined in 1992 is based on the cost of a bundle of goods, where the share of food is 68 percent. The bundle was devised by the Russian Institute of Nutrition of the Academy of Sciences and by WHO. It became the official standard for poverty measurement in Russia in March 1992. For details, see Gontmakher and others (1995, pp. 43-52) and World Bank (1995e, p. 15). 11. The official poverty line was used in a detailed study of poverty in Russia (World Bank 1995g) and in Klugman (1997). However, in both cases, this "central" per capita poverty line was scaled to take into account economies of size and the lower expenditure requirements of children. This explains why the headcount for Russia found in Foley (1997) is about 40 percent, that is, about 10 percentage points lower than the one pre- sented below in table 5.1. Table 5.1. Estimated Poverty Headcount and Poverty Deficit in 1987-88 and 1993-95 Using HBS Income 1993-95 data Total number Shortfall Total Average Poverty of the poor as % of poverty income Type and headcount (%) (millions) poverty deficit as % per capita year of Country 1987-88 1993-95 1987-88 1993-95 line of GDP Elasticity, ($PPP pm) data Balkans and Poland 5 32 3.6 22.4 28 2.2 0.5 193 Bulgaria 2b 15 0.1 1.3 26 1.1 0.3 282 A:93 Poland 6 20 2.1 7.6 27 1.4 0.4 213 SA:I/93 Romania 6b 59 1.3 13.5 32 5.4 0.7 123 M:3/94 Central Europe <1 2 0.1 0.4 25 0.1 0.1 348 & Czech Republic 0 <1 0 0.1 23 0.01 0.01 411 M:1/93 Hungary' 1 4 0.1 0.4 25 0.2 0.2 266 A:93 Slovakia 0 <1 0 0.0 20 0.01 0.01 332 A:93 Slovenia 0 <1 0 0.0 31 0.02 0.01 481 A:93 Baltics 1 29 0.1 2.3 33 3.1 0.6 204 Estonia 1 37 0.02 0.6 37 4.2 0.7 172 Q3:95 Latvia 1 22 0.03 0.6 28 2.3 0.5 213 Q4:95 Lithuaniac 1 30 0.04 1.1 34 2.9 0.5 212 A:94 Slavic republics 2 52 3.5 112.1 39 4.8 0.5 170 Belarus 1 22 0.1 2.3 26 1.2 0.5 197 Q1:95 Moldova 4 66 0.2 2.9 43 7.0 0.6 115 A:93 Russia 2 50 2.2 74.2 40 4.2 0.6 181 Q3:93 Ukraine 2 63 1.0 32.7 47 6.9 0.5 136 M:6/95 Total without Central Asia 3 43 7.2 137.2 31 3.1 230 1993-95 data Total number Shortfall Total Average Poverty of the poor as % of poverty income Type and headcount (%) (millions) poverty deficit as % per capita year of Country 1987-88 1993-95 1987-88 1993-95 line of GDP Elasticityo ($PPP pm) data Central Asia 15 66 6.5 30.7 47 9.8 0.5 113 Kazakhstan 5 65 0.8 11.0 39 9.2 0.7 115 A:93 Kyrgyz Republic 12 88 0.5 4.0 68 64.4 0.2 75 Q3:93 Turkmenistan 12 61 0.4 2.4 40 7.7 0.6 124 A:93 Uzbekistan 24 63 4.8 13.3 39 12.4 0.6 118 A:93 Total transition 4 45 13.6 168.0 35 3.5 215 Comparators Brazil 33 48.3 44 4.4 0.3 466 A:89 Colombia 35 11.6 40 5.4 0.3 360 A:92 Ecuador 35 3.9 31 4.4 0.5 219 A:94 Paraguay 44 2.1 51 8.1 0.3 266 A:95 Malaysia 31 18 5.1 3.6 29 0.8 0.3 403 A:87,95 Turkey 31 16.7 33 3.8 0.5 255 A:87 United Kingdom 1 <1 0.6 0.5 good A:88,92 Note: "Slavic republics" includes Moldova. Poverty line = 120 international dollars per capita per month. A = annual data; SA = semiannual data; M = monthly data; Q = quarterly data. 10/93 means that the data refer to October 1993. 1/93 means that the data refer to the first half of 1993. Poverty rates were calculated using World Bank software POVCAL. For poverty headcount, the number of the poor, poverty deficit as percentage of GDP, and income per capita the regional means are weighted averages; for elasticity and average shortfall, the regional means are unweighted averages. a. Elasticity is percentage point change in poverty headcount divided by percentage change in income (around the poverty line). b. 1989 data. c. Income does not include consumption-in-kind in 1993-95. d. Estimate based on the ratio between equivalent adult units and per capita measures. Source: Transition economies: calculated from household budget surveys presented in appendix 4. For caveats and data biases, see appendix 1. For other details regarding calculations, see appendix 5. For comparator countries, the sources are as follows: for the United Kingdom, Family Expenditure Survey data as reported in United Kingdom Central Statistical Office (1991, appendix 1) and United Kingdom Central Statistical Office (1994, appendix 1); for Turkey, Turkey Statistical Yearbook 1990, pp. 206-207; for Malaysia, Ahuja (1997, Annex 2); for Brazil, Psacharopoulos et al. (1992); for Colombia, World Bank (1994b, vol. 1, p. 1 and vol. 2, Annex 1); for Ecuador, World Bank (1995d, vol. 2, p. 4); for Paraguay, Encuesta de Hogares 1995. EUROPE AND CENTRAL ASIA A.708O 'o MONTHLY INCOME PER CAPITA ($PPP) AC CA INCOME PER CAPITAA RC TIC C EA N ($PPP PER MONTHI) SKaa 1,. . .. .. . .. .. . o ________________________________________ IO~~~~~~~~~ 2O~~~~~~~~~~~~ 4O~~.. ............. TV 40 ar pI o S~~~~~~~~~~~~oeat p~~~~~~~~~~~~~~~~~~~. AA~ ~ ~~~ ) RUSSIAN FEDERAIION~~~................... ~~~ C [ATYLA o.*. ESTOIB~............................ (.11 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ......... ............. ..XN S \ .'~~~~ ~~...~~:.: ~~~ 'i-'- __________________________........ .... .. _ . .... ... / A. STIS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. ........ ........ edi~~~~~~~~~~err...ane.....o...n P ... ... .TAa M . .A . . P . a ........ Sea A~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.. .e . A . A.pe .~ pp ....... .' p~~~~~~~~.... R S IN.. ..... Poverty 71 caveats (see appendix 1), it is safe to conclude that poverty increased greatly in these countries. The Baltic countries started the transition with low pov- erty rates, close to those of Central European countries. Their poverty rates recorded very sharp increases reaching almost 30 percent of the population in 1994-95. In Ukraine and Russia, poverty rates increased even more. Russia's poverty headcount is estimated to have risen from 2 percent to 50 percent of the population. Finally, the Central Asian countries began the transition with relatively high poverty headcounts (in double digits for every country ex- cept Kazakhstan). Data from 1993 show poverty rates over 60 percent for all countries, including what is probably a too-high 88 percent for the Kyrgyz Republic. The composition of poverty has also changed. There are now many more poor in the states of the former Soviet Union, as compared with Eastern Eu- rope, than there were at the beginning of the transition. In 1987-88, about as many poor people lived in Eastern Europe as in the European part of the U.S.S.R. In 1993-95, however, for each poor person in Eastern Europe there were five living in the European part of the former Soviet Union. If the Cen- tral Asian countries are included, the preponderance of poverty in the former Soviet Union compared with Eastern Europe becomes even greater (six-to- one). Elasticity of poverty with respect to income Differences in how much poverty has increased do not depend only on how much real income has declined and how much income inequality has risen, but also on the absolute level of a country's income. This is because we use the same absolute yardstick to measure poverty. Richer countries will have lower poverty headcount than poor countries, even if their incomes declined by the same percentage. The Czech Republic, Hungary, the Slovak Republic, and Slovenia had relatively high average incomes both before the transition and in 1993-95. With the poverty line of $PPP 120 per capita per month, there were very small increases in poverty because almost no one's income was below that level, either in 1987 or in 1993-95. The situation is different in poorer countries, where the same percentage decline in income led to mas- sive increases in poverty as many fell below the poverty line. In these coun- tries, the poverty line lies in the region of "dense" income distribution, as can be seen by comparing Moldovan income distributions to that in the Slovak Republic (see figure 5.3 ). The equation 5.6 shows that the elasticity of the poverty headcount with respect to a given uniform percentage decline in income (that is, a decline without a change in the income distribution curve) will vary among coun- tries. Elasticity (as defined here: it is in effect semi-elasticity), 5.6 dF(z) - dF(z) z = f(z) z dz/z dz 72 Income, Inequality, and Poverty during the Transition Figure 5.3. Income Distributions in Moldova and Slovakia, 1992 12 10 A 75 8 Moldova 0 120 200 400 600 800 1000 $PPP per capita per month Source: Countries' household budget surveys. is equal to the product of the density function around the poverty line, and the poverty line. The poverty line is the same for all countries. The value of the density function at z, however, will be higher for poorer countries (com- pare points A and B in figure 5.3 for z=$PPP120), and elasticity will tend to be higher for them. (Note that the rule C above refers to the change in headcount for a given absolute decrease in income in rich and poor countries.) Indeed, elasticities in Balkans, Poland, and the former Soviet republics are invariably around 0.5 to 0.6, meaning that an across-the-board decline in in- come of 10 percent raises the poverty headcount by 5 to 6 percentage points. For example, in Russia every 10 percent decline in income will make an addi- tional 5 percent of the population, or about 7 million people, poor. In richer countries, elasticities are much smaller, around 0.1. How do these values compare with those in other countries? Are socialist countries likely to have higher elasticities because of a more compressed in- come distribution? Squire (cited in World Bank 1993, p. 41) estimates the elas- ticity of poverty headcount with respect to expenditures to be 0.24 after con- trolling for the initial headcount. His equation is AHC = -0.24 growth in mean expenditure - 0.01 HCo (where HCo = initial headcount). Because initial pre-transition headcounts were small in the countries studied here (0.04 on average), the second term in Squire's equation would be close to zero. Squire's equation would then yield elasticities of around 0.24, which is lower than that observed here for all countries except Central Europe. In a study of ap- proximately twenty countries, Ravallion (1993, p. 7) finds that the elasticity of poverty rates to income (when income alone is included in the equation, Poverty 73 that is, assuming that distribution does not change) is about 2.4. When his equation is re-expressed in terms of headcount semi-elasticity as here, its value is around 0.1, which is again much lower than what is found in transi- tion economies. Another important question is what will happen to elasticity as the economy recovers. Is there likely to be a symmetrical movement that, accordion-like, after having first pushed people below the poverty line, then raises them above it? Whether this occurs depends on the shape of future growth. If in- come growth occurs first at the top and the middle levels of income distribu- tion, then inequality will continue to rise and poverty will remain stable. This seems to have been the case in Poland, the only transition economy in which real GDP has grown for four consecutive years (1992-95). While GDP grew by 5 percent between 1992 and 1994 (and real personal incomes grew by 4 percent), the poverty rate in 1994 was still slightly higher than in 1992, as inequality seems to have risen. But when growth "trickles down," sharp declines in poverty could result, as relatively large segments of the popula- tion now hovering around the poverty line are pulled above it. Several ele- ments suggest that such an optimistic scenario is likely. Most of the poor in transition economies do not represent a distinct "underclass" as they do in Latin America (see the section below, Who Are the Poor?): their educational achievements are not much lower than those of the rest of the population; their access to social services; their ownership of consumer durables and apartments is also close to that of non-poor segments of the population."2 The declines in their incomes are recent and are not yet reflected in a marked deterioration of their asset ownership. If it takes a long time for income growth to "trickle down," however, these relatively favorable elements will be lost. Alternative poverty calculations Another implication of the poverty rate's high elasticity with respect to in- come is that mistakes in income reporting are likely to have a substantial impact on calculated poverty rates. Underreporting of income seems to be widespread in transition economies. Thus, both inadequate income report- ing and "bunching" of the population around the poverty line lead to a great variability in calculated poverty rates. The rates "move around" a lot: they 12. Trying to "guess" who is poor in transition economies based on asset ownership or access to social services fails (see Dupre 1994 on Russia). A Polish survey of recipients of social assistance shows that almost 20 percent own cars and 60 percent own color televisions (see Polish Central Statistical Office 1993, p. 21). Aside from automobiles and videocassette recorders, a recent World Bank study found, "no strong correlation [in Ukraine] between housing conditions or ownership of most consumer durables and the frequency of poverty" (World Bank, 1996b). 74 Income, Inequality, and Poverty during the Transition are not robust. For example, if the elasticity around the poverty line is 0.5 to 0.6, then a 20 percent underestimation of income, which is quite likely, will lead to an overestimation of the poverty rate by between 10 and 12 percent- age points. Because of uncertainty regarding the "true" poverty rates for the years 1993-95, they are recalculated here in two additional ways. First, rates were recalculated using macroeconomic incomes rather than data from HBSs whenever macroeconomic data showed higher incomes (see table 3.4); then they were recalculated using expenditures instead of income. Estimates based on macro ("adjusted") income data are referred to as INCOME2; estimates based on expenditures are referred to as EXPEN. Table 5.2 shows that, using macroeconomic data instead of HBSs, the over- all average poverty headcount drops from 45 (as in table 5.1) to 40 percent. The total number of the poor goes down from 168 rnillion to 147 million. The most important changes occur in Romania and Uzbekistan, where estimated poverty rates decline by between 20 and 24 percentage points. Headcounts in Poland and Russia decline by 6 points. Poland, Romania, Russia, and Uzbekistan account for virtually the entire decline in the number of the poor. This is because income adjustments in Central European countries had little effect on poverty rates, because the poverty rates were low and the elastici- ties small. In terms of regional poverty rates, the picture changes somewhat. There is a sharper split between the Balkans, Poland, and the Baltic countries (except Romania and Estonia), on the one hand, which have poverty rates of about 20 percent, and the Slavic and Central Asian republics of the former Soviet Union, on the other hand, where about half the population is poor. The increase in the number of poor, compared with the situation before the transition, is tenfold: from about 14 million to 147 million (see figure 5.4). Table 5.3 shows a different set of poverty estimates, where expenditures per capita are used instead of incomes per capita, for the eight countries for which both HBS income and expenditure data are available for the same years. In seven of these countries, the use of expenditures results in lower poverty headcounts than did the use of HBS income. For Belarus, Poland, Romania, and Russia, reported expenditures are significantly higher (between 19 and 30 percent higher) than reported INCOME1, and the headcounts are between 8 and 11 percentage points lower (see the two last columns in table 5.3). For example, Russia's headcount goes from 50 to 39 percent; Romania's goes from 59 to 48 percent, and Poland's goes from 20 to 10 percent. The Kyrgyz Repub- lic and Ukraine are in a class of their own, with reported expenditures in these countries more than twice reported incomes. When expenditures are considered rather than incomes, more than one-third of the population in these countries, ceases to be poor! For the three Slavic republics of the former Soviet Union, where the transition has brought about a huge increase in in- formal sector incomes, the average poverty headcount goes down dramati- cally from more than 50 percent, when unadjusted survey data are used, to 34 percent, when expenditure data are used. The latter figure probably offers a more realistic estimate of poverty. Poverty 75 Table 5.2. Estimated Poverty Headcount and Poverty Deficit in 1993-95 Using a Higher (Macro) Income instead of HBS Income Total Total Poverty Shortfall number poverty head- as % of of the deficit Income count poverty poor as % of adjustment Country (%) line (millions) GDP (%) Balkans and Poland 22 27 15.6 1.2 Bulgaria 15 26 1.3 1.1 0 Poland 14 27 5.3 0.9 +15 Romania 39 28 8.9 2.4 +26 Central Europe 1 28 0.2 <0.1 Czech Republic <1 26 0.0 0.0 +5 Hungary 2 33 0.2 0.1 +27 Slovak Republic <1 20 0.0 0.0 0 Slovenia <1 33 0.0 0.0 +12 Baltics 29 33 2.3 3.1 Estonia 37 37 0.6 4.2 0 Latvia 22 28 0.6 2.3 0 Lithuania 30 34 1.1 2.9 0 Slavic republics and Moldova 48 38 104.0 4.1 Belarus 22 26 2.3 1.2 0 Moldova 66 43 2.9 7.0 0 Russia 44 38 66.1 3.3 +10 Ukraine 63 47 32.7 6.9 0 Total without Central Asia 38 31 122.1 2.5 Central Asia 53 44 25.0 7.5 Kazakhstan 62 38 10.6 8.2 +4 Kyrgyz Republic 86 67 3.9 57.7 +8 Turkmenistan 57 39 2.2 6.7 +6 Uzbekistan 39 32 8.3 4.4 +44 Total transition 40 35 147.1 2.8 Note: Poverty line = 120 international dollars per capita per month. HBS incomes are raised across the board (thus leaving inequality unchanged) by the macroeconomic income minus HBS percentage difference from table 3.4 (if macroeconomic incomes have declined more than HBS, adjustment = 0). Source: Calculated from the data presented in appendix 4. For caveats and data biases, see appendix 1. 76 Income, Inequality, and Poverty during the Transition Figure 5.4. Estimated Number of Poor Before the Transition and in 1993-95 (millions) 1993-95 1987-88 Russian Federation 66 Baltics 2 ~~~~ ~~Central Europe 0.2 Balkans, Poland 16 Central Asia 25 6.5 Other Slavic 38 Total=147 million Total=14 million Note: 1993-95 information is based on "adjusted" HBS data (INCOME2 set). "Other Slavic" includes Moldova. How Much Is Needed to Cover the Poverty Deficit? A relatively shallow poverty Tables 5.1-5.3 indicate that poverty in transition economies, while widespread, is relatively "shallow." This means that the average income of the poor is not substantially below the poverty line. Poverty shortfall (that is, the percent- age by which the average income of the poor falls below the poverty line) in the Balkans, Poland, the Baltic countries, and Central Europe is around 30 percent. This means that the average income of the poor person is about $PPP 2.8 per day. Poverty is deeper in the Slavic republics of the former Soviet Union and in Central Asia, where the average shortfall is about 40 percent, and the average income of the poor is thus $PPP 2.4. Broadly speaking, pov- erty shortfalls increase with higher poverty headcounts. This means that as poverty widens it also becomes deeper which, of course, puts a double pres- sure on the poverty deficit."3 13. In general, there is no reason why this should be the case. The average shortfall in any given country tends to decrease as the poverty headcount goes up (see box 5.1). If the headcount is small, only the very bottom of income distribution (the "down and Poverty 77 Table 5.3. Estimated Poverty Headcount and Poverty Deficit in 1993-95 Using HBS Expenditures EXPEN Shortfall Total compared to Poverty as % of poverty INCOMEI head- poverty deficit as Head- Country count (%) line % of GDP Elasticitya count Mean' Poland 10 20 0.5 0.4 -10 +30 Romania 48 34 4.9 0.7 -11 +19 Hungary 7 20 0.3 0.3 +3 0 Estonia 34 28 2.9 0.7 -3 +2 Slavic republics 34 34 3.3 0.4 -17 Belarus 14 23 0.7 0.4 -8 +21 Russia 39 44 3.7 0.4 -11 +23 Ukraine 26 37 2.3 0.4 -37 +110 Kyrgyz Republic 55 46 27.2 0.5 -33 +132 Note: Poverty line = 120 international dollars per capita per month. a. Elasticity is percentage point change in poverty headcount divided by percentage change in income (across-the-board). b. The difference between EXPEN and INCOMEI from table 3.3. Source: Calculated from household budget surveys presented in appendix 4. For caveats and data biases, see appendix 1. The situation in transition economies may be contrasted with that in Latin American countries. In Latin American countries, using a lower poverty line of $PPP 60 per capita per month, average income shortfalls are approximately 40 percent.'4 The average income of a poor person in Latin America is there- fore only $PPP 1.2 per day, or half the average income of a poor person in the Commonwealth of Independent States15 and even less compared with a poor person in Eastern Europe. Widespread but shallow poverty has several policy implications. About 2.8 percent of GDP (using the INCOME2 data set)16 would be needed to elimi- nate poverty, assuming perfect targeting, that is, assuming that transfers are received only by the poor and in the exact amounts needed to bring them up to the level of the poverty line. Poverty deficits as a percentage of GDP vary widely among the countries. The deficits are only 0.1 percent of GDP in Central European countries. In out") are included, and they are likely to be very poor. As the headcount increases, the poor are the "ordinary folk" whose income is often relatively close to the poverty line. 14. See Psacharopoulos and others (1992). 15. The Commonwealth of Independent States (CIS) comprises all but Baltic repub- lics of the former Soviet Union. 16. Poverty deficits as a share of GDP are calculated in relation to countries' actual GDPs in corresponding years. 78 Income, Inequality, and Poverty during the Transition Figure 5.5. Estimated Annual Costs of Poverty Elimination Assuming Perfect Targeting, 1993-95 Russian Federation 5.8 Balti Central Europe 0.03 "'/" "' ~Balkans, Poland 1.5 Central Asia 2.1 Other Slavic 39 Total = $13.6 billion Note: Current $ billion, INCOME2 data. "Other Slavic" includes Moldova. Figure 5.6. Estimated Monthly Spending Required to Lift an Average Poor Person out of Poverty, 1993-95 16 14- 12 - En 10 - 0 0 g 0 Balkans Central Baltic Russia Other Central Total and Europe Slavic Asia Poland Note: Current dollars using INCOME2 estimates. The poverty line is $PPP 120 per person per month. "Other Slavic" includes Moldova. Poverty 79 Belarus, Bulgaria, Lithuania, and Poland, the poverty deficit is approximately 1 percent of GDP. In Latvia and Romania, it is about 2 percent of GDP. In Russia, the deficit is 3.3 percent of GDP."7 In the Central Asian republics, Moldova, and Ukraine, the deficits are even higher.'8 These figures may be contrasted with the poverty deficits of 3.8 percent of GDP in Turkey, 4.4 per- cent in Brazil, and 5.4 in Colombia, figures which were obtained using the same poverty line of $PPP 120 (see table 5.1). Eliminating poverty in Russia would cost about what it would in Turkey, relative to GDP. Relatively small poverty deficits (under 3 percent of GDP) in most transi- tion economies appear to carry a favorable message for anti-poverty poli- cies. It seems, at first glance, that elimination or substantial alleviation of poverty is feasible. However, before estimates can be made of actual resources needed to eliminate poverty, these numbers must be multiplied by a factor of about 3 to account for realistic estimates of "leakage."'9 Then, even in coun- tries where estimated deficits are about 3 percent of GDP, the welfare spend- ing required becomes an "impossible" 9 percent of GDP. How much would be needed in current dollars, assuming perfect target- ing, to eliminate poverty in all eighteen transition economies? The annual total for all countries is estimated at between $13.6 and $17.3 billion, depend- ing on the income assumptions used (the former with INCOME2 data; the latter with INCOMEI data). Figure 5.5 shows the regional breakdown as- suming higher income and thus lower poverty deficit. As much as 70 percent of the total $13.6 billion needed, would be spent in the Slavic republics of the former Soviet Union and Moldova, with Russia alone accounting for $5.8 billion per year. The Balkans and Poland would need $1.5 billion per year, and the Central Asian republics would need $2.1 billion per year. The re- quirements of the other two regions, the Baltics and Central Europe, are small: the Baltics would require an estimated $300 annually, while the Central Eu- ropean countries would require only $30 million per year. In the case of the Baltic countries, this is because the population is small, even if the headcount ratios are high (the total population of the Baltic countries is less than that of Bulgaria). Central Europe's needs are very modest because poverty rates there are low. The dollar costs of poverty elimination per one poor person (figure 5.6) are relatively small in Central Asia (about $7 per month), slightly higher in the Slavic republics of the former Soviet Union and the Balkans (about $8 per 17. If expenditures are used, almost the same poverty deficit is obtained for Russia (3.7 percent of GDP). 18. If expenditures rather than income are used to calculate the poverty deficit in Ukraine, however, the deficit drops from 6.9 percent of GDP to only 2.3 percent. 19. See the subsection 'Leakage" and actial cost of poverty alleviation below. 80 Income, Inequality, and Poverty during the Transition month), and high in the Baltics and Central Europe ($14 per month). This is chiefly because most currencies in the Commonwealth of Independent States are more undervalued with respect to purchasing-power levels, than are those in Eastern Europe and the Baltics. (This is rapidly changing, however, as wit- nessed by the brisk real appreciation of the ruble in 1992-95.) While the total number of the poor in Russia alone is almost four times as high as in Eastern Europe and the Baltics combined, the estimated dollar costs of poverty elimi- nation are only three times as high. This is not because the poor are less poor in Russia-on the contrary, Russian poor's average shortfall below the pov- erty line is greater-but rather because the Russian poverty line in current dollar terms is relatively low.20 How should a given amount of internationalfunds be distributed? If the objective is to minimize the number of poor in the combined transition economies with a given amount of international funds, in US dollars, then the best approach would be to help those who are nearest to the poverty line. And yet, the situation is not so simple. Those nearest to the poverty line are not necessarily those with the lowest $PPP shortfall, but those with the low- est actual dollar shortfall. For example, the average poverty shortfall from the $PPP 120 poverty line is 33 percent in Hungary and 67 percent in the Kyrgyz Republic (see table 5.2). Is it less expensive, then, for the international com- munity to help a poor person in Hungary? No-because the Hungarian 33 percent $PPP shortfall translates into $19 per month, while to fill up the 67 percent shortfall in the Kyrgyz Republic requires only $10 per month. The reason for this is simple: although in PPP dollars the gap is greater in the Kyrgyz Republic, in current dollar terms it is much smaller. This is because a dollar is worth much more in the Kyrgyz Republic than it is in Hungary. If countries are ranked by how inexpensive it is to help the average poor person to reach the poverty line, then Uzbekistan, Belarus, and Moldova top the list with less than $6 per person per month.2' They are followed by Slovakia, Bulgaria, Romania, and Lithuania, which require less than $8 per 20. In 1993, the same poverty line ($PPP 120) amounted to $21 in Russia compared with $51 in Poland, $57 in Hungary, and $60 in Estonia. 21. These rankings are only approximate because they disregard differences in in- come distribution among countries. Although the average dollar shortfall from the pov- erty line in one country may be less than in another, it does not follow that it would be cheaper to help all the poor from the first country. There may be some poor from the second country who are very close to the poverty line and who would be very inexpen- sive to help. Ideally, all the poor from all the countries would be ranked by the size of the shortfall expressed in current dollars. Poverty 81 person per month. Russia comes next with exactly $8. The most expensive countries are Estonia ($22) and Slovenia ($24). Assuming that the interna- tional community has a given amount of funds F and that it is concerned with a representative poor in each country, whose distance from the poverty line is equal to the average shortfall, the optimal strategy would be to help first the poor in "less expensive" countries and then, gradually, to move to more "expensive" countries until funds are exhausted. Under this scenario, it is unlikely that "expensive" countries would be helped at all. A different objective could be envisaged. If what matters is to reduce the level of deprivation among the poor, and deprivation is assumed to be an increasing function of the shortfall, then reducing the poverty of a very poor person by x dollars matters more (that is, it yields greater utility) than reduc- ing the poverty of a less poor person by the same amount. Put another way, the marginal utility of money is a decreasing function of income. If depriva- tion is a quadratic function of the distance from the poverty line, then the objective function, guiding the allocation of funds, becomes: 5-7~ ~ ~ ~i kt £- E, ( yij -> d,)' 5.7 di L 1'( where z = the poverty line in $PPP (same across the countries), k = number of countries, Pi = number of poor people in country i, yij = income of j-th indi- vidual in country i in PPP terms, d.. international transfer to that individual, E, = purchasing power of a dollar in i-th country (that is, the reverse of the country's price level relative to the world).2 The objective is to minimize overall deprivation. Equation 5.7 must be minimized under conditions of a given international amount of dollar funds (F), and only for poor people. This means that those whose income is above z are not included in the mini- mization of 5.7. These two conditions are written: I 'Idj = F (fund exhaustion) (z - yj - dIJE,) 2 0 (irrelevance of the non-poor). The second condition ensures perfect targeting because the income of a poor person after the transfer cannot exceed the level of the poverty line. Assuming, for the sake of simplicity, that only the representative poor in each country is addressed (that is, disregarding differences in countries' in- 22. The minimization of equation 5.7 is equivalent to the minimization of Foster- Green-Thombeck's function P2 for all transition economies combined. 82 Income, Inequality, and Poverty during the Transition come distributions among the poor), the objective function simplifies, and we can write the Lagrangean:23 5.8 mdin L = (z - Yi - 2X d - F) where y, = average poor person's income in PPP terms. z, E1's and F are given. The first-order condition is:24 i -2Ei (z - y- diE,) + = 0 and the solution is: 5.9 d. E (z - yi - E In order that all di's be positive, X must be less than or equal to the mini- mum value of 2(z-y,)E, (conditionA). Let the country with the minimum value of 2(z-y,)E, be called "the binding country" (because it binds the value of X from the above). The "binding country" may be denoted by b. X, the mar- ginal change in deprivation if the amount of available funds, F, changes in- finitesimally, reflects how binding the funds constraint is. A high value of X means that the constraint is fairly binding (that is, the marginal utility of an additional dollar is high). From equation 5.9, it can be observed that di's are not necessarily posi- tively related to Ei: countries with more depreciated currencies (high E) will not necessarily receive more funds per person. This is because the depth of poverty also influences di. If poverty of a representative poor (z- y,) is deep, then gains from reduced hardship may be significant (since the objective func- tion is squared) even if the "bang for the buck" (E,) is small. The condition for di and E, to move in the same direction is that k be greater than the country's (z-y,)E, (condition B).25 This condition may or may not be satisfied. 23. Note that in the case where Ei=1, the problem reduces to: mdin L k=X' (z - y, - di)2 + k ( d - F) which is minimized when (z - y- d,) = 2 that is, when the poverty deficit after transfers for each country is the same. This is equiva- lent to the condition derived by Kanbur (1987) namely that to minimize P2 across several categories (countries, social groups, and so on) one needs to equalize P1's. 24. The second order condition is satisfied: 2Ei2>0. Poverty 83 Figure 5.7. Optimal Allocation of Funds by Countries if the Objective Is to Minimize Deprivation of the Poor Russian Federation 4 Balkans, Poland 6 Central Asia 17 Other Slavic 30 Total funds = 100 Note: "Other Slavic" includes Moldova. Based on INCOME2 data. Now, using the actual values for eighteen transition economies, the opti- mal dI's are calculated. Optimal di's, or transfers per representative poor per- son, are then multiplied by the number of poor in each country to obtain the optimal allocation of international funds. The optimal distribution of funds, assuming the objective function 5.8, will be such that 46 percent of the funds would be allocated to Russia, 28 percent to Ukraine, 8 percent to Kazakhstan, and 4 percent each to Romania (the highest in Eastern Europe) and the Kyrgyz Republic (figure 5.7). The remaining thirteen countries would share 10 per- cent of the funds. The optimal allocation of money would thus be very heavily concentrated on five countries. Appendix 6 shows the exact distribution by countries.26 The lesson here is threefold: * In general, it is less expensive for the international community to help the poor in countries where the exchange rate is low in comparison with its purchasing parity. * It is probably less expensive to help the poor rapidly because the real exchange rates in the CIS countries still seem to be below their long- 25. For d, and E. to be positively related we must have: 8d - = E2 (z _ Yi _ x ) > ° From the last relation, it directly follows that 1 must be greater than (z-y1)E,. 26. The binding country is Slovenia. The countries for which condition B is satisfied are: Poland, Bulgaria, Hungary, Czech Republic, Slovakia, Slovenia, Latvia, and Estonia. 84 Income, Inequality, and Poverty during the Transition term equilibrium-that is, they are still rising. It is easy to imagine a situation where real incomes in domestic currency and nominal dol- lars rise, the poverty rates decline, but the overall cost of poverty alle- viation from an international perspective does not fall or even rises, as the effect of currency appreciation offsets the effect of the decreasing number of the poor. * Since funds are limited, it is useful to try to explicitly model what the objective function is because different objectives lead to different re- sults. If the objective is to reduce the number of the poor, then the distribution of the poor (that is, how many are near the poverty line) will matter a great deal. Countries with much shallow poverty will be helped first. If a representative poor is targeted (for example, because data on the distribution of the poor are not available) and our objec- tive is to minimize the number of the poor or the poverty deficit, then countries with low domestic price levels will benefit, because the cost of "vaulting" an average poor above the poverty line is relatively low. If the objective is minimization of hardship, then the depth of poverty plays a more important role, and the amount of transfers is determined through the interaction between the (squared) depth of poverty and the "bang for the buck" that can be obtained. "Leakage" and actual cost of poverty alleviation All these calculations assume perfect targeting. Relaxing the assumption of perfect targeting increases estimated outlays needed to "solve" poverty by a factor greater than 2-2.5, which are the numbers derived from the Western experience.27 This can be shown as follows. Data on the distribution of social assistance reported here (see table 5.9 below) show that the bottom quintile receives 28 percent of total social assistance in transition economies, and 42 percent in market economies. Assuming that the bottom quintile is targeted in both cases, it follows that to arrive at an estimate of the actual cost of pov- erty elimination, the poverty deficit needs to be multiplied by a factor of 2.4 in market economies (100 divided by 42) and by 3.6 in transition economies (100 divided by 28). An implication of the shallowness of poverty in transition economies is that the poor are not a distinct underclass. A lack of distinct poverty features implies that the costs of identifying the poor and the amount of "leakage" can be high.28 If people who are poor cannot be identified easily through 27. See, for example, data quoted in Atkinson (1995, pp. 29-30) or Sawhill (1988, p. 1101). 28. This, in addition to relative inexperience of social assistance offices in transition countries in administering such programs, probably explains their lower targeting effi- ciency as compared with market economies. This is indirectly confirmed by the fact that there is no difference between the two sets of countries in the targeting efficiency of unemployment benefits, where eligibility criteria are dearer (see table 5.9). Poverty 85 certain categorical criteria (for example, place of residence, age, ethnic group, occupation, family size), and if their income cannot be gauged accurately (because it is barely below or above the poverty line and because it includes much in-kind income and income from the informal sector), then the costs of anti-poverty policies rise. The ratio between overall spending and the amount of cash actually "delivered" to the poor rises. The choice between (i) relying on growth as the primary anti-poverty instrument, or (ii) relying on welfare, then, moves in favor of the former. In conclusion, shallowness of poverty by itself represents an argument in favor of growth-based (as against redistributive) policies as the toolsfor poverty reduction. Formally, the expected cost of poverty alleviation can be represented as the product of the "real" poverty deficit (PD) and a factor (greater than 1) that represents "leakage" (or payments to the non-poor), "spillover" (or pay- ments to the poor in excess of what they need to reach the poverty line), and administrative costs. Empirical results in Poland show that both "leakage" and "spillover" increase when (a) the percentage of the poor increases, and (b) the poor are less distinct from the rest of the population. An increase in the percentage of potential claimants-if the distinctness of the poor from the rest of the population is assumed to be fixed-means that the probability of errors of targeting goes up.29 The proportion of mistakes will be greater if, for example, half the population is poor than if only 5 percent is poor. The probability of such errors further increases if income differences between the poor and the non-poor are small, that is, if the poor are not distinct. As shown in equation 5.10, the "effective" poverty deficit as a percent of GDP (PD*/GDP) is equal to the "calculated" PD/GDP "grossed up" by a factor L, which is a positive function of the headcount index and the "shallowness" of poverty (1 /s). Both a high headcount index and shallow poverty are characteristics of transition economies and increase the cost of poverty alleviation. 5.10 PDP GDP (1 + L(HC,s)) GDP GDP How to Explain Increases in Poverty? Letting both income and distribution change The descent into poverty is the product of two forces: lower income and greater income inequality. This is illustrated in figures 5.8 and 5.9 with the example of the Czech Republic. Individuals are ranked according to their per capita monthly disposable income, and the height of the ordinate indicates the in- come level for each decile at constant prices. In 1993, real income for all but the top decile was less than it was in 1988. The percentage decline in income was greater for the poor: the lowest income decile lost 24 percent in real terms, while the ninth decile lost only 12 percent. Meanwhile, the top decile gained 20 percent. The Gini coefficient rose from about 19.4 to 26.6. 29. See Milanovic (1995a, pp. 43-45). 86 Income, Inequality, and Poverty during the Transition Figure 5.8. Distribution of Income in the Czech Republic, 1988 and 1993 0 - - Uniform percentagedecline 1988 ------ 1993 1 2 3 4 5 6 7 8 9 10 Income decile Note: In 1988 crowns. Vertical axis in logarithms. Source: For 1988, Microcensus; for 1993, Survey of Economic Expectations and Attitudes. Both reported in Veeemik et al. (1994). If 900 crowns per capita per month is taken to be the poverty line for 1988, an amount equal to the poverty line actually used in Czechoslovakia during that year30 then approximately 6 percent of the population was poor in 1988 (see point A in figure 5.9 which "zooms" on the bottom quintile from fig- ure 5.8). If income declines across the board by 12 percent, which indeed was the actual average decrease in real income between 1988 and 1993, then the percentage of the poor rises to about 10 percent (point B). In addi- tion, however, the income distribution curve itself changed. The actual 1993 distribution is given by the line Actual 93. The headcount ratio now jumps to 17 percent (point C). Thus, both the overall decline in income and the change in distribution combine during the transition to raise the number of poor. When income distribution becomes more unequal, as it did during the transition, the real income of the poor has to rise dramatically-in this case it would have to more than double; see the ratio between AF and FG in figure 5.9) just to maintain the poverty rate constant. Such an increase in 30. Czechoslovakia differed from other East European countries in that its poverty line was not a mere "accounting" poverty line but was used in social policy (see also section Should OECD-like Social Assistance Be Introduced in Transition Economies? in chap- ter 6). In 1988, the Czechoslovak per capita poverty line (for a family of four) was 875 crowns per month, or the equivalent of 210 intemational dollars. Poverty 87 Figure 5.9. Decomposing the Change in the Poverty Rate: Parts Due to Change in Income and Change in Distribution C.# 1200- Actual 1988 c ~~~~~~~~~~~~~Unfform S ~~~~~~~~~~~~~~percentage X / _ _ < ~~~~~~~~~~~~ decline - A , / ctual 1993 600 6 10 17 20 Cumulative percentage of recipients Note: AB=due to decline in income, BC=due to distributional change. Source: For 1988, Microcensus; for 1993, Survey of Economic Expectations and Attitudes. Both reported in Vecernik et al. (1994). the income of the poor cannot, of course, take place while overall income declines (that is, it cannot occur at the same time as income distribution "worsens"). A notable feature of all transition economies however is a si- multaneous decline in income and an increase in inequality; both effects contributed to rising poverty in these countries. In contrast, in the 1980s in Latin America, when poverty also increased, these two negative changes occurred simultaneously in only two countries: Brazil and Peru (Morley 1994). Decomposition into growth and distribution effects The increase in poverty can be broken down into a portion due to declining income and a portion due to more unequal distribution.3' The typical pattern 31. The decomposition procedure used is the one described by Kakwani and Subbarao (1990), and Kakwani (1995, pp. 38-39). The formula for the change in poverty between period 0 and period 1 due to income is 2 ( lo Poo + Pil Pol) (Continued on the next page) 88 Income, Inequality, and Poverty during the Transition in transition economies seems to include first a period of quick decline in output with relatively small changes in the distribution of income, and then a second period in which inequality increases, the GDP bottoms out and, eventually, rises again. Poverty should be driven up by declining incomes during the first period and by increased inequality in the second (at least for as long as growth is not sufficient to offset the impact of more unequal distri- bution).32 This pattern is illustrated for Poland and Russia during the period 1990-94 (figures 5.10 and 5.11). In Poland during 1990-93 almost all the in- crease in the poverty headcount (right-hand side axis) was due to the decline in real income (solid area in figure 5.10). However, beginning in 1992, the increase in the Gini coefficient and the associated change in the distribution also begin to add to poverty (the shaded area). By 1994, rising incomes began to pull poverty down while poverty due to a widening income distribution increased. In Russia, the income decline was, of course, more severe: in 1994, real income was less than half what it was in 1988. The increase in the poverty headcount was also much sharper. Here again, however, widening income inequality becomes gradually more important in driving poverty up (see the shaded area in figure 5.11). In 1993 and 1994, the growth and distribution effects each explain about the same proportion of the overall increase in pov- erty. Interestingly, even a slight decrease in the Gini coefficient in 1994 (com- pared with 1993) did not stop the growing role of income distribution in explaining poverty. This suggests that while Gini might have decreased be- cause of some reduction in the share of the rich, the distributional change continued to be unfavorable to those close to the poverty line and pushed some down into the ranks of the poor. (Continuedfrom the previous page) where P,. is the poverty rate with the income level from period i and the distribution from period j. The change in poverty due to change in distribution is: D = 2 (Pol- PmO + PIl - P10). It can easily be seen that P,1 - Poo = Y + D. Downscaling all incomes is equivalent to raising the poverty line. To calculate Plo the poverty line is shifted upward (because income between 0 and 1 has declined) using the formula: PL PLo = IL 01+r where PLO* is the new poverty line, PLO the original poverty line and r = rate of income decline (negative). Similarly, to calculate P0, we redefine PL, as: PL;= PL, (1 + r). 32. The same pattem is observed by Comia (1994, p. 39) and UNICEF (1995, Table 1.2, p. 11). Poverty 89 Figure 5.10. Breakdown of the Increase in Poverty Headcount between Growth and Distribution Effects in Poland, 1990-94 100 - 35 hIncome.-' 80 - Gini 3 8g 60- .. 2 Q An~I.......................... . . . ......... . ......2 0~~~~~~~~~~~~~~~~~~1 S 20 X, , I , O5 1990 1991 1992 1993 1994 Note: 1987 real per capita income = 100. Real per capita income = solid line (left axis). Gini coefficient = broken line (right axis). Headcount increase compared to 1987 (right axis). Growth effect: solid area; distribution effect: shaded area. Figure 5.11. Breakdown of the Increase in Poverty Headcount between Growth and Distribution Effects in Russia, 1990-94 120 - 70 110 60. Income 100 - ........- ...... ... . 50 . 80 - .. .......40 62 0 . ... . -..-- ----------- ----\- -----,- ll--44? ..... .. ... ... ....... .... .. .... .... ... ... ... ... ........ ,:30 L^ 60 - 30 g Gmin ~40 . .. ...........20 20 10 1990 1991 1992 1993 1994 Years Note: 1988 real per capita income = 100. Real per capita income = solid line (left axis). Gini coefficient = broken line (right axis). Headcount increase compared to 1988 (right axis). Growth effect: solid area; distribution effect: shaded area. Table 5.4. Explaining the Increase in Poverty Headcount between 1987-88 and 1993-94 Constant AiNCOME AGINI PPP INCOME Equation (t-value) (t-value) (t-value) (1988) ALIB k2 (F) SE A 24.0 (1.5) -0.63 (3.1)a 0.50 (0.8) -0.0041 (1.3) 0.70 (11.8) 14.2 B 1.7 (0.4) -0.28 (1.6) 1.33 (2.6)b 0.44 (6.5) 12.9 C 22.6 (2.1) -0.75 (4.5)a 0.28 (0.5) -30 (2.0) 0.74 (14.4) 13.2 D 25.2 (3.l)b -0.47 (3.2)a 0.94 (2.4)b -41 (3.2)- 0.68 (11.0) 9.7 Note: AHC (change in poverty headcount) is the dependent variable. It is the difference between the 1987-88 poverty headcount from table 5.1 (column 2) and the 1993-94 poverty headcount calculated using the macroeconomic INCOME2 data (table 5.2). AINCOME is the change in population real (macroeconomic) income from table 3.4, column 3. AGINI is the difference between the 1987-88 and 1993-94 Ginis from table 4.1. PPP INCOME is annual per capita household income in 1988 obtained from HBSs and expressed in intemational 1990 dollars. ALIB is the change in the liberalization index between 1989 and 1994 (maximum value of the index = 1, minimum value = 0). t-values are in parentheses. a. Significant at 1 percent. b. Significant at 5 percent. Poverty 91 Cross-country analysis: economic policy and increase in poverty The previous analysis of poverty shows that the increase in poverty was driven by two strong forces: a decline in population incomes and an increase in in- equality. This can be studied more formally by estimating the first-difference equation 5.11 across all transition economies: 5.11 AHC = fct (AINCOME, AGINI, PPP INCOME) where AHC = change in poverty headcount between 1987-88 and 1993-94, AINCOME = percentage change in real income over the same period, AGINI = = change in Gini points over the same period, and PPP INCOME = $PPP income per person (from household surveys) in 1988. PPP INCOME is used to control for the fact that poverty headcounts for different countries are cal- culated using the same real poverty line of $PPP 4 per day per person. Given the same percentage decrease in income and the same increase in the Gini coefficient, the richer country's headcount will change less because fewer people will fall below the given (real) poverty threshold. In other words, a 10 percent decline in income will push fewer people below the $PPP 4 line in a relatively rich country such as the Czech Republic, and many more people in the poorer Kazakhstan (see also equation 5.6). Equation A in table 5.4 shows the results of running equation 5.11 over fifteen transition economies.33 All coefficients have the predicted signs: lower income and higher Gini increase poverty; higher initial (1988) income level reduces the increase in the headcount. However, only the coefficient of AINCOME is statistically significant; AGINI and PPP INCOME are not.34 Each percent decrease in real population income is associated with an increase of the poverty headcount by 0.63 points. This is virtually the same (semi-) elas- ticity as found in the section, Elasticity of poverty with respect to income. Equation B takes into account the fact that estimates of poverty increase are not equally accurate. Here, observations are weighted using an explicit indicator of survey quality. 35 Now the AGINI coefficient becomes statistically significant while the importance of AlNCOME diminishes. Is there a relationship between the increase in poverty and the speed of reform, that is, independent from, and in addition to, the effect that the speed 33. Macroeconomic income data are not available for Bulgaria, Estonia, and Latvia (see table 3.4). 34. Multicollinearity reduces the significance of the coefficients: the partial correla- tion between AINCOME and AGINI is 0.6 and between AINCOME and PPP INCOME it is -0.5. This means that income loss has been accompanied by increased inequality and has been greater in poorer countries. 35. The indicator ranges from 4 = perfect survey to 1 = unreliable survey. Values are obtained as 4 minus the estimated bias for 1993-94 surveys given in table A1.4. Thus, the value for Poland is 3.5, for Bulgaria 2, and so on. 92 Income, Inequality, and Poverty during the Transition of reforms may have on real income and the Gini coefficient (which are al- ready included in the regression).36 In equation C, an estimate is added for the speed of reforms (LIB), expressed as the change in the liberalization in- dex, which ranges from 0 (no reform) to 1 (full-scope reform).37 In this equa- tion, which is identical in formulation to equation A, the direct effect of re- forms on the change in poverty is not significant. However, if a weighted OLS formulation is used to account for the quality of observations (equation D), ALIB shows that faster reforms reduce poverty. Moreover, with this equa- tion, all coefficients become statistically significant: 1 percent of decrease in real income is associated with an increase in the headcount of 0.47 percent- age points; 1 Gini point increase in inequality is associated with a little less than I percentage point increase in headcount, and each 0.1 point "increase" in reform reduces the poverty headcount by 4 points.32 Adjusted R2 is almost 0.7. Who Are the Poor? Poverty by social class The section in chapter 4, Disparity among social groups, demonstrated that the approximate order of gain by socioeconomic group during the transition was as follows: pensioners > workers39 2 farmers. Given this, it would be reasonable to expect the reverse ordering in terms of poverty rates. This is illustrated in figure 5.12, which shows relative poverty rates for the main so- cioeconomic groups. Relative poverty rates are obtained by taking the actual poverty rate for one group-for example, workers-and dividing it by the average rate for all social groups (that is, by the country average). Values smaller than 1 indicate that a given socioeconomic group has a smaller pro- portion of the poor than the country average. This approach is used here because the data in figures 5.12 and 5.18-5.25 come from a variety of sources that use different poverty lines. Expressing poverty rates in relative terms makes comparisons possible." Workers' households have, in all countries, lower poverty rates than the average. Poverty incidence among pensioners is also lower than the average, except in Bulgaria and Estonia. In Belarus and the Slovak Republic, the inci- 36. That is, speedy reforms might lead to an increase in inequality which, of course, would already be reflected in the Gini coefficient. However, that effect seems negligible as indicated by low correlation coefficients between ALIB and AGINI (-0.15), and ALIB and AINCOME (-0.13). This also suggests orthogonality of the right-hand side variables. 37. The index is calculated by de Melo, Denizer, and Gelb (1996). 38. The effect is independent of countries' income level. If countries' PPP income level in 1988 is induded in regression D as a control, the same results hold. 39. Unless otherwise specified, workers include both white-collar and blue-collar workers who are employed in the public, private, or mixed sectors (but not the self- employed and not those employed in the small-scale private sector). Poverty 93 Figure 5.12. Social Group and Relative Poverty Rates, 1992-95 Workers . . Farmers -_-_-_-_-_- Pensioners Self-employed ...._*____ Unemployed 0 1 2 3 4 Relative poverty rate *Poland Romania IM Hungary ]Romania E Bulgaria FII Belarus 0 Russia | Estonia Note: Poverty rate equal to country average = 1. Source: Calculated from: Poland 1993: World Bank (1995, Table 2.5). Romania 1994: World Bank (1995a, Table 3.1, p. 17). Hungary 1992-93: T6th and Forster (1994, Table 5, p. 34). Slovakia 1992: calculated from Microcensus 1992, p. 30. Bulgaria 1992: calculated from HBS 1992, p. 109. Russia 1993: World Bank (1994a), Klugman (1995). Belarus 1995: World Bank (1995c, p. 36). Estonia 1995: World Bank (1996, p. 13). dence of poverty among pensioners is even less than among workers' house- holds. In all countries, farmers' poverty rates are between 1½2 and 2 times the average (three times the average in Estonia). The self-employed, often thought to be among the "winners" in the transition, have lower poverty rates than the average in Hungary and Poland, about average in Belarus and the Slovak Republic, but, interestingly, higher than the average in Esto- 40. The resulting comparability, while satisfactory, is not full. This is because the level of the poverty line used will also influence the relative incidence of poverty. To take an extreme example, suppose that the poverty line used is so high that virtually all house- holds in a country are poor. Then all socioeconomic groups will have the same poverty incidence and the relative poverty rates will be unity for all. Fortunately, the sources used here all employ broadly comparable poverty lines ranging between $PPP 100 and $PPP 200 per capita per month. 94 Income, Inequality, and Poverty during the Transition Figure 5.13. Poverty Headcount by Socioeconomic Group in Poland, 1987-94 60 - Unemployed. 50 ... 40- 0 o 30- = ,, - ~~~~~~~Farmers --_ > 20 - ixed 0 Pensioners - lo - - Workers Self-employed 1987 1988 1989 1990 1991 1992 1993 1994 Note: Poverty line = 1993 minimnum pension (= official poverty line) indexed by the cost-of- living. Source: Calculated from Polish HBSs (various years). nia and Romania. This probably reflects issues of classification (what house- hold are classified as "self-employed") as well as the lack of homogeneity noted above among the self-employed, a group that includes both small-scale sub- sistence farmers and rich private entrepreneurs. Finally, households with at least one unemployed person have poverty rates of between 13; (Belarus, Ro- mania, and Russia) and four times (Poland) greater than the average. From this information the following conclusions can be drawn: the unem- ployed and farmers are, in all countries, more likely to be poor than the aver- age person; pensioners' likelihood to be poor is about average; and workers' households are slightly less likely. No regularity emerges regarding the posi- tion of the self-employed. In addition to a comparison between countries, historical single country data can be used. Historical data are preferable because they allow for the tracking of changes in poverty, but they are also more difficult to find. For Poland, however, detailed and relatively consistent data are available on the incidence of poverty by social group going back to the late 1970s. In figure 5.13 they are shown dating back to 1987. The incidence of poverty among farmers' and mixed (worker-farmer) households has increased by more than it has among workers and pensioners. Farmers, who traditionally have had high poverty rates, still have the highest incidence of poverty (except for the unemployed), but they are now closely followed by mixed households.4" 41. During the stagnation of the 1980s, however, mndxed households fared better than others (Milanovic, 1992). Poverty 95 Table 5.5. Relative Poverty Ratesfor Different Types of Unemployed Households With more With one un- With long- than one Household employed term un- unem- head un- Country at least employed ployed employed Estonia 1995 2.2 3.3 Hungary 1993 1.0 3.3 4.9 3.7a Poland 1993 1.7 2.6 3.2 4.0 Russia 1992 1.5 2.5 Note: Long-term unemployed is unemployed for more than a year. a. Becomes 9 if head of household is long-term unemployed. Source: Calculations for Estonia from World Bank (1996); Hungary from World Bank (1995b); Poland from World Bank (1995); and Russia from World Bank (1995a, p. 17). The self-employed and the unemployed are two socioeconomic groups that have become more important during the transition, and are, for the first time, included in Polish HBSs. These two groups have the lowest and high- est incidence of poverty, respectively. The absence of data for the years prior to 1993 does not allow a comparison of the position of these two groups be- fore and after the transition. Anecdotal evidence suggests, however, that the incidence of poverty among the self-employed has always been low. On the other hand, the current unemployed once belonged to either workers' or mixed households4l and their poverty incidence was probably not too dis- similar from the average of these groups. Thus, during the transition their position must have deteriorated markedly. Unemployment and poverty The link between unemployment and poverty is clear. The higher the rates of unemployment and, particularly, the higher the share of the long-term un- employed (that is, those who are unemployed for more than a year), the greater the poverty. The long-term unemployed either lose their entitlement to un- employment benefits or receive only a fraction of earlier benefits. A low rate of unemployment and a virtual absence of long-term unemployed make the correlation between unemployment and poverty weak in the Czech Repub- lic.43 But the situation is different in countries where unemployment rates are greater and the share of the long-term unemployed is high. As shown in table 5.5, in Poland and Hungary in 1993, households with a long-term un- employed member had poverty incidences that were 2.6 and 3.3 times higher 42. Only very few current unemployed would have been farmers. Polish agricul- ture was private, and it is unlikely that a farmer would abandon agriculture to register as an unemployed without right to benefits. 43. A point made by Professor Jifn Vecernik at the Third Central European Forum held at the Institute for Human Sciences, Vienna, 21-23 January 1994. 96 Income, Inequality, and Poverty during the Transition than the average, respectively; households with two or more unemployed members were 3.2 and 4.9 times more likely, respectively, to be poor than the average; and households headed by an unemployed person were about 4 times more likely to be poor than the average. In Poland and Bulgaria, 30 percent of poverty could be attributed to unem- ployment. In Hungary, the share is even 60 percent (see table 6.1 below). In Russia, where the unemployment rate is much lower, this is true for only 11 percent of the poor.44 Even if unemployment does not continue to rise in transi- tion economies, the share of the long-term unemployed will probably still in- crease, and the link between unemployment and poverty will become stronger. How the working population became poor The working poor represent a sizable proportion of those who were pulled below the poverty line during the transition. A study of poverty in Poland finds that in 1993 60 percent of the poor were working poor (World Bank 1994a, vol. 1, p. iv); similar results are obtained for Russia, where 66 percent of the poor are working poor (World Bank 1994a, p. 15). Those whose rela- tive wage position appears to have deteriorated during the transition were primarily manual workers (in declining industries) and low-skilled clerical staff who often had only a vocational education or less. The latter were prob- ably the most overstaffed professional category in socialist countries. The streamlining of production, together with a reduced need for various report- ing functions related to the bureaucratic management of the economy, re- duced demand for this category of worker. In Poland, for example, the posi- tion of people with only a vocational education has deteriorated more than that of any other educational group (World Bank 1995, p. 81). Two examples illustrate how working people have slid below the poverty line. The first is that of Ukraine, and to some extent Russia, where even the minimum conditions necessary for working people to stay ahead of poverty are no longer satisfied. The second is that of a richer country, Poland, where the minimum conditions are satisfied even if many working people are still poor. The situation in Ukraine and Russia is depicted in figures 5.14 and 5.15. The average wage and average pension are expressed in terms of a constant real per capita poverty line, which is taken to be the 1993 official Russian poverty line (also virtually equal to the poverty line discussed above of $PPP 120). At minimum, the average wage should be twice the (per capita) pov- erty line. This means that the "typical" dual-income working couple that is, a couple that has two children and is earning the average wage, should not be poor. Not all such dual-income, four-member families would remain above the poverty line, however, because approximately 60 percent of workers nor- 44. The same figure for Canada in 1990 was 28 percent (calculated from Statistics Canada, 1991, pp. 154-64). Poverty 97 Figure 5.14. Average Wage, Pension, and per Capita Income as a Ratio of the Poverty Line in Russia, 1987-96 6 4 3 ~~~~~Income\ … 2 _ ~~~.-_P~ err sion ~ ~~~~~ ~ 1~~~~~ . - .,rr_*; 1-----------------------------------'-- ,.-.'----.---.------- O I I , I I . 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Note: The poverty line is the official Russian Ministry of Labor minimum subsistence line for 1993. In 1993, the average subsistence minimum per person was Rs. 20,562, or $22 per month, or $PPP 120. This line is indexed using consumer price index. Average income comes from nacroeconomic income data. All amounts are monthly averages. Figure 5.15. Average Wage, Pension, and per Capita Income as a Ratio of the Poverty Line in Ukraine, 1987-95 6 5 4 .3-- Income _ 0 . . . . . I,. Pension,._-' '-_.." 1987 1988 1989 1990 1991 1992 1993 1994 1995 Note: The poverty line for Ukraine is the official Russian Ministry of Labor minimum subsistence line for 1993 (average value). This line is indexed using Soviet and Ukrainian (as appropriate) CPI index. Average income comes from macroeconomic income data. All amounts are monthly averages. 98 Income, Inequality, and Poverty during the Transition Figure 5.16. Cumulative Wage Distribution in Ukraine (1992) and Poland (1993) 100* 8; A 0 ~60 40 - - - -I - - - 20 ~~~Ukraine U Poland . 10 40 50 70 100 200 Wage (mean wage = 100) Note: Horizontal axis in logarithms. mally earn less than the average wage. But because there are offsetting ele- ments such as women who are systematically low earners marrying men who are high earners, family allowances, income-in-kind,45 and so on, the average wage-to-poverty-line ratio of 2 to 1 seems a satisfactory approxima- tion. In Russia and Ukraine, the ratio declined from about 4 (and even 5 in Russia, due mostly to high wages paid to workers in Siberia) before the tran- sition, to about 2 in 1995 and 1996. (In Ukraine, the ratio has been slightly below 2 since 1993.) Thus, a dual-income couple earning the average wage cannot be sure of staying above the poverty line in Ukraine, and barely man- ages to do so in Russia. Because the average wage-to-poverty-line ratio in Ukraine in 1995 is only 1.8, only those earning approximately 10 percent above the average wage should be able to stay out of poverty. As figure 5.16 shows, fewer than 20 percent of the employed fulfill this condition (point A). Similar conclusions emerge if the relationship between the average pen- sion and the poverty line is considered. The average pension should, at a minimum, be equal to the poverty line: one-person or two-person pension- ers' households with no dependents and no other sources of income should be able to avoid poverty. With that objective in mind, governments often link the minimum pension to the official poverty line, as in the Czech Republic, Hungary, Poland, Russia, and Slovakia. The "average" pensioner should, then, 45. Workers' families receive about 20 percent of total income from non-wage pay- ments. Poverty 99 be able to avoid poverty. This, however, is not always the case. The mini- mum pension that is linked to the poverty line may be the minimum full- entitlement pension. Disability, social, family, or early-retirement pensions may be below the minimum pension. The vagaries of slow indexation, or delays in pension payments, may further push some pensioners (at least tem- porarily) below the poverty threshold, and the overall average pension may barely stay at the subsistence level. In Russia, this has been the case since 1992 (see figure 5.14). In 1995 and 1996, because of huge pension arrears, the paid-out average pension was 10 to 20 percent below the official poverty line.46 In Ukraine, the average pension has been less than the poverty line since 1993, and in 1994-95 it was approximately 40 percent below the poverty line. Thus even the minimum pension requirement identified above is no longer satisfied in Russia and Ukraine. The average per capita income in Ukraine, as estimated using macroeco- nomic sources (see figure 5.15), is equal to the poverty line since 1993. Put- ting aside for the moment possible underestimations of macroeconomic in- come, this fact alone (combined with a log-normal income distribution) ensures that approximately 60 percent of the population lives below the pov- erty line.47 In Poland, the wage requirement is satisfied, as the average wage in 1996 was about 234 times greater than the Polish official poverty line (see figure 5.17). Who are the working poor in Poland? The Polish per capita poverty line was 35 percent of the average wage in 1995 and 1996. This means that if only one member of a "typical" family of four is employed earning the average wage, the family will fall short of the poverty threshold (because the family income needed to stay above the poverty line is 35 percent times 4, or 140 percent of the average wage). Even if economies of scale in consumption, lower needs of children, and additional sources of income are accounted for, such a family is clearly hovering near the poverty line. If the spouse is out- side the labor force, or his/her unemployment benefit has expired, or even if she is receiving an unemployment benefit'48 but the earning of the working spouse is less than the average wage, the family is in a precarious situation. Translated into practical terms, this means that to stay ahead of poverty, a typical Polish family has to have two wage earners. While this was typical under Communism, it no longer is, as participation rates have declined and many have become unemployed. Consequently, many families in Poland with only one employed member will probably be poor or near poverty. But this is not all. If both husband and wife are employed and each earns approximately 70 percent of the average wage or less, they too will be poor. 46. The average pension includes all supplementary pension payments. 47. The mean of the log-normal distribution is at approximately the 60-65th per- centile. 48. The unemployment benefit was a flat 35 percent of the average wage. 100 Income, Inequality, and Poverty during the Transition Figure 5.17. Average Wage, Pension, and per Capita Income as a Ratio of the Poverty Line in Poland, 1987-96 4 4 < X~~age 3 Incomex Pension . 0 1 I 1 1 1 1 I 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Note: The poverty line is the 1993 minimum pension equal to $68 per month or almost $PPP 170. It is indexed using the cost-of-living index. Average income comes from macroeconomic income data. All amounts are monthly averages. Figure 5.16 shows that about 30 percent of the employed in Poland fall into this category (point B). The working person's situation in Poland can be contrasted with that of his counterpart in the United Kingdom or the United States. The minimum wage in the latter two countries is a little more than $30 per day.49 This is three (in the United States), and four times (United Kingdom), greater than the per capita poverty line for a four-member household.50 This means that, for a typical family, only one member of the household need work at the minimum wage for the whole family to be near the poverty line.51 Virtually all "typical" families with one or more employed members earning more 49. This assumes an eight-hour work day and twenty-two work-days per month. 50. The U.S. federal poverty line for a four-member household is about $10 per per- son per day. The U.K. Income Support is about $5 per person per day plus about $3 for the housing allowance. 51. The U.S. federal poverty line is among the highest (in PPP terms) in the world. It differs from European poverty lines, however, because it does not represent a "guaran- teed minimum income." It defines which individuals and households are eligible for federal assistance (for example, housing benefits, Medicaid, food stamps). The states de- fine their own-lower or higher-effective poverty lines for the allocation of state-fi- nanced benefits. Poverty 101 Figure 5.18. Household Size and Relative Poverty Rates, 1993-95 1 :............ ,~3 ...:::-: ......... 0 s , ............ 5Q . . .. ............... . .. ------.. , O~~~ .... . .................. ............... 6+._ _ _ _ _ _ _ _ _ _ _ _ 0 0.5 1 1.5 2 2.5 3 Relative poverty rate *Poland Hungary L' Czech Republic S Slovakia FIBelarus t Russia E'Ukraine Note: Poverty rate equal to country average = 1. Source: Calculated from: Poland 1993: World Bank (1995). Hungary 1992-93: T6th and Forster (1994, Table 4). Czech Republic and Slovakia 1993: Veeemik et al. (1994). Belarus 1993: Roberti (1994, Table 3). Russia 1993: Center for Economic Conjuncture (1994). Ukraine 1995: Family budget surveys 1995 reported in Ukrainian Committee for Statistics (1996). than the minimum wage will be above the poverty threshold. The position of minimum-wage workers in the United States and the United Kingdom vis-i-vis their countries' poverty lines therefore, is similar to that of an average-wage worker in Poland vis-a-vis the Polish poverty line.52 In both cases, if only one member of the household is employed at that wage (the minimum in the United States and United Kingdom, and the average wage in Poland), the household is in a precarious position. Poverty by size and type of household Poverty also varies as a function of size and type of household (see figure 5.18). In general, larger households tend to be poorer both in per capita terms 52. Note that, in addition, the poverty line is approximately $PPP 10 per person per day in the United States, $PPP 8 in the United Kingdom, and $PPP 6 in Poland. 102 Income, Inequality, and Poverty during the Transition Figure 5.19. Relative Poverty Rates for Single-parent Households, 1993-95 2.5 - 2 2- ~ ~ ~ ...... . ...... ... ... .. 0.5 0 Belarus 93 Poland 93 Hungary 93 Estonia 95 Note: Poverty rate equal to the country average = 1. Source: Calculated from: Belarus: Roberti (1994, Table 3). Poland: World Bank (1995, Table 2.17). Hungary: World Bank (1995b, p. 22). Estonia: World Bank (1996). (meaning that total household income rises more slowly than the number of household members) and in equivalent terms (when economies of scale and lower consumption needs of children are taken into account). The position of larger households is the same in transition economies. Poverty rates for five- member households are everywhere between 1.5 and 2.5 times higher than the average. On the other hand, for one- and two-person households, pov- erty rates are less than one-half the average almost everywhere. Exceptions include declines in poverty rates in the Czech Republic and Hungary from three-person to four-person households. Not surprisingly, single-parent households also have a higher-than-aver- age incidence of poverty (see figure 5.19). Poverty by age The high correlation between household size and poverty implies that chil- dren will be one of the most poverty-stricken groups. This is shown in figure 5.20. Children under the age of fourteen are between 20 and 70 percent more likely to be poor than the average person. In all eight countries shown here, there is a remarkable similarity in the way in which poverty rates decline with age. Poverty rates for people of retirement age (55 for women, 60 for men) are only about one-half the country average in Hungary and Poland, Poverty 103 Figure 5.20. Age and Relative Poverty Rates, 1992-95 Children AA A = A . Adults < 30 .... - . . . . -'.. . . . . . . Middle age Retirement age 0 0.5 1 1.5 2 Relative poverty rate Russia * Poland E Hungary 9 Bulgaria 0 Latvia II Belarus El Czech Republic |]Estonia Note: Poverty rate equal to the country average = 1. Source: Calculated from: Russia Q2/1993: McAuley (1994, p. 37). Poland 1993: World Bank (1995, figure B, p. iii). Hungary 1992-93: Andorka and Speder (1994, table 2). Bulgaria 1992: HBS 1992, calculated from table 4, p. 111. Latvia 1992: HBS 1992, calculated from p. 23 and p. 26. l3elarus 1995: World Bank (1995c, p. 36). Czech Republic 1993: Vecemik and others (1994, tables IV/3a and IV/3b, pp. 21-22). Estonia 1995: World Bank (1996, p. 17). and even less in the Czech Republic and Slovakia,s3 a finding consistent with findings above regarding the relatively favorable position of pensioners in Eastern Europe. Among the countries shown here, only in Estonia do the elderly have higher-than-average poverty rates. This is due to low and flat (i.e. almost equal for all) pensions in existence in Estonia since 1992. The most dramatic reversal in relative poverty rates between the old and the young probably occurred in eastern Germany in the wake of the 1990 Unification. In the German Democratic Republic, the aged had relatively high poverty rates because the pension-wage ratio, as in the Soviet Union, was 53. Slovakia is not shown in figure 5.20 due to lack of space. 104 Income, Inequality, and Poverty during the Transition Figure 5.21. Age and Relative Poverty Rates in Eastern Germany, 1990 and 1992 2 - 1 0.5 ~ 1990 ~0.5- Under 16 16-30 3145 45-60 61-75 over 75 Age Note: Poverty rate equal to eastern German average = 1. Source: Calculated from Krause (1992, reported in Heimerl [1993]). low (about 35 percent). Pensioners in the former German Democratic Repub- lic "gained" from the Unification, however, because their pensions were first converted from Ostmarks into DMs at a favorable one-to-one rate, and then the pensioners were absorbed into the more generous West German pension system.T4 As shown in figure 5.21, almost no eastern German pensioners in 1992 were among the poor. By contrast, the relative poverty rate among chil- dren in 1992 was about the same as prior to Unification. The figure 5.21 also shows that in 1992, poverty rates were steadily decreasing with age, as they were in other transition economies. Before Unification, however, poverty rates peaked for the very young and the very old. 54. For example, at the time of the Unification the average income of a pensioner household in West Germany was 90 percent of the country-wide average while in East Germany, it was 76 percent (Heimerl, 1993, p. 77). 55. Vedernik (1994a) finds that in the Czech Republic the rate of return to one addi- tional year of education has increased from 3.5 percent before the transition to 6.2 percent after the transition. Similarly, in comparing Czech and Slovak earnings distributions for 1984 and 1993, Chase (1995) finds the rate of return to education to have risen by between 2 and 3 percent. World Bank (1995) and Rutkowski (1996) find that in Poland this rate went up from 6.4 percent in 1989 to 7.5 percent in 1992. Rutkowski (1995, p.29) also finds increasing returns to education in Bulgaria. Poverty 105 Figure 5.22. Education of Household Head and Relative Poverty Rates, 1993-95 Primar ~~. . . . . . . . . . .... ...........v Y. . . . . . . . . . . . . . . . . Vocational . . . . . . . . . . .. . . . . . . Secondary University _:.,.. 0 0.5 1 1.5 2 Relative poverty rate L Poland 11 Hungary L] Czech Republic I Slovak 1 Estonia [ Belarus E Romania Note: Poverty rate equal to the country average = 1. Source: Calculated from: Poland 1993: World Bank (1995, Table 2.8). Hungary 1993: World Bank (1995b, p.24). Czech Republic and Slovak Republic 1993: Vecernik and others (1994, Tables IV/3a and IV/3b, pp. 21-22). Estonia 1995: World Bank (1996, p. 14). Belarus 1995: World Bank (1996a, pp. 6-7). Romania 1994: World Bank (1995a, Table 3.1, p. 17). Education and poverty The educated, not surprisingly, are less likely to be poor. In addition, increas- ing returns to education have probably further strengthened the link between high levels of education and low levels of poverty.55 As can be seen from figure 5.22, university education virtually guarantees in every country but the Slovak Republic and Belarus that a household will not be poor.56 House- holds whose head has only an elementary education are between 20 and 60 percent more likely than the average household to be poor. Vocational edu- cation yields an average probability of being poor. Secondary education re- duces the probability to about half the average. 56. In these two countries, the probability is half the country average. 106 Income, Inequality, and Poverty during the Transition Figure 5.23. Size of Locality and Relative Poverty Rates, 1993-95 Village . . . . . . . .. . . . . . . . . ., Small city Medium city . 1, . .....-:: ...... .......-: -- ---:::::::::1.. Large city . ..:;. .....~ . ... ... 0 0.5 1 1.5 2 Relative poverty rate * Poland 0 Hungary m: Czech Republic n Slovakia [ Ukraine Note: Poverty rate equal to the country average = 1. Source: Calculated from: Poland 1993: World Bank (1995, Table 2.16). Hungary 1992-93: T6th and others (1994, Table A 1.2). Czech Republic and Slovak Republic 1993: Vecemik and others (1994, Tables IV/3a and IV/3b, pp. 21-22). Ukraine 1995: World Bank (1996b, p. 22). Hungary displays the sharpest decline of poverty with education level. In Belarus and Slovakia there is an increase in the incidence of poverty as one moves from household heads with vocational education to household heads who are secondary school graduates. This might suggest a slower restructur- ing as typically demand for workers with vocational education declines dur- ing the transition. Regional aspect of poverty Poverty rates decline with increase in the size of the locality. Low rates of unemployment and the concentration of highly skilled people in capital cit- ies, increasing returns to education, and the finding above that farmers expe- rience higher levels of poverty than other groups, suggest that larger cities would be richer. Figure 5.23 confirms it for four East European countries. Poverty 107 Unfortunately, reliable data do not exist for the rest of Eastern Europe and countries of the former Soviet Union, except for Ukraine, where, on the con- trary, urban poverty is more prevalent. The importance of home-consump- tion in total population income in Ukraine, and particularly in rural areas, is the key reason for lower poverty in Ukrainian villages. Whether this feature may be shared by other countries of the former Soviet Union is likely but impossible to confirm because of lack of data. Poverty and gender Figure 5.24 exhibits poverty rates by age and sex. Data from Bulgaria and Russia (the only two countries for which data is available) show that poverty rates for men decrease as they retire, while for women they remain unchanged. Figure 5.25 contrasts age profiles of poverty in female-headed and male- headed households in Poland. Female-headed households have a higher in- cidence of poverty than male-headed households, regardless of age. The dif- ferential rises with age, however. As in Bulgaria and Russia, the poverty rate for people over 70 years old continues to decrease for male-headed house- holds, and rises for female-headed households. This finding may be due to the fact that women receive lower pensions. It explains, in part, why overall poverty rates tend to be higher for women than for men. It would also seem that most of the overall reduction of poverty with age (such as noted in fig- ure 5.20) may be due to lower poverty among males. Figure 5.24. Sex, Age, and Relative Poverty Rates in Bulgaria and Russia, 1992-93 0.9 _,8 - -_ .e 0.7 - 0.6 0.5 Active Retired Note: Poverty rate equal to the country average = 1. Source: Calculated from: Russia Q2/1993: McAuley (1994, p. 37). Bulgaria 1992: HBS 92, table 4, p. 111. 108 Income, Inequality, and Poverty during the Transition Figure 5.25. Poverty Ratesfor Male- and Female-headed Households in Poland, 1993 1.4- 1.4 __ ~Female 1.2 Mal > 0.8 0 A 0.6 0.4 0.2 0 . l l <30 30-39 40-49 50-59 60-69 70+ Age Note: Poverty rate equal to the country average = 1. Source: Poland 1993: World Bank (1995, Annex Table 5.11). The Incidence of Cash Social Transfers In this section, the extent to which cash social transfers are focused on the lowest income groups is analyzed. The analysis will not address pensions except for income-tested state pensions because, strictly speaking, pensions are not an anti-poverty transfer. They are a deferred labor income. Although in all pension schemes there is sorne redistribution in favor of poorer people (in the sense that pensions do not exactly mirror contributions made during a person's active life), pensions are nonetheless "earned" incomes in the same sense as wages. Their main function is income-smoothing over a person's lifetime. The second reason not to discuss pensions here is somewhat more compli- cated. Pensions account for between 70 and 80 percent of total cash social transfers. They also represent most of the income received by pensioners: on average, more than 80 percent. The result is that, if the average level of pen- sions in a country is relatively low, pensioners will tend to be poor. The poorer the pensioners, the better targeted pension spending will appear simply be- cause most pensions will be received by the poor. Thus, the "good" poverty- focus of pensions in countries where pensions are low, as in Russia, will be misleading. Other more extreme and erroneous conclusions regarding tar- geting may be drawn if a comparison is made between the number of pen- Poverty 109 Table 5.6. Concentration Coefficients of Family Benefits before the Transition (1988-89) and in 1993-95 Change in Country 1988-89 1993-95 targeting Eastern Europe -20 -17 Bulgaria -17 +5 Worse Hungary -20 -18 Same Poland -28 -22 Worse Romania -9 -19 Better Slovakia -28 -30 Same Baltics -12 Estonia -12 Latvia 4 -12 Better Slavic republics -2 Russia +15 -2 Better Ukraine -1 Note: Family-related benefits include monthly and regular family allowances, one-time child allowances (given at birth), matemity allowances and other family-related transfers. Individuals are ranked by their household's disposable per capita income. The regional averages are unweighted. Source: Bulgaria is calculated from 1989 HBS and 1995 Gallup survey; Estonia, Hungary, Latvia, Poland, and Slovakia from countries' HBSs presented in appendix 4; Romania from data given in World Bank (1995a, Table 2.3, p. 8); and Russia from Russian Longitudinal Monitoring Survey, Round 4 October 1993-February 1994; Ukraine from World Bank (1996b, Table A9). sion recipients who are poor before and after receiving the pension. For an important transfer like pension, most recipients would obviously be poor without it. Pensions would then appear to be extremely well-targeted as they would lift a large proportion of their recipients out of poverty. The same could be, however, said about wages, because without wages almost all work- ers would be poor. It therefore makes sense only to look at the targeting of transfers (1) whose implicit or explicit function is poverty alleviation (for example, only income-tested pensions should be included). If it is not en- tirely clear that (1) is satisfied-as, for example, in the case of a universal family allowance-then the following condition must hold: (2) the size of transfer must be such that it cannot decisively influence the position of the recipient household. This condition is needed to avoid including transfers such as pensions which will always lift many recipients out of (pre-pension) poverty. Family benefits satisfy condition (2). For almost no family are they likely to be a major source of income so that they determine the family's position in the income distribution curve. Then if it is established that family benefits are paid mostly to the poor, one can infer that they are well targeted. Table 5.6 compares the targeting of family benefits before the transition and in 1993-95. Targeting is approximated by the concentration coefficient:57 57. For definition see footnote 9, chapter 2. 110 Income, Inequality, and Poverty during the Transition the lower its value, the better the targeting (with negative values indicating a negative correlation between the absolute amount of a transfer and the rank of a household according to disposable income). Family-related benefits have negative concentration coefficients every- where, except for Bulgaria in 1995 and Russia before the transition. These benefits appear not to have become more pro-poor during the transition. This is rather surprising because children and larger families have become poorer (relative to the rest of the population). Benefits that are, by definition, tied to children should show more focus on the poor simply because their recipients have slid down the income distribution ladder. The fact that this has not happened, and that the reverse seems to have happened in Bulgaria and Po- land, is probably due to the changes in the eligibility rules, for example, in- troduction of income testing that has led to some poor families being refused the benefit or to the declines in take-up rates among the poor. Table 5.7 shows the targeting of a new transfer (the unemployment ben- efit) that did not exist before the transition, and an "old" transfer (social as- sistance) that has gained in importance since the onset of the transition. Un- employment benefits are well targeted-with concentration coefficients, in Eastern Europe and the Baltics, all exceeding -15 in absolute amounts (Rus- sia and Ukraine, where unemployment benefits, even if low, are received by better-off households rather than the poor, represent exceptions.) This is a good, although not a surprising, result because the unemployed are often among the poor, and payment of unemployment benefits is governed by clear rules, so there are relatively few classification errors (for example, a poor Table 5.7. Concentration Coefficients of Unemployment Benefits and Social Assistance, 1993-95 Country Unemployment benefits Social assistance Easter Europe -26 -17 Bulgaria -36 -3 Hungary -28 -25 Poland -18 -13 Romania -17 +9 Slovakia -29 -53 Baltics -31' -5 Estonia -37 -16 Latvia -25 +10 Lithuania -10 Slavic republics +27 +13 Russia +33 +25 Ukraine +21 +2 Note: Data for Slovakia are from 1992. Individuals are ranked by household's per capita disposable income. The regional averages are unweighted. a. Does not include Lithuania. Source: As given in table 5.6, plus Lithuania as reported in Cornelius and Weder (1996). Poverty 111 unemployed person who is denied a benefit). These results show that unem- ployment benefits do have an important anti-poverty role to play, and that their progressivity is similar in most transition countries reflecting in turn similarities among countries in the position of the unemployed and the size of the benefit. How well the system of poverty alleviation performs is best assessed by looking at social assistance. Social assistance (or welfare) is a discretionary benefit the targeting of which reflects both the rules that govern it and the ability of social assistance offices to implement these rules, that is, to income- or means-test potential claimants. Social assistance benefits should be more focused on the poor than any other kind of transfer because they are the only transfer whose function is specifically poverty alleviation. Good targeting may be difficult to achieve, however, because social workers lack the rela- tively clear rules that govern the payment of categorical benefits, such as family or unemployment allowances. Also, social assistance as defined in statistics and household surveys is often a "mixed bag" that includes not only discretionary cash or in-kind payments, but also many heterogeneous benefits such as expenditures on nursery homes for the aged, mental institu- tions, benefits paid by enterprises to their poor workers, and enterprise-pro- vided free meals. Both the extent of these "other" social assistance benefits and their statistical coverage vary from country to country, which makes re- sults regarding social assistance transfers particularly difficult to compare. These elements explain, at least in part, why social assistance in virtually all countries seems less well targeted than unemployment benefits. The Slovak Republic, however, stands out for its good performance, while Latvia, Ro- mania, and Russia stand out for their poor performance. Did transfers become better targeted during the transition as GDPs de- clined and both poverty and inequality increased? Unfortunately, it does not seem that there was any systematic improvement in targeting since the tran- sition. Pre-transition data are not available for most countries of the former Soviet Union, but in Eastern Europe, all non-pension transfers combined ap- pear to have become only slightly more pro-poor: the overall concentration coefficient moved from an average of -13 before the transition to -18 in 1993- 95 (see table 5.8). This improvement is due primarily to the introduction of relatively well-targeted unemployment benefits. Distribution of non-pension benefits in the three Baltic countries is close to a flat per capita amount given across income distribution. In Russia and Ukraine, the incidence of non-pen- sion transfers is, paradoxically, pro-rich: only family benefits are "neutral" (almost equal per capita); other cash transfers are pro-rich. The absence of significant improvements in targeting shows that social transfers have not contributed much toward mitigating increases in poverty since the onset of the transition. This is in contrast to Chile-a country that also experienced a sharp decline in income-where since the mid-1970s very focused transfers have played an important role in checking the spread of poverty (see Graham 1994, chapter 2). Similarly, better targeting of transfers 112 Income, Inequality, and Poverty during the Transition Table 5.8. Concentration Coefficients of Non-pension Cash Social Transfers before the Transition (1988-89) and in 1993-95 Change in Country 1988-89 1993-95 targeting Eastern Europe -13 -18 Bulgaria -6 +2 Worse Hungary -13 -16 Better Poland -17 -22 Better Romania -10 -36 Better Slovakia -14 -21 Better Slovenia -16 -17 Same Baltics -2 Estonia -6 Latvia -7 +3 Worse Lithuania -4 Slavic republics +16 Russia +13 +14 Same Ukraine +23 Note: Non-pension transfers indude family-related benefits, unemployment benefits, social assistance, sickness benefits, scholarships, and miscellaneous cash and in-kind transfers. Source: Bulgaria is calculated from 1989 HBS and 1995 Gallup survey; Estonia, Hungary. Latvia, Poland, Slovenia, and Slovakia from countries' HBSs presented in appendix 4; Lithuania from data in Lithuanian Statistical Yearbook 1994-95; Romania from data given in World Bank (1995a, Table 2.3, p. 8); Russia from Russian Longitudinal Monitoring Survey, Round 4, October 1993- February 1994; Ukraine from World Bank (1996b, Table A9). played a role in the United Kingdom by offsetting the impact of declining overall cash social transfers and increased inequality. Between 1979 and 1989, the share of social transfers in the income of the British population went down from 17 to 13 percent; meanwhile, the concentration coefficient of transfers improved from -20 to -30 (Milanovic 1994, table 2, p. 181 and figure 5, p. 190). Table 5.9 contrasts results for transition economies with those for advanced market economies (Targeting is approximated by the percentage of transfer that reaches the bottom quintile of income distribution.) It must be noted that market economies are not a homogeneous category. There is much vari- ance among them in the focus of non-pension transfer payments. Some, such as Australia and Chile, have extremely narrowly targeted social assistance, unemployment benefits. Others, such as the United Kingdom and the United States, have very focused social assistance but not unemployment benefits. Finally, in the Northern European countries (the Netherlands, Norway, and Sweden), all three types of transfers are unfocused. Table 5.9 shows, first, that social assistance is much better targeted in market economies: in six of twelve market economies, the bottom quintile receives at least 40 percent of social assistance. Only one transition economy (out of eight) meets this level. Second, there is virtually no difference between market and transition coun- tries in the targeting of unemployment benefits. Third, all non-pension trans- Poverty 113 Table 5.9. Percentage of Social Assistance, Unemployment Benefits, and Non-pension Cash Social Transfers Received by the Bottom Quintile of Population All non- Social Unemployment pension cash Country (year) assistance benefits social transfers Transition economies 28 (16) 29 (19) 22 (8) Slovakia (1992) 52 37 31 Estonia (1995) 36 56 26 Bulgaria (1995) 36 46 19 Hungary (1993) 35 33 29 Poland (1993) 29 26 25 Romania (1992) 23a 25 25 Russia (1994) 6 8 12 Ukraine (1995) 6 0 8 Market economies 42 (17) 27 (15) 23 (6) Australia (1989) 78 50 30 United States (1991) 70 15 19 United Kingdom (1991) 55 29 33 Chile (1990) 51 57 31 Finland (1991) 43 20 26 West Germany (1984) 40 26 14 Belgium (1992) 35 38 23 Ireland (1987) 35 34 28 Netherlands (1987) 31 11 18 Norway (1986) 27 13 21 Switzerland (1982) 25 20 16 Sweden (1987) 21 10 9 n.a = not available. Note: Shares for transition and market economies are unweighted averages. Standard deviations are shown between brackets. Countries are ranked according to the share of social assistance accruing to the lowest quintile. Individuals are ranked by their household's disposable per capita income. a. Also includes other unspecified allowances. Source: Market economies: calculated from individual household data from LIS (Luxembourg Income Study) using unemployment compensation (variable V21), means-tested social transfers (variables V25 and V26), and all government cash transfers except pensions (sum of variables V16 through V26 minus V19). Chile was calculated from Schkolnik and Aguero (1993, p. 245). All non-pension cash social transfers for Chile include income-tested state pensions. Transition economies: sources are as given in table 5.6, except for Romania: HBS 1992. fers are about equally focused in the two types of economies. In both cases, the poorest quintile receives only slightly more than its population share. Even the within-group variability in targeting seems to be about the same as shown by the very similar standard deviations. The only area where transition economies clearly seem to lag, in terms of targeting, is therefore social assistance. If, for the sake of simplicity, the poor- that is, those whom policymakers want to reach-are assumed to occupy exactly the bottom quintile, and only the bottom quintile, then little over 114 Income, Inequality, and Poverty during the Transition one-fourth of expended social assistance in transition economies can be said to reach them. The poverty deficit-calculated on the assumption of perfect targeting-would then have to be multiplied by a "gross-up" factor of 3.6 (that is, 100 divided by 28) to get an estimate of the actual amount of money needed to eliminate the poverty deficit (see equation 5.10 above). By con- trast, the gross-up factor in market economies is less than 2.4. Unlike social assistance, targeting of unemployment benefits cannot be much improved. Unemployment benefits are relatively "passive" instruments: they are paid after certain requirements are satisfied-for example, if some- one has been dismissed, or is without a job and looking for one, or has contri- bution record. These requirements do not generally include a low income. In addition, the gap between transition and market economies in progressivity of unemployment benefits is almost non-existent. Of course, unemployment benefits could be income-tested, as they are in Chile,58 but this does not seem a politically acceptable option in Eastern Europe and in countries of the former Soviet Union. How much could better targeting of non-pension transfers contribute to reducing poverty? There the illustrative analysis must proceed by several steps. First, non-pension cash and in-kind transfers normally account for only 5 percent of household disposable income in transition economies. From equa- tion 4.2 it is clear that with a concentration coefficient of -18 (equal to the average value for Eastern Europe in 1993-95, as per table 5.8) and a share of 5 percent, non-pension transfers reduce the overall Gini coefficient by slightly less than 1 point (-18 times 0.05 = 0.9). If targeting were improved so that it reached the OECD level of social assistance targeting (that is, a concentration coefficient of -30),59 the overall Gini would be reduced by 1.5 points. The gain from better targeting would be 0.6 Gini points. Second, how would a Gini reduction of 0.6 translate in terms of poverty reduction? According to table 5.4 (equation D), each one-point decline in the Gini is associated with about a 1 percentage point reduction in poverty. Achieving the OECD level of target- ing could therefore be expected to reduce the poverty headcount by approxi- mately 0.6 points. Third, because the estimated poverty headcount in transi- tion economies is about 40 percent (see table 5.2), approximately one in sixty-six poor people (0.6 divided by 40), would cease to be poor thanks to better targeting of social transfers. In terms of actual people, it would mean that about 2.2 million people in transition countries would escape poverty. 58. Which explains a very high share received by the bottom quintile. 59. The average concentration coefficient of means-tested social assistance for mar- ket economies included in table 5.9 is -31. 6 Selected Issues in Social Policy Should OECD-like Social Assistance Be Introduced in Transition Economies? Probably the key issue that faces policymakers designing safety net programs in transition economies is choosing between the following two concepts of social assistance. The first is the OECD or, more narrowly, the northwest European concept. Its key features are (1) there is an official poverty line;' (2) income below the poverty line and assets below a certain minimum are a sufficient condition for welfare eligibility; and (3) social assistance offices try, in principle, to cover the entire gap between the poverty line and income, and that gap is filled with cash. Such a concept can be termed the minimum income guarantee (MIG) system.2 The overall cost of means-tested benefits in OECD countries is typically between 0.75 and 1.5 percent of GDP and benefits account for about 2 percent of household disposable income.3 No transition economy has such a system.4 The second concept of social assistance is what exists in transition econo- mies-shorn of the enterprise-based welfare where still present. The system differs from MIG in the following respects. First, although income testing is 1. Examples include the French revenu minimum d'insertion, the minimum pension in Finland and Norway, and the survey-based (that is, calculated fiom actual expendi- tures of the bottom decile or quintile of the population) poverty lines in Belgium, the Federal Republic of Germany, and Sweden (see Veit-Wilson 1996, pp. 45-47). Differently, there may be a de facto, even if not de jure, official poverty line. For example, the British government is, for political reasons, adamant that the Income Support Line is not an official poverty line, even if it functions as such. 2. In reality, not all the gap will be identified (that is, the take-up rate is less than 100 percent) or filled (because of mistakes in assessment). The MIG system can be imple- mented in different ways: through direct payments that bring recipients up to the pov- erty line, a universal negative income tax (that is, the same amount is paid to everybody and then is subject to taxation), or a specific negative income tax to the poor (like the earned-income tax credit in the United States). 3. OECD countries included here are the "old" OECD countries: Western Europe, Northern America, and Japan. More recently, Hungary, the Czech Republic, Poland, South Korea, and Mexico have joined OECD, but their welfare systems differ from those of other OECD countries. 4. Social assistance in the former Czechoslovakia and in today's Czech Republic and Slovakia comes closest to it (see Vecernik 1991, p. 2 quoted by Sipos 1992, p. 38). 115 116 Income, Inequality, and Poverty during the Transition an integral component of the system, having an income less than the poverty line is a necessary but not sufficient condition for receiving social aid. Addi- tional criteria must also be fulfilled. These criteria are related to households' low earning capacity (almost zero elasticity of labor supply): single-parent status, presence in the household of handicapped or elderly members; or "dysfunctionality": alcoholism in the family, drug abuse, mental incapacity. For example, Poland's 1990 law on social assistance lists eleven such addi- tional criteria, at least one of which must be present, in addition to low in- come, before a household can be eligible for social assistance (Fijalkowski 1992, pp. 63-85); Latvian 1994 law lists four additional criteria (Latvia, 1994, articles 53 and 81). The Czech Republic's 1991 social assistance law also stipu- lates that beneficiaries must be unable to "increase their income due to their age or health situation" (Vernik 1994, p. 7) the wording which is identical to the one used in the Ukrainian law on social help to the "underprovisioned" families (Ukraine Council of Ministers, 1995, pp. 44-45). Second, social assistance in transition economies is viewed as a temporary relief and is often provided in-kind-for example, through hot meals or food vouchers, drugs, child care assistance, payment for kindergartens, payment of utilities and rent, provision of wood and coal, and so on. Third, categorical social assistance in transition economies plays a more important role than under MIG-for example, through state pensions for the aged, family allowances, additional allowances for large families, milk and food for school children, and special assistance for regions with a high con- centration of poor. Although most OECD countries have categorical programs in addition to MIG, the scope and financial importance of these programs are smaller, particularly when compared to countries of the former Soviet Union, where practically all social assistance is (still) based on indicator (categori- cal) targeting.5 Fourth, social assistance in transition economies does not aim to cover the entire difference between the poverty line and actual income. The amount covered depends on the judgment of local social assistance workers. Such systems may be called income testing as screening (ITS) systems to indicate that income is tested in order to screen applicants, but that there is no mini- mum income guarantee. MIG and ITS systems are similar respectively to the "Type A" system, which seeks to eliminate poverty, and the "Type B" sys- tem, which only seeks to alleviate it, as defined by Sipos (1994). ITS systems exist in virtually all countries in Eastern Europe and the former Soviet Union. Officially, the Czech Republic, Estonia, Hungary, Poland, and 5. For example, in Lithuania in 1994 there were no less than ten types of family allowance (childbirth grant, child care benefit, adoption and foster parents' allowance, single-mother benefit, preschool child benefit, military-family child benefit, alimony ben- efit, and so on; see Cornelius 1995). A World Bank study of Ukraine (1996b, p. 49) lists twelve types of family benefits and ten types of social assistance (for example, social pensions, housing subsidies, cash allowances for Chemobyl victims, food subsidies for Chemobyl victims, funeral aid, and transportation subsidies). In Hungary, in 1995, there were 35 kinds of family and social assistance benefits (Sipos, 1995, Annex 1). Selected Issues in Social Policy 117 Table 6.1. The Composition of the Poor, 1993-95 Hungary Bulgaria Estonia Poland Russia Belarus 1993 1994 1995 1993 1993-4 1995 Total "categorical" poor 77 76 73 62 56 44 Unemployeda 60 30 24 30 11 5 Pensioners, 3 35 9 6 26 21 People in single- parent fanilies 3 2 32 1 3 3 Childrenb 12 9 8 25 16 15 Other poor 23 24 27 38 44 56 Note: All poor = 100. Poverty line is $PPP120 per capita per month except in Belarus (about $PPP100). a. All individuals living in poor households whose head is a pensioner or where there are unemployed are included in these two groups, respectively. b. All children living in poor families where the head is not a pensioner, unemployed, or a single parent. Source: Hungary: HBS 1993; Bulgaria: HBS 1994; Estonia: HBS July-September 1995; Poland: HBS 1993; Russia: RLMS (Russian Longitudinal Monitoring Survey) Round 4 (October 1993 to February 1994); Belarus: New Household Budget Survey 1995. Slovakia have ITS systems, where low income plus family dysfunctionality or zero elasticity of labor supply are the requirements for social assistance. In Russia and Ukraine, a "weaker" form of ITS system is applied. Although social workers are supposed to observe official poverty lines (which in a coun- try as vast as Russia vary by region), much is left to the workers' discretion; moreover, implementation depends on local availability of funds (see Foley and Klugman 1997).6 To guide a policymaker choice between these two social assistance con- cepts (MIG and ITS), four criteria are proposed. First, it is necessary to ascertain the correlation between poverty and char- acteristics that are both observable and difficult to hide or alter. The stronger the correlation, the stronger the case for ITS, which relies more heavily on categorical benefits. Thus, in countries where family size and age are good predictors of poverty, categorically based benefits (such as family allowance and minimum pension) can be efficient instruments for combating poverty. As shown in table 6.1, in Bulgaria, Hungary, and Estonia unemployment ben- efits, the minimum pension, and family allowances may be expected to reach more than three quarters of all poor. In Belarus and Russia, where the work- ing poor are more numerous, categorical benefits are not likely to be as effi- cient. The same, of course, applies to the correlation between poverty and other characteristics, such as age and region. For example, in countries such as 6. A regionally diverse system does not exist only in large countries such as Russia and Ukraine. Latvia, too, has a regionally based social assistance system, with the result that the poor in different parts of the country are treated differently. 118 Income, Inequality, and Poverty during the Transition Bolivia and Chile (see Grosh 1994; Graham 1994; World Bank 1995f, chapter 1), where poverty is strongly regionalized and the distances between the poor and rich areas are relatively great, governments subsidize staple foods that are sold only in poor areas. A regional approach may be relevant for large countries with low population densities and a heavy regional concentration of inefficient industries. Russia would be an obvious candidate. Second, income-testing must be feasible for MIG to be chosen. If the corre- lation between poverty and easily observable household characteristics is weak, thus favoring adoption of MIG, one needs to check if it is possible to determine incomes with sufficient precision. During the transition, personal incomes have become difficult to monitor. Many people are engaging in un- reported economic activities; the "gray economy" is blossoming. The rela- tionship between reported and actual income becomes weak. If many in the private sector operate on the margins of legality, then certifications produced by their employers (for example, vouching that workers' wages are suffi- ciently low so that they can qualify for various benefits) may be worthless for determining actual incomes.7 The lower the correlation between reported and actual income, the weaker the case for MIG because both effectiveness (the number of poor reached by programs) and efficiency (the proportion of money disbursed to the poor) will be low. In the section, Decline in real popu- lation income, in chapter 3, it was noted that income underreporting seems to be particularly great in the Slavic, Central Asian, and Caucasian republics of the former Soviet Union. The case against using MIG will be strong in these countries. The third criterion relates to financial feasibility. In all transition econo- mies, a new transfer has already emerged: unemployment benefits. Is there room for yet another transfer: MIG-based social assistance? This is an em- pirical question. If other types of transfers (pensions, family allowances, and so on) are scaled down, then universal welfare may be introduced. But how costly will universal welfare be? The answer to this question hinges on how high the guaranteed minimum income is pitched, how many people are likely to fall below that line and by how much, and what are the expected take-up rates, and the "leakage." As discussed in the section, How Much Is Needed to Cover the Poverty Deficit?, in chapter 5 the "true" poverty deficit in transition countries amounts, on average, to 9 to 10 percent of GDP. Clearly these costs cannot be financed. Only in some Central European countries where the pov- 7. For example, private firms in some transition economies pretend to employ people who then "work" for six months until they qualify for unemployment benefits and are then promptly "fired." The employer's only cost would be, say, 45 percent payroll tax for six months. If the wage replacement rate is 60 percent and the duration of unemploy- ment benefits is a year, then both the employer and employee can make a profit. 8. Note that this discussion does not address the issue of whether "leakage" is more or less under MuG than it is under ITS. The point is simpler: if "leakage" is assumed to be the same under both systems, then MIG programs will be more expensive because they aim to eliminate the entire poverty deficit, whereas ITS attempts only to alleviate poverty. Selected Issues in Social Policy 119 erty deficit-to-GDP ratio is relatively low can the guaranteed income be a financially feasible option. It is almost certainly not financially feasible in the Soviet successor states.8 The fourth criterion is the ability of local administrations to implement centrally mandated schemes. Even if MIG is feasible on the basis of the three previous criteria, it may be impossible to implement because of administra- tive weakness. Or it may be impossible to implement in some less developed parts of a country. Then a case could be made for simpler local schemes. Based on the above discussion, two types of conclusions can be made: con- clusions regarding elements that favor the selection of either system and con- clusions regarding information needed to make an informed choice. When is a MIG-like system to be preferred? (1) If the correlation between household characteristics and poverty is weak, (2) if determination of house- hold incomes is reliable, (3) if the poverty gap is not large, and (4) if local administration is reasonably capable. In market economies, for example, the last three criteria are satisfied, and the correlation between household char- acteristics and poverty is probably the same as in Eastern Europe.9 In Central Asian countries, on the other hand, none of the last three criteria is likely to be satisfied. Because the choice between MIG and ITS is an empirical question, the in- formation needed to make an informed choice becomes relatively easy to define. To address the first criterion-the correlation between household char- acteristics and poverty-household surveys and poverty profiles are needed. Using this information, it is possible to learn whether poverty is strongly related to some household characteristics or not (as has been done for ex- ample in the section, Who Are the Poor?, in chapter 5). To address the second criterion-can household incomes be ascertained?- information regarding the functioning of personal income taxation, the esti- mated size of the unrecorded economy, the relationship between incomes and expenditures (based on household surveys), is needed. To address the third criterion-is universal welfare financially affordable?- a calculation of poverty deficits for a set of "reasonable" $PPP poverty lines is required.10 To address the fourth criterion-is local administrative capability satisfac- tory?-information is needed from those with first-hand experience in deliv- 9. The unemployed and pensioners accounted for 47 percent of all poor in Canada in 1990 (calculated from Statistics Canada 1991, pp. 154-64), a percentage almost the same as in Poland (see table 6.1). 10. It is not necessary to spend too much time calculating sophisticated social minima. These are long and costly exercises, subject to much controversy, and may be irrelevant. If the poverty deficit is too big (more than 2 percent of GDP) for almost any "reasonable" range of poverty lines (say, between $PPP 60 and $PPP 120 per capita per month) then it does not make sense to investigate what the "best" poverty line might be. Complicated poverty line exercises should be postponed until a case can be made that the poverty line lies within the financially feasible range. 120 Income, Inequality, and Poverty during the Transition ering social assistance. This is an important task. Even if it is established that MIG does not make sense in a given country, there are still at least two rea- sons to look at the local administration: to learn whether local administra- tions would be able to administer even simple local schemes and to help the central administration define such local schemes. Transfer Payments to Different Groups of Recipients A key problem encountered when designing social transfers is to balance the needs of individuals with disincentive effects (mostly on the supply of labor) that can arise due to "over-generous" social payments. The disincentive effects of transfers on different types of recipients can be considered using the following example. Let there be three categories of people. A are people with a zero elasticity of labor supply (Ls) such as inva- lids, the handicapped, and social pensioners. Group B are poor (unemployed) people with a positive Ls who are net recipients of government transfers and who, at the margin, have to decide between receiving government transfers and working. Group C are working people, net tax payers, whose labor sup- ply is a function of net wage (equal to gross wage minus tax to finance social transfers to A and B). Let there also be the restriction that the disposable (post-transfer and post-tax) income of A and B must be lower than the dis- posable income of Cs. Labor decisions of As are unaffected by transfers. Their supply of labor is inelastic."' The supply of labor by Cs will be reduced, however (the labor supply curve will shift from Lso to Lsl, figure 6.1) because they have to pay taxes to finance transfers to As. The output loss is X(L*) - X(L), where X = output, L* = the equilibrium labor supply of As in the absence of taxation, and L = equilibrium labor supply with taxes, where, implicitly, the price of output = 1 and output depends on L alone. The deadweight loss represented by the cross-hatched area in figure 6.1 is the cost of the scheme. The gain is the utility gain of As. If the utility gain is greater than the deadweight loss, then the move (that is, the progressive transfer from the richer Cs to the poorer As) is acceptable. Cs have to pay additional taxes to finance transfers to Bs, however. The analysis for Cs is the same as before. Bs' work effort will decrease because they receive transfers and the elasticity of their labor supply is positive (fig- ure 6.2). Their labor supply schedule shifts to the left because Bs receive some income-unemployment benefits-without working (that is, their consump- tion of leisure increases with higher income). The amount of output loss will depend on the elasticity of labor demand (Ld) for Bs. If Ld is fixed, then B's work effort and, thus, output will stay the same. If Ld is perfectly elastic and 11. If the elasticity of the labor supply is zero, the welfarist perspective (where utility of leisure is taken into account) and the non-welfarist perspective (where leisure is ig- nored or does not matter) coincide (see Kanbur, Keen, and Tuomala 1995). Selected Issues in Social Policy 121 Figure 6.1. The Effect of Taxes on Supply of Labor of the Working Population Ls Ld/ Wage LsO B A 0 L L* Labor Note: OB = gross wage = OA (net wage) + AB (taxes). the wage rate for Bs is given, then there will be maximum output loss (see movement from Q to P in figure 6.2). Consequently, when the demand for Bs' labor is fixed, there is no differ- ence in the analysis of C-financed transfers to As and Bs because in both cases, output produced by the latter two groups is not affected. For all in- tents and purposes, then, As and Bs can be treated as the same group. If the demand for Bs' labor is elastic, however, transfers will lead to a decrease in Bs' work effort and output. The argument could then be made that to mini- mize output loss, transfers to Bs should be reduced to below the level paid to As, or should even be eliminated altogether. In practical terms, much will depend on who the poor (the Bs) are. The greater the labor demand and supply elasticities of Bs, the greater the output loss and the stronger the argument for giving low transfers to Bs. If, for ex- ample, labor demand for skilled and unskilled unemployed is the same, and labor supply of the skilled less elastic, then the argument can be made that skilled Bs should receive higher transfers than unskilled Bs, for example, through past earning-related unemployment allowances. This point illustrates the importance of assessing the elasticities of labor demand and supply when designing social transfers for various types of re- cipients. A strong case can be made for transfers to As, or to those Bs who have obsolete skills and for whom labor demand is inelastic. The case is not so strong for Bs whose labor supply and demand are elastic. 122 Income, Inequality, and Poverty during the Transition Figure 6.2. Labor Supply of the Poor (with and without Transfers) Wage L L,(y0) L, p Q Labor Note: y, = income without transfers. y, = income with transfers. Guaranteed Minimum Income and the Supply of Labor When people are paid the exact difference between their income and the social minimum (i.e., guaranteed minimum income is in place), their dis- posable income does not change if they work because the additional earn- ings are "taxed away" through lower social transfers. A typical situation is illustrated in figures 6.3 and 6.4. Up to point A, which is equal to the guar- anteed minimum income, all earned income is taxed at the rate of 100 per- cent; between points A and B (where B is the income level at which social assistance is entirely phased out), the marginal tax rate is less than 100 per- cent but still in excess of the statutory tax rate. Only if gross income is greater than B does the effective tax rate become equal to the statutory rate. A typi- cal marginal tax schedule then looks as in figure 6.4: lowest incomes are the most heavily taxed. Labor supply of those with lowest income is likely to be reduced, particularly up to point B because the marginal tax rates are high. The reduction in labor supply represents an additional "cost" of guaranteed minimum income. All things being equal, the cost increases if the guaranteed social mini- mum and the minimum wage are very close (and if the country is poor, the two cannot be very different). Thus, for example, if a person's maximum Selected Issues in Social Policy 123 Figure 6.3. Gross and Net Income if There Is a Minimum Income Guarantee Gross income /o Net income Statutory tax MIG //I 0 A B Gross income Note: The net income schedule is shown by a bold line. OA = minimum income guarantee (=MIG). B = social assistance phased out. attainable earned gross income is barely in excess of MIG, he will be subject to an almost 100 percent marginal taxation throughout. Gains to working are miunimal. Browning (1993, pp. 9-10) argues that the costs, i.e., foregone out- put due to lower labor supply, are substantial and that they increase when income distribution is compressed. In a poor country, the costs in terms of GDP increase for two reasons. First, direct costs increase because more people fall below the poverty line than would in a richer country with the same poverty line. The expenditures needed to close the poverty gap are greater. Second, because the guaranteed social minimum is not much less than the minimum wage, people have weak incentives to work, and foregone output is greater. The relationship between the poverty line and minimum wage is illus- trated for six transition economies (arranged from left to right in ascending order according to their average dollar manufacturing wage) in table 6.2. If the poverty lines that are currently used to screen social assistance applicants 124 Income, Inequality, and Poverty during the Transition Figure 6.4. Marginal Tax Rate in the Presence of a Minimum Income Guarantee Tax rate 100% Statutory tax rate 0 A B Gross income Note: OA = guaranteed minimum income. B = social assistance phased out. (see line 3) were used as the guaranteed minimum income (gapfill lines), then returns to working at the minimum wage, compared to receiving welfare, would be substantially negative in Russia, only 5 percent in Estonia, 35 per- cent in Poland, and from 50 to 60 percent in Bulgaria and Hungary. For a "typical" family (of two adults and two children) the returns to both adults working would be even less. Only in Hungary, where wages are relatively high (almost 4.5 times higher in dollar terms than in the former Soviet Union), and thus the wedge between the minimum wage for both adults and the pov- erty line for the entire family is relatively wide (50 percent), would it be worth- while to work. In other countries, even if MIG system were financially fea- sible, disincentive effects would be strong. These results may be compared with results for market economies, where retums to working are between 120 and 190 percent in Portugal, France, and the United Kingdom. The last two have a MIG-type system. The data also show why, using the current federal poverty line, a MIG-type system would be difficult to implement in the United States: the retums to work would be about the same as in Hungary. Table 6.2. Relationship between Poverty Lines and Wages, 1992-94 (in current dollars at current exchange rates, per month) Estonia Russia Bulgaria Poland Hungary Portugal France U.K U.S. early 1994 April 1994 June 1993 end-1993 Q2/93 1993 1992 mid-1993 1993 Poverty line for 1 adult (PLI) 22 45 29 68 72 90 383 285 614 Poverty line for 2-adult, 2-children household (PL4)a 59 121 78 184 194 243 825 750 1,230 Minimum wage 23 15b 45 92 115 255 1,000 825c 930 Universal family allowance (FA) per childdR 9 7 9 9 32 17 175f 60 0 Income-to-poverty line ratio Minimum wage to PLi 1.1 0.3 1.6 1.3 1.6 2.8 2.6 2.9 1.5 Both adults at the minimum wage plus 2 FAs to PL4 1.1 0.4 1.4 1.1 1.5 2.2 2.8h 2.49 1.5h Average manufacturing wage 92 110 122 240 325 560 1,600 1,750 1,950 Note: Poverty line for one adult. Russia: Ministry of Labor subsistence minimum; Bulgaria, Poland, and Portugal: the minimum pension; Estonia: official subsistence benefit; France, revenue minimum d'insertion; United Kingdom: income support and housing benefit; United States: the federal poverty threshold (the effective poverty line is less and varies by state). a. The poverty line for the standard four-member household is calculated using the OECD equivalence scale, where requirements of a four-member household are equal to 2.7 adult units except for the United States, United Kingdom, and France where the actual four-member family poverty lines are used. b. The official minimum wage in Russia was less than $10 but few were paid that little (0.2-0.3 percent of workers). The official minimum wage is used mostly as a scalar to calculate various social payments. An estimated effective minimum wage is used here instead. c. There is no official minimum wage in the United Kingdom. An effective minimum wage is used here. d. Family allowances vary with the number of children and their ages. An approximate average value was chosen in each case. e. To compare working and not working, only the universal part of the family allowance is relevant because the income-tested part is lost if working. f. Spending on universal family allowance is about a half of total family allowance expenditures (see d'Agostino and Trombert, 1992, p. 180). Only a half of total in inputed here. g. Becomes 2.7 if family credit (allowance received by low income working families) is included. h. Does not include family tax credit received if working (earned-income tax credit in the United States and France). Source: Transition economies: World Bank data; France from d'Agostino and Trombert (1992); United Kingdom: United Kingdom Central Statistical Office (1994, pp. 71, 184); United States: US Bureau of the Census, CD-Rom, Income and Poverty 1993, Table 1; Portugal, personal communication by Luisa Ferreira. Data on average manufacturing wage from Institut der deutschen Wirtschaft reported in German Brief, June 5, 1993. 126 Income, Inequality, and Poverty during the Transition Informal Sector and Pension Reform Transition economies presently function as two-sector economies. On the one hand, there is a shrinking formal sector on which are assessed all pension contributions and other taxes. On the other hand, there is the informal sector, which, by definition, pays no taxes and whose workers do not accrue any social security (including pension) rights. The situation in transition economies is not unique. In many developing countries there is also a sharp division between the formal and informal sec- tors. There are, however, two important differences between transition econo- mies and developing countries. First, transition economies face a much greater current liability in the form of pensions that must be paid. In developing countries, few people-mostly public sector employees-have accrued pen- sion rights: current public sector workers, who may account for less than 10 percent of the labor force, are taxed to pay for pensions of past public sector workers, who are similarly few in numbers. In transition economies, on the other hand, a shrinking labor force in the formal sector must pay for pension rights accrued by almost 90 percent of the former labor force. Second, transi- tion economies have a more unfavorable demographic structure with a higher percentage of older people and a correspondingly smaller percentage of people of working age. The model In this section, a model is presented that shows the link between growth of the informal sector and the need for pension reform. The economy consists of two sectors: the formal and informal sectors. Both sectors produce the same good, which they price the same (it is the numeraire); both use the same technology; and both have the same demand for labor and face the same labor supply. They are identical in all respects, except that in the formal sec- tor there is a tax on wages that is used to finance current pensions. In the formal sector, workers are paid Mf which is take-home cash wage plus aT where T is the payroll tax and a is the parameter that indicates how much of these taxes current workers expect to get back in the future in the form of pensions. In other words, the actual wage is composed of two com- ponents: the current cash wage and the expected value of the future pension. For example, if current taxation is perceived simply as yielding an equiva- lent net present value in the form of (future) pensions, the tax is in effect a deferred wage and a takes the value of 1. More generally, a shows the ex- pected net present value of future pensions to be obtained from paying a dollar in current taxes. If workers perceive current taxation as likely to yield nothing in terms of future income, then a = 0 and Tf is a pure tax.'2 12. Gramlich (1996, pp. 64-65) shows a ratios for U.S. workers of different age co- horts and household sizes. Many "baby boomers" are likely to receive less than what they paid. This is true for all single workers (with a=0.8 for those with low earnings to a=0.4 for those with steady high earnings) as well as for most married workers (although their a is higher). Selected Issues in Social Policy 127 In the informal sector, workers are paid only in cash, a money wage Mj, which we assume equal to the sum of M and T in the formal sector. Suppose now that for workers in the forma{sector a=l. They are, at the margin, indifferent between current and deferred wage. The labor supply curves in the formal and informal sector will be the same, as would employ- ment and wages. In order to have the pension system in balance, it is neces- sary simply to solve for such a T7 which is sufficient to pay for all current pensions. The level of equilibrium T7, whether high or low, would have no effect on the formal sector labor supply because workers view their current taxes simply as equivalent deferred wage. Now let a be less than 1. This could happen for a variety of reasons. Cur- rent workers could believe that demographic trends are so unfavorable that the state would be unable to impose a tax rate on future workers that is high enough to pay for their pensions (this is the situation in the United States). A variant of this is the belief that because of a future disequilibrium in the so- cial security system, the state would renege on its current promises and would, for example, income test future pensions. Workers can also perceive the state to be fundamentally unstable and untrustworthy. This is the case in many transition economies. Political and social changes in the last several years have been dramatic: not only have the states repudiated their implicit and explicit promises, they have also conducted confiscatory policies-for ex- ample, wiping out nominal savings and imposing high inflation taxes on the population. The credibility of these states is low. Additional uncertainty has been created by the political turmoil that gave birth to many new states. A given state may not even exist when it is time for a worker to claim his or her pension rights, an unpleasant experience that many pensioners in the former Soviet Union and Yugoslavia have already had to face. In general, a may be thought of as a function of the expected future tax rate. If workers expect that the future tax rate, needed to finance their future pensions, will be very high, they are more likely to believe that such a tax would be unsustainable and that it would ultimately lead to the government's reneging on its obligations. The relationship between the current a and the future tax rate t* (future variables are denoted by asterisk) is depicted in figure 6.5. There is a threshold tax rate t* below which a = 1. As the ex- pected future tax rate increases, workers expect to lose some of their contri- butions, and a declines. At some t*=t* the whole current contribution is thought of as a pure tax. In addition, independent factors (for example, lack of confidence in the state) may also shift the relationship so that a declines more sharply. The curve may shift inward (see the southwesterly movement shown by the ar- row in figure 6.5). When a becomes less than 1, a part of t is perceived as a pure tax. There is a wedge between gross and net wages. The labor supply curve in the formal sector shifts leftward from AlAl to A2A2 (figure 6.6). The equilibrium formal sector employment declines, and the equilibrium gross wage rises (the new equilibrium moves from 0 to E in figure 6.6). The relative size of the formal sector (compared to the informal sector) shrinks. 128 Income, Inequality, and Poverty during the Transition Figure 6.5. Relationship between the Expected Future Tax Rate on the Formal Sector and Current Value of Contributions Current value of contributions (a) 0 Future tax rate (tf*) tf,min tjffla Note: Current value of contributions shows the share of current contributions that workers expect to receive through future pensions. The new formal sector equilibrium will depend on how important the de- crease in a is, how elastic labor supply is with respect to both wage and ax, and how elastic labor demand is. At any event, it is clear that the equilibrium gross wage in the formal sector, Gf will now be greater than the equilibrium wage in the informal sector, but that the money wage in the formal sector M will be less than money wage in the informal sector." 3 In addition, T must be such that current pensions are paid out. This is shown in equation 6.1, where P is the current average pension, N is the num- ber of current pensioners, and N is the number of current formal sector work- ers: 6.1 TpN1=P Np. Since N1 and G depend on cc, and Tf = tfG where t is the tax rate on current gross wages in the formal sector, and p = ratio between current pension and gross wage, the following is obtained: 6.2 tf G/oa) Nf(a)=-pG/ua) Np. Both p and Np are fixed in the short run as denoted by the bars in equation 6.2. The first is fixed by law; the second is fixed by demographics and past pension practices and rules. 13. Note that the gross (and net) wages in the two sectors were initially the same (when a was 1). Selected Issues in Social Policy 129 Figure 6.6. Shift of Formal Sector Labor Supply Curve as a Function of Current Discount of Contributions Gross wage A2 (atf, then it follows from equation 6.3 that the future equilibrium will depend on the future a. Current workers clearly cannot guess what future workers' expectations will be. So it can be assumed that current workers as- sume that future workers' a* will be the same as their current a. To obtain the current equilibrium it is then necessary to solve equations in 6.4 for tfB tjand a: tf R (a) = p (present equilibrium) 6.4 If R*(a) = PT (future equilibrium) a =f(t') (determination of a). Let the solution be t , t * (tj'>t because R* ~~~~~~~~~~Income , o 2000 N X , I xpenditure i 1500 1000- 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Ventile Note: Ventile is 5 percent of the population. Source: Polish 1993 Household budget survey. ferred to as "adjusted" or "corrected" income data (INCOME2 data). The adjustment, however, is a crude one, in the sense that it raises everyone's income by the same percentage. In the absence of data on the pattern of underreporting, however, this is the only available solution. Per capita versus equivalent units International comparisons of poverty are complicated, to say the least. The choice of units of analysis (per capita or per equivalent adult) may also lead to differences in results. First, the use of a per capita poverty line, rather than an equivalent-scale derived poverty line, exaggerates poverty. This is because with an equivalence scale the needs of additional family members (often chil- dren) do not rise in proportion to number of people as they do when a per capita measure is used. The use of a per capita poverty line exaggerates pov- erty even more, then, in countries with a larger average family size.13 For 13. Coulter, Cowell, and Jenkins (1992) show that the poverty headcount charts a U- shaped pattern, first decreasing and then rising, as the equivalence scale moves from 0 (full economies of scale, where only total household income matters) to 1 (per capita calculations). The same results are obtained by Michael Forster (1993, p. 21) in an empiri- cal study of thirteen OECD countries. Description of the Surveys Used and Data Problems 151 example, Mamie and Micklewright (1993) compare Ukraine and Uzbekistan using the same Soviet FBSs for 1989. They find that of the 38 percentage point difference in the headcount index between the two republics obtained using a per capita measure, the larger household size in Uzbekistan accounts for 14 percentage points. There are several reasons why the per capita line is used here. First, data from all countries are published in that form. Second, econo- mies of scale in consumption under socialism were typically less than in market economies because the main source of such economies of scale (hous- ing, utilities, and so on) were heavily subsidized; and this is still true, al- though not to the same extent."4 Third, the use of per capita poverty compari- son allows for an easy transition from per capita comparisons to GDP per capita comparisons. Quarterly versus yearly data A final problem concerns the period over which data are collected. Normally, surveys are designed so that households report (that is, keep track of or re- call) income and expenditures for one quarter or one month."5 These data are then "blown up" for the entire year. The same thing is done by another (quar- terly or monthly) set of households, until four or twelve such sets are added up to obtain the final annual survey results. Under conditions of high infla- tion, however, data collected during different months, and even weeks and days, represent wholly different real quantities of goods and services and cannot be summed up without being adjusted for inflation. Statistical agen- cies, however, sometimes do not make such adjustments, or if they do, they sometimes make them inadequately. Under such conditions, quarterly data on income or expenditure distribution-from which poverty figures are cal- culated here for several countries of the former Soviet Union-are preferred because they refer to a shorter time period and imply about the same real command over goods and services. The usual drawback of the short-period data-namely, that they overestimate income inequality and poverty (because the shorter the time period, the more people who will report extraordinarily low and high incomes)-is then more than offset by the advantage of using data where the same reported money amounts represent approximately the same real quantities. 14. If the price of all goods that provide economies of scale in consumption is zero, then economies of scale no longer matter. 15. To increase the response rate (which was a main source of bias), Polish house- hold surveys began requiring households to track their income and expenditures for one month rather than three months. The response rate increased from 65 to 80 percent (see Kordos 1994). 152 Income, Inequality, and Poverty during the Transition How Biases Affect Comparability between Pre-transition and Transition Years Ideally, a household survey would: (a) be representative of the country as a whole; (b) gather information regarding annual income and expenditures; (c) use disposable income or net expenditures as welfare indicators; (d) in- clude consumption-in-kind; and (e) define both income and expenditures "correctly," that is, the way they are defined in economic theory. Note, how- ever, that even the "best" achievable survey could still contain biases. The "best" survey would still probably understate the two tail-ends of income distribution, because the poorest and the richest segments of society are typi- cally undersurveyed, and it would also underestimate some sources of in- come, such as property (which is routinely underestimated by up to 50 per- cent in industrial countries) and entrepreneurship. No transition-year survey used here meets all five criteria (see table A1.5) but several come close. Bulgarian and Slovenian surveys have only a slightly incorrect definition of income; adjustments could not be fully made to ac- count for this as individual data were not available. Hungary's 1993 survey does not include consumption-in-kind in income,16 and semi-annual rather than annual data were available for Poland. Among the East European data, the Czech data are the least satisfactory. Both the poverty headcount and inequality are overestimated, because the time period of the survey is so short (see table A1.5). In Hungary, the absence of consumption-in-kind leads to a slight increase in inequality and poverty, because consumption-in-kind is generally greater in poorer households. Poland's poverty headcounts may also be slightly overestimated, because the pre-harvest period only-that is, the first six months of the year-was included-and incomes are generally higher among the agricultural popula- tion during the second half of the year. In Bulgaria, the use of gross income reduces the poverty headcount and inequality to the extent that PITs are progressive. PITs are small, however, as most taxes are deducted at the source, so the downward bias is negligible. Finally, Slovak data exclude pensioner-headed households with economi- cally active members. It is not clear how this exclusion might affect poverty and inequality, because it is not clear if excluded households are richer or poorer than the average, or how they are distributed along the income spec- trum. In conclusion, with the exception of Czech and Slovak data, the quality problems and the conceptual differences between the various East European data sets are not significant. 16. It includes consumption-in-kind in expenditures. Description of the Suirveys Used and Data Problems 153 Table AI.5. Survey Defects: Pre-transition and Transition Years Country Pre-transition Transition Poland Incomplete coverage Semi-annual data of recipients Hungary No consumption-in-kind Czech None Monthly data Slovakia Not fully representative Slovenia Income definition problem Bulgaria Income definition problem; gross income instead of disposable Romania Not fully representative Monthly data income definition problem Estonia Quarterly data Latvia Quarterly data Lithuania Gross income; no consumption-in-kind Russia Quota sample, not fully Quarterly data Belarus representative; Quarterly data Ukraine income definition problem Monthly data; large underestimate of income Kyrgyz Quarterly data; large underestimate of income MoldovaF Kazakhstan | Quota sample, not fully Turkmenistan representative; Uzbekistan income definition problem Note: The surveys assessed here are the same ones listed in Tables AM.l, A1.2 and A1.3. Source: Author's estimates. For data from countries of the former Soviet Union, the problems are more serious. At most, Soviet surveys satisfied only three of the five criteria listed above: use of annual data, inclusion of consumption-in-kind and use of dis- posable income.17 The new, improved surveys suffer from too short an ob- servation period (between one month and one quarter), and in Kyrgyz Re- public and Ukraine, there is large-scale income underreporting (see table A1.5). What can be said, then, regarding non-systemic biases in estimating changes in the poverty headcount and in inequality between 1987-89 and 1993-95 17. Even if, strictly speaking, income was defined as gross, the difference between gross and net income was negligible. 154 Income, Inequality, and Poverty during the Transition Table A1.6. Bias in Estimating Change in Poverty and Inequality During the Transition Country Poverty headcount bias Inequality bias Poland Unclear Overestimates increase Hungary Overestimates increase Czech republic Slovakia Unclear Unclear Slovenia [ None Bulgaria Romania Estonia Latvia Lithuania Overestimates increase Russia Belarus Ukraine Kyrgyz republic Moldova Kazakhstan Turkmnenistan Unclear Uzbekistan Note: The surveys assessed here are the same ones listed in Tables Al.1, A1.2 and A1.3. Source: Author's estimates. (see table A1.6)? For Poland, the exclusion (in 1987) of the non-agricultural private sector and of army and police personnel yielded too-low inequality estimates; the measured increase in inequality during the transition will there- fore be overestimated. For Hungary, absence of data on consumption-in-kind in transition years will lead to an overestimation of both income-poverty and income inequality. For the Czech Republic, increases in both poverty and inequality will be overestimated because monthly rather than annual data were used in 1993. For Bulgaria and Slovenia, no biases canbe discerned a priori. In Romania and the former Soviet Union, pre-transition coverage was biased toward "average" or "typical" households, thus underestimat- ing both inequality and poverty. In the countries where the quality of sur- veys has improved, then comparisons over time will show increases in pov- erty and inequality that are greater than actual. Where the quality of surveys has remained unchanged (Moldova and Central Asia, except the Kyrgyz Re- public), it is unclear which way the bias will go. What do these biases imply, in a nutshell, for the estimation of poverty rates? In the past, income data were reliable for those covered in a survey, but the surveys' coverage was biased. They excluded the poorest and over- Description of the Surveys Used and Data Problems 155 represented average, "standard" households. As a result, poverty rates were shown to be lower than they were in reality. Improvements in survey design mean that current data are more representative of the population as a whole. On the other hand, however, households now underreport income more than they did in the past,18 so that while survey design is now adequate, income reporting is now biased. Unfortunately, rather than offsetting each other, both elements (more representative surveys and income underreporting) tend to show the increase in poverty rates to be greater than actual. The gap between current (high levels of) poverty and pre-transition (low levels of) poverty, while certainly large, then appears even larger. 18. In Poland, in the past, wages and pensions reported by survey participants al- most perfectly matched the macroeconomic values. Currently, they underestimate mac- roeconomic values by 10 to 20 percent (see Kordos 1994). Appendix 2 Decile Shares of Total Income In tables A2.1-A2.18, decile shares are calculated from the original HBS data given in appendix 4. All the detailed sources are given there. Note: A = annual data; SA = semi-annual data; M = monthly data; Q = quarterly data. For example, M:10/1993 means that the data refer to Octo- ber 1993, Q:1/1995 that the data refer to the first quarter of 1995, and SA:I/ 1993 that the data refer to the first half of 1993. Table A2.1. Belarus Decile A:1988 Q:1/1995 First 4.47 3.38 Second 6.01 5.32 Third 6.99 6.38 Fourth 7.89 7.31 Fifth 8.80 8.30 Sixth 9.76 9.34 Seventh 10.83 10.58 Eighth 12.12 12.03 Ninth 13.92 14.46 Tenth 19.20 22.88 Table A2.2. Bulgaria Decile A:1989 A:1993 First 4.49 2.80 Second 6.14 4.31 Third 7.05 5.54 Fourth 7.86 6.65 Fifth 8.67 7.74 Sixth 9.53 8.90 Seventh 10.51 10.25 Eighth 11.75 12.03 Ninth 13.59 14.95 Tenth 20.42 26.84 156 Decile Shares of Total Income 157 Table 2.3. Czech Republic Decile A:1988 M:1/1993 First 5.41 4.60 Second 6.51 5.90 Third 7.41 6.60 Fourth 8.11 7.30 Fifth 8.91 8.00 Sixth 9.91 8.90 Seventh 11.01 9.90 Eighth 11.11 11.40 Ninth 14.11 13.90 Tenth 17.52 23.50 Table 2.4. Estonia Decile A:1988 Q:3/1995 First 4.20 2.06 Second 5.79 4.07 Third 6.88 5.41 Fourth 7.91 6.68 Fifth 8.93 7.83 Sixth 9.99 9.03 Seventh 11.15 10.48 Eighth 12.49 12.50 Ninth 14.24 16.03 Tenth 18.41 25.90 Table 2.5. Hungary Decile A:1987 A:1993 First 4.96 4.40 Second 6.46 6.22 Third 7.31 7.22 Fourth 8.08 8.03 Fifth 8.85 8.79 Sixth 9.66 9.60 Seventh 10.58 10.51 Eighth 11.72 11.79 Ninth 13.37 13.65 Tenth 19.00 19.79 158 Income, Inequality, and Poverty during the Transition Table 2.6. Kazakhstan Decile A:1988 A:1993 First 4.05 3.07 Second 5.42 4.42 Third 6.46 5.60 Fourth 7.47 6.72 Fifth 8.51 7.84 Sixth 9.63 9.06 Seventh 10.90 10.50 Eighth 12.43 12.39 Ninth 14.56 15.45 Tenth 20.57 24.94 Table A2.7. Kyrgyz Republic Decile A:1988 M:10-11/1993 First 5.14 0.80 Second 5.43 1.89 Third 6.14 3.20 Fourth 7.02 4.23 Fifth 8.05 5.58 Sixth 9.23 7.16 Seventh 10.62 9.17 Eighth 12.35 12.01 Ninth 14.76 16.96 Tenth 21.26 39.18 Table A2.8. Latvia Decile A:1988 Q:4/1995 First 4.46 3.18 Second 5.98 4.86 Third 6.99 6.09 Fourth 7.92 7.14 Fifth 8.86 8.12 Sixth 9.85 9.14 Seventh 10.95 10.30 Eighth 12.24 11.81 Ninth 14.00 14.26 Tenth 18.73 25.09 Decile Shares of Total Income 159 Table A2.9. Lithuania Decile A:1988 A:1994 First 4.59 2.38 Second 6.05 3.83 Third 7.01 5.08 Fourth 7.91 6.25 Fifth 8.81 7.43 Sixth 9.77 8.70 Seventh 10.85 10.21 Eighth 12.13 12.20 Ninth 13.91 15.51 Tenth 18.96 28.41 Table A2.10. Moldova Decile A:1988 A:1993 First 4.27 2.58 Second 5.69 3.80 Third 6.80 4.97 Fourth 7.77 6.15 Fifth 8.70 7.40 Sixth 9.67 8.80 Seventh 10.77 10.49 Eighth 12.17 12.75 Ninth 14.32 16.42 Tenth 19.86 26.65 Table A2.11. Poland Decile A:1987 SA.1/1993 First 4.08 3.20 Second 5.63 5.17 Third 6.63 6.33 Fourth 7.56 7.37 Fifth 8.50 8.40 Sixth 9.52 9.49 Seventh 10.69 10.78 Eighth 12.12 12.38 Ninth 14.21 14.87 Tenth 21.06 22.00 160 Income, Inequality, and Poverty during the Transition Table A2.11. Romania Decile A:1989 M:3/1994 First 4.08 3.52 Second 5.89 5.22 Third 6.96 6.25 Fourth 7.99 7.21 Fifth 8.91 8.17 Sixth 9.84 9.27 Seventh 10.82 10.59 Eighth 12.13 12.32 Ninth 14.22 14.95 Tenth 19.16 22.51 Table A2.13. Russia Decile A:1988 Q:3/1993 First 4.14 1.64 Second 5.77 3.22 Third 6.82 4.19 Fourth 7.79 5.05 Fifth 8.77 5.99 Sixth 9.80 7.09 Seventh 10.95 8.49 Eighth 12.31 10.66 Ninth 14.19 14.14 Tenth 19.45 39.52 Table A2.14. Slovak Republic Decile A:1988 A:1993 First 5.34 5.70 Second 6.55 6.85 Third 7.37 7.58 Fourth 8.14 8.27 Fifth 8.92 8.96 Sixth 9.76 9.70 Seventh 10.69 10.54 Eighth 11.82 11.57 Ninth 13.40 13.06 Tenth 17.99 17.78 Decile Shares of Total Income 161 Table A2.15. Slovenia Decile A:1987 A:1993 First 4.55 4.34 Second 5.94 5.76 Third 6.87 6.69 Fourth 7.74 7.57 Fifth 8.63 8.47 Sixth 9.59 9.44 Seventh 10.68 10.57 Eighth 12.02 11.97 Ninth 13.94 14.04 Tenth 20.05 21.16 Table A2.16. Turkmenistan Decile A:1988 A:1993 First 5.12 2.72 Second 5.38 3.98 Third 6.08 5.15 Fourth 6.96 6.29 Fifth 7.99 7.48 Sixth 9.19 8.79 Seventh 10.60 10.36 Eighth 12.36 12.47 Ninth 14.82 15.91 Tenth 21.51 26.85 Table A2.17. Ukraine Decile A:1988 M:6-7/1995 First 4.38 1.51 Second 5.93 2.76 Third 6.92 3.92 Fourth 7.83 5.07 Fifth 8.75 6.27 Sixth 9.73 7.62 Seventh 10.82 9.24 Eighth 12.14 11.45 Ninth 13.99 15.20 Tenth 19.51 32.74 162 Income, Inequality, and Poverty during the Transition Table A2.18. Uzbekis tan Decile A:1989 A:1993 First 3.83 2.99 Second 5.14 4.29 Third 6.24 5.47 Fourth 7.26 6.61 Fifth 8.26 7.77 Sixth 9.33 9.04 Seventh 10.58 10.55 Eighth 12.21 12.54 Ninth 14.80 15.74 Tenth 22.35 25.00 Appendix 3 Change in the Poverty Deficit Due to a Uniform Slide in Income The poverty deficit is, by definition, equal to PD =|z(z-y) f(y) dy which is shown in Figure A3.1 as the sum of rectangles such as ABEz and CDFz. Obviously, the poverty deficit will be equal to the summation of all such rectangles for all values of yS C l > o m N 0 C 00 (ON ~~Cl en ~ CA . N -4 CO N LIj C tt 1 CD 0 <~~~~~ R^ ~ ~~~~ L^' c oe 01 t CA41 Cl4 Cl~~~~~~~~~~~~~~~L *D q 2 ~ ~~ ~~ ~~ ~~ ~~~~~N It 1- a? C V)~~~~~I .O 4 Cl O O S Cl Cl Cl Cl C4 Cl ;0 It~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I a~~~~~~~~~~~~~~~~~O .o L N£ ) LI) LI) CO 0 0 4.- ~ ~ ~~~~. z L C b ~ ~~~ C> Cl o4 ClNo tCl LI) CO0 0 to o - U j!.,>_ N o 00 o 0o ( R: ;tE t m O O 0 t]t ¢ e | ¢....... Uo Q o o £ o zy o ..... - =aa - P .. ..... BB. ............B 0 8 : :g :g, EE E ! g g gg ., .! g ,g, . ,,,! ,,| , lE,! . , E ,, . g,B ,' - , , . .BB......................B O z J _ Bai. ili i- 8 ig. ~~~~~~~~~~~~~~~~................... ........... . E i Eg-ggg...........EErgB 'i B X ; ,::! tt -S,Z,~~~~~~~~~~~~~~~~~~~. .................................................. .- ..",....'-.."_. .""" '.... 9 'BEiEE Ei # # ffi $ Cl _ gga B ~~~~~~~~~~~~~~~~~~.>.g....<............................ ........... ........ - .......BBB.. > gL, k E ....................................... s = = ; B B B '' < 8 a g f = ' M ' B ; f 8 ' B B a0................ ............ ,O !i O p m ~~~~~~~~~~~................................ . ...........=lmBaB'"BB.-'g 184 Average per Poverty capita Country Survey headcount Average Exchange Poverty line income [Poverty line] period (in %) shortfall Elasticity Gini rate ($ pm) ($ pm) CSSR 15011 1988 A 0.05 26.3 0.0 19.5 14.3 35 129 Slovakia 13,428J 1993 A 0.12 19.6 0.01 18.3 30.9 36 100 c I~u1~Bulgar 191 1989 A 1.4 31.7 0.05 23.3 2.2 34 98 Ul BU 1993 A 14.6 25.8 0.32 34.3 27.6 30 71 1989A 5.8 15.0 0.22 23.3 73 11 25 March 1994 59 32 0.7 28.6 1601 28 29 (Continued on thefollowing page) Table A5. (Continued) Average per Poverty capita Country Survey headcount Average Exchange Poverty line income [Poverty line] period (in %) shortfall Elasticity Gini rate ($ pm) ($ pm) 1987A 0.07 46.3 0.00 21.5 373 78 401 1993 A 0.3 30.8 0.01 25.1 113.2 73 293 < :lRus00gilX4::: -:gg:::l00 1988 A 1.5 22.3 0.07 23.8 2.2 25 72 Rus 3Q/1993 49.7 39.6 0.57 48.0 1,025 21 32 (july prices) *Ukgltkrfi&|ggXt ilBBllllllll 1988 A 1.9 21.7 0.09 23.3 2.2 25 65 June and 1 62.9 46.7 0.51 47.4 142,000 21 24 mi11iox~1 week of July 1995; (June prices) Average per Poverty capita Country Survey headcount Average Exchange Poverty line income [Poverty line] period (in %) shortfall Elasticity Gini rate ($ pm) ($ pm) Bel.arus 1541 1988 A 1.1 23.7 0.05 22.8 2.2 25 71 Belarus 1l9g $$ ---l1Q/1995 22.3 25.6 0.53 28.4 11,525 19 32 (March prices) N Mo 1988 A 3.5 8.2 0.22 24.1 2.2 25 61 ;o - N1993 A 65.9 43.4 0.55 36.5 1.6 13 12 Xs ( 1988 A 0.9 23.5 0.04 23.0 2.2 25 81 .st ( . .3Q/1995 37.2 37.1 0.7 35.4 11.1 60 86 (July prices) (Continued on the following page) Table A5. (Continued) Average per Poverty capita Country Survey headcount Average Exchange Poverty line income [Poverty line] period (in %) shortfall Elasticity Gini rate ($ pm) ($ pm) 1988 A 0.7 22.5 0.03 22.5 2.2 25 78 4Q/1995 21.7 27.6 0.49 31.0 0.536 58 101 1988 A 0.8 24.0 0.04 22.5 2.2 25 75 1994 A 29.8 33.6 0.48 37.3 4 26 46 1988 A 4.6 8.9 0.27 25.7 2.2 25 61 1993 A 64.9 38.7 0.66 32.7 1,168 23 22 Average per Poverty capita Country Survey headcount Average Exchange Poverty line income [Poverty line] period (in %) shortfall Elasticity Gini rate ($ pm) ($ pm) Uzbeldslan [5& 1989 A 23.8 19.8 0.60 28.2 3.4 16 28 1993 A 62.6 38.9 0.64 33.3 951 11 11 1988 A 12.0 10.4 0.59 26.0 2.2 25 47 lz [11111 Oct-Nov. 88.0 68.0 0.20 55.3 7 16 9 1993 (October prices) lenistan t541 1988 A 11.8 10.1 0.60 26.4 2.2 25 47 Turktnenistin 1993 A 60.9 40.0 0.61 35.8 892 33 34 (Continued on thefollowing page) Table A5. (Continued) Average per Poverty capita Country Survey headcount Average Exchange Poverty line income [Poverty line] period (in %) shortfall Elasticity Gini rate ($ pm) ($ pm) Expenditure-based Measures 1993 9.8 19.9 0.36 30.6 17,300 51 116 6 months March 1994 47.6 33.8 0.72 32.8 1601 28 35 M Jul-Sept. 95 33.7 28.4 30.7 11.1 60 87 (July prices) 3Q/1993 39.4 43.8 0.44 49.6 1025 21 39 (July prices) UkrE1 g;llilll#0-SgJune and 1 25.7 37.2 0.35 43.8 142,000 21 50 nil y-0SNgS -