Report No. 26011-YU Serbia and Montenegro Poverty Assessment (In Two Volumes) Volume II: Main Report November 13, 2003 Poverty Reduction and Economic Management Unit Europe and Central Asia Region Document of the World Bank VOLUME TWO: TABLE OF CONTENTS ACKNOWLEDGEMENTS ...................................................................................................................... vu.. INTRODUCTION ................................................................................................................................... ... vi11 CHAPTER 1 Policies.MacroeconomicPerformanceand Living Standards . ....................................... 1 1.1.GeneralEconomic Developments inthe 1990s........................................................................ 1 1.2. Changes InThe Distribution Of PersonalIncomes InThe 1990s: Poverty Dynamics.......................................................................................................................................... 2 1.3. Living Standards in 1990s......................................................................................................... 4 1.5. Montenegro: A Unique Transition.......................................................................................... 1.4. Outcomes of Reforms................................................................................................................ 8 11 1.6. Serbia and Montenegro: GrowthProspects............................................................................. 12 1.7. Poverty Projections................................................................................................................. 15 CHAPTER 2 Serbia and Montenegro: Poverty in 2002 2.1. Serbia And Montenegro: Poverty And PovertyLines............................................................ . ....................................................................... 19 19 2.2. How Many Poor InSerbia?Vulnerability And Non-Income Poverty ................................... -25 2.3. How Many Poor InMontenegro In2002? InternationalPoverty Comparisons...................... 27 2.5. RegionalDisparities.Rural Poverty........................................................................................ 2.4. Poverty And Displacement...................................................................................................... 31 34 2.6. Poverty Profile: Who Are The Poor?...................................................................................... 36 2.8. Poverty InThe Life Cycle: Youth, Old Age, Disability.......................................................... 2.7. Understanding Poverty: Decompositions And Multivariate Analysis .................................... 38 45 2.9. Gender And Poverty................................................................................................................ 50 2.11.Inequality In2002.MainDrivers Of Inequality..................................................................... 2.10.Ethnicity And Poverty........................................................................................................... 52 53 2.12.Conclusions ............................................................................................................................ 57 CHAPTER 3 Labor Market and the Poor 3.1. Challenges Ahead: Employment Reallocation........................................................................ . ............................................................................................. 59 59 3.2. Restructuring: How Many are Already Affected ................................................................... 63 3.3 Workers' Reactionto Underemploymentand The Threat of Lay-Offs .................................. 66 3.4 Labor Market Data InSerbia And Montenegro ...................................................................... 68 3.5. Unemployment: Its Characteristics and Dimensions .............................................................. 75 3.6. The Working Poor: "Good" And "Bad" Jobs ......................................................................... 79 3.7. Private Sector Development: How It Can Serve The Poor?.................................................... 80 3.8. Rural Incomes And Poverty .................................................................................................... 82 3.9. Conclusions............................................................................................................................. 85 CHAPTER 4 Education: Policies to Combat Poverty . ......................................................................... 4.1 Context .................................................................................................................................... 87 87 89 4.3. Children's School Enrollment And Completion..................................................................... 4.2. Access To And Quality Of Schooling..................................................................................... 91 4.4. Secondary School Enrollments AndThe Poor........................................................................ 94 .'........................................ 98 4.6. Conclusions And Questions For Research............................................................................. 4.5. Public And PrivateEducation Spending....................................... 98 CHAPTER5 Health Care Reformandthe Poor . ............................................................................... 5.1 Introduction: Overview Of The Health Sector....................................................................... 101 101 5.2. Population Health Status....................................................................................................... 102 5.3. HealthcareProvision And Access To Healthcare ................................................................. 105 5.4. Health Services Utilization.................................................................................................... 5.5. HealthCare FinancingAnd Expenditure .............................................................................. 106 108 5.7. Conclusions........................................................................................................................... 5.6. Barriers To Access ................................................................................................................ 111 112 CHAPTER 6 Protecting the Poor . ........................................................................................................ 113 6.1 Social Protection System In Serbia And Montenegro........................................................... 113 6.2.The Role O f NGO................................................................................................................. 123 6.4. PensionReform: Approaches And Issues............................................................................. 6.3. Incidence Analysis Of Key Transfers ................................................................................... 125 6.5. MediumTermFramework for Public Finances .................................................................... 134 136 6.6. Protecting The Vulnerable And Improving The Targeting................................................... 137 CHAPTER 7 Monitoringand Evaluation . ........................................................................................... 141 7.1 Components Of A MonitoringSystem................................................................................... 141 7.2. Serbia .................................................................................................................................... 142 7.3. Montenegro ........................................................................................................................... 143 7.4. Challenges InStrengthening Feedback Mechanisms ............................................................ 143 References................................................................................................................................................ 145 ANNEX I .Measurement of Poverty .................................................................................................... 151 ANNEX I1 .Statistical Tables and Figures ........................................................................................... 162 ANNEX I11 .Glossary of Terms ............................................................................................................. 170 Tables Table 1.1.Poverty in Serbia and Montenegro in2000 and 2002 ................................................... Table 1.2. Key Economic Indicators and GrowthProjections for S A M . 2000-5........................... 10 Table 1.3. Consolidated Social Expenditures in GDP. Serbia ....................................................... 13 14 Table 1.4. Poverty Projections for Serbia. 2002-6 ......................................................................... 15 Table 2.2. Overview of Multidimensional Poverty Indicators....................................................... Table 2.1. Serbia and Montenegro: Overview of Poverty Lines. 2002......................................... 21 24 Table 2.4. Poverty in Serbia inKey Dimensions ........................................................................... 27 Table 2.3. Poverty Incidence Based on Absolute and RelativeLines inSerbia............................. 26 Table 2.5. Multidimensional Poverty Indicatorsfor Montenegro.................................................. Table 2.6. HouseholdComposition and PovertyNulnerability of IDPs and Refugees..................29 33 Table 2.7. Poverty Rates inMontenegro by Region...................................................................... Table 2.8. Poverty by Region and Type of Location inSerbia in 2002......................................... 34 35 Table 2.9. Poverty Among Adults According to Education inSerbia in2002 .............................. Table 2.10. Poverty According to the Employment of Household Members inSerbia in2002....42 36 Table 2.11. Observed and SimulatedPoverty Rates by Regions ................................................... 37 Table 2.12. Observed and SimulatedPoverty Rates by Educationand Employment Status .........43 Table 2.13. Comparison of Serbia and Montenegro with Selected Eastern European Countries ..54 Table 3.1. S A M Employment and Unemployment........................................................................ Table 2.14. Structure of Total Income and Sources' Contribution to Total Income Inequality.....56 Table 3.2. Unemploymentin Selected Transition Countries. 1993-2002...................................... 72 Table 3.3. Poverty According to Socioeconomic Status in Serbia. 2002....................................... 73 76 Table 3.4. Characteristics of the Unemployed in Serbia: Registered versus Strictly DefinedL O 7 7 Table 3.5. Characteristics of Rural Households. by quintiles of the Rural Population..................84 Table 4.1. EstimatedDistance (inkms) between Household Residence and Closest School........90 Table 4.2. EstimatedDistance (inkmrs) between HouseholdResidence and Closest School.......90 Table 5.2. MainReasonsfor Not UsingHealth Services............................................................. Table 5.1. Distanceinkmfrom Health Care Facilities. Medians and Means (inbrackets) ........106 Table 6.1 Social ProtectionExpenditures as a share of GDP ,selectedECA countries..............111 114 Table 6.2. Public Expenditures on Pensions in selectedcountries to GDP ................................. 115 Table 6.3. Number of Pensioners and Average Pensionby Type of Pension Fund..................... 118 Table 6.5. Main Child Benefits inSerbia, 2003 ........................................................................... Table 6.4. The Government Support for the PoorFamilies with Children in Serbia...................116 119 Table 6.6. Serbia: Public Expenditures on Social Assistance ...................................................... Table 6.7. S A M : Coverage of Population and the Poor by Social Protection Programs, 2002 ...120 122 Table 6.8. Coverage of Main Social ProtectionPrograms inSerbia, 2002.................................. Table 6.9. Share of Households Receiving Social Transfers in selectedECA countries .............126 127 128 Table 6.11.Poverty among Pensioners inSerbia.......................................................................... Table 6.10.Coverage of Programs in Serbia................................................................................. 135 Table A2. Poverty Line................................................................................................................ 158 Table A1.MinimumFood Baskets .............................................................................................. 157 Table A3.Indicators of Poverty and Population Vulnerability inMontenegro (2003) ................161 Table Al.l. Serbia and Montenegro: Overview of PovertyLines in2000 and 2002 .................162 Table A1.3. Determinates of Poverty in Serbia in2002............................................................... Table Al.2. RegionalVariation of Multidimensional Poverty Indicators ................................... 162 163 Table A1.4. Probability of Spending on Education (household level)......................................... 165 Table A1.6. Predictors of ReportingChronic Disease and Acute Symptoms for Ages 40> ........166 Table Al.5. Conditional on Non-zero household expenditure on education............................... 168 Table A1.7. Probability of Spending on Health DuringLast Month ........................................... 170 Figures Figure 1.1SAM: Evolution of Output. Employment. Unemployment. Average net Wage and Average Pension. 1989-2000 ........................................................................................................... Figure 1.2 Change inthe Poverty Headcount, 1990-2000............................................................... 2 3 Figure 1.3 Share of Wages inHouseholdIncomes in SelectedTransition Economies.................... Figure 1.4 Number of Employed and Pensioners inS A M 1989-2000............................................. 3 Figure 1.5 Structure and Evolution of Real HouseholdIncomes by Source, 1990=100..................6 7 Figure 1.6 Poverty Rates in Serbia (2002-2015) with Various GrowthRates of Consumption and 16 Figure 2.1 OverlappingPoverty Dimension................................................................................... Inequality........................................................................................................................................ Figure 2.2 Poverty Rate inMontenegro as a Function of Poverty Lines ....................................... 28 30 Figure 2.3 Absolute Poverty Incidencefor Selected Countries inSEE, ComparableData, Relative to Serbia's Poverty Rate................................................................................................................. Figure 2.4 Decomposition of Difference inConsumptionbetween Top and Bottom Quintiles....31 38 Figure 2.5 Simple versus Partial Correlation inPovertyAnalysis ................................................. 40 Figure 2.7 Poverty Rate and Composition of Poverty by Age Groups in Countries of SEE.........49 Figure2.6 RelativePoverty Rates for Children versus the Elderly ............................................... 47 Figure 2.8 Poverty Rate Relativeto the Average Poverty Rate by Gender of HouseholdHeadin ... Figure 3.1 Serbia: Differences in Survey and Registry Unemployment, 2002 ............................. Serbia: Observed versus Simulated................................................................................................ 51 69 Figure 3.2 Labor Market Flows...................................................................................................... 70 Figure 3.3 ParticipationRates, 1995-2002..................................................................................... 74 Figure3.4 L OUnemployment Ratesfor Adults (Population above 15), by Gender.................... 75 Figure 3.5 Change inAge Specific Unemployment Rates. Cumulative 1995-2002...................... 78 Figure 3.6 Non-agricultural Self Employment as Percent of Total Non-agricultural Employment81 Figure 4.1 Public KindergartenAttendance by Age Group. Poverty and UrbadRuralResidence92 Figure 4.2 Timing of Primary School Enrollment ......................................................................... 93 Figure 4.4 Monthly HouseholdEducationExpenditure (mean amount by consumptiondecile) ..95 Figure 4.3 Enrollment at the SecondaryLevel. Children Aged 15 to 17....................................... 97 Figure 5.1 Outpatient Visits to Doctors inState Institutions (Last Month) ................................ 107 Figure 5.2 HouseholdHealth Expenditure by ConsumptionDecile............................................ 110 Figure 6.1 Concentration Curves for Selected Social ProtectionPrograms................................. Figure 6.3 Coverage. Targeting and Adequacy of the Main Social ProtectionPrograms ...........133 Figure 6.2 Coverage: Targeting and Adequacy of Social Insurance and Social Assistance .......130 Figure 6.4 Changes inthe Demographic Composition and Dependency..................................... 134 135 Figure 6.5 Trend in Consolidated Public Revenues and Expenditures ........................................ 137 Figure A1 Engel Curve ................................................................................................................ Figure A2 Probability of Non-zero household expenditure on Education .................................. 158 Figure A3 HouseholdExpenditure on Education - OLS Predictions .......................................... 166 167 Figure A4 Chronic and Acute Illness by Poverty......................................................................... 169 BoxesBox 1.1War and Displacement........................................................................................................ Box 1.2InformalSector................................................................................................................... 5 7 Box 1.3 What i s Pro-Poor Growth? ............................................................................................... 17 Box 2.1 Old and New Approaches to Measuring Poverty inSerbia .............................................. 22 Box 2.3 Decomposing the Poverty Gap......................................................................................... Box 2.2 Cross-country Poverty Comparisons ............................................................................... 32 39 Box 2.4 Example of Roma Settlement inMontenegro................................................................... 53 Box 3.1 Costs of Postponing Difficult Reforms: The Case of Bulgaria ....................................... 60 80 Box 4.1 Why EducationMatters for Poverty Reduction................................................................ Box 3.2 Gender and Work inS A M................................................................................................ 89 Box 4.2 Equity of Access to Education: The HumanRightsPerspective..................................... 94 Box 6.2 Pension System: Who i s Out?........................................................................................ Box 6.1 Introduction to the Evaluationof Social Policies ........................................................... 131 136 ACKNOWLEDGMENTS This Poverty Assessment report is the result of collaboration in poverty analysis between Serbian and Montenegrin counterparts and the World Bank. H.E.Gordana Matkovic, the Minister of Social Affairs of Serbia, made possible the collaborative work on the report by her constant support, direct involvement in the analysis and policy guidance. Mrs. Snezana Mijuskovic, Deputy Minister of Labor and Social Affairs of Montenegro and the Head of PRSP Management Unit, and PRSP ETFLeader Prof. Gordana Djurovic were the key Government counterparts in Montenegro, while the ISSP-CEED, NGO headed by Petar Ivanovic was the main working level partner. The first draft of the report was produced between March and May, 2003 and discussed with the Governments in June-August, 2003. The compressed schedule was made possible by the successful preparatory work (Poverty Survey Project No. PO74904), co-financed by the Dutch Trust Fund, and the EC Food Security Program. Ministry of Social Affairs of Serbia played a prominent role in the survey conceptualization, design, management, funding, coordination with other Government agencies, including the Republican Statistical Office, and the dissemination of results. The SMMRI (Belgrade) conducted the first round of the survey inMay-June 2002; Srdjan Bogosavljevic, DragisaBjeloglav and Aleksandr Zoric played a key role in the design, data consistency checks and in the initial analysis of results. The ISSP- CEED in Montenegro, represented by Dragana Radevic completed the surveys in 2002. Crucial contributions on the World Bank side to the survey's success were provided by Mamta Murthi, Philip O'Keefe, KathleenBeegle and Michael Lokshin. The report relies on analytical work by Serbian and Montenegrin counterparts: Gorana Krstic (poverty profile for Serbia), Bosko Mijatovic (social policy assessment), Biljana Bogicevic (social protection), Branko Milanovic (inequality and social assistance analysis), Dragana Radevic (poverty inMontenegro), Vladimir Vukojevic (social impact analysis), Mihail Arandarenko, Branko Jovanovic (labor market) and Alexandra Posarac. The team i s also indebted to many government officials from Serbia and Montenegro, inparticularto the staff of PRSPunits, for comments and suggestions. The report was prepared in the World Bank (ECSPE) by Ruslan Yemtsov with key inputs by Cem Mete (health and poverty, education for poverty reduction), Cornelia Tesliuk (social protection) and Lazar Sestovic (macro policies). Ardo Hansson, Henri Gordon, Loraine Hawkins, Jim Stevens, Mamta Murthi, Dena Ringold, Taies Nezam participated in the analysis. Emily Evershed was an editor of the report. Helena Makarenko, Judy Wiltshire and Jim Lynch undertook the report processing. Rory 0'Sullivan and Marina Petrovic from the World Bank Belgrade office provided crucial link to policy dialogue. Zana Ivanovic seamlessly organized numerous missions and timely translations required for fully collaborative work. The team i s wholeheartedly grateful to Nancy Cooke (Lead Country Officer for S A M ) for her support and encouragement. Asad Alam (Poverty and Gender Sector Manager, ECSPE) provided advice, quality control and overall supervision. The report was prepared under the guidance of Orsalia Kalantzopoulos (Country Director, ECCU4). The peer reviewers were Jeni Klugman (Lead Economist, PRMPR) and Salman Zaidi (Senior Economist, SASPR), who contributed substantially to the report's analysis and presentation. i INTRODUCTION The Poverty Assessment is the first output followed by additional data collection and of a multi-year program of analytical, analytical work aimed to fill the remaining monitoring and capacity building poverty work knowledge gaps and monitor changes in poverty adopted by the World Bank to assist the situation in2003-2006. Governments of Serbia and Montenegro in the development and implementation of their The report is organized in two volumes. Poverty Reduction Strategies (PRSP). This Volume One (Executive summary) summarizes report i s aimed at informing the Government's the Report content and covers the agenda for Poverty Reduction Strategy as well as the poverty reduction in SAM and its two Bank's Country Assistance Strategy (Transition constituent republics, with an emphasis on Support Strategy for Serbia and Montenegro). common aspects. Volume Two (Main report) provides detailed results of poverty analysis and The report contains the poverty diagnostics the policy context for developing the poverty based on the 2002 data for Serbia and reduction strategy. It focuses partly (but not Montenegro, the most recent at the time of exclusively) on Serbia. Annexes to the Volume writing the report. The in-depth analysis Two give a through description of the data used presented in the report aims to provide an inpreparingthe report, including the most recent understanding of who are the poor and why (2003) findings from Montenegro, discuss there are poor, and the detailed assessment of the methodologies for poverty measurement and access by the poor to key social services and the provide a glossary of technical terms used inthe labor market. The report identifies knowledge report. In-depth coverage of the situation in gaps that remain to be filled in the future work. Montenegro using 2002 data was presented in a An impressive pace of reforms in Serbia and joint World Bank-ISSP background paper by Montenegro requires often updates of the D.Radevic (ISSP) and K.Beegle (The World knowledge base and data. This report will be Bank). .. 11 1. POLICIES,MACROECONOMICPERFORMANCEAND LIVINGSTANDARDS This chapter describes living standards in a country recovering from a decade of economic mismanagement. Poverty increased in SAM compared with the late 1980s, while many problems were pushed underground, or solutions postponed, or empty promises given. The new Government in Serbia decisively broke with the past and began long-delayed reforms, but many challenges remain. Inview of these challenges, the governments have adopted an over-reaching goal of poverty reduction. This chapter suggests the central importance of economic growth to poverty reduction. Chapter also argues that growth alone i s not sufficient for accelerated poverty reduction but that growth should be associated with distributional changes - with a more rapid growth of the incomes o f the poor and near-poor than of higher incomes. 1.1. GeneralEconomic Developments in the 1990s 1.1 The former Socialist Federal Republic of Yugoslavia (SFRY) probably had the best pre- conditions among the former socialist countries for a rapid transition to a market economy, owing to some incremental reforms that were started in the early 1970s and that continued in the 1980s. Inthe early 199Os, the republics that constitutedthe SFRY became independent states and Serbia and Montenegro created the Federal Republic of Yugoslavia (FRY). The first years were the hardest years for the FRY in its short history owing to a number o f internal and external political and economic shocks, followed by poor policies of the government of that period. The results o f these circumstances were a deep and sharp output decline, hyperinflation, a rise in unemployment, a huge internal and external public debt, and an increase in the informal sector. The impact on the living standards was obvious - poverty increased significantly and spread to all segments of society. 1.2 GDP declined sharply and remained at less than half of the pre-transition levels. While many factors contributed to output decline, the main reasons were poor macro and micro economic management, the loss of the markets of other former SFRY republics, and international sanctions. Figure 1.1 shows the macroeconomic developments in Serbia and Montenegro (SAM) from 1989to 2000. 1.3 The decline was fairly general, with the private sector showing considerable resilience. The state (social) sector of the economy suffered more from these shocks than didthe private sector, which proved to be more flexible: in the early 1990sit contributed only around 15 1 percent of total output and reached 30 percent of GDP in 1994. The decline in output was also greater in industry than in services and agriculture (agriculture output in 1990 was 10percent of GDP, while in 1994 it was 31.4 percent of GDP). In 1994 the new stabilization package (which mainly relied on monetary policy) was adopted, and the FRY economy began a recovery. The stabilization measures were later supported by the positive impact of the ending of the war in Bosnia and Croatia, and by the partial lifting of international sanctions. The output recovery of this period (the 1994-98 average yearly growth was around 5 percent) was halted by the Kosovo crisis in 1999 (GDP in 1999declined by around 18 percent). Figure 1.1: SAM: Evolutionof Output, Employment, Unemployment, Average Net Wage and Average Pension, 1989-2000 (1989=100) 150 -o- realwage 125 - 100 GDP 75 - -employmeat 50 A unemployment 25 -X- -averagepension 0 1989 1990 1991 1992 1993 1994 1995 1996 1991 1998 1999 Zoo0 Note: GDP excluding Kosovo from 1999. Staff estimates for non-registered employment and LO-consistent unemployment prior to 1994. Source: Staff estimates, National Bank of Serbia, Federal Statistical Office: Serbian Ministry of Social Affairs. 1.2. Changes in the Distribution of Personal Incomes in the 1990s: Poverty Dynamics 1.4 Informationon poverty in SAM (as in any country) i s basedon (i) level data on the micro distributionof welfare inthe population and (ii) poverty line, reflecting an absolute minimum the required to sustain the functioning of individuals in a society. Micro data on welfare are obtained through sample surveys, during which households are asked to answer detailed questions on their spending habits and sources of income. The advantages and shortcomings of the methodologies adopted are described in Box 2.1 of this report, which highlights a number of problems with these data. However, even with limitations, these series provide useful information on poverty trends. 1.5 Figure 1.2 shows a sharp increase in poverty in 1999-2000 as compared with 1990 or with 1997-98. Two particularly interesting features emerge. The first i s the huge impact of 2 hyperinflation in 1993 on poverty: the poverty rate almost doubled in a single year. The second i s a rapid and strong increase in poverty in 1998-2000. Figure 1.2: Changeinthe Poverty Headcount, 1990-2000 (poverty in 1990=100%) 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Note: Usingthe FSOfood basket andHBS data onhouseholdincomes; without Kosovo; data includebothSerbia andMontenegro. Sources:For 1990-99,SMMRI based on FSO HBS primary data andPosarac (1997); for 2000, B. BogiCeviC(2001). 1.6 Behind this general picture, a strong redistribution was taking place. An unparalleled fall inthe role of wages as a source of incomes occurred. The forces that ledto a decline inthe role of earnings in Yugoslavia were similar only to those which ledto the collapse of the formal labor market in several CIS countries (Figure 1.3). Figure 1.3: Share of wages inhouseholdincolrps inselectedtransitionecononies 0% 10% 20% 30% 40% 50% 60% 10% Source: For all countriesexcept Yugoslavia ,World Bank (2000), MakingTransitionWork for Everyone, Chapter 4; FRY- FSO Statistical Yearbook, 2000. 3 1.7 The World Bank report "Breaking with the Past" (World Bank [2001 b]) discusses channels of redistribution during this period. Changes in the structure of the incomes of the population reflected the double impact of falling employment and real wages. Figure 1.3 uses HBS data to compare the total share of earnings in the incomes of households in Yugoslavia and other transition economies. 1.3 Living Standards in the 1990s 1.8 Employment was falling more or less steadily but lagged behind the fall in output. The level of employment did not follow output proportionally (see Figure 1.1), thus adjustment to the lower level of economic activity was made mainly through lower wages, especially during the hyperinflation of 1993. However, by 2000 real wages recovered somewhat and employment continued its slow and sure decline. Overall, the adjustment to both wages and employment by the year 2000 was much less than the GDP fall would suggest, pointing to the building up of pressuresfrom hoardinglabor. 1.9 Policies were put in place to keep employment levels artificially high. This labor hoarding was an explicit government policy, which effectively outlawed any layoffs. It led to the stalled restructuring of large socially owned companies and to an increase in hidden unemployment (a large number of the employed were on "forced holidays," and some estimates were that 150,000-250,000 workers annually came under this status). The main impact of such a policy was that labor mobility was blocked, since redundant workers hadno real opportunities to change their qualifications, acquire new skills, and enter another more profitable area of activity. 1.10 Even though the decree prohibiting layoffs was annulled by the end of the 1990s, not many companies used this opportunity to cut the number of redundant workers, but either kept wages at a very low level or paid no wages at all. This type of response led to a significant decrease in the statistically measured average wages (because the total wage bill was divided by the entire number of officially employed, includingthose on forced holidays with minimumpay) and also led to an increase in the informal sector. Studies (Krstic 2000) show that many of the affected workers found a job on the side, but continued to use the advantage of having official employment in some state companies because of their broad social guarantees (pensions, health insurance, etc.). 1.11 The unemployment rate increased, even when measured by survey data. Unemployment in Serbia and Montenegro was a problem even before the transition and had structural and long-term characteristics. However, unemployment was not accurately measured, as the analysis in Chapter 3 of this report shows. Figure 1.1 uses consistent estimates of employment and unemployment based on estimates for the years with missing LFS data. The fall inemployment was accompanied by a flow out of the labor force. 1.12 In the early 1990s wages declined sharply in real terms. The steepest decline in real wages was in the period of hyperinflation 1992-93 (-49 percent and -61 percent, respectively) and in 1999, when real wages declined 15 percent. The protracted Yugoslav hyperinflation of 1993-94 has had no equal in recent economic history. At the peak of hyperinflation the average 4 monthly wage fell to less than US$5. Living standards were further harmed by the freezing of foreign currency savings in the state banks and also by failed pyramidal financial schemes in which thousands of people took part. These two events wiped out hardcurrency savings (if any). 1.13 The decline in wages as measuredby the cost of living index is also evident in a more direct measure of real wages: the ratio of average net wage to the current cost of the FSO food basket (on the composition of the basket, see Chapter 2 and Annex I). On average in 1989, one net wage was sufficient to buy 1.25 baskets; in 1993 one average net wage was sufficient to buy less than one-third of this basket. Monthly data are even more alarming: thus, in January 1994, when inflation reached its peak, one average net wage was sufficient for only 6 percent of the basic food basket. After January 1994 there was at first rapid, and then (after 1995) relatively slow, growth in purchasing power, which stopped in the second quarter of 1999 with the beginning of armed conflict associatedwith the Kosovo crisis (see Box 1.1). During most of the 1993-95 period, a family of four with two average wage earners could not afford basic food Ibasket. This was also the case in 1999-2000. BOX 1.1: War andDisplacement. I The social costs of conflict include refugees and internally displaced people. The FRY'Sengagement inconflict with other Yugoslav successor states (both directly and indirectly) contributed to the flow of refugees. The struggle between the Yugoslav National Army and the Croatian forces for control of territory prompted an initial flow of Serb refugees to the FRY. These numbers increased dramatically between May and August of 1995, when approximately 150,000 peoplefled Croatia for Serbia (reaching a total of 290,000 as of 1996). Similarly, fighting in Serb-dominated areas of Bosnia-Herzegovina triggered refugees, in numbers estimated at 230,000. NATO intervention in Kosovo triggered an exodus of approximately 230,000 IDPs of Serb and Roma ethnic origin from that region. Around 12,000 amongthe refugees from Kosovo are "double refugees," that is, they fled from Croatian Krajina and were settled inKosovo. At the peak, the number of displaced was as high as 7 percentof the population in Serbia and Montenegro. This number has beenreducedsince then with the integration of some refugees and the internationally mediatedreturn. The total cost to the budget of hosting refugees was estimated at around 1.5 percent of GDP annually. This influx had several repercussions on the population, related to housing, the labor market, and access to social services. The majority of refugees were absorbedin the families of the domestic population. While this was judged to be a better outcome for their potential prospects than living in collective centers, it also had significant costs for the receiving families. IDPs living with relativedfriends often moved amongtheir circle to limit the burden. Lack of continuity in housing, while redistributing the burden, limited the possibility of utilizing a network of contacts for jobs. The pressure of IDPs on the labor market led to some social cleavages. In the addition, IDPs were sometimes blamed for the pressure on public services and criticized for receiving aid for which other vulnerable families were not eligible. Sources: UNHCR census of April and May 1996. Steering Committee of the Government of Serbia for Drafting the National Strategy, Secretariat and UNHCR, and the Economics Institute of Belgrade, National Strategy for Resolving the Problems of Refugees, Expellees and DisplacedPersons, Belgrade,2002. CPMCPS, The Social Situation inSerbia, 2001. 1.14 The shrinking budget of social transfers had spread to more people. Pension, health and employment funds were in misbalance owing to lowered revenues (decreased economic activity and the shift to the informal sector) and growing expenditures. Particularly, difficult situation existed with regard to the pension and health funds. The number of pensioners was constantly growing, and since registered employment was declining during the same period, the dependency ratio was worsening. The result i s shown inthe Figure 1.4. 5 Figure 1.4: Number of Employed and PensionersinSAM 1989-2000. (thousands) 3,000 2000 employed (in thousands) 1,500 OTotal number of pensioner: SAM (in 000) 1,000 ~~01989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Source: Federal Statistical Office and PensionFunds. 1.15 Real paid out pensions were hard hit. Duringthe period under discussion, the average pension declined significantly in both nominal and real terms, thus in 1999 it was only around one third of those from pre-transition period. Average pensions due were kept at a relatively high level compared to average wages, but they were irregularly paid, and in some cases pensioners were not paid in cash but were compensated with coupons for electricity, wood and coal. The monthly pension was often split and paid in smaller parts. In addition to that, during whole period of nineties, there was almost not even one year with all 12 monthly pensions being paid. Significant inequalities between pensioners widened. Beneficiaries from the Agriculture Pension Fund had between two (in 1994) and five times (in 2000) lower average pensions than beneficiaries of the other two pension funds (for the employed and self-employed). 1.16 The fall inwages was accompanied by a collapseinthe real value of social transfers. Figure 1.5 documents important shifts in the role of different income components over time. The second panel of the figure 5 takes the real value (compared to the cost o f the poverty basket) o f main sources of income for an average household in 1990 as 100 and reports the subsequent evolution of each household income. 1.17 Informal sector incomes were growing in real terms, albeit from a low base. Figure 1.5 makes clear that formal sources o f income, especially wages from regular employment and the social assistance were hit hardest over the period 1990-2000, in particular in the last two years. On the other hand, informal earnings (see Box 1.2) and consumption in-kindincreased in real terms. 1.18 Inequality, however, remained stable. It i s well documented for CIS countries that such shifts away from the formal sources of income normally lead to a rapid increase in inequality. This didnot happen in Serbia and Montenegro. Data from the HBS (reported by Posarac [19971 'EconomicsInstitute(2000), Survey of Pensioner Householdsfor Nutrition and Living Conditions 6 and Bogicevic et a1 [2001]), and other sources (such as LFS data on earnings) suggest that inequality remained constant: in 1990 the Gini index for per capita incomes was 0.28, and in 2000 it was 0.27. This stability can partly be accounted for by a forceful compression of all salary differentials in the public sector (as documented in Krstic [2002]), increased reliance on production for own consumption (as a coping strategy) and a hostile business environment that suppressed private sector activity, especially self-employment and entrepreneurship (key drivers of inequality change in transition elsewhere). Therefore, moderate inequality comes at a huge cost. Figure 1.5: Structure ofhouseholdincomesinFRYaccording Evolutionofrealhorseholdincomesbysource, to HBS, by sources 1990=100 -Totalcash incom 100% 8Consunption inkind -Regularenp!aymnt 90% 0Otherincom 80% -+-Temparary 70% 8Netincomfmmsmllshop enploymnt or faming 60% +Pension and 0Renittancesfmmabmad dhability insurance 50% -C-Othersocial 40% Other socialtmnsfers transfen 30% - - OPension anddisability 0 Renittancerfrom abmad 20% insurance +Net incomf" 10% IBITenpomy enploymnt smllshopor nr, faming Consunptwn in 1990 2000 0Regularenploymnt 1990 1994 1996 1959 zoo0 kind Notes: HBS data without Kosovoand Metohija, 1996 and 2000 data for the first six months; all values are deflated usingthe cost of the minimum consumptionbasket. 1Sources: FSO, Statistical Yearbook and Bulletin, FSOUpdates. BOX I. 2: InformalSector I Informal sector issues have achieved great prominence in the public debate in Serbia and Montenegro. Although an informal market existed in socialist Yugoslavia and remains a common feature o f all transition economies, several factors have reinforced it existence in Serbia and Montenegro. These factors include: the reduction in state employment, a distrust of state and private banks, the high taxation on income, and the high contributions to social and health insurance, with the lack o f enforcement in the legal and taxation systems. Widespread criminality by paramilitary units associated with the regime began with wartime looting and continued with smuggling and racketeering. The relative absence of the rule o f law allowed for state actors to utilize public sector enterprises for private profit. The size o f the informal sector is notoriously difficult to measure with any precision. Although a large literature exists, the subject is still quite controversial and there are disagreements about the definition of shadow economy activities, the estimation procedures and the use o f estimates in economic policy. Krstic (2002), in the most comprehensive study on the subject in Serbia and Montenegro to date, points to a number o f contradictions in the different methods used to assess informality in other countries and their limited applicability to the FRY context (currency approach, energy consumption and national accounts statistical discrepancy method). Krstic also dispels some of the common perceptions o f the exceedingly large size of the informal sector in Serbia and Montenegro. For example, it is inaccurate to consider many o f the typical informal sector activities, such as subsistence agriculture or street trade, as additions to the officially measured GDP, as it incorporates estimates of such activities. The most accurate estimate produces a magnitude of around a quarter o f total GDP (35 percent of the officially measured GDP), which is still very significant but is close to the levels observed inmany SEE and even some EUcountries. Source: Krstic (2002); Bemabe UNDP, Human Development Report: Yugoslavia 1997; Schneider and Enste(2000) 7 1.19 Conclusion: Failed facelift. Living standards in S A M were kept high artificially through the running down of the capital stock o f the country, the unrealistic pricing policies of public services and basic food products, a tolerance o f the accumulation of arrears to state enterprises and funds, a tolerance o f the growing informal sector of economy etc. Finally, an important role was played by remittances and pensions from abroad, which were a major source of a survival strategy for many households. All of these factors led to unsustainably high overall levels of consumption, but they didnot help prevent the decline inliving standards. 1.4. Outcomesof Reforms 1.20 The initial conditions were extremely difficult. The legacy of the regime and its policies from the 1990s created difficult initial conditions for the new reformist governments - very low level of living standard, accumulated macroeconomic imbalances, a devastated infrastructure, almost stopped industry, shortages o f basic goods and energy, etc. In breaking with this difficult legacy, the Federal and Republican governments have followed a two-pronged approach combining stabilization measures with structural reforms aimed at roviding basis for the short-term recovery o f the economy and for long-term sustainable growth. !? 1.21 Macroeconomic stability was achieved. After the initial spontaneous liberalization o f the previously tightly controlled prices of public services and basic goods, retail prices rose by a cumulative 48 percent in October-November 2000, bringingthe inflation in 2000 to 115 percent (year end). The National Bank of Yugoslavia (now the National Bank o f Serbia) during the previous two years operated with a tight monetary policy, which resulted in a broadly stable exchange rate, a decline in inflation (to 39 percent at end - 2001, and further to 14.2 percent at end - 2002) growth of official foreign currency reserves (from USD 524 million at end - 2000 to US$2.3 billion at end 2002), a reduction in inflationary expectations and an increase in private savings. 1.22 Although inflation decreased over time, there was an important one-time increase in the cost of living owing to the removal of the disparities in prices for basic goods and services (such as electricity, transport and some basic food staples). The exchange rate policy, which kept Serbian currency almost at a fixed level for two years, resulted in a real appre~iation.~ 1.23 The tight monetary policy, aimed at macroeconomic stabilization, was supported by deep changes in public financial management and other aspects of fiscal policy. The cash-based general government fiscal deficit (before official grants), however, increased from 1percent of GDP in 2000 to an estimated 4.3 percent in 2002. This was because the expenditure savings from the structural reform program were initially outweighed by the more realistic budgeting o f commitments (i.e., lower accumulation o f arrears) and the bringing on budget o f some quasi- fiscal activities (e.g., through higher energy prices and explicit subsidization which contained the losses of state-owned power companies). More recently, the deficit has been affected by an increase in debt service payments. These tighter macroeconomicpolicies were supportedby IMFprograms (Stand by Arrangement approved inJune 2001, and ExtendedArrangement approvedinMay 2002) and with large donor support. Since Montenegro adopted the euro as the legal tender at an earlier date, the National Bank of Yugoslavia was responsiblefor monetarypolicy only in Serbia (excluding Kosovo). Table 1inAnnex Ishows the change inthe current cost of fixed subsistence basketsbetween2002 and 2000. 8 1.24 Better fiscal management positively affected the welfare of the population. Better functioning of the fiscal system in Serbia i s reflected in increasing tax revenues with lowering tax rates and increased regularity in the servicing of all obligations to public funds recipients: salaries, pensions, repayment of old public debts to citizens (from loans for Serbian economic rebirth, non-paid pensions from the 1990s, frozen foreign currency savings, etc.), payment of the costs for repairing the electric power supply system, payment for procured agricultural goods, road reconstruction, etc. A reduction in certain taxes has had an important direct impact on the living standard. The most important impact has been the lowering o f the payroll tax and the personal income tax, and the elimination of retail taxes on basic food products (meat, sugar, edible oil, etc.). 1.25 Progress in structural reforms has been impressive. During the first two years, the National Bank o f Serbia took important steps to restructure the financial sector, especially the banking sector. Some o f the main measures were: closure of the four largest loss-making banks, changes in legal bases, privatization of some banks. As a result o f the improved environment, a few foreign banks entered the market and some local banks merged. This allowed private foreign currency deposits to grow (from US$365.8 million in December 2000 to US$639.5 million in December 2001, and further to US$783.3 million in December 2002), and consequently made access to credit easier. 1.26 However, much important action on the structural front is still needed. Competition and hardbudget constraints were made more prominent on the Serbian market owing to a more liberal foreign trade regime, the liberalization of domestic prices, and accelerated privatization. Nevertheless, certain monopolies still enjoy artificially highmargins. 1.27 The chosen method o f privatization-a combination o f direct sale through auctions and tenders-guarantees--also benefits from this process to insiders, since 30 percent o f the shares go automatically to the employed. 1.28 The reform has begun to bring benefits. The real GDP of S A M has rebounded from a decline in 1999, to a 5.5 percent growth in 2001 and a growth o f around 4 percent in 2002. Industrial production showed only a modest recovery of 1.7 percent in 2002, after stagnation in 2001. Services (especially trade) have recorded faster growth rates inthe last two years; followed by agriculture. Growth rates for agriculture largely depend on weather conditions and are also affected by the fact that that the level o f irrigation i s very low and there is low use of fertilizers and other factors of modern agriculture. The growth o f retail trade i s due to liberalization, to the growth of personal incomes, and to the restructuring o f the banking sector, which provides more credits to both trading companies and consumers. It can be observed worldwide that the growth in agriculture and inthe service sector is generally pro-poor, fueling consumption and the micro enterprise sector. 1.29 Wages in Serbia and Montenegro grew inboth nominal and real terms duringthe last two years (real growth was 13.3 percent in 2001 and 24.6 percent in 2002). This growth was fairly equal across sectors, with somewhat bigger increases in the financial services and the public sector (public administration, education andhealth). 9 1.30 Thus, economic developments in SAM have had straight repercussions on household welfare: the incomes of those in employment have increased, and the benefits o f transfer recipients have been strengthened with one-off repayments of arrears. Inaddition, as described in Box 1.2, there has also been an important informal part o f the economy not covered by official statistics. Hence, aggregate macroeconomic developments suggest that there was some reduction inpoverty. This is indeedwhat data are showing. 1.31 Comparing poverty based on survey data in2000 and in 2002 i s not easy, as explained in Box 2.1 (see next Chapter). Only some comparable data exist, but these datasets for 2001-2002 are not available outside the Federal Statistical Office. Table 1.1 puts together rough estimates for 2000-02 based on the staff analysis o f published sources. These are necessarily very approximate figures. For 2002 more accurate data and improved methods to set the poverty line were produced based on SLS, these improved data are described in Chapter 2. Figures in Table 1.1are not intendedtherefore to provide an accurate assessment of poverty for 2002, butroughly assess its dynamics compared to 2000. It i s also very important to note that the analysis o f PRSP team reported in Krstic (2003) and in the draft PRSP shows more dramatic improvement in the living standards between 2000 and 2002: the incidence of poverty measured based on the official MinimumFood Basket has dropped from 36.5 to 14.5 percent of the population. Though Krstic (2003) i s basing her estimates on survey data, unlike figures reported in Table 1.1, she compares data obtained with HBS for 2000 and SLS for 2002, - i.e. juxtaposing two surveys with very different design and coverage. It i s therefore unclear whether a reduction in poverty rate from 36.5 to 14.5 i s a result of economic development or a product of improvement in data quality. Figures in Table 1.1 represent more conservative estimates. But the fact that two different estimates (reported in Table 1.l.and from Krstic[2003]) point in the same direction suggesting clear progress inpoverty reduction, is important. Table 1.1:Poverty inSerbia and Montenegro in2000 and 2002 (based on the WFP poverty lines and per capita expenditures) Total Percent Number Total Percent Number Population poor of poor Population poor of poor Serbia 7,747,000 12 936,000 7,498,000 9 643,000 IDPsand Refugees* 723,000 25 181,000 608,000 19 113,000 Montenegro 654,000 11 72,000 663,000 8 59,000 Serbia and Montenegro 9,124,000 13 1,189,000 8,769,000 9 809,000 Sources: Figuresfor 2000 are > Public administration reform Regional and international cooperation. 1.43 The implementation of these policies should lead to the creation of a better business environment that would help to accelerate the growth o f the existing private sector and would encourage the start of the new private companies--mainly small and medium enterprises. Also if such policies are implemented, more foreign direct investment would be attracted and the regional importance of the economy would be raised. 1.44 In particular, the Stabilization and Association process creates an important anchor for policy and boosts its credibility. Favorable trade regime granted by the EU will stimulate exports. In addition, significant resources made available as part of this process will help to improve the governance and business environment and attract FDI. Table 1.2: Key Economic Indicators and Growth Projections for SAM, 2000-5 Realannualper capita growthrates(%, 2000 prices) Grossdomestic productat marketprices 5.0 5.4 3.9 3.9 3.9 3.9 Gross Domestic Income 5.6 4.0 4.0 4.0 Total consumption 12.4 7.5 5.1 1.1 1.4 Private consumption 14.2 6.3 3.0 2.0 2.3 Services 100 100 100 100 Nationalaccounts (as % of GDP) Total Consumption 102.7 107.3 107.0 103.5 101.2 99.4 Gross domestic investment 14.2 13.6 16.1 16.5 17.2 17.8 Government investment 3.1 1.6 3.4 3.4 4.0 4.4 Private investment 11.1 12.1 12.7 13.1 13.2 13.4 Exports (GNFS)" 29.6 23.7 20.7 20.1 21.9 23.1 Imports (GNFS) 46.5 44.6 43.8 40.1 40.3 40.3 Gross domestic savings -2.7 -7.3 -7.0 -3.5 -1.2 0.6 Gross national savings 10.3 9.1 7.2 7.6 9.4 11.2 PublicFinance (as % of GDPat marketprices) Current Revenues and Grants 39.6 43.6 41.2 41.7 41.0 Current and Capital Expenditures 40.2 47.5 45.1 45.4 44.5 Consolidated Public Sector Balance -0.6 -3.9 -3.9 -3 7 -3.5 GNPper capita (US$, Atlas method) 990 Gross Domestic Income 5.6 4.0 4.0 4.0 Note :excluding Kosovo Source: Staff estimates based on RSMX model. Source: World Bank (2003), Annex 2. 1.45 Projections from the Public Expenditure and Institutional Review (World Bank 2002c) show that with the medium term constant yearly growth rates of 4 percent, the sustainable deficit 13 i s only about 1percent of GDP. Projections in Table 1.2 confirms that even with strong effort, SAM's macroeconomic sustainability will remain fragile and vulnerable to external shocks. Any slippage could put the country at additional risk and impair creditworthiness prospects. To sustain at least 4 percent growth rates the sustained progress in fiscal consolidation will be particularlycritical. Increasedinvestment mainly depends on increasing domestic savings. Lower savings and investment, although allowing higher consumption, will not only compromise future growth rates but will also reduce the chances of export recovery. And lower than projected exports can also make the maintenance of macroeconomic stability and the servicing of foreign public debt more difficult. But even putting this requirement aside, there i s little doubt that productivity has to grow, and that this can be achieved only through investment. 1.46 The key factor as outlinedby the projections is therefore a reduction of consumption as a share of GDP to allow higher investment. And this i s a key factor that influences the outlook for the living standardandpoverty. 1.47 The importance of growth is further highlighted by the extremely tight budgetary envelope. As illustrated by Table 1.3 for the case of Serbia, direct poverty alleviation programs are too small in their size to have a definitive impact on poverty. These are also projected to remain at their current levels according to the PRSP (see on this in Chapter 6). This data also suggest quite a limitedscope for the redistribution in favor of the poor through the budget. Table 1.3: Consolidated Social Expenditures inGDP, Serbia Consolidated public expenditures, total 39.5 45.9 45.9 Housing construction and utilities 2.6 3.6 3.4 Health care 4.6 5.2 5.3 Education 2.7 3.4 3.8 Social protection, ow: 14.5 17.6 18.0 Pensions 11.0 13.4 13.0 Child welfare 1.4 1.3 1.3 I Financial assistancefor low income families 0.3 0.4 0.4 I Source: MainChapters are from PRSP, detailed breakdown o f social protection expenditures i s from Ministry of Social Affairs for child welfare and social assistance data, Ministry o f Finance and Economy, Memorandum on the budget and the economic and fiscal policies for the budget year 2003 and 2004. Note: Consolidated public expenditures include all expenditures at central level (the Budget of Serbia, Republican Pension and Disability Insurance Fund, Republican HealthProtection Institute and Unemployment Insurance Fund) and at the local level (autonomous provinces, cities and municipalities); *estimated 1.48 Main factor that influences the outlook for the living standard and poverty i s therefore the economic growth, it patters and its impact on the poor. 14 1.7. Poverty Projections 1.49 The economic projections underlying the development strategy of Serbia show that the growth of private consumption in 2003-06 may vary between 2 percent per year and 2.3 percent per year (Table 1.2). The poverty reduction that can be achieved in the future will follow directly from this consumption growth (Table 1.4). Table 1.4: Poverty Projectionsfor Serbia, 2002-6 (percent poor inthe population) 2002 2003 2004 2005 2006 ~ GDPgrowth rate 4.0 4.0 e 4.0 4.0 4.0 Growth rate o f private consumption 6.3 3.0 2.0 2.3 2.3 Optimistic poverty projection, pro-poor growth a 10.6 8.9 8.0 6.9 6.3 Baseline poverty projection, equally shared growth 10.6 9.4 8.7 8.1 7.5 Pessimistic poverty projection, pro-rich growth 10.6 10.0 9.7 9.2 8.9 Notes: a Improving distribution (Gini index falls by one percent per year, consumption o f the poor grows by 17 percent against 10 percent on average). Distributionally neutral growth (consumption o f an average poor grow as fast as the average, 10 percent over the 2003-06); Worsening income distribution (Gini index increases by one percent per year, consumption o f the poor increases by only 2 percent over the entire period versus 10 percent average); actual; e in 2003, real GDP is expected to grow by about 3.5 percent, with the slight slowdown relative to 2002 reflecting the effects o f drought and a modest fiscal contraction. Source: Staff estimates based on SLS data and Government Serbia (PRSP 2003). 1.50 Ifprivateconsumption increasesat the impliedrate of 2 percent, poverty will fall by the year 2006 (PRSP time horizon) to 7.5 percent of the population, and this change, if any, will probably be difficult to measurewith the data at hand. Only when growth i s combined with more than proportional gains to the poor i s noticeable poverty reduction (to the headcount index of 6.3 percent) possible. Different growth paths offer dramatically different benefits for the poor. If the distributiondeteriorates (poor will lag behindthe middle in terms of gains from growth), poverty will fall by only 1.5 percentagepoints by 2006, and this change, if any, will be difficult even to measurewith the data at hand. 1.51 Indiscussing the strategic objectives of poverty reduction andthe means to achievethem, it is worthwhile separatingtwo deeply related, but still distinct, issues: 0 How to maximize growth potential while ensuring its economic sustainability 0 How to maximize the benefits of a given growth rate for the poor (i.e., how to ensure its pro- poor character). As the statistical confidence interval (95 percent) around the measured poverty rate for the SLS 2002, the largest household survey in Serbia that provided baseline data for PRSP, is k1.2 percentage points. 15 1.52 Figure 1.6 explores both dimensions for various long-term growth scenarios.6The figure first shows what kind of difference growth will make for poverty: 2 percent growth rate of consumption will cut the current poverty incidence in half by 2015, and 6 percent annual growth rate of private consumption per capita will eliminate absolute poverty completely. Figure 1.6: Poverty RatesinSerbia (2002-2015) with Various Growth Ratesof Consumptionand Ineaualitv -No growth -t- 2% p a. growthof consumption anduorsenigndistrbution(by 1% p a ) A 2 % p a wthunchmged lnquallty -+- 2% p a pro-poorgrowth - - - P6 percent (dlstnbution lnproves by 1% - a) dstributtonallyueuval growth 5 percent dstnbutiody neutral g r o - ~ Note: Projections assume either that consumption rises for all of the population with the same growth rate, or a combination annual average growth rate o f 2 percent and increasing I decreasing inequality by 1percent per year. Source :Staff estimates using SLS for Serbia and baseline poverty line (see Chapter 2). 1.53 Figure 1.6 extends the time horizon of projections beyond 2006. This is necessary. The only reason why the consumption growth rates are below GDP growth rates in the period 2003- 06 i s to allow the foundations for future growth. It i s important to asses the kinds of benefits that such strategies may bring beyondthe PRSPtime horizon. 1.54 The figure highlights the importance of achieving the pro-poor character of growth, which i s the only way to maximize the benefits of a relatively modest expected growth in consumption for the poor. It should be noted that baseline growth process will be pro-poor in a simplistic sense (see Box 1.3 for a discussion): consumption for the poor will grow as much as the average person's consumption. With this assumption, poverty incidence will be halved between 2002 and 2015, thus the key MDG goal will be achieved. But when growth is pro- poor in a narrow sense (i.e., the poor grow faster than the rich) that faster poverty reduction i s possible. Taking2 percent per year as an baseline growth rate of consumption, it can be seen that the pro-poor growth, in which the poor will participate with slightly more than their current share in the national "pie," will achieve a reduction in the poverty rate by 70 percent by 2015. We use LSMS data 2002 as the basic tool. To project the impact of economic growth with constant inequality, we simply multiply everyone's consumption by a given growth rate and compare the simulated level o f per capita consumption with the constant poverty line. To simulate the effect of a given percentage change ininequality (measured by the Giniindex), we use the following simple representation for the proportionate shift inthe Lorenz curve for the distribution of per capita consumption Y*~,,by calculating a value o f p, such as y*i =(1+ p)[( 1-a) yi curve: a new level of per capita consumption y for household i to get a l+a change inthe Gini, with the Lorenz + p(yi )], where p is the mean. The new Lorenz curve i s then givenby L*(p*)=( 1-a)L(p)+a, and the change in Giniis Gini(y*)=(1-a)Gini(y). 16 As slight worsening of the distribution will slow down the reduction in the poverty rate by half to 34 percent. It should be noted that the resulting increase in inequality in the case of such worsening will be quite moderate, with the Gini index for per capita consumption reaching a level of 0.35, a level observed in several economies of CEE (see Milanovic (2003) for international comparison of inequality and Chapter 2 of this report). 1.55 Finally, an important message from Figure 1.6 i s that poverty will remain a concern in Serbia for a long time to come. Even with pro-poor growth, the incidence of poverty that Serbia will have by the year 2015 will be higher than that which Croatia had in 1998 (around 4 percent). Poverty will remain high inthe regional context beyondthe PRSP completion. BOX 1.3: What Is Pro-Poor Growth? The relationship betweeneconomic growth andpoverty has been studied extensively yet has remainedin debate. Most of the evidence suggeststhat there is a strong correlation betweengrowth and poverty reduction. A recentstudy by Dollar and Kraay (2000) has attracteda lot of attention. Using the cross-country regressionon a sample of 80 countries over four decades, the study has arrived at the conclusion that positive economic growth benefits the poor to the same extent that it benefits the whole economy. Many researchers and policy analystshave concludedthat if the economy grows, the benefits will trickle down to the poor. This is pro-poor growth inabroad sense: any growth is good for the poor if they gain. There havebeen many dissenting voices. It was shown that the measureof the gain to the poor -the growth rate of the lowest quintile - may give distorted and inconsistentresults hiding an increaseinthe severity of poverty or even poverty itself. Most important, researchers havenoticed that cross-countryregressionscan only depict the averagepicture. Individual country experiencesof the relationship betweengrowth andpoverty are by far more complex. Ravallion and Chen (2002) and Kakwani et al. (2003) call for a more disaggregatedapproachto poverty, focusing on the issue of inequality. Pro-poor growth is about the interrelation of growth, inequality andpoverty, which i s clearly atrue inter-relationship, where each side plays an active role. Ravallion and Chenproposedan idea of a "growth incidence curve" showing how the growth rate at a given place inthe distribution (percentile) varies across the whole distribution. They proposedto use the meangrowth rateof the poor calculatedwith this methodology as a measure of pro-poor growth. An alternative conceptualization was proposedby Kakwaniet al. with the methodology of using the poverty equivalent growth rate, which takes into account not only the magnitude of growth, but also how much benefit the poor receive from growth. The idea i s rather simple. Taking apoint inthe past one may project the evolution of poverty under the assumptionthat inequality is stable (Le., that the poor will gain as much from actual growth as the averageperson). This change inthe poverty rate may be comparedwith the actual dynamics of poverty to see whether growth is more or less pro-poor. Alternatively, one may calculate how muchgrowth is needed to achievean observedreduction of poverty with the initialpattern of distribution. Both approaches (Ravallion andKakwani) provide useful insights into individual countries studied (Korea, Thailand, and China), but they require access to householdlevel data as comparableover time as possible to (i) set the benchmark, (ii) performanceover time. The simple approachby Dollar and Kraay (2000) may rely on monitor aggregatepublished data, but a true assessmentof whether the growth is pro-poor requires access to survey data. Source: Dollar andKraay (2000), RavallionandChen (2002), and Kakwani et al. (2003) 1.56 Conclusion. The key challenge for poverty reduction in the future i s very tight path dictated by macroeconomic factors. Under the current scenario of 4 percent growth of GDP and stable income distribution, poverty will change noticeably only if pro-poor character of growth i s achieved over 2003-06. If the country i s to eradicate absolute material poverty by the year 2015, then: (i) policies must deliver a higher GDP growth rate, or (ii) poor have to benefit from the growth at least at the rate of the GDP increase, if not more, or (iii) a combination of these two 17 positive outcomes must take place. The remainder of this report aims to answer the question of how to achieve these results, focusing on the second aspect: how to ensure that the poor benefit handsomely from growth. 18 2. SERBIAAND MONTENEGRO:POVERTYIN2002 This chapter describes the poverty situation in Serbia and Montenegro on the basis of the most recent data collected in households surveys in 2002. The analysis starts by setting the criteria for defining poverty and its key dimensions. Based on these definitions, specific to each Republic, the report assesses the overall extent of poverty in each dimension. The household level data are then used to analyze the profile of poverty by moving from regional correlates to household characteristics (related to asset ownership, market participation and social inclusion), and further to intra-household issues (such as age and gender). In full accordance with the diagnostics presented innational PRSPs, the chapter finds that the extent of poverty in S A M i s significantly lower than previously estimated on the basis o f deficient data. But many people are found to be concentrated just above the poverty line. The profile of the poor suggests that poverty i s related to long-term factors, such as education and regional disparities, and that therefore reducing poverty will be difficult. T o address the deeply entrenched causes o f poverty, a comprehensive cross-sectoral development strategy i s needed. 2.1. Serbia and Montenegro: Poverty and Poverty Lines 2.1 This report is mostly (but not exclusively) about material p ~ v e r t y .The focus is on ~ material poverty not because other aspects o f well-being are less important. Rather, this choice i s dictated by two considerations: (i) material poverty empirically i s found as the key determinant o f other aspects of welfare, such as health, longevity, and self-esteem; and (ii) measures o f non- income poverty are not well developed. In addition to baseline indicators aimed at measuring material poverty, this report presents data on those non-material dimensions o f poverty that were possible to compile. 2.2 To measure material poverty, the key i s collecting accurate household data. Box 2.1 describes features o f the surveys (SLS in Serbia and ISSP survey in Montenegro) that were conducted in Serbia and Montenegro in 2002 and gives reasons why they are believed to be superior to the data from the official Household Budget Survey. Based on the collected data, the welfare indicator was developed to assess the living standard of each household. In both Republics the decision was to use total consumption, rather than incomes. Consumption i s viewed by the local researchers as more accurate from a data collection standpoint. First, income The term "income" poverty i s sometimes usedas a synonym for material poverty. As discussedbelow, it does not imply the use of income rather than consumption to measurepoverty, but rather correspondsto the usual meaning, of this term to distinguish it from other, non-income dimensions of poverty (see Laeken EuropeanCouncil Resolution). 19 i s highly seasonal in the rural areas. Second, it may be difficult to get accurate income reports from households, given the large gray (informal) economy. Under-reporting of incomes will lead to an over-estimation of the poverty rate. In addition, consumption i s a better measure than income for reflecting the level of the living standard of a household from a theoretical perspective (see Deaton andZaidi [2002]). 2.3 The consumption aggregate included all current consumption expenditures (no investment, business or durable purchases), the imputed value of in-kind food and non-food consumption based on local prices, the imputed flow of services for durables and the imputed rents for owner-occupied housing based on the information on local rental markets (this latter element was innovative and i s not used elsewhere to date). A special price index was developed to account for regional price variations. Finally, to account for differences in needs of different demographic groups, an equivalence scale was developed (see Krstic [2003]). 2.4 Using the consumption data, the team developed a subsistence minimum (poverty line) which i s internally consistent with the welfare measure itself. Defining this minimum, the team adopted simultaneously relative and absolute approaches. The poverty line in an absolute sense i s the level of consumption for people at which they are able to buy the goods necessary for their survival. The poverty line can also be definedinrelative terms as some proportionof the mean or the median income or consumption. For example, the European Commission defines the European poor as all people whose consumption expenditures or income falls below 60 percent of the median in the country where they live. 2.5 These two concepts are far from being equivalent, even though they may well coincide at some point of time in a particular country. * The debate on this issue has been going on for a very long time among economists - see, for instance, Atkinson (1998) or Ravallion (1992), and i s not settled. However, there does not seem to be a strong reason why it should be settled in favor of one or the other alternative; these concepts are simply aimed at describing and analyzing different issues. Physical poverty i s about the capacity to buy food and all the goods necessary for the fulfillment of basic physical needs. Relative deprivation describes a social phenomenon that may be considered as beingvery close to the more modem concept of "social exclusion." 2.6 Relative and absolute poverty therefore are not exclusive concepts. They describe different conditions and may in some instances call for different policies. If physical poverty i s very widespread in a country, growth-enhancing policies must be given high priority. If physical poverty i s less important but if social poverty and therefore "social exclusion" i s found to be excessive, then redistribution policies are called for. 2.7 The absolute poverty lines are intended to measure the extent of absolute deprivation according to the standards of a particular country. The poverty lines follow the consumption patterns of the population and are basedon objective evidence from the survey, not onjudgment. To anchor absolute poverty lines in the economic reality of each country, the report team * See Bourgignon (1999). This is the "linen shirt" argument that Sen (1983)borrowed from Adam Smith. Ineighteenth century England, a peasant unable to afford to wear a linen shirt would not participate in social events inhis village, even though he and his family might not be undernourished. Inthis sense, poverty arises any time an individual cannot afford doing, or "functioning" inthe words of Sen as "most" people do inthe society in which helshe lives. 20 adopted an approach proposed by Ravallion (1992). This approach consists of establishing the local cost of the minimumfood basket that meets key nutritional requirements and follows the consumption patterns of the populations. The cost of this minimum basket can be used to measurethe extent of extreme poverty." To establish the full, baselinepoverty line, the report uses the survey data from each country to find out at which level of total consumption (Le., both food and non-food) the minimumfood requirements of all household members will be met by the actual food spending of the household (see Krstic [2003] and Annex Ifor details). This poverty line has a very clear meaning. All households with total consumption per equivalent unit below the poverty line suffer from material deprivation: either they lack adequate nutrition, or, to meet basic food requirements they must cut back on other essentialneeds. Table 2.1: Serbia and Montenegro: Overview of Poverty Lines, 2002 (all values are expressed& per month dinars andeuros) Values ConceptandDefinition (m)Montenegro Serbia (Euros) Extreme poverty: consumption (without imputed rent)per capita below 1,901 41 Absolute poverty (baseline): full consumption per eq. unit* below 4,489 107 Vulnerability to poverty (upper): consumptionper eq. unit* i s below a thresholdof 6,733 161 Vulnerability to poverty (lower): consumptionper eq. unit* is below a thresholdof 5,507 132 EURelative poverty: below 60% of the medianequiv.** income without imputations 5,497 166 ECA Relative poverty: full consumptionper equiv unit*** below 50% of the median 5,652 164 Notes: *National equivalence scale (For Serbia: scale gives the weight of 1 to the first adult ,0.81 to all subsequent adults, the weight of only 0.24 to each child under 6 years and 0.75 to all other children under 18; for Montenegro: per capita ). ** EU equivalence scale: 1.0 to the first adult, 0.5 to other householdmembers aged 14 or over and 0.3 to each child. The resulting figure is attributed to each member of the household, whether adult or child.*** ECA equivalence scale: householdsize in the power of 0.75. Sources: Staff estimatesand datafrom SLS (2002) andHHS ISSP(2002). 2.8 To a significant extent this approach to setting the absolute poverty line i s relative: poverty i s defined on the basis of the reality of each country and of what i s revealedby a survey. These needs are obviously higher if the level of development of a country i s higher. 2.9 The relative line has no objective anchor. One standard, to set the relative line at 50 percent of median consumption, was adopted by the World Bank report for ECA, "Making Transition Work for Everyone" ([World Bank 2000 a], lower line). To make comparison with other countries in Europe, another standard developed by EC i s 60 percent of the median. Inall cases, an appropriate equivalence scale was used to account for differences in household composition. lo theWorldBankmethodologyasoutlined Using inRavallion (1992),food poverty line is compared to total per capita consumptionto estimate extreme poverty. This methodology provides for international comparability of extreme poverty data. I t is based, however, on insights from poorest countries where the number o f extremely poor allows in-depth assessmentof measurement assumptions and does not coincide with nationally specific definitions o f extreme poverty used inSAM. 21 BOX 2.1. OldandNew Approachesto MeasuringPovertyinSerbia The work on poverty inthe FRY dates back to the 1980s ,when the topic was new inthe world literature. Inthe work by Posarac (1991) for the World Development Report of 1990 and Milanovic (1991) the HBS survey data from Yugoslavia were used for the first time to describe poverty incidence by republics and the profile. The latest attempt to apply this methodology is very recent (Bogicevic et al. 2001). The first chapter of this report relied on these series to describe the evolution o f living standards. However, in these latest publications, the limitations o f this methodology for measuring poverty were discussedindetail. Specifically it was shown that: HBS data did not offer accurate measuresof welfare The FSO food basket was not an adequatepoverty line. Yugoslavia established a representative household survey (HBS) conducted quarterly by the FSO since 1984. The sample was based on the latest census available; therefore, for 1990-2002 it did not include refugees and IDPs. The food subsistence basket developed more than a decade ago (65 items) by the FSO was anchored to the actual diet of a consumer in the lowest decile, but it was significantly "blown upward" to meet "rational" rather than minimal criteria. The tests o f its adequacy presented in Krstic (2003) show that it significantly differs from actual consumption patterns. In parallel with the basket an equivalence scale was developed that takes into account the difference in needs o f families o f different sizes and composition in meticulous detail including types of job o f adults, etc. The equivalence scale, however, was anchored exclusively to food needs and therefore i s inappropriate to measure the overall extent o f welfare. Despite these shortcomings, these data remained for a long time the only information on living standards available. And given that the primary HBS data were not accessible to researchers outside the statistical office, they were also hard to improve. In 1999, the WFP developed an alternative poverty line. It was anchored in the minimum caloric intake (differentiated for summer and winter months), included allowances for non-food items and was based on the actual consumption patterns o f the poorest 40 percent of the population. The WFP monitored the current cost of this basket in 1999-2002. The disadvantage of the WFP methodology is its reliance on a per capita equivalence scale, an approach that makes it impossible to correct for economies o f scale. The new stage in the debate and analysis o f poverty began with the Poverty survey in initiative in 2001-2. The new survey, SLS (A%, Anketa o zivotnom standardu) in Serbia and HHS survey in Montenegro were different from HBS inmany respects: The survey sample was updatedusing the most recent Census (Serbia) or listing operation (Montenegro) The survey instrument was developed inan open participatory process involving key stakeholders The survey instrument allowed the capture of non-income dimensions o f poverty The instrument allowed for collecting detailed and accurate data on consumption Expandedlabor market modules solicited more detailed informationand informal sector activities The survey collected detailed informationon pension and non-pension social benefits The survey instruments recognized informal payments for public services 0 The protocol of the surveys provided for data quality These data will be made publicly available to use their full potential. This report adoptsthese new data and poverty lines as baseline data on poverty. Although we believe these data are more accurate than the previous numbers, the advantage o f having this information for this report comes with a cost. The new data are totally incomparable with the pervious HBS data. The issue is not simply the application o f the right poverty line or the correct welfare aggregate. It i s also the quality and comprehensiveness of the primary data themselves. Certain choices and assumptions had to be made to make them usable. Usually the validity of these assumptions is tested in use over time. The SLS round data is the first, pilot use o f this methodology. Given that in future applications some details may be revisited and improved, it may mean that the poverty analysis apparatus presented in this report is not carved in stone and may be revised later on when the practice o f applying these methods will reveal their strengths and weaknesses. Thus, the profile and certain findings may be revised in the future. Source: Krstic (2003) 22 2.10 Finally, subjective perceptions may be used to capture subjective deprivation. Subjective poverty i s an important complementary source of poverty data that i s extensively usedto analyze various determinants of well-being and social mobility (Milanovic [2000]). The indicator i s obtained by comparing actual income or consumption as reported by a household with the household-specific minimumthreshold that is considered a poverty line (Annex I). 2.11 Material deprivation i s the conventional objective measure of poverty. However, it does not necessarily accord well with people's own assessment of their economic welfare, as Ravallion and Lokshin (1999) demonstrate. They find that many non-income factors influence the subjective poverty. Adding other explanatory variables (such as educational attainment, health status, average income in the area of residence, and employment status) to household income substantially improves the ability to predict people's assessment of their economic welfare. Health and education are important aspectsof well-being independentof their impact on income levels. Being unemployed or being afraid of losing one's job also lowers self-rated welfare, even controlling for householdincome. 2.12 The first step in going beyond a static measure of who i s below a poverty line at a given moment of time i s to look at the changes in poverty status over time. Poverty, absolute or relative, would not be a problem if it were known to be purely transitory, that is, limited for all poor to very short periods of time. This means that the only measuring physical poverty and social deprivation i s insufficient. They must be made dynamic. With such a perspective, the link between poverty analysis andthe concept of inequality appears more clearly. 2.13 With an inter-temporal definition, absolute poverty would correspond to a situation in which an individual would remain permanently, or at least for a very long time below some physical poverty line. This might be for two reasons: (i) limited income mobility, and (ii) slow economic growth, which would render the prospect for poor people to cross the poverty line more distant. It may be shown that such an inter-temporal definition of absolute poverty (i.e., how many people are below the poverty line and for how long) i s not independent of the initial importance of relative deprivation, or more generally of inequality. Kakwani et al. (2002) and Chen and Ravallion (2002) notedthat under the assumption of a constant distribution of relative incomes and a constant rate of growth it would take more time to eradicate absolute poverty in a more inegalitarian economy than in an economy with lessrelative deprivation. 2.14 Panel household data would allow a direct measurement of chronic poverty and studies of its profile. Such data will become available in Serbia later in 2003 with the second round of SLS. At the time of the writing of this report, this measurementwas not yet possible. 2.15 Different aspects of poverty - income and non-income - interact and reinforce each other in ways that often exacerbate the deprivation that poor people face. Poor health outcomes and low educational achievement not only decrease well-being, but also limit people's income- earning potential. Lack of voice and participation feeds exclusion and biases the distribution of public spending against the poor. Operationalizing this general concept of the multidimensional characterization of poverty the 1998 Eurostat Task Force on Social Exclusion and Poverty statistics defined poverty, as : "... a dynamic process.. which ends up with persistent multiple (deprivation) disadvantages. Individuals, households and spatial units can be excluded from access to resources like employment, health, education, social or political life." (Eurostat 1998: 23 25).' InternationallyendorsedMillennium Development Goals (MDGs) are also related to this concept. 2.16 Defining poverty as multidimensional, however, also raises the question of how to measurethese different dimensions. There are no strict and agreed upon standards that would fit every country. This chapter uses the survey data and other sources of information to obtain individual or household-level measures of deprivation in the following dimensions, national development priorities as reflected in the national PRSPs, to develop the following set of indicators: education poverty, health poverty, housing and citizenship rights poverty, employmentpoverty, and rights poverty (see Table 2.2). Table 2.2: Overview of MultidimensionalPoverty Indicators Concept and Definition Realization Serbia Montenegro Education poverty: Individual is education poor if he or she is above 15 years old, is not in school and has only uncompletedprimary educationor no education at all J Housing and citizenship rights poverty: Households with uncertain citizenship status and lor illegal and temporary occupants, households without documents confirming their ownership of J housing. Health poverty: Individuals of working age who in the month preceding SLS were suffering from a major physical ailmentsprecluding their normal independent functioning J Housing conditions poverty: Households not connected to tapped water and using latrines, households living in a building unsuitable as dwelling or living in partly destroyed home and J * living inovercrowded dwellings with more than 3 personsper room Employment poverty: Lacking social inclusion, defined for working age adults as being not * employedcontinuously for over 2 years, but able and willing to work Sources: Staff estimates, for exact values and details of derivation and definition, see Krstic (2003), Radevic and Beegle (2003) Applied, * appliedwith modifications. 2.17 Among these dimensions, employment poverty-especially its association with long- term unemployment-is particularly important. Unemployment deprives people from a critical means of contributing to and integrating into society. It can also deprive them of access to key social networks which are often necessary to getting a job and to otherwise participating in economic life. This can push people into an unemployment trap and can lead to growing social exclusion and marginalization. Unemployment i s thus a cause of material poverty, but it i s also an indicator of a different poverty dimension. ''See Atkinson et al. (2002). The Laeken European Council in December 2001endorseda first set of 18 common statistical indicators for social inclusion, which will allow monitoringina comparable way of Member States' progresstowards the agreedEUobjectives. These indicators needto be consideredas a consistent whole reflecting a balancedrepresentation of EU social concerns. They cover four important dimensions of social inclusion (financial poverty, employment, health and education) which highlight the "multidimensionality" of the phenomenonof poverty. 24 2.18 Such multidimensional definitions of poverty, although largely based on judgment and prevailing norms, are increasingly gaining prominence in poverty analysis because they help with understandingnew aspects of poverty, such as: a Householdsthatarenotincomepoormaybe poorinotherdimensions The same households may suffer from multiple deprivations, thus constituting "the core" of the poor. 2.19 Like poverty itself, vulnerability to poverty i s also a multi-dimensional concept. In the dimensions of income and health, vulnerability i s the risk that a household or an individual may experience an episode of income (consumption) or health poverty over time. But vulnerability exists in many other dimensions as well. For a child, it may be the risk of being pulled out of school. It may be the risk of beingthe victim of a natural disaster, of crime, or of arbitrary use of power by the state or bribery or violence. 2.20 Indicators of vulnerability reflect both the extent of exposure to shocks and the ability to cope with them. Unfortunately, the SLS data analyzed in this report were not yet a panel, and dynamic measures of exposure to risks at the household level were impossible to develop. Vulnerability in this report is measuredby exposure to economic and non-economic shocks. The lower the current consumption, the higher such an exposure is. As arbitrary cut-offs the report uses two vulnerability lines: the first i s close to the usual definition of vulnerability and i s 50 percent above the baseline poverty; the second i s the narrower definition in which the upper limit of the lowest quintile is set as a threshold (and the incidence therefore is assumed to be 20 percent). This definition has to be revised as panel data become available (late 2003). 2.21 Finally, poverty can also be measured based on perceptions of the population. The report uses SLS data as well as data from other sources to trace the extent and evolution of subjective poverty and perceptions regarding the future. 2.22 Each of these methods has its own advantages and disadvantages, but together they well represent the extent and different forms of poverty and vulnerability to poverty. 2.2. How Many Are Poor in Serbia? Vulnerability and Non-Income Poverty.'2 2.23 The absolute poverty rate is moderate (see Table 2.3 for the incidence of poverty derived using different poverty lines on the basis of the 2002 SLS data), affecting about 10 percent of the population. 2.24 But extreme poverty is non-negligible, affecting 2 percent of the population according to the World Bank methodology. l3Even with limited statistical precision of this estimate the very existence of extreme poverty in SAM i s a surprising finding. In neighboring countries where similar methodology was applied (Bosnia and Herzegovina and Croatia), such ` 2This sections draws heavily on Krstic (2003). l3 The extremepoverty estimatereportedinTable 2.3 is based on the World Bank methodology, 25 poverty has been found to be nil. On the other hand, the rate i s not as high as inKosovo (where extreme poverty was 11percent based on the 2001 national LSMS - World Bank [2002a]) and Albania (4.7 percent according to the 2002 LSMS -World Bank [2003b]). 2.25 Poverty estimates presented in the Table 2.3 must be taken as low-bound estimates of poverty. First, certain groups are not covered by the household survey by design (such as institutionalized population). Such groups, e.g. IDPs in collective centers, may be particularly poor. Second, the survey team made all attempts to cover all strata of the population, but some were extremely hardto reach in a general population survey. Different groups o f Roma minority i s a particular case in point, as well as some DPs in private accommodations. The most recent data collection.round (2003) in Serbia included a special survey of Roma, and its future results need to be carefully assessed to arrive at a more accurate poverty estimate for the country. In Montenegro inclusion of these groups in a recent survey produced estimated o f poverty that are 20 percent higher at the national level. Finally, the statistical confidence interval (95 percent) around the measured poverty rate for the SLS 2002 i s k1.2 percentage points. Table 2.3 reports the exact figure o f 10.6 percent; it would be more accurate to say that poverty in Serbia i s estimated to be between 9.4 and 11.8 percent. Lack of precision may particularly affect the extreme poverty figure. But even 9 percent for absolute material deprivation incidence produced with the existing data tells that poverty i s a very serious problem. Table 2.3: Poverty Incidence Based on Absolute and Relative Lines inSerbia (percent of population and statistical confidence intervals) Absolute Absolute Relative (v2 of Relative (60% Subjective Vulnerable Extreme Baseline the median) of the median) to poverty Poverty rate 2.3 10.6 9.0 20.2 44.9 32.6 Statistical confidence range k 20.5 k1.2 kl.l k1.6 k2.0 kl.9 percentage points Source: SLS 2002. Per capita consumption for absolute poverty line, per equivalent adult consumption (OECD scale) for relative poverty lines. Statistical precision i s estimated as 95 percent confidence intervals with standard errors computed with stratified sampling design (Kish effect) correction. The headcount index measuresthe population whose consumption is less than the poverty line. 2.26 Relative poverty is moderate, 20 percent according to EU methodology (Table 2.3). How do the relative poverty figures lineup against those for other transition countries? Serbia fits inthe upper-middle range of poverty incidence among transition economies: its relative poverty i s significantly higher than that in Bosnia and Herzegovina, the Czech Republic, Hungary, Slovenia, Croatia, Romania and Albania, but lower than that inPoland, Bulgaria, Macedonia, the Baltic States, and CIS c~untries.'~Compared to countries in the EU, relative poverty i s as high as inthe countries with highest relative incidence (Portugal and Greece). ~ l4Using the World Bank study Making Transition Workfor Everyone: Poverty and Inequality in Europe and Central Asia, 2000, we used here consumption as more accurate measure of welfare. Applying a standard European Commission definition o f relative poverty (household's total disposable monetary income with no account o f indirect social transfers, transfers paidto other households, receipts inkindand imputed rent for owner-occupied accommodation), the relative poverty defined based on 60 percent o f the median is found to be 22 percent - higher than anywhere inEUcountries, above Greece and Portugal (each has 21 percent incidence). 26 2.27 As section 2.1 pointed out, the comparison between relative and absolute poverty matters. Inthis respect, both are equally worrying. Not only i s absolute poverty a problem, but relative poverty points to the existenceof significant inequalities affecting the poor. 2.28 So far we have used only the simplest and most common measure of poverty: the headcount index. This measure, however, does not tell us whether the poor are only slightly below the poverty line or whether their consumption falls substantially short of the poverty line. The headcount measure also does not reveal whether all the poor are about equally poor or whether some are very poor and others are just below the poverty line. To highlight the complex distributional aspects of poverty, one measures the depth (measured as poverty shortfall or as poverty gap) and the severity of poverty. The poverty depth measures show how badly off the poor are - how far below the poverty line their consumption levels are. The poverty gap in Serbia using the baseline definition of absolute poverty i s equal to 2.2 percent (s.e. 0.17). A corresponding measureof an average shortfall of 20 percent means that mean consumption of the poor falls 20 percent short of the poverty line. Poverty severity i s a measure closely related to the poverty gap but giving those further away from the poverty line (the poorest) a higher "weight" in aggregation than those closer to the poverty line. Its level in is found to be 0.8 percent (see. 0.07). These data suggest that the depth and severity of poverty are not extreme, consistent with the absence of extreme (food) poverty. Table 2.4: Poverty inSerbia inKey Dimensions: (percent of populationbelow corresponding poverty threshold) Poverty rate Notes Material (consumption) Baseline Absolute Poor 10.6 Extreme Poor 2.3 Non-incomedeprivation: Housingand citizenshiprights poor 8.7 Housingconditions poor 16.2 Healthpoor 3.6 Education poor 17.3 Only adults Employment poor 3.3 Only adults Non-income deprivation (at least I ) 36.5 Non-incomepoor (at least 2) 8.5 Non-income extremepoor (3 and more) 1.3 Combinationof non-income and income dimensions Material poor OR poor inat least one non-income dimension 30.3 Material poor AND at least one non-income dimension 6.7 Non-incomepoor AND incomepoor 2.3 Non-income extremepoor OR income extreme poor 3.4 Non-income extremepoor AND income extreme poor 0.0 Source: Staff estimatesbasedon definitionreportedin Tables 2.1 and 2.2 SLS (2002). 2.29 Multidimensional poverty indicators (Table 2.4) can be combined to provide further insights into the characteristics of the poor and the causes of poverty. The application of the poverty concept to non-income dimensions of well-being using the SLS revealed that poverty spans all aspects of human capabilities: 27 0 An astonishing 17percent of adults incanbeconsidered "education poor" 0 Health poverty affects almost 4 percent of the population 0 Housing conditions for 16 percent of the populationare below the poverty standard 0 Over 3 percent of working age adults are poor with respectto employment 0 Finally, and most important almost 9 percent of the population suffers from a lack of basic rights protection. 2.30 Simultaneous deprivation in several dimensions i s rare, as Table 2.4 suggests. This i s a confirms that extreme poverty inthe country is not widespread. But as poverty consists of non- overlapping, or only partially overlapping groups (see Figure 2.1), a significant share of the population i s poor in at least one dimension. As many as 30 percent of all adults in Serbia are poor in at least one of the poverty dimensions: material consumption, education, health, employment, housing or property rights. Figure 2.1: Overlapping Poverty Dimension 2.31 Five conclusions stem from the application of this approach to measuring poverty in Serbia. 3 Itispossibletomeasurepovertyinitsdimensionsanditcanbemonitored a PovertyinSerbiaexistsineverydimension a Itismostseriousinthedimensionsofmaterialdeprivation, education,housing,vulnerability to poverty and rights 3 Thisstrategytoaddresspovertyhastobemulti-sectoralandaddressmanygroups a Prioritiesaretobedevelopedineachsectoronwhatmattersforpoverty. 2.3. How Many Poor in Montenegro in 2002?International Poverty Comparisons 2.32 Table 2.5 reports the poverty rates for Montenegro based on the comparison of consumption with a minimum standard (poverty line) for the population. The table reveals that consumption poverty affects 9.2 percent of the population, who live below the absolute poverty line and there i s no extreme poverty. This figure may be an underestimate because the sample misses some of the most vulnerable populations (Roma and IDPs). Recently completed UNDP- sponsored special survey of Roma, IDPs and refugees in Montenegro demonstrated that the poverty rate for the entire country may be revised upward by as much as a fifth (from 9.4 percent to 12.2 percent), once these groups are fully incorporated in the survey sample (ISSP- UNDP [2003]). The special section in Annex Iin this Volume (p. 159) provides some details on these new findings. 28 Table 2.5: Multidimensional Poverty Indicators for Montenegro Indicators % of % ofconsumption- population poorpopulation Consumptionpoverty extremepoverty 0.1 (statistically nil) absolutepoverty 9.4 100.0 economically vulnerable 36.4 100.0 Relative poverty ECA standard(50 percentof medianconsumption,eq. adults) 3.7 37.1 EUstandard(60percentofmedianstandard, no imputedrents, eq. adults) 11.0 71.8 Education poverty 16-24years: not in school and did not attend secondary school 4.7 13.5 Health poverty any illnesshnjury inlast 30 days that precludedusualactivities or disabled 6.4 6.1 Housing poverty drinking source for dwelling i s not piped water or dwelling has no bathroom 13.1 25.O Dwellinghas less than 10mzper person 8.2 22.7 Employmentpoverty ages 16-65: not working but ready to work if havejob opportunity 22.3 39.7 Lucking consumer assets No telephone 9.7 14.6 No television 3.7 5.7 No washing machine 7.8 22.7 Source: HouseholdSurvey5 and 6 reportedinBeegleandRadevic (2003). 2.33 In sum, data suggest that the depth and severity of poverty are also not extreme. To highlight the complex distributional aspects of poverty, one measures the depth (measured as poverty shortfall or as poverty gap) and the severity of poverty. For the Montenegro Household Survey 5 and 6 the poverty gap usingthe baseline definition of poverty i s equal to 1.3 percent. A corresponding measureof an average shortfall of 13.7 percent means that the mean consumption of the poor falls 14percent short of the poverty line. Poverty severity i s a measureclosely related to the poverty gap but giving those further away from the poverty line-the poorest-a higher "weight" in aggregation than those closer to the poverty line. Its level in Montenegro i s found to be 0.3 percent. 29 2.34 Additional poverty measures reported in Beegle and Radevic (2003) show that 4 percent of the population lives in households with total expenditure below the value of the minimum food basket, indicating that there i s no measurable extreme poverty. However, more than one- fifth of households spent more than 60 percent of their resourceson food. 2.35 As shown in the Table 2.5, about 5 percent of young adults in Montenegro can be considered "education poor," meaning that they are not in school and that they finished their education with primary school. The measure of health poverty affects 6 percent of the population. Housing conditions for 13 percent of population are below the poverty standard of pipeddrinking water or bathroom inthe dwelling; 8 percent of the populationhas less than 10m2 per person. Over 22 percent of working age adults are poor with respect to employment. While there i s only one household in the sample with no electricity, about 10 percent of the population lives in households with no telephone. The second column in the table shows the poverty indicator rates for the population of the consumption-poor and shows that there i s only a partial overlap between the consumption-poor and those deprived in other dimensions. However, for most indicators, the rate i s much higher among the consumption-poor. 2.36 Since poverty i s not necessarily very deep, one can expect that small changes in the poverty line could have a magnified impact on the proportion of the population in poverty. The effect on the poverty rate from changes inthe poverty line i s presentedinFigure 2.2. The figure shows that an increase of the poverty line significantly increases the percentage of the poor. As poverty line increaseof 20 percent would double the poverty rate. Figure 2.2: Poverty Rate inMontenegro as a Function of Poverty Lines 90 80 70 8 6 0 ; 5O h30 20 10 0 I 50 75 100 125 150 175 200 225 250 275 300 325 350 375 Povertyline Source: Beegle and Radevicbasedon ISSPsurvey 5 and 6. 2.37 Four conclusions stem from the analysis of the poverty situation inMontenegro. Itis possibleto measurepoverty inits dimensions and it canbe monitored 3 Poverty inMontenegro exists in every dimension; absolute poverty, though lower than in Serbia, i s a source of concern 30 =Povertyi s most serious, and i s greater than in Serbia, in the dimensions o f vulnerability and employment z Thiscallsforpoliciestopromoteeconomic growthleadingtotheexpansionofemployment. 2.38 How does poverty compare across the two constituent states of SAM? And how can it be compared with the situation in other countries? Box 2.2 discusses the issue of international poverty comparisons and the methodology developed for this report. 2.39 Figure 2.3 reports the result of a comparison of absolute material poverty across several countries in SEE. Figure 2.3 :Absolute Poverty Incidence for Selected Countries inSEE, Comparable Data, Relative to Serbia's Poverty Rate Serbia Montenegro Bosniaand Macedonia Croatia Bulgaria Herzegovina Source :staff estimates based on SLS data for BiH (2001), Serbia (2002) and Bulgaria (2001); HBS 2000 data for Macedonia andHBS 1998 data for Croatia, and ISSP2002 survey data for Montenegro. 2.40 The internationally comparable data will be used further in the report to assess the characteristics of the poverty profile in each country. 2.4. Poverty and Displacement 2.41 Compared to the previous (1996) registration o f refugee~,'~there was a 30 percent decline, as some refugees returned to their countries o f origin while others integrated locally, But those who remain inthis status are increasingly marginalized. Inaddition to refugees, more recent IDPs remain stalled (see Box 1.1for background). l6 The majority of the displaced were ethnic Serbs (90 percent of the refugees and 76 percent o f the IDPs), indicating that linguistic and cultural assimilation should not present great difficulties. However, recreating the social networks that facilitate access to employment or social services are often not easy. The Roma l5UNHCR and the Commissioner for Refugees of the Republic o f Serbia. 2001, "Refugee Registration inSerbia." l6UNHCR and the Commissioner for Refugees of the Republic o f Serbia. 2000, "Registration of Internally Displaced Persons from Kosovo and Metohija." 31 constituted the next largest ethnic group among the IDPs (11percent). However, it i s believed that a large number of Roma IDPs have not been includedin the registration. As with the local Roma population, they seem to suffer from a disproportionate amount of economic and social exclusion. BOX 2.2: Cross-countryPovertyComparisons Comparing the pattern o f poverty across countries immediately raises the question of how to set the poverty line and what data to use. ASan absolute line, the international poverty lines o f `1 (or two) dollar (s) a day ` line has been extensively used ever since the 1990 World Development Report. Such lines have to be converted into local currencies using Purchasing Power Parity (PPP) exchange rates. PPP rates are derived from the International Comparison Programme's price surveys carried out by the OECD. S A M still does not have recent PPP figures. And even when PPP rates will be available, setting it to one or two dollars is arbitrary. Therefore, such poverty rates will provide little guidance on the extent of the poverty problem ineach particular country. To understand the positiono f S A M vis-&-visother countries in this report, the same methodology is applied to derive national poverty lines in other countries' datasets. How does this methodology work? The team took surveys that were comparable as possible, applied to the raw data the same definition of consumption everywhere - including imputed rents. Food poverty lines were developed based on actual consumption patterns o f each country. And finally the level o f total consumption at which food spending for a household exactly met this minimum food requirement was established using the Engel curve approach. Source: Annex I,andWorld Bank (2000a), MakingTransition Work for Everyone. 2.42 The majority of refugees (49 percent) lived in Vojvodina, 30 percent lived in Belgrade and the reminder lived in other regions of Serbia. Compared to the previous registration, the number of refugees in Vojvodina and Belgrade seemed to be increasing--perhaps in an attempt to find better economic opportunities. Most of the IDPs, on the other hand, were in Central Serbia (64 percent), with another 28 percent in Belgrade, and only 6 percent in Vojvodina. As with other displaced populations in the region, the displaced seemed to be concentrated in areas bordering the former conflict zones or the capital city-not evenly across the Republic. l7 This has put enormous pressure on the local economic and social infrastructure, leading to some resentment among the local population. For example, in the municipality of Kursumlija, which borders Kosovo, there were 26 IDPs for every 100 local residents; and 11percent of Belgrade's population i s composedof IDPsand refugees. 2.43 With respect to accommodation, the SLS for Serbia showed that most of the refugees rented an apartment (44 percent) or lived with family or friends (one-third). Only about 18 percent owned their accommodation. Compared to the 1996 registration, refugees seemed to be moving away from collective centers and from sharinghomes with family and friends to renting and buying their own accommodation. This may be a result of their ability to gain better employment andor to sell their properties in Croatia or BiH. Similarly, the majority of IDPs also rented an apartment (41 percent) or lived with family or friends (40 percent). Although only 5-7 percent of the displaced live in collective centers, living conditions in these institutions are especially harsh. 2.44 According to the SLS 2002 for Serbia, the displaced had a higher poverty and vulnerability incidence. While about 10 percent of the resident population were poor, about 17 percent of the IDPs and 21 percent of the refugees were poor. The difference between the IDP l7 World Bank. 2003 (Draft). "Conflict-Induced Displacement." 32 and refugee populations might be explained by the fact that IDPs received three times more humanitarian assistance. 2.45 Interestingly, as shown in the Table 2.6, both the poverty and the vulnerability percentages fall when displaced persons live with residents, which reflects more the pattern o f poverty found among the residents. This seems to suggest that as displaced populations move from households that they share with family and friends to their own accommodation, they become poorer and more vulnerable. This i s an alarming findingbecause that seems in fact to be the dominant pattern among refugees, as indicated above. Table 2.6: Household Compositionand PovertyNulnerability of IDPsandRefugees Percent Poor Percent Vulnerable Householdcomposition Residentsonly 10.3 19.6 IDPsonly 15.9 26.4 Refugeesonly 22.0 31.8 Residentsand IDPs 9.7 19.8 ResidentsandRefugees 10.9 23.7 Total 10.6 19.4 Source: Staff estimates basedon SLS 2002. 2.46 Displaced seemed to fare worse with respect to employment. While 8 percent of the residents were unemployed, 11 percent of the IDPs and 19 percent of the refugees were unemployed. The large difference between the two displaced populations could be due to the fact that some Kosovar IDPs were able to maintain their previous employment status--if not their salary. 2.47 The outcome i s surprising given the fact that the displaced had more education than the resident population; this was especially true for IDPs. More residents have either no elementary education or incomplete elementary education (18 percent compared to 10 percent for IDPs and 14 percent for refugees), more refugees have complete secondary education (54 percent compared to 43 percent o f residents), and more IDPs have a university education (12 percent compared to 7 percent of residents). 2.48 The recently completed survey of refugees and IDPs in Montenegro with sampling data provided by the humanitarian agencies and census of displaced, found that they face poverty risk which i s three times higher than for the general population. Deprivation in housing and employment i s particularly pronounced (see Annex Iof this Volume for details). 2.49 The Government's National Strategy for Resolving the Problems of Refugees18 focuses on the following measures to ameliorate the living conditions of refugees: repatriation and integration, which includes closing collective centers and supporting housing and employment schemes. With respect to IDPs, the government seems to consider repatriation as the only solution and does not have a separate strategy for them. Agencies engaged with displaced populations, such as UNHCR, while not recommending special programs for refugees and IDPs, '*Governmentof Serbia, June 2002, National Strategy for Resolvingthe Problemsof Refugees. 33 advocate that the poor not be treated as an undifferentiated group. Thus, the displaced should be treated in a similar way to residents with respect to employment, education, health and other social services. However, with respect to housing, they should be treated as a vulnerable group with specific needs. These agencies stress that the humanitarian phase is over. They urge development actors, including the government, to step in. An indication of this i s the fact that all food assistance provided by UNHCR/WFP and ICRC to refugees and IDPs, respectively, even those incollective centers, will end over the next year. 2.5. Regional Disparities:Rural Poverty 2.50 Both Serbia and Montenegro are countries with large and persistent regional disparities. These disparities are most clearly reflected in the large differences in poverty rates across regions. It i s striking, as Table 2.7 reveals, that a country as compact as Montenegro had 1:2 differences in poverty rates between the poorest and the richest regions. Regional differences inthe poverty rate inSerbia are even greater. 2.51 The determinants of regional disparities can be divided into three broad groupings: (i) endowments, comprising the stock of human and physical capital (including the supporting infrastructure); (ii) the business and economic environment, which i s affected by the inherited production structure but also by economic policies; and (iii) the institutional and political frameworks, which condition both policies and their outcomes. Table 2.7: Poverty Rates inMontenegro by Region Montenegro North Center South Povertyrate: HeadCount 9.4 14.9 6.5 6.8 Percentof all poor 100.0 54.0 30.5 15.5 Poverty gap 1.3 2.2 0.9 0.7 Severity of Poverty 0.3 0.5 0.2 0.1 Source: Radevic and Beegle (2003) ISSPHousehold Survey 5 and 6. 2.52 Table 2.8 reports even greater differences across the regions in Serbia and between urban and rural areas. The poverty rate used in this report varies from a low of 6.9-6.8 percent in urban Vojvodina and Belgrade to a high o f 22.7 percent in rural South Eastern Serbia. The variations in distributionally sensitive measures i s even greater. 2.53 The large disparities in regional poverty rates-echoed by disparities in regional unemployment rates and earnings-suggest that labor immobility may be an issue. The l9 UNHCR-WFP, April 2003, Joint Assessment MissionReport. 34 unemployed and workers are often unwilling to move to regions that have better job prospects because of problems with finding affordable housing, the cost of relocating, the risk of rupturing social support networks, and uncertainty about their ownjob prospects. Inother cases they worry about abandoning elderly parents for whom they are responsible, and for whom no alternative affordable care solutions are available. Table 2.8: Poverty by Regionand Type of LocationinSerbia in 2002 (Dercent of Dormlation) \I I I Vulnerable, Poor, % of Relative Share in Share Poverty Poverty %of population poverty the amongthe depth severity population rate population poor Belgrade total - 15.0 7.9 -25.5 21.1 15.8 1.5 0.5 Urban 13.2 6.9 -34.9 17.2 11.2 1.2 0.4 Rural 22.9 12.2 15.1 4.0 4.6 2.9 1.o Vojvodina total - 18.4 8.8 -17.0 27.1 22.5 1.9 0.6 Urban 16.0 6.8 -35.8 15.4 9.8 1.3 0.4 Rural 21.5 11.5 8.5 11.7 12.7 2.6 1.o West Serbia 23.9 13.5 27.4 11.2 14.2 2.8 0.9 Urban 22.7 12.1 14.2 4.3 5.0 1.9 0.5 Rural 24.7 14.4 35.8 6.8 9.3 3.4 1.2 Centr. Serbiatotal 19.5 10.2 -3.8 17.3 16.6 2.2 0.7 Urban 15.2 6.9 -34.9 8.5 5.5 1.4 0.5 Rural 23.7 13.2 24.5 8.8 11.1 3.0 1.o EastSerbia total. - 17.4 10.1 -4.7 9.3 8.9 2.3 0.8 Urban 14.3 9.2 -13.2 4.4 3.8 2.0 0.7 Rural 20.2 10.9 2.8 4.9 5.1 2.6 1.o SESerbia total - 29.8 16.6 56.6 14.0 22.0 3.6 1.2 Urban 21.3 10.0 -5.7 6.7 6.3 2.2 0.7 Rural 37.7 22.7 114.2 7.3 15.7 5.0 1.7 Overall 20.0 10.6 100 100 2.2 0.8 Urban- overall 16.0 7.8 -26 56 42 1.5 0.5 Rural - overall 25.1 14.2 +34 43 58 3.2 1.1 Note: The relative poverty risk i s calculated as the growth (drop) percentage of the poverty index of each group in relation to the averagepoverty index of the entire population. Source: Krstic (2003)based on SLS 2002. 2.54 Table 2 in the Annex I1suggests that regional differences are not limited to material poverty. But it also shows that poverty in other dimensions does not necessarily follow the similar pattern to material poverty. It highlightsthe fact that different regions may be poorest in each dimension, and poverty risks are therefore widespread. 2.55 Sections 2.5 and 2.6 attempt to account for most important factors that explain such wide regional disparities. 35 2.6. Poverty Profile: WhoAre the Poor?' 2.56 There are several poverty correlates that normally are reported in the recent poverty analysis conducted for PRSP (see, for Serbia, Krstic in Bogicevic et a1.[2003]). It i s important to note, however, that some of the observed correlations can be driven by measurement assumptions (an issue that we will address in section 2.7) and that some require explanations themselves, as, for instance, the regional differences described inthe previous section. 2.57 Two of the poverty correlates are absolutely prominent in Serbia: employment and education. Both have a robust effect on key measurement assumptions and provide clues about the causes of poverty. 2.58 Table 2.9 presents connections between education level, poverty and vulnerability, as definedfor this report. It is easy to draw a conclusion that the share of the poor declines with the education level. Individuals who have not completed elementary school have the highest poverty risk. The poverty risk of this group was twice more than the population average, and the depth and severity were considerably higher inrelation to other education levels. Table 2.9: PovertyamongAdults Accordingto EducationinSerbia in2002 (populationover 15 years of age; percent) Vulnerable, Poor, %of Relative Share inthe Share among Poverty Poverty %of population poverty rate population the poor depth severity population Uncompleted elementary school 36.1 21.5 102.5 17.5 35.4 5.1 1.8 Elementary school 25.4 14.3 33.9 23.3 31.2 3.1 1.o Secondary school 15.1 6.9 -35.1 47.5 30.8 1.2 0.3 Jr. college 7.8 2.9 -72.8 5.1 1.4 0.4 0.1 University 5.6 2.0 -81.4 6.6 1.2 0.4 0.1 Total for adults 20.2 10.6 100 100 2.2 0.7 Source: Krstic (2003) 2.59 On the other hand, the highly educated (junior college and university) individuals were not exposedto a poverty risk and this risk was below average. Only 2.9 percent of those who had completedjunior college and 2 percent of those with university degree were poor. Such poverty distribution according to education indicates that education pays off since the labor market rewards education. A very similar picture i s obtained for level of vulnerability according to education level. 2.60 The well-being of individuals depends not only on their own labor market position but more broadly on the degree of contact with the world of work of their household. This i s 2oThis section i s basedon Krstic (2003) and Radevic and Beegle (2003) 36 why to properly measure the link between the labor market and poverty it i s necessary to consider the household as a single unit and its overall employment profile. The European Commission widely uses the definitions of "jobless" households and "active age" households as a key indicators covering the employment dimension. Jobless households are those in which one could expect (on age grounds) at least one member to be economically active, but where no one works. The focus of this indicator i s therefore on the cumulative negative impact, at the household level, of lacking contact with the world of work. The specific objective of this indicator requires that non-active age households be first correctly identified(those households where all members fall into any of the following categories: children aged less than 18; persons aged 18-24 in education and inactive; and persons aged over 65 and not working). It i s also important to use a consistent definition of employment that allows a full reflection of the entire spectrum of labor market activities ( L O criteria). The results are presented in Table 2.10. For Serbia, slightly over 8 percent of people were living in non-active households and 6 percent in jobless households. Table 2.10: Poverty According to the Employmentof Household Members inSerbia in 2002 (percent of population) % of Poor, % of Relative Share inthe Share among Poverty Poverty population population poverty rate population the poor depth severity Non-activehousehold 28.6 15.4 45.06 8.1 11.8 3.26 1.11 Jobless household 35.4 23.I 117.53 6.2 13.4 6.29 2.37 Working household 18.1 9.2 -12.95 85.7 74.6 1.85 0.60 Total 20.0% 10.6% 100% 100% 2.2% 0.8% Notes: "Working" categoriesare defined on the basisof the L O standarddefinitions. Non-active age household in which i s a householdwhere one could expect - on age grounds - not a single member to be working (where everybody falls into one of the following categories: aged less than 18 years old, aged 18-24 and in education, aged 65 and over). A jobless household is an active age household which has no working members. A working household i s an active age household which has at least one employed member (defined according to L O criteria). Source: Staff estimates based on SLS 2002 2.61 Table 2.10 shows that both factors are strongly correlated with poverty, but no work in a working age household more than doubles its poverty rate. But an equally important message in this table is that not all of the jobless are poor, and poverty is not only a lack of employment. The majority of the poor in Serbia are inworking families. This finding i s in stark contrast with EU countries and with many more advanced CEE countries, where poverty is concentrated among the unemployed and inactive. 37 2.7. Understanding Poverty: Decompositions and Multivariate Analysis 2.62 The fact that we find a correlation between poverty and education (or employment) does not necessarily imply that there i s a causal relationship between the two. There could also be reverse causality (the poor are less able to invest in education), or there could be some "third" variable that i s related to both low education and poverty (such as less educated families living in more economically depressed areas). 2.63 To be able to see the importance of each factor in determining the poverty status of a household, the report team applied two types o f analytical approaches to poverty: decomposition and simulation. The first approach consists of finding the key factors that explain why the poor are different from the rich and relies on simple comparisons (see Box 2.3). The second approach uses multivariate techniques to assess the importance o f each factor inisolation. 2.64 Decomposition. Box 2.3 describes the methodology for assessing the gap between the rich and the poor in their level o f consumption. Figure 2.4 applies this approach to a number of countries to highlightthe differences and also to show important common characteristics. Figure 2.4: Decompositionof Difference in Consumption between Top and Bottom Quintiles, Total Gap =loo% Serbia Montenegro Macedonia Croatia Hungary Poland Russia I-40% -20% 0% 20% 40% 60% 80% 100% Sources: for Croatia, Hungary and Poland , World Bank, Making Transition Work for Everyone (World Bank 2001a). For Macedonia, staff estimates based o f HBS 2000. For Serbia -SLS, Montenegro -1SSP-CEED HHS. 2.65 The "skills" component, which is the difference in what an average employed poor adult and rich adult makes on the labor market, clearly dominates the whole picture. Most of the gap between the rich and the poor can be accounted for by different levels o f pay, which are in turn a function of several factors, education playing a primary role. In Serbia in particular this component accounts for as much as 63 percent of the gap. 2.66 Offsetting the effect o f the smaller share of labor earnings per adult, the bottom quintile has a higher share of non-labor ("transfer") income to total income in all countries. In all 38 countries this component i s composed o f the three largest elements: transfers, imputed rents and financial incomes. Serbia has one of the most powerful (after Hungary) inequality- reducing non-labor components among countries in the comparison (the gap i s reduced by 20 percent). Further decomposition of this component (not shown) across countries traces the key role o f transfers, with imputed rents accounting for a relatively smaller part. 2.67 The factor "mouths," which captures the ratio of adults to total household size, i s everywhere, except in Montenegro, positive, indicating that households in the top quintile have a lower dependency ratio than households in the bottom quintile (in Montenegro the reverse i s true, richer households have higher dependency rates than the poorer ones). 2.68 The top quintile "works" more than the poor, and has more income earners per adult (this is particularly striking in both Macedonia and Montenegro), indicating that in these countries non-employment i s an important driver of differences between the rich and the poor. BOX 2.3: Decomposingthe PovertyGap. I s poverty a result of having more dependents, lower labor market participation, or lower earnings among those who participate in the labor market? The answer i s all of the above but to different degrees. Consumption per equivalent adult can be decomposed into five factors: measures o f dependency ("mouths"), labor market participation ("labor participation"), labor earnings per income earner ("earnings"), share of other income in total income ("other income"), and ratio of consumption to income ("shadow or unreported incomes"), as the following equation shows: This equation can be transformed into logs to determine the share each factor contributes to consumption. The difference in mean equivalent consumption between the top and bottom quintiles i s equal to the sum o f the difference of the means o f each factor between the top and bottom quintiles. For consistency and comparability with other countries, the OECD equivalence scale i s used. Source: MakingTransitionWork for Everyone (WorldBank 2001 a) 2.69 Finally, "shadow income" - the gap between reported income and household consumption -accountsforalowof18percentinCroatiatoahighofover40percentin Serbia (almost as much as inRussia) of the difference in mean log consumption between the top and bottom quintiles. The importance of this factor may be indicative of the highly informalized and nontransparent nature of markets inthese countries. 2.70 This comparison offers a clue to what matters for poverty. Employment is important for reducing the poverty (especially in Montenegro), but in itself it is not enough: the low earning potential of the poor i s to blame for their lower living standards. And earnings are closely correlated with the measures of human capital. Last, but not the least, the rich are richer than the poor not because they benefit more from the informal sector - they are richer because they have more productive employment, inboth the formal and the informal sectors. 39 2.71 However informative the simple decomposition is, it does not tell us why the poor earn so much less than the rich and why they have much lower chances to benefit from the informal economic activity. 2.72 Multivariate analysis of poverty i s called on to address such types of questions. Figure 2.5 recapitulates the rationale for this approach. The poverty profile shows simple correlations between householdcharacteristics and poverty. For example, we learn that the unemployed have a higher risk of poverty than the employed, and thus there i s a correlation between poverty and unemployment. At the same time, we also know that poverty i s correlated with low education: those with primary education or less have a high risk of poverty. But these two factors - unemployment and education - are not independent of each other. What i s the true link between employment and poverty: are the unemployed poor because they are without work, or are they are poor because they have a low education level? If the latter i s true, even if they find employment, this employment i s likely to be in a low payingjob, so that their risk of being poor would not be reduced. If the partial correlation between unemployment and poverty i s much smaller than the corresponding simple correlation, this would imply that much o f the relationship between unemployment and poverty can be accounted for by lower level of education, and not by the lack of employment per se. 2.73 All of this may seem easy to understand in a particular case of education and employment: at the end, one knows that a low education level often means low productivity and low earnings, thus employment i s not "solving" the poverty problem for this group. The value added in this approach is the ability to estimate the magnitude o f these effects, and to estimate notjust three, but multiple, links and correlations at the same time. Figure2.5: Simple versus Partial Correlation inPoverty Analysis Simple correlation: IUnem$1loymint Poverty Simple correlation v - k Partial correlation Simple (controllingfo correlation Partial correlation (controllingfor unempl.) 2.74 A multivariate analysis of household consumption in Krstic (2003) reproduced in the report's Annex IT also shows that most household characteristics-education, employment status, location, household demographics-are significantly associated with consumption levels in the 40 expected directions. Although important determinants of poverty remain unidentified (for example, personal ability/preference), the model which i s used identifies important factors relatedto poverty. 2.75 Education i s a significant determinant of poverty, even controlling for other factors. The household head with the lowest education level has a smaller consumption and a higher poverty risk in relation to persons whose household head has a secondary level education (reference variable), given other characteristics. On average, the households with university educatedheads recorded consumption that was higher by one-third than that in households whose heads completed only secondary school. This effect i s relatively high in relation to other transition countries. Observed in different quintiles of the consumption distribution (as reported in Krstic [2003]), it i s higheramong the poorest. 2.76 The share of unemployed in a household significantly decreases consumption and increases the poverty risk in relation to the share of employed. An increase in the share of the unemployed in a household, given unchanged household size, decreases the consumption level by 23 percent in average. Also, an increase in the share of the employed in a household working in a private registered firm, makes their consumption go up compared to those employed by sociallyhtate owned enterprises (by 18 percent on average). Consequently, the poverty index declines. This effect grows if we move from the poorest toward the wealthiest part of the population. 2.77 Location plays an important role. Households in rural areas of Southeastern Serbia have low consumption, even if we control for other factors. These households have lower consumption by 14 percent on average in relation to households in the urban areas of Belgrade (the reference variable). A similar but somewhat smaller effect on consumption can be noticed for households from the urban areas of Southeastern Serbia. 2.78 All factors seem to matter somewhat. How can we measure more exactly their relative role? The full battery of factors employed by Krstic (2003) cannot be used for poverty rate simulations for the following reason. As shown in Lokshin and Jovanovic (2002), there i s a strong correlation between unobservable characteristics (such as personal abilities, health, entrepreneurship, etc.) and work inthe highest paid occupations in the private sector. Assuming (reasonably) that higher pay means higher consumption, inclusion of too detailed indicators of employment into the model violates the assumptions about the regression and introduces a strong bias inits predictions. 2.79 There i s no straightforward way to get around this problem. In a three-step approach that yields easy-to-interpret results and that uses all available information on the relationship between household characteristics and equivalent consumption, first, a reduced form e uation with a very limited set of household characteristics using an OLS regression i s estimated.91 Second, this regression i s usedto simulate per capita consumption if one of these key characteristics would be constant across the population (set to its mean value). Third, using a baseline poverty In(eqcons,) = %+aleduc,taz lfpart,+a, region,+Q dependency, + ajvuln, + EL where eqcons, denotes equivalent consumption of hosueholdi, educ, denotes averageyears of education per adult in the household, Ifpart,i s a measureof labor force participation of a household, region, denotes a set of regional dummy variables, dependency, consists of dependencyrates for children andelderly, and vuln, are control variables for specific factors of vulnerability (displacement status anddisabled ina household). The error term is denoted by E, and the cr, are vectors of coefficients to be estimated. Results are reported in the Annex. 41 definition, poverty risks are predicted for each household using this simulated measure of equivalent income. The simulated poverty rates (in fact, the probability of being poor for a household given its characteristics) are purged from any effect running through all included factors, and they therefore show a partial relationship. 2.80 It is informative that the predicted risks of poverty after controlling for key poverty factors remain significant. A household headed by a post-secondary graduate with an average employment profile and demographics will face still significant odds o f 5 percent o f being poor. 2.81 Table 2.11 shows the results from an application of this approach to studying regional variations o f poverty risks. The first column reports the observed poverty rate in relation to the national mean. It shows, for example, that Belgrade has a poverty rate lower than the national by one-quarter, while the South East has a poverty rate above the national by almost 60 percent. Rural households have higher poverty, and urban households have lower poverty. 2.82 After controlling for differences in the demographic situation i s undertaken, the structure of the poverty rates remains very similar (column 2). The poverty rate for rural areas increases compared to the observed rate. The simulated poverty rate in Western Serbia i s lower than the observed one indicating unfavorable demographic profile (higher dependency and older population). But the overall simulated poverty rates after equalizing all households with respect to their demographics mirror the observed differences inpoverty rates between regions. Table 2.11: Observed and SimulatedPoverty Rates by Regionsafter Controlling for Key Factors (poverty rate for group relative to average rate, inpercent) Observedpoverty Controlling for Controlling for Controllingfor Controlling for ratdnational demographics Employment education all Location Belgrade -25% -23% -20% +6% +12% Vojvodina -17% -15% -18% -16% -18% West Serbia +28% +14% +16% -1% -2% Centr. Serbia -4% -3% -3% -12% -13% EastSerbia -4% -7% -5% -20% -20% SE Serbia +57% +59% +57% +52% +48% Urban- overall -26% -21% -24% -0% +0% Rural-overall +34% +36% +32% +0% -1% All households reference reference reference reference Reference Source: staffestimatesfrom the SLS data. 2.83 Once controls for labor market status are introduced, the picture o f relative rates remains unchanged. It means that regional differences in poverty are not explained by differences in regional unemployment or employment rates. 2.84 But once the controls for education are used (i.e., households are equalized in terms of their schooling, and their consumption i s "purged" from the effects o f education, a dramatic shift in regional poverty rates occurs. Now there are no more differences between rural and urban 42 areas, and the poverty rate for Belgrade flips on the other side. What does this mean? It shows that the average level o f educational attainment in the capital is far above to the rest of the country that if Belgrade were to be left with average skills, with its poor demographics it would do worse than the average. The amazing positive effects of education are undone partly by the poor employment record in the city, as the last column suggests. There i s something about the capital that probably reflects the impact of Belgrade being the capital city, with all that this implies in terms of receiving more of everything and having more new job opportunities. West Serbia, on the other hand, shows a significant improvement, which points to lower educational attainment in this region as a key poverty factor. But this unusual effect aside, not much i s happening overall to the rest of the country. 2.85 The last column o f Table 2.11 shows that (with the notable exception o f Belgrade and Western Serbia) regional effects tend not to go away once controlling has been done for education, for the labor market status of households and for demographics. These factors all contribute to explaining some parts, but not all, o f the observed regional differences. 2.86 This is an important conclusion. Regional differences have to be addressed as a separate policy dimension. They will not wither away with economic progress. It should be noted that the reverse i s true for the ruralhrban divide. This can be fully redressed by improving access to education inrural areas. Indeed, this i s easier said than done, but Chapter 4 will look at this issue to asses whether the educational differential can be addressed inthe foreseeable future. Table 2.12: Observedand SimulatedPoverty Ratesby Educationand Employment Status of the Household Head after Controllingfor Other Factors (poverty rate for group relative to average rate, in percent) Observed for poverty Controlling ControllingControlling for rate/nationai employment education for for a11 Educationof HouseholdHead No schooling-unfinished elementary school +94 +lo2 +86 -4 Elementary school +72 +33 +29 -0 Vocational school -1-2 years +45 +18 +3 -1 Second. 3 years vocational school -14 -14 -15 -1 Second. 4 years vocational school -65 -36 -34 +1 Gymnasium -66 -49 -45 +7 Post secondary -68 -55 -54 +3 University -95 -70 -68 +5 EmploymentStatusof HouseholdHead Inactive +27 +35 +12 -4 Unemployment (ILO) +76 +79 +94 -3 Employment (LO) -26 -26 -13 +3 Source: Staff estimates basedon SLS 2002. 43 2.87 For the next stage of the analysis the team took up another question: What i s the relative role of employment (unemployment) and education as correlates of poverty? Applying the same methodology, one can assess what would happen to the poverty risk by the level of education, once employment i s controlled for, and vice versa. Finally, once the full set of controls i s introduced, are there residual effects similar to ones observed for regions (Table 2.12). 2.88 The second column of Table 2.12 shows that the differences in poverty risk by level of education of the household head become slightly smaller when we control for the partial effects of demographic characteristics and location. In other words, the relationship between higher poverty and lower education i s due in part to the fact that those with lower education have less favorable demographic characteristics and location. 2.89 But this is not always the case. Take the first row as an example. We see that the relative poverty risk for the poorest households (households where the head has only an unfinished elementary education) increases once the effects of demographics and geography are wiped out-- from +94 percent to +lo2 percent. This means that the poorest families tend to have demographic and location features that would of themselves be associated with lower levels of poverty. Once we take their "favorable" demographic factors out, their risk of poverty increases even compared to the observed very high rate. Compare this to the case of those with post- secondary education: their relative poverty rate remains lower than average, but their advantage falls when we control for demographics and location. 2.90 The differences in poverty risk across education groups also becomes smaller when we control for labor market status in column 3 of Table 2.12. The relative poverty risk for those with elementary education, for example, falls from +72 percent to +39 when we control for labor market status. This suggests, again, that part of the observed correlation between poverty and low education comes about because those with lower education are more likely to be unemployed. The most interestingpart, however, i s the striking difference in the ways in which employment influences poverty risk for vocational school graduates. For lower level vocational school, the pattern i s very stark: once controls for employment are introduced (i.e., once the effect of the high unemployment rate for this group i s taken out), the poverty risk falls dramatically to just about the average (+ 3 percent). For middle level vocational school, there is no change once controls are introduced in their poverty risk. And for upper vocational school, the pattern i s very similar to all post-secondary groups with significant effects coming from locatiorddemographics and from the labor market. 2.91 Finally, when we control for all factors at once we find that the relative poverty risks become "inverted," with the relative risk for those with less education being less than the average, that of university graduates being close to the average, and that of advanced vocational being actually higher than the average. What this means is that the residual or "unexplained" part of poverty is actually more negative for those with a higher level of education than for those with a lower level of education. Since we know that the variance of incomes tends to increase with education, this simply means that as education goes up, it i s harder to predict (or explain) why a family with an educated head would be poor, as it is likely to reflect very household- specific reasons (e.g., health problems). Another way to think about this i s to turn it around: poverty among those with low education i s explained by observable factors such as the 44 combination of their demographic characteristics, their labor market status, the region they live in, andthe fact that those with lower education commandlower incomes on thejob market. 2.92 As we did in the case of education, we can examine the pathways through which labor market status i s associated with poverty (shown on the lower panel of Table 2.12). We can see that, unlike the case o f education, controlling for demographic and location variables has a significant impact on the relative risks of poverty according to the labor market status of the unemployed (who seem to have relatively favorable demographics). 2.93 Controlling for education i s also very insightful. The unemployed ( L O definition) show a much greater partial risk of poverty than observed (+94 percent as apposed to +76 percent, Table 2.12). This means that their personal education characteristics are reinforced by the poor educational attainment of their family members who live with them. Thus, based on their family backgrounds, they have to be even poorer than they are. Controls nevertheless suggest that there are very strong direct effects o f unemployment on the well-being of a household. Thus, the relationship between labor market status and poverty appears to reflect the direct link between labor market status and household income. It is only when we control for all these factors simultaneously that the relative poverty risks tend to equalize themselves across labor market and employment categories, as inthe case of categories by education. 2.94 Conclusion. The message from these simulations is that there is no single key cause of poverty. The causes of poverty are multifaceted. It i s a combination o f "misfortunes" that makes people poor. Therefore, there must be multiple "prongs" in a strategy to fight poverty, not only because it i s multidimensional, but also because the causes of material poverty cannot be reduced to a narrow set of problems. But three factors stand out: regional disparities, education, and employment. Regional differences have to be addressed as a separate policy dimension. They will not wither away with economic progress. It should be noted that the reverse i s true for the ruralhrban divide. It can be fully accounted for by improving access to education. Both education and employment have a strong independent impact on poverty, and matter equally for poverty. When the handicaps o f poor education and non-employment are combined, the odds that a household i s poor are overwhelming. They are the key factors which influence poverty at the household level. For the poorest categories, demographic and location characteristics in fact work in their favor, mitigating against poverty which i s entirely a factor of their double employment-education handicap. But at the same time the simulation shows widespread vulnerability to poverty: even the households that are well-positioned on the labor market and that have good levels of education face non-negligible risks o f poverty. This suggests that the strategy for reducing poverty must have three prongs: (i) to address regional disparities, (ii) to focus on the education and employment o f the poorest, and (iii) to reduce exposure to risk for everybody. 2.8. Poverty in the Life Cycle: Youth, OldAge, Disability. 2.95 Considerable attention in the key PRSP poverty analysis paper by Krstic (2003) and in the draft PRSP document i s devoted to the demographic correlates o f poverty. 2.96 The elderly are thought to be less able to find alternative coping mechanisms due to age and health. Pensioners with access to land are frequently unable to use it to produce 45 supplemental food, and fewer elderly participate in the informal economy. According to one survey conducted inNovember 2000, over half of all pensioners lived below the poverty line and did not have enough money for food.22 Of the pensioners interviewed for the WF'P study, nearly half claim that medicine i s not available or i s too expensive for them. 2.97 Youth between the ages of 15 and 24 account for approximately 15 percent of the population. The decade of war disrupted education and limited the opportunities for employment. 2.98 Children also are believed to be at great risk for poverty. Data from the SLS on poverty for children and the elderly as reported in Krstic (2003) have already proved to be a source o f controversy and have ignited a heated discussion about the validity of the results. The finding that children o f pre-school age are much less poor than the rest of the population, as shown inthe basic analysis, while the elderly have the highest poverty risk, may well be driven by the specific equivalence scale adopted for the measurement o f welfare. 2.99 This is why it is important to carefully examine the risk of poverty with alternative measurement assumptions before jumping to conclusions. This section offers a careful examination o f the relative poverty rates for the elderly and children in comparison with other countries. Academic studies show that any findings about the poverty o f a demographic group are sensitive to assumptions about economies o f scale (see Lanjouw et al. for a review). If no economies of scale are assumed, children-who tend to live in large families-are normally at higher risk of poverty than the elderly-who tend to live in small families. But if substantial economies of scale are assumed (for example, an economies-of-scale parameter of 0.5), the elderly tend to be at higher relative risk of poverty. 2.100 What would the poverty rates of children versus the elderly look like if we applied different economies of scale? Annex Ishows that the methodology developed to set the poverty line for Serbia allows full flexibility with respect to the equivalence scale. The report team tested the robustness o f the demographic profile of poverty by applying two alternative scales: per capita and the OECD equivalence scale (usingthe database described inBox 2.2).23 2.101 Because the total poverty rates vary by country, the raw data on poverty rates by group are not perfectly comparable. A group whose incidence of poverty is 9 percent in Croatia has a very high risk o f poverty relative to the 3 percent poverty rate o f the Croatian population as a whole. In contrast, a group in Serbia having the same 9 percent poverty rate as our comparator in Croatia would have a lower poverty risk than the Serbian population as a whole, which has an 11 percent incidence o f poverty. Therefore, for each group its poverty rate i s divided by the average national poverty rate to assess whether it i s higher or lower than average. This ratio i s sometimes called the relative poverty risk index (Foster and Toth [19991). 22 World FoodProgramme, Vulnerability Assessment of Pensioner Households in Serbia. 23 The "Serbian" scale gives the weight of 1to the first adult, 0.81to all subsequent adults, the weight of only 0.24 to each child under 6 years and 0.75 to all other children under 18. The OECD equivalence scale gives 1.0 to the first adult, 0.5 to other household members aged 14 or over, and 0.3 to each child below 14. 46 Figure 2.6: Relative Poverty Ratesfor Childrenversus the Elderly (incountries of SEE with different equivalence scales, percent to averagepopulationpoverty rate) A. Serbia B.Montenegro per capita Serbianscale cecd scale percapita Serbian scale oecdscale -1001 *SOG 19under 6 Bunder6 mberuccn6and 18 -0q mberueen6ffld 18 Oorcr65 Oorer65 -SO'& .100'1 C. Bosnia and Herzegovina D.Croatia Der cauita . . Serbianscale oecdscale per :aptla rcrbiansale occd scale rl00l -50'2 8under6 Eunder 6 TO'* mknrecn6md18 mberuecn 6and I 8 Oobrr65 ~- Oo\er65 -SO% -1001 E.Bulgaria F.EYRof Macedonia per capita Serbianscale wcd scale per capita Serbianscale Decdscale *100ct 11001 -50% r5Og 0under6 under 6 41 mbetueen6and18 41 m b e l n e e n 6 d 18 Oo\cr 65 Oovcr 65 -50% -501 -100Ci -1001 Source: Staff estimates based on household survey datasets. 2.102 Figure 2.6 provides the results of this comparison. The figure shows sensitivity of results by age group to the assumptions about scale economies and equivalence scale in the household. A very unusual result in Serbia to the effect that small children have a low poverty risk, is completely driven by the assumptions about equivalence scale. 47 2.103 The Figure 2.6 also shows that the demographic profile of poverty i s quite unique in each country. Serbia and Montenegro share certain common characteristics: whether the poverty risk i s higher for both children and the elderly i s entirely dependent on equivalence scale assumptions. InBulgaria, for example, children are poorer than the average person regardless of the scale applied. InCroatia, this i s true for the elderly, and inMacedonia the elderly always fare better than average. Figure 2.6 suggests that the differences in poverty risks by age in both Serbia and Montenegro are not very sharp, as they are, for example, in Bulgaria or Croatia. Thus, overall, in SAM the demographic composition is less of a guideline than elsewhere for targeting and for policymakers inthe area of social policy. 2.104 The figure also makes the point that the Serbian scale adopted for the baseline measurement o f poverty should not be used in the analysis of poverty by age groups: it i s rather unlikely that the poverty situation of a household would change so dramatically depending on whether it has a child who i s 4 years old or one who i s 10 years old. In almost every country (with the remarkable exception of Bulgaria), this scale produces similar results regarding differences inpoverty rate by the age o f children. 2.105 One problem with these results i s that all of the scales tested depend on both the differences in need between demographic groups (equivalence scale per se) and economies of size (how much less i s needed for an "additional adult"). A convenient manner o f assessing assess this sensitivity of results to assumptions about differences in need across demographic status i s to reconstruct demographic profiles of poverty with alternative specifications o f the single parameter that captures the degree of "scale" economies. 2.106 In addition, as shows inKlugman, Micklewright and Redmont (2002), it is also extremely important to assess the contribution of each group to total poverty (i.e., the share o f children and elderly among the poor). To compare poverty profiles across countries, the analysis so far used the poverty rates directly generated by data. To make such data comparable and meaningful across countries, we need to consider equalizing the national poverty rate, to, say 20 percent for all countries and for all variations of scale parameter, as suggested by Lanjouw et al, observing how this relative poverty rate changes as the scale parameter i s allowed to fall from 1to 0. 2.107 Figure 2.7 reports the results. For example, in panel A, for Serbia, we observe that with a per capita measure of consumption, with no economies o f scale (per capita), the incidence of poverty for the elderly i s 20 percent--the same as for the population as a whole. The incidence of poverty for children i s 25 percent .When there are some economies o f scale, at the level implied by the Serbian scale we can see that poverty among the elderly increases above that o f children. With higher economies poverty rate of children declines further, and poverty among the elderly increases. By the time the scale parameter reaches the level implied by the OECD scale, poverty among the elderly is 37 percent, while among children it is 19percent (for an average person it i s 20 percent by assumption). 2.108 The lower part of the panel for each country shows that the composition o f the poor changes as well. For Serbia, again, the per capita scale implies that 25 percent of all poor are elderly, while OECD assumptions move the share o f the elderly among the poor to 40 percent. 48 Figure 2.7: Poverty Rate and Composition of Poverty by Age Groups inCountries of SEE (with different assumptions about economies of scale; percent) A. Serbia B.Montenegro Serbia: Poverty of ChlldmnVI. ElderlyWith Dlllennt Economlwdof Montenegro:Poverty of Children VI. ElderlyWith Diflannt Ewnomiu SUI0 Of scale Serbia:Sham of Children and ElderlyWith Dillemnt Ewnomles of Montenegro: Shamof Children and Elderly With DMonnt Economies scale amongthe Poor of Scala amongtha Poor ...... ___ . . . . . . . . . . .............. I C. Bosnia D.Croatia Bosnlaand Herzegovina: Povertyof ChlldmnVS. ElderlyWith Cmatla:Poverty of Chlldnn Vs.Scale ElderlyWith Different Economlerof DillerentEconomiesof Sale Croatla:Sham of ChildrenVI. ElarlyWith Dlllerent Economlesof Scah amongthe Poor ............................. Source: Staff Estimates basedon Datasetsdescribed in Box 11.1 Theta=O means largest economies of scale . 49 2.109 Figure 2.7 shows that the demographic profile o f poverty when the rates are set at a comparable level across countries i s remarkably consistent. There are also some interesting similarities. In both Serbia and Croatia the reversal of a higher poverty risk for children occurs at very low levels of economies of scale (theta close to one), and the elderly tend to be poorer than the average person regardless of the measurement techniques. They also tend to dominate as a share of the poor. Montenegro and BiH, on the other hand, have children remaining at an elevated risk of poverty compared to elderly even with significant implied economies o f scale. However, they contribute to poverty, a relatively little amount, slightly over 20 percent, which i s not surprising, given on average the low fertility rates. 2.110 Overall, the results call for caution. Given our inability to precisely observe the degree of economies of size in consumption for a household, the question boils down to how sensitive conclusions regarding the relative poverty of the elderly to that of the young are to the presence of economies of size. If one has to make highly unrealistic allowances for economies o f size before there are any re-rankings between these two population sub-groups, then statistical results based on the per capita (or close to per capita) assumption can probably be accepted. If,however, only mild deviations from the zero economies o f size assumption result in sharp re-rankings, then there is clearly reason for caution in interpreting the results. Such robustness i s observed in Montenegro andBosnia and Herzegovina, but not in Serbia or Croatia. 2.111 The basic conclusion is that poverty is not caused by age. There are other factors that influence whether a household i s poor or not, and the analysis relevant for policies must focus on these other factors. 2.9. Gender and Poverty 2.112 The transition has resultedina greater burden on women as they engage inproducing and processing more food within the household. According to the 2001 survey, approximately 60 percent of households produce their own food and 33 percent make and/or repair their own clothes, work which i s generally performedby women. 24 2.113 Households that are headed by a single elderly female are perceived as being at highrisk, because they are usually dependent on pensions and transfers from children and other relatives. 2.114 Overall, female headed households account for 16 percent o f the population in Serbia and for about 10percent in Montenegro - compared to less than 4 percent in CEE countries and less than 2 percent in the CIS countries. The poverty data used in this report show a slightly (but statistically significant) elevated risk of poverty for households headed by women in both Serbia and Montenegro. Usingmultivariate techniques described in Section 2.6 for the poverty profile analysis, the report finds that the observed slight disadvantage of female headed households hides a set of disadvantages that women face in the society. The summary of the results is given inFigure 2.8. 24CPNCPS, The Social Situation in Serbia, January 2001 50 Figure 2.8: Poverty Rate Relative to the Average Poverty Rate by Gender of Household Head inSerbia: Observed versus Simulated Controlling Observed for Controlling Controlling poverty demographics for Controlling for all of the ratelnational and location employment for education above -20% -15% -10% -24% -g L -5% L w f ---lo% s t -5% .--x -15% -20% Male headed 0Femaleheaded Source: staff estimates based on SLS. 2.115 The figure shows that a female headed household faces a slightly higher observed poverty rate than does a male headed household. But once controls are introduced for differences in their human capital characteristics, employment status, and location, it becomes clear that a female headed household should have a lower poverty rate than a male-headed household. 2.116 It means that similar households headed by males have much higher poverty risks than average. Thus, women are over-represented in households with unfavorable demographic structure (e.g., high dependency rate) located in depressed regions (rural areas). Thus, the observed higher rate o f poverty among female headed households i s in some part due to their specific demographic composition and unfavorable location. Controlling for human capital characteristics, employment status, it becomes clear that among these factors of disadvantage o f female-headed households, education plays the most important role, followed by employment. Thus, economic factors play a key role inexplaining the elevated poverty risks of female-headed households. 2.117 These findings are consistent with the results o f the gender wage gap that finds it to be one o f the smallest among countries in transition. Krstic and Reilly show that, if controlling i s undertaken for other factors, women are paid approximately 10 percent less than men in SAM. Butthis gap is almost entirely due to the discrimination component. 2.118 Figure 2.8 highlights the fact that there are various other (non-wage) handicaps that single females face in the society that make them more vulnerable. It means that women heads of households have less education, are less active than men in the labor market and have more dependants. 51 2.10. Ethnicity and Poverty 2.119 Serbian nationalism under the MilogeviC regime eroded the position of minorities, triggering inresponse their alienation from the state. While the current (draft) Law on Minorities and efforts by the government have addressed some of these issues, the policies of the previous decade (or longer for Kosovo Albanians and the Roma) have left legacies of social exclusion and disempowerment which have directly contributed to poverty and have undermined the social networks o f the poor. 2.120 One measure o f social exclusion i s the majority or dominant nationality's reluctance to interact with minority groups. A survey of the extent to which different forms of interaction (living in the neighborhood, working together, friends, kinship) were acceptable found that there is a sense of hostility felt by many. 25 In addition to changes in attitudes, existing bonds have been frayed. The wars have had a destructive effect on ethnically mixed marriages, triggering divorce for many. Among IDP couples in mixed marriages, there has been a tendency to live apart, each with his or her own ethnic group. 2.121 In a survey conducted in Vojvodina inMarch 2001 o f Romanians, Slovaks, Croats, Serbs and Hungarians, respondents were asked about their opinions on the extent of discrimination in employment policy and in the labor market.26 Thirty-nine percent of Croat and Romanian respondents, 26 percent of Hungarian and 24 percent Slovaks considered ethnic discrimination to be present, and nearly half of these respondents considered that there was not an equal opportunity to obtain management positions. 2.122 In general in the country, according to various reports, living conditions for the Roma, whether IDPs or locals, are generally far below standard. They tend to be on the peripheries of urban areas or villages, and are usually built of cardboard, sheets of metal or plastic or other low- quality materials. Only a third of these dwellings have indoor plumbing. Only 10 percent o f the Roma attend school beyond elementary school; as o f 2001, there were no Roma language schools in the FRY. Illiteracy and drop-out rates are higher among Roma women as girls take on family responsibilities at an early age and also marry young. Unemployment i s high, with only 20 percent of the Roma of working age employed for a salary, and salaries are lower than the average salary.27 Discrimination and violence against the Roma further marginalize the community (see Box 2.4). 2.123 Until very recently there were no solid data on the poverty of the Roma in SAM. The ISSP survey in Montenegro found that the poverty rate among Roma exceeds 50 percent using the same baseline definition, thus i s almost five times the average incidence (ISSP-UNDP [2003]). The on-going analysis o f Roma survey by M O S A in Serbia will provide robust assessment of poverty for this minority. 25 Institute for Sociological Research at the Faculty of Philosophy, "Society inCrisis." 26 Helsinki Committee for HumanRights inSerbia, "Minorities and Refugees inVojvodina, "2001. 27 UN-OCHA, "Assessing the Needs of the Roma Community in the Federal Republic of Yugoslavia (excluding Kosovo)." LFS data offer more preciseestimates. They are believed to under represent the Roma, but still it i s informative that their ILOunemployment rate as measuredby LFS data was as high as 34 percent, and only one in three working age adults was employed. 52 BOX 2.4: Example of RomaSettlement inMontenegro N o group in the society of Serbia and Montenegro is believed to be subject to as severe discrimination as are the Roma. One example is a story of a Roma community inMontenegro told by an international advocacy group - a community living in abject poverty for years without much notice from the state. Around 90 persons from 20 the Romani families, mostly displaced from Kosovo and currently settled in the Lovanja settlement, struggle daily just to survive. Lovanja is located in the Tivat Field (Tivatsko polje), in the territory of Kotor municipality, on the beautiful Montenegrin coast of the Adriatic Sea. Roma of Lovanja live on the edge o f a local garbage dump in "substandard" housing conditions in self-made huts. The settlement does not have a supply o f potable water or an electricity supply, and i s under threat o f flooding in heavy rainfall. The closest medical facility i s in the town o f Kotor, around 8 km away from the settlement, and there are no public transport connections. Reportedly, the local authorities had decided to relocate the settlement to a more humane environment in 1999,but nothing has happened to date. About one-half of the settlement's inhabitants are under the age o f 18, and none o f the children attends school. The way o f making a living for the Lovanja Roma i s the collection of scrap materials and occasional manual labor paid by the hour. According to the Secretariat for Displaced Persons of Montenegro, Lovanja is categorized as an "unofficial centre for displaced persons." Such unofficial camps largely outnumber official camps, and provide self-made temporary shelters for a majority o f an estimated 20,000 Roma IDPs from Kosovo inMontenegro. Sources: The EuropeanRomaRightsCenter (ERRC), citing the Tivat-basednon-governmentalorganization "MARGO' - Association for Help and Support to MarginalSociety Groups,see: http://errc.org/publications/letters/2002/montene~o~an~10~2002.shtml. 2.11. Inequality in 200: Main Drivers of Inequality2' 2.124 An important contribution made by the recent economic literature has been to show that poverty and distribution issues were not pure distributional problems (that is, they were not just the problemof how to divide a cake of given size. There are many reasons to believe that, on the contrary, the size) of the cake depends precisely on the way it i s divided in society. As argued in section 2.1 inequality in living standards i s in itself an important indicator of an important dimensionof poverty ina broad sense. 2.125 Table 2.13 shows that Serbia's income and inequality lie at about the average for a group of Eastern European transition economies. Montenegro seems to be closer to the upper values of inequality in material well-being. Particularly interesting are the comparisons with the neighboring countries which have had surveys that have been very similar in design and in definition of aggregates as those of Serbia. 29 The range spans Gini values from 26 to 36, with Serbia's (income) inequality o f 33 around the mid-point o f the range. 2.126 As inequality measures can be quite sensitive to the definition of a welfare measure, in international comparisons every attempt was made to use definitions that are as comparable as possible. 30The bottom panel of the table uses exactly the same definition of every element of the consumption aggregate and primary survey data. It generally confirms the conclusion taken from income inequality comparison. Serbia fits squarely in the middle range of countries, 28This section is heavily basedon Milanovic (2003). 29Since inequality is not a unique concept, there are many statistics that measure inequality. The Gini coefficient is perhaps the best-known inequality statistic. It ranges between 0 (perfect equality) and 1(complete inequality).The Giniis most sensitive to inequality inthe middle of the distribution. 30Nevertheless, the data come from different surveys, from different years. Hence, the resulting inequality statistics may not always be perfectly comparable, and this should be kept inmind. 53 showing no sign of particularly low or high inequality; inequality in Montenegro appears to be slightly higher. Table 2.13. Comparison of Serbia and Montenegro with SelectedEasternEuropean Countries (countries rankedby the Gini coefficient; calculated on per capita basis) Counay (Year) Income or consumption per Ginicoefficient capita p.a. (in US$) Income (with imputed rents where possible) ~ Hungary (1999) 1,800 26 Slovenia (1998) 4,900 26 Bulgaria (1999) 820 33 Serbia (2002) 1,480 33 Macedonia (2000) 1,205 34 Croatia (1998) 3,200 35 Estonia (2001) 1,600 38 Montenegro (2002) 1,568 39 Bosnia and Herzegovina (2001) 1,445 39 Russia (2000) 1,000 40 Kosovo (2000) 662 58 Albania (2002) 58 Consumption (everywhere with imputed rents) Bosnia and Herzegovina (2001) 1,912 26 Croatia (1998) 3,854 21 Serbia (2002) 1,910 28 Kosovo(2000) n.a. 28 Bulgaria (2001) 1,159 28 Montenegro (2002) 2,916 29 Albania (2002) ma. 30 Macedonia (2000) 1,049 31 Source: Milanovic (2003) andstaff estimatesbasedon SLS data. For Albania, Kosovoand Croatia income i s without imputed rent. 2.127 As this table suggests, the citizens of Serbia and Montenegro are not at all equal interms of their standards of living, and there are significant differences both among the rich and among the poor. This i s much more clearly highlighted by decile ratios (showing how many times the consumption of a person at the ninetieth percentile - "poorest among the rich" - i s higher than the consumption of a person at the tenth percentile - "richest among the poor"). This ratio i s highest in Serbia and Montenegro than in any other country for which comparable data exist (Radevic and Beegle). For most countries it i s in a rage between 3 and 4, while in Montenegro it is as highas 5.8, and inSerbia it is even higher (6.7). This inequality which really matters for the poor, puts SAM closer to those countries in ECA with the most unequal distribution of consumption - Turkey and Russia (where the ratio is also close to 7). 2.128 This observation leads to a very important conclusion: of all inequality measures used to describe the distribution of incomes, the Gini index has the most ambiguous links to poverty. It i s therefore the least useful of indices in monitoring changes in distribution, which matters for the poor. 2.129 As the key concern i s poverty, the inequality measures need to have a closer association with the movements of poverty. This i s not the case with the Gini index, whose only advantage 54 is wide use and relatively straightforward interpretation. The Gini index is clearly most sensitive to changes in the distribution in the middle and may completely overlook changes affecting the poor. 2.130 Access to survey data also allows the computation of standard errors for the Gini index (following the methodology proposed by Jenkins). This robust comparison for consumption- based measures shows no significant difference between the listed countries, except, for Macedonia, which appears to have higher levels of inequality. Again, the Gini index proves to be a quite meaningless measure even for international comparisons of inequality. 2.131 The literature calls for further caution in using survey data for assessing the level of inequality. Ravallion and Mistaen, for example, show that differences in the response rate in a sample survey between the well-off and the poor may introduce a very large bias in inequality measurement. They also show that this bias i s negligible for assessingpoverty rates. 2.132 Moreover, apparent significant differences between income and consumption for countries in the table calls for caution in this comparison. The design of the questionnaire might have significant effects on measured inequality. For example, income inequality for Bosnia and Herzegovina, Kosovo and Albania, where SLS aimed at collecting data on consumption and omitted some income components, shows a level o f Gini far above that recorded for income, while measured expenditures exceed reported incomes significantly (such a difference i s over 17 percent in Serbia, but i s as much as 85 percent inMontenegro). 2.133 It i s noteworthy that across all countries in South Eastern Europe and the CIS, measures of inequality based on incomes exceed those based on consumption. Given the large unreported income, the measures of income inequality for these countries are clearly less accurate than consumption-based indices. Thus, for monitoring, measures based on consumption are o f primaryinterest. 2.134 Rather than looking at the Gini index per se, which tells us little about the economic factors affecting the poor, it i s more informative to attempt to see beyond the measured levels of inequality into its components. Total householdincome can be broken down into its components. Table 2.14 breaks down total income into its 11 components (sources) and shows their contributions to inequality (on a per equivalent unit basis). 2.135 As expected, the most important source i s labor income which accounts for 45.5 percent of total income. Since labor income i s distributed almost exactly as total income ( the concentration coefficient i s 33, almost the same as total income's Gini o f 32.2), labor income also contributes a little less than half o f total inequality. It i s very informative for the labor market analysis to see that non-agricultural self-employment income represents about 2 percent of total earnings and 1 percent o f total income. This i s significantly lower than in any other country in Europe and rises the question o f access to self-employment opportunities, which this report will address in Chapter 3. 2.136 Next come three sources which are about equally important: pensions (16.3 percent of total income), income in-kind (12.8 percent) and net income from agriculture (11.4 percent). Both pensions and income in kind are distributed more equally than total income. It is interesting 55 to note that consumption in kind, although somewhat more equally distributed than total income, i s not an equalizing source (Le., the poor do not receive more than the rich). Net income from agriculture has a fairly highconcentration coefficient (46) and this explains more than 16 percent o f total inequality, pointing to significant inequality inrural areas. 2.137 When these four key income sources are added (labor, pensions, consumption in kind and net agricultural income) one accounts for 85-86 percent o f total income and of total inequality. Other sources are fairly small to have much o f an impact on inequality. Only unemployment benefits are distributed in a pro-poor fashion3' in the sense that poor households receive more o f them in absolute amounts than do rich households. Their concentration coefficient is negative, indicating that they reduce total inequality. Table 2.14: Structure of Total Income and Sources' Contribution to Total Income Inequality (1) (2) (3) (4) (5) Amount in Sharein Concentration Contribution Contribution dinarsper total coefficient to Gini to total equivalent income (2)*(3) inequality (in unit (%I %) (4)kotal Gini Total laborearnings 3976 45.5 33.0 15.0 47.O Incomefromproperty 331 3.8 55.3 2.1 6.6 Net farm income 999 11.4 46.1 5.3 16.5 Pensions 1427 16.3 24.9 4.1 12.7 Unemploymentbenefits 36 0.4 -13.6 -0.1 -0.2 Socialassistance 30 0.3 30.8 0.1 0.3 In-kindconsumption 1121 12.8 27.3 3.5 11.0 Stipends 14 0.2 12.8 0.0 0.1 Healthbenefits 19 0.2 27.8 0.1 0.2 Imputeddurablesconsumption 412 4.7 30.0 1.4 4.4 Imputedrent TOL~ 382 4.4 10.4 0.5 1.4 income 8747 100.0 32.2 32.2 100.0 Source: Milanovic (2003) basedon SLS. All amounts are population-weighted in dinars per month per equivalent unit. 2.138 Finally, it should be noted that neither imputed income from durables nor imputed rent have much of an impact on inequality. Imputed income from durables has almost an identical distribution as overall income, implyingthat the distribution of durables is about the same as the distribution o f income. Imputed rent seems to be pro-poor in relative terms with the concentration coefficient of only 10.4. Thus, when we add housing and durables, their distribution i s less skewed than the distribution of income, an interesting and perhaps unexpected conclusion since we normally expect wealth to be more unequally distributed than income. However, this can be explained by the privatization of housing to their tenants o f the housing stock (which under socialism was allocated rather equally) . 31 A source is pro-poorinrelative terms ifthe share of that source inthe poor's income is greater thaninthe income of the rich. 56 2.12. Conclusions 2.139 By the year 2002 Serbia and Montenegro faced a complex poverty situation. Material (consumption) poverty was clearly a problem with at least 10 percent of the population in absolute poverty in both Republics and sizable numbers in relative poverty. But even more worrying finding i s that those who were vulnerable to poverty by being just above the poverty line were twice as numerous as the poor. 2.140 The most striking features of poverty in Serbia and Montenegro are the regional differences in poverty rates, and the extremely tight link of poverty with employment and education. There are also clear signs of the strong ethnic correlates of poverty, with the Roma facing severe disadvantages. Therefore, two challenges face policymakers and societies. One i s related to the threat of vulnerability to poverty. The second challenge i s related to the "structural" nature of poverty. 2.141 The strongest and closest correlate o f poverty i s education. Poorly educated individuals make up the majority o f the poor and have the highest poverty risk. The labor market i s strongly interacting with education, reinforcing its effects especially for particular groups that suffer from skills deficiencies. These groups have the lowest chances to benefit from growth. 2.142 These conclusions are based on the available data. Significantly more accurate analysis will be possible when the new panel data will become available (for Serbia). Why are panel data important? It i s commonly thought that the transiently poor-those who escape poverty in some years-are most likely to benefit from an improvement in economic conditions. They are vulnerable to economic shocks but have characteristics that allow them to escape poverty. In contrast, the persistently poor-those who are observed to be poor in every period o f observation-are thought to be an underclass that is unlikely to be able to take advantage of new economic opportunities. Following families over time through panel surveys makes it possible to examine the stability, or security, o f their economic situations and to develop the exact definitions of vulnerability thresholds. Panel data will also significantly improve the accuracy o f multivariate analysis, making it possible to control for some unobservable characteristics and thus to arrive at clearer answers regardingthe role of different drivers of poverty. 2.143 Finally, new data are critical to establish a better understanding of the ethnic dimension of poverty, especially the disadvantages faced by the Roma. Such data have already been collected in Serbia and Montenegro and the process o f their analysis will shed light on very important determinants of poverty. 57 3. THE POORAND THE LABORMARKET There i s little doubt that a well-functioning labor market i s essential for Serbia and Montenegro, as well as for other countries in transition, if the progress towards a market economy i s to be sustained and if the level of poverty i s to be reduced. Labor mobility, a skilled workforce, an entrepreneurial spirit, modem labor laws that balance employer flexibility with workers' rights, and an adequate social safety net are all key requirements for success of the poverty reduction strategy. But translating these principles into action requires good labor market data. The labor market in S A M i s a source of controversy. Many researchers point out various contradictions in the available data. They also highlight the dangers of over simplistic assessment of the link between labor market and poverty. This chapter attempts to convey the full richness o f activities that people engage in SAM. The chapter takes a critical review of the available evidence with the new data at hand (SLS, G- 17 Labor market adjustment survey and 2002 LFS), identifies current problems and provides an outlook for the future. 3.1. Challenges Ahead: Employment Reallocation. 3.1 It is generally believed that the slow pace of enterprise restructuring is at the heart of many of the labor market problems. Without hard budget constraints, enterprise managers and workers tend to behave in ways that run counter to longer-run economic interests (see World Bank 2002b for a summary). As already noted in Chapter 1, many firms appear to keeping on too many workers, running into wage arrears, using unpaid administrative leave. This leads to an inefficient use of resources, sub-optimal levels of human resource investment, and poor social protection for many of the affected workers. The restructuring issue has many dimensions, but the concern i s with its consequences for poverty and the potential role of policy reform in promoting the "creative destruction" process. (Box 3.1 gives an example, from another country, of the costs of postponing reforms. 3.2 There i s general agreement among labor market analysts in Serbia that restructuring and privatization would inevitably see a fall in total employment and a rise in unemployment. It i s recognized that destroying jobs in the state sector would be easier than creating them in the private sector, and that this would give rise to unemployment, some o f which might be persistent. Within this framework, there are disagreements as to how long and how severe the effects of transition - in particular the persistence of unemployment - would be.32This section will review the evidence in this matter as it concems Serbia and will not specifically discuss these issues as See World Bank (2002 b) and Boeri (1999)for a useful summary o f similar problems inother transition economies. 59 regards Montenegro, since, given the data limitations, these ca.nnot be addressed at the time of writing. 3.3 Although there i s general agreement that enterprise restructuring i s an essential catalyst for improving labor market functioning, it should be recognized that this i s a complex phenomenon. The success of enterprise restructuring depends on many factors, in particular on the business environment. The key framework for the on-going process o f restructuring in Serbia was set by a number of new laws. IBox 3.1. Costs Of PostponingDifficultReforms: The Case Of Bulgaria I Inthe 1990s, Bulgaria was avoiding decisive steps instructuralreforms, postponing the most difficult decisions. Often this was justified by referring to the extremely high unemployment rate. The imbalances grew untilthey were out of control, leading to a full-fledged macroeconomic crisis in 1997. After this crisis, macroeconomic policies were brought under strict control with a currency board regime, and a number of bold steps have been taken to achieve the long awaited progress in reform. But despite an impressive performance, which included large FDI inflows and deep restructuring, unemployment was still on the rise and by 2001 had reached 19 percent o f the labor force. Why did unemployment continue to grow despite the positive signs in the economy? The recently completed Poverty Assessment for Bulgaria found an explanation for this paradox in the long-delayed restructuring. As a result, all o f the employment reallocation in the economy began to take place at once when reforms started in earnest. The new demand from growing firms has yet to absorb the enormous amount o f redundancies from labor- shedding enterprises. On the other hand, deep structural imbalances between demand and supply on the labor market emerged which were hidden by a suppressedlabor adjustment. The Bulgarian scenario shows that the consequences of postponing key restructuring steps out of the fear of high open unemployment may, ironically, lead to even higher unemployment. I Source: World Bank. Bulgaria Poverty Assessment, 2002. 3.4 The new Privatization Law, adopted in June 2001, attempted to incorporate the lessons learned from the experiences of other transition economies. A wave o f privatization of socially- owned companies i s currently under way. All socially-owned enterprises are sorted into three groups according to their size. The large enterprises are sold through international tender, while the small and medium-size companies are auctioned off. Investors are invited to purchase 70 percent o f the shares in the company, while 15 percent o f the shares are given to the employees of the company and the remaining o f 15 percent i s put into an investment fund whose shares will be distribute to the population, at a later date. 3.5 The privatization also allows for the alternative process, which will be applied to 40 or so companies that the government judges are in need o f "sanitation" before undergoing the privatization process. A model example was set in 2001 with a major restructuring o f the car company Zastava of Kragujevac, with almost half of the employees (15,000 out o f 33,000) declared redundant. Approximately half of those made redundant opted for severance pay, and half chose to be retrained under a four-year fixed term contract with a separate unit called Zastava -Employment and Education, with the remuneration of 40 percent of their previous pay. 3.6 To facilitate restructuring, another important piece of legislation has been passed - The Labor Code (December 2001, Official Gazette of the Republic of Serbia, no. 70/2001). This 60 legislation i s one of the main institutional pillars o f reform and i s considered to be a necessary accompaniment to the Privatization Law. 3.7 The new provisions o f the Labor Code promote flexibility and are meant to fight informal employment by facilitating alternative employment, such as work at home, work outside the business premises of the employer and fixed term contracts. Wage setting, especially collective bargaining, was also considerably liberalized. Unions are required to prove their membership threshold, which i s set at 15 percent of employees in a firm, and 10 percent at branch and national levels. Although in firms where representative unions exist collective bargaining i s mandatory, reaching and signing collective agreements i s not. 3.8 Accompanying the Labor Code, the new Employment Law was passed in the fall of 2002. Unlike its predecessor, the new Law does not stipulate the criteria for the identification of surplus workers (such as work results, qualification, health and social condition of the employee andthe employee's family). The decision lies solely with the employers, who are allowedto use any criteria that they find suitable. Although employers are fully independent inthe identification of surplus labor, they are required to cooperate with the local employment office in designing the measures for reemployment. 3.9 The Law brought about important changes, offering less generous unemployment assistance schemes, in terms of duration and in terms of benefits. The replacement ratio was set at 60 percent of the average wage of the unemployedinthe last three months and may not exceed the last average reported net wage in the economy. After three months the replacement ratio drops to 50 percent. Together with other provisions, these changes are expected to discourage moral hazard behavior on the part o f the unemployed and to move the incentives structure away from reliance on assistance for as long as possible to the active seeking o f a job. Thus, the new labor market policy strategy developed by the Serbian Labor Market Bureau (LMB) in 2002 places emphasis on active rather than passive measures. 3.10 In order to make the restructuring of the privatized enterprises easier, a additional regulations for "surplus labor" have been enacted. An enterprise can declare a portion of its labor force "surplus" for technological, economic and organizational reasons and that would make the redundant workers eligible for severance pay.33The Law stipulates that the minimum severance pay should be between two and five salaries of the employee, depending on the employee's work experience. The intention of the legislator was that the program for surplus labor should be closely related to and/or a part o f the restructuring of the firms and other legal entities, -- the procedure for which is defined by the special regulation. The restructuring, in turn, i s supposed to precede the privatization. 3.11 It is clear that the new provisions of the labor legislation are intended inthe first place to encourage foreign and domestic strategic investors to buy firms offered on tenders and auctions. 33An employer with more than 50 employees who intends to cancel individual labor contracts with more than 10 percent of its employees during one calendar year for the above mentioned reasons, is obliged to prepare a "program for surplus labor" in cooperation with the employment Bureau (LMB).The programrequires that the informationbe given about the employees affected, specifically: their job posts, skill structure, and age structure, the measures undertaken for their re-employment, and the duration o f advance notice (not shorter than three months), according to Article 114. of the Labor Code. 61 Still, incumbent workers in to-be-privatized firms enjoy a certain degree of protection through the legal provisions regulating surplus labor.34Social Program gives lists the methods of "solving the socioeconomic status" of surplus workers. Surplus workers can choose one o f several options: (i) termination of employment and registration at the LMB,with the usual set o f rights (unemployment benefits, health, pension and disability insurance, etc.); (ii)increased unemployment benefits with the obligatory additional training; (iii) severance pay with fairly generous entitlements. 35 3.12 The "high-case" scenario of restructuring may look like the following. The new privatization law aims to introduce efficient corporate governance and to generate significant budget revenues by attracting strategic foreign investors for large enterprises and mostly domestic capital for small and medium enterprises. To facilitate privatization and investment, the Labor Law relaxes hiringand firing procedures, introduces flexibility in contractual employment forms, and breaks up the union monopoly in bargaining. Some large loss-making enterprises are restructured before privatization; redundant workers are provided with severance pay equal on average to 18 months' average net salary in the economy (the rule usually applied i s 100 euros per year of work experience); some of these workers use the money to start up their own small business, while the rest use it for survival until a new job i s found or retirement age i s reached. The government facilitates the creation o f small and medium enterprises to absorb the surplus labor influx from the state sector. The rest o f the surplus workers and new entrants (school leavers) are being absorbed gradually but at a fairly high rate through the revival of privatized firms andnew foreign and domestic direct investment. 3.13 The speed of economic recovery and the timely completion o f the privatization process are crucial in this scenario. If, however, privatization i s delayed for either political or economic reasons, the labor market situation may worsen further. Restructured enterprises could again become sources of labor unrest. Workers who are already laid off could be at risk for experiencing very long unemployment spells, permanent welfare (wage) losses even after re- employment, and impoverishment. 3.14 A middle-ground alternative might be the existence of an autonomous source of labor market restructuring, a sort of internal dynamics that i s independent of privatization. In order to explore whether there i s such a dynamics, we need to turn to the available data. As a response to lasting (over 10 years) economic uncertainty and, to various shocks, employees in Serbia have 34 In March 2002 the Ministry of Labor prepared a document (with an unclear legal status) called the Social Program, in which the rights o f surplus workers are further extended beyond the Labor Code regulation. Social program (full name: Social Program for Employees Becoming Redundant during the Process of Enterprise Restructuring and Preparation for Privatization, Bankruptcy and Liquidation), authored by the Ministry for Labor and Employment and the Labor Market Bureau, further specifies the procedure for handling surplus labor. It does not have the power of a legal act; however, given that its text received the official endorsement o f the government, it i s highly unlikely that managers of firms entering the privatization, program would be willing to ignore its content, at least without the silent approval of the government. 35 Surplus workers can choose between two options: either to receive ten average (net) wages in Serbian industry according to the last available monthly statistics, or to receive 6,000 dinars (approximately. 100euros) for each year o f work experience. In practice, it means that workers with more than 12-14 years' experience, depending on the average wage, will opt for the latter, while those with less experience will accept the first offer. Those who receive severance pay have forfeited their right to pecuniary unemployment benefits; however, they are eligible to receive other services from the LMB. 62 developed a number of coping and risk mitigation strategies that help them adjust to the imminent risk of enterprise closures and mass layoffs. Understandinghow workers affected by a partial adjustment of employment have survived i s important to buildinga realistic picture of the future status of living standards andpoverty when reforms havebeenimplemented infull. 3.2. Restructuring:How Many Have Been Affected? 3.15 Estimates of the extent of forthcoming reallocation vary dramatically. In 1996 the estimate based on detailed data on workers compensation by sectors was 600,000 workers, (30 percent o f the employed under consideration according to Posarac [19971). This labor hoarding was partially explained by regulations in the context o f sanctions (up to 1998, firms were banned from laying off). But the key explanation for unwillingness to shed labor i s the process of privatization, favoring insiders and soft budget constraints. The most recent direct estimate of the extent of excess labor uses the number o f workers on administrative leave (close to 150,000) as a proxy for the forthcoming restructuring. In 2000, economic trends for Montenegro estimated excess labor to be in the range of 1525,000 workers (10-15 percent o f employees), of whom 10,000 were on "waiting lists", and 20,000 were on forced unpaid leave. Many companies are under bankruptcy procedure (almost 30,000 companies in Serbia, employing 940,000) and many o f these companies report wage arrears. Thus in March 2003 almost 220,000 employees have not received wages (approximately 10percent o f all employed). 3.16 How much restructuring has gone on since the adoption of the new law? A full account is difficult to give since data are lacking for many important dimensions, and, as the next section will show, registration data are largely inaccurate. Establishment level data are not detailed enough to capture any changes inemployment at the enterprise level. 3.17 Three different surveys are used to assess the degree of ongoing adjustment: the SLS survey, conducted in Serbia in May-June 2002 with a sample of close to 20,000 individuals, the key data source for this report; the LFS survey, a regular ongoing annual survey conducted in October of a sample of about 11,500 individuals in Serbia and Montenegro (2002 and 2001 data are used); and a small-scale Labor Adjustment survey conducted by the G-17 in May o f 2002 of a nationally representative sample of 2,500 respondents. 3.18 These data combined show that by the end o f 2002 not much had happened yet regarding employment. Many of the previous practices of hoarding labor were still very much present, but fear o f the forthcoming restructuring was truly overwhelming. 3.19 The G-17 survey conducted in M a y of 2002 picked up the early start in employment restructuring. According to the survey, for May 2001-May 2002, overall, no more than 2.5 percent of workers (about 50,000) had actually been laid off during the year. These layoffs were very concentrated: while only about 22 percent o f salaried workers witnessed layoffs in their workplaces, in enterprises where layoffs did occur, 30 percent of labor resources were shed, on average. The share o f workers that saw their colleagues dismissed was highest among the majority state-owned enterprises (as much as 44 percent). Few of the workers, however, have become unemployed as a result of layoffs: at the time of the survey, only about one-third o f the 63 new unemployed (those who were unemployedfor less than 12 months) had been laid off, which amounts to an estimate of around 20,000 dismissed workers during the year entering the ranks of the unemployed and staying there. The majority (30,000) either found new employment or retired from the labor market. 3.20 This number has to be compared to the officially reported figure of around 40,000 dismissed in layoffs (redundancies). These numbers are not very large, but they suggest some intensification of the process of separations, especially given the working histories of all respondents: overall, as many as 320,000 able-bodied adults (less than 10 percent) experienced a closure of enterprise and dismissal at least once intheir entire working career. 3.21 The profile of workers that are affected directly i s very distinct. A majority among them describe themselves as "unskilled" or "semi-skilled" workers. It i s interesting to note that workers in firms undergoing privatization reported as much layoffs as a share of their employment as didworkers in state-owned or already privatizedenterprises. 3.22 Incontrast to the limited scale of ongoing layoffs, the fears have been overwhelming. As many as 62 percent of workers were afraid of future layoffs at their firms and as many as 23 percent believed that they would have difficulties finding employment if they were affected. These concerns were clearly related to the level of skills: 36 percent of unskilled workers were afraid of remaining unemployed after dismissal, while only 20 percent of skilled and clerical workers had similar fears. The latter set o f figures points to moderate optimism on the issue of layoffs: three-quarters o f workers considered themselves re-employable. 3.23 Incontrast to actual adjustment inemployment, adjustments inhours and pay, according to the G-17 survey in May 2002, was a predominant reaction of enterprises to tightening budget constraints. Some 40 percent of workers experienced delays during the previous year in payment o f wages, and 10 percent were at some point intime on forced administrative leave for economic reasons. 3.24 The SLS data were not specifically focused on labor market adjustment, but these data, owing to their much larger sample, can nevertheless provide more accurate estimates of those who went from being laid off to unemployment. 3.25 The SLS data also give better "stock" o f point-in-time estimates o f the share of workers affected by various indirect forms of employment adjustment. In a single week of May 2002, barely 1 percent of all workers had been subject to such an adjustment in hours, but a significantly higher 3 percent was found among workers employed at SOEs. Wage arrears have been the key form of adjustment: in May 2002, as many as 11percent o f all workers had back wages owed to them with delays exceeding one month, and these delays, again, were concentrated among SOE employees (20 percent were affected). For SOE workers, both forms of adjustment were often combined, and those workers on forced holidays also represented half of the cases of wages owed to workers. 3.26 Finally, the LFS data (2001-02) offer a unique opportunity to monitor some changes over time. Again, the survey i s not designed, as the G-17, to address a particular issue o f lay-offs, but it is nevertheless informative as it shows some trends over time. Consistent with the SLS 64 findings, the LFS shows the relatively minor role of forced leave as a labor hoarding strategy: in October 2002, only 3 percent of salaried workers were affected by the forced holiday. It is also interesting to note that this share dropped somewhat (from 4 percent) compared to 2001, and that according to the LFS such a practice affected many more workers in Montenegro (10 percent of employees). It also shows that 4 percent o f workers in Serbia did not receive their wages for the last month of work (this i s only a partial measure o f wage arrears and shows a new "inflow" o f affected workers). What i s important i s that the share o f workers affected by arrears increased compared to 2001, and that arrears were quite concentrated among state (socially) owned enterprises with almost 7 percent of their employees not being paid on time. It also shows that wage arrears and forced leave were often combined, especially in the SOEs where half o f those put on leave were also not paid. 3.27 The LFS data can also help clarify the impact of layoffs on unemployment. Focusing on the new unemployed in Serbia (those with unemployment for less than one year), the LFS data suggest that those dismissed in layoffs are still a minority. The total number of "fresh" unemployed in Serbia increased by 5 percent between 2001 and 2002, and one-third were those dismissed in layoffs. This gives a figure of around 20,000 dismissed workers becoming unemployed inone year - a flow estimate very similar to the one obtained with the G-17 survey. 3.28 The LFS data also highlight another tendency which tends to be overshadowed by concerns over layoffs: the severe cuts in the number o f workers on fixed term contracts that occurred between 2001 and 2002 and their inflow into the unemployed. These workers accounted for only 7 percent of the salaried employed in 2000, and many of them were women. By 2002, their share dropped to 5 percent. Thus, adjustment to employment when it is necessary seems to shy away from a "core" group o f employees and tends to be concentrated in marginal groups with little voice. This outcome is just opposite to what the new labor market regulation gears to. 3.29 Conclusions. Despite widespread fears about the massive scale of layoffs, the intensity of layoffs barely increased in 2002. Instead, wage arrears were widely practiced to avoid open dismissals and possible unrest. In short, it appears that in 2002 nearly 10 percent of the labor force was either without work or else was subject to some measure of hours/pay adjustment All in all, estimates of about 150,000 workers "being at the margin" are plausible and possibly represent a lower bound of the estimate o f excess labor. This suggests that significant employment adjustment is still to come. 3.30 Data from the three surveys cited above provide interesting insights into what is happening in the labor market. But their deficiency i s very clear: these data do not follow laidoff workers over time. Therefore, there i s no opportunity to assess the full complexity o f transition in the labor market triggered by layoffs. The data allow us to trace back the histories only of a minority o f those who have been affected by layoffs (and who were unemployed at the time o f the survey), but are not sufficiently complete to capture the main tendency. Tracer studies or panel data analysis have to be used to provide further guidance for policymakers. 65 3.3 Worker Reaction to Underemploymentand the Threat of Layoffs 3.31 The practices of putting workers on unpaid leave or of paying wages late were used to avoid direct layoffs in the past. As the previous section shows, these practices are still widespread. How are workersreacting to thesepractices?Have they already adjusted tofuture mishaps? 3.32 The G-17 survey provides a unique opportunity to see whether actual layoffs, or the fear of other forms of adjustment to employment are forcing workers to explore alternative strategies. The survey also collected information on the subjective assessment of respondents' living standardsandquite detailedinformation on consumption expenditures. These data can be usedto assess the impact of the labor market situation on the perceptions of poverty in Serbia. As shown by Krstic (2002) in the most comprehensive study of the labor market in S A M available to date, the key strategy o f workers reacting to forced adjustment in hours was additional work in the informal sector. Overall, the G-17 survey suggests that around 10percent of salaried workers are engaged in all kinds of additional secondjobs. Among those affected by wage arrears, this share i s significantly higher and reaches 40 percent. Among those affected by forced leave, the situation i s very different, as suggested by the data. Only a tiny minority among them i s able to secure a secondjob as a coping strategy inresponse to a reduction in hours. 3.33 This difference may be explained by two quite disparate reasons, and without a more detailed purposefully sampled survey it i s impossible to support either of them. One reason could be that forced leave as opposed to arrears i s an ultimate "last resort" strategy that i s practiced by enterprises located in very depressed areas, where not much alternative employment i s available. Another quite opposite explanation i s that workers are "rotated" in the pool o f those put on forced leave and each spell o f involuntary slack i s relatively short. The workers simply don't have time to find an alternative employment. These two explanations lead to different conclusions about the welfare implications of arrears, and possible policy responses, and therefore require further examination. 3.34 What are the consequences o f arrears and forced leave for poverty? They vary dramatically. Workers experiencing wage arrears do not have any higher incidence o f poverty thando other workers intheir sector. Results regardingthe incidence of the holding of secondary jobs explain why. On the other hand, workers on forced leave suffer from an approximate doubling of the poverty risk as opposed to other workers. However, as we have seen, these workers represent only a minority among those affected by adjustment. 3.35 Therefore, the majority o f workers have developed certain strategies to cope with open and hidden redundancies. But these strategies may be quite fragile and very region-specific. It i s telling that the predominant form o f additional employment is salaried unregisteredemployment, and that very few attempt to set up a business or become self-employed. 66 3.36 While layoffs and other forms of labor adjustment may not in themselves be a leading cause of poverty, the fear of layoffs has very significant effects on the perceived well-being of workers. Even controlling for household consumption level, ownership o f assets, education, location, and demographic profile of households, workers in fear of layoffs had drasticall low perceptions o f their own living standardthan didworkers with safer employment prospects.T6 3.37 All of the surveys reviewed above suggest that despite late wages, payment in kindand forced holidays, many workers have been slow to leave their jobs. Workers frequently remain at their jobs in the hope that they will eventually receive their back wages. They fear that if they leave they will never receive them. Others still believe that their enterprise may revive and become viable again. Workers over 45, who know their chances of finding new employment are very low and who understand the enormous transaction costs of relocating, have every reason to cling to enterprise jobs as long as possible. Despite late salaries, people also continue to work to preserve routine, to fulfill a sense of duty, and, through their work collectives, to maintain access to information, public resources, and professional contacts. 3.38 The G-17 survey shows that workers who are affected directly are becoming more active in searching for employment. Those who work in firms already undergoing reductions in staff have a 50 percent higher rate o f secondary job holding than do other workers. Fears o f future layoffs increased secondary job holding only marginally. Still, less than one-third of all workers in imminent threat of losing their jobs are doing anything to manage this risk. The analysis of the G-17 survey data shows that the key response of workers i s fear. It i s therefore unrealistic to assume that workers will suddenly become more active on the labor market and take action to meet the challenge o f labor reallocationthemselves. 3.39 Policymakers are well aware of this problem, as the discussion in section 3.1 suggested. The severance benefit package was designed to provide incentives for a more active response, specifically to take on the risk o f self-employment or own business. 3.40 Itis therefore instructive to look at the SLS data for a (small) subset of workers who have received severance pay and to study their particular behavior on the labor market. About half of the workers who reported receiving severance pay remained unemployed at the time o f the survey, and another half had found employment, mostly in the private sector, and some had moved to agricultural activities. But only 4 percent tried to set up their own business. When workers moved to employment, their poverty rate was low, but those who remained unemployed, had a high poverty rate. Thus, severance pay, according to these extremely limited and early data, were not providing much of a shelter against poverty, nor were they stimulating entrepreneurship. Section 3.6 addresses specific barriers that may explain this outcome i s not surprising. ~~ 36Results of multinominal logit estimates using G-17 data focusing on a group of workers which evaluated their living standards as "bad" or "very bad." Results available on request. 67 3.4. Labor Market Data in Serbiaand Montenegro 3.41 Understanding the labor market requires good data, and such an understanding i s at the heart of a poverty reduction strategy. It i s even more important given the recent experience o f transition. Economic and social transformations in ECA countries over the last decade have put their economies on higher, and more sustainable, growth paths. Intoo many cases, however, the changes have also led to higher unemployment and to more inequality and exclusion. In Central and Eastern Europe, unemployment jumped to about 15 percent of the labor force in the early phases of the transition to a market economy and, despite strong economic growth in the second half of the 1990s and early 2000s in some countries, the unemployment rate has often remained well above 10 percent (or even close to 20 percent in Poland, the Slovak Republic, and Bulgaria), and labor force participation rates have declined steadily. These developments made growth in these countries less pro-poor than it should have been. It i s therefore essential, given the need for pro-poor growth in SAM, to get as objective, clear and regularly updated a picture o f the labor market as possible. 3.42 Nevertheless, labor market data in S A M are a source of confusion. This section attempts to sort out some of these apparent differences, first by finding more appropriate measures of employment and unemployment, and then by assessing the scale o f flows on the labor market. It i s useful to have a picture of the general framework in mind, and this i s presented on Figure 3.2. At this time the analysis will focus on establishing the right scale for each of the main labor market states, represented in the figure by rounded shapes: unemployment, employment and the out-of-labor force (inactivity). 3.43 Some methodological remarks are helpful. Each person in a county's working age population has to be classified as belonging to one of the states. Those who hold secondary jobs have to be accounted for, but should not confuse the basic framework. Otherwise, double counting will render the data meaningless. 3.44 The second principle i s the priority o f active states in the classification. If a pensioner is working according to the usual labor market data standards, the pensioner has to be classified as employed, and not inactive. Those unemployed who report "additional" work have to be counted as employed on their job. The second active state i s unemployment. If a student i s looking for employment, and i s ready to start working, he or she i s exercising pressure on the labor market participants and therefore has to be considered as unemployed, not as a person outside the active population. 3.45 Unemployment.There are large discrepancies among various sources on the employment situation in Serbia and Montenegro. According to administrative data, the number o f 68 unemployed i s about twice as high as in the LFS data (close to 800,000). However, it i s believed that the LFS data are more likely to approximate reality owing to various incentives to register - mainly the availability of pension and health insurance. As a result, a large fraction of the registered unemployed are not genuinely unemployed (i.e., they either have a job in the informal sector, or are not actively looking for work and are available for work). But what i s the real correspondence between the registrations and the actual economic state? Are the real unemployed the same people who are registered or are they totally different? The Labor Force Survey (LFS), which has been undertaken regularly by the Federal Statistical Office since 1995 (in 1994 it was only a pilot survey) in combination with the SLS gives a unique opportunity to answer this question. The result i s reported inFigure 3.1. Figure 3.1. Serbia: DifferencesinSurvey and Registry Unemployment, 2002 , 100,000 350,000 ILO Unemployed 250,000 Register Unemployed according to LFS llL0 unemployed I 750,000 registeredat LMB 200,000Registeredunemp. ILO Employed I 300,000 Registered unemp. ILO Inactive Note: Dataon registeredunemployed are extrapolations, as only 550,000 are reported to be registered accordingto SLS; LFS registrationdata are closer to the administrative figure, but they relate to a different month. Numbersare rounded to give an approximate representation . Sources: LFSand SLS. 69 Figure 3.2: Labor Market Flows m ,/ No benefits I I \ Privatized B. Public admin. D e novo Private ............................................................... I I Unregistered businesses/work Informal out of I Occasional ~ ~ d l ~ I workers Education Retirement, I HouseworkI I Emigration I Disability Source: Adapted from Dutz et al, (2002). 70 3.46 Figure 3.1 shows that only one-third o f those registered as unemployed in the LMB, according to the 2002 survey data, are truly (LO) unemployed; the remaining two-thirds are either working or inactive according to ILO standards. In fact, one-quarter o f all registered unemployed are actually working (predominantly as unregistered workers, but also as farmers) and 40 percent are not actively looking for a job or have no intention o f working and consequently are considered inactive according to ILO standards. This distribution i s close to those observed in Bosnia and Herzegovina (see Labor Market Study, 2002) and Croatia (see Bisogno 2001). 3.47 The larger number of registered unemployed persons than the L O unemployed can be attributed to the underlying incentives to register as unemployed in the LMB, and the fact that the LMBs have relatively little capacity to carry out effective monitoring that would ensure that the worker i s actually seeking ajob or i s actually without ajob. The incentives to register come not so much from the fact that it is necessary to register to collect unemployment benefits in cash since, on average, only percent of those registered in the LMB receive the cash benefit. Even those who do not collect this benefit need to register to have access to a broad range o f other benefits, including: health insurance, a continuing record of pension contributions, social assistance in cases of qualifying for it, subsidized child care, subsidized local transportation, and access to several active labor policy programs. 3.48 This analysis partly explains why the unemployment rate reflects the most unreliable data in the labor market statistics for SAM. Other reasons for this come from the definition of employment. 3.49 Official employment data are not more reliable than the registered unemployment figures. There are at least three different sources of information for these figures: (i) aggregated employment reports of enterprises; (ii)estimates (unpublished procedures) based on establishment surveys; and (iii) LFS survey data. Summary employment numbers (the sum of full-time employment reported to statistical authorities) for all enterprises i s available now for a very limited range of firms (200 employees and more) and gives a number on the order o f 1.3 million. employees. Establishment surveys are supplementing this number with data on SMEs that give magnitudes near 0.6 million. The total o f about 1.9 million is for registered employment, or "the employment" in SAM. 3.50 For some unclear reason, farmers are not considered as "employed" and therefore are not taken into account. And, of course, all other "unregistered" activities (such as work in a family enterprise) are outside of employment. On the other hand, while relying on the "solid" evidence of the register, the source combines it with estimates based on a periodic survey o f small enterprises. 3.51 The so-called registered unemployment rate i s based on these administrative sources. The LMB provides the number of unemployed that register in that office, while the register of the Statistical Office and its estimate o f employment at SMEs provides the total of employed persons. Table 3.1 shows how the two are combined to obtain a 32 percent unemployment rate for 2002. 71 3.52 The LFS shows a much larger employment figure, which i s not surprising. It shows informal and part-time employment. It also qualifies all those without work between active jobseekers and inactive dependants. But in doing so it also makes some rather unusual assumptions. Most of these assumptions come from treating many workers as "active" as opposed to "employed" (see Table 3.1 for details). Thus, some publications omit completely farmers and family workers ("FRY" definition). More recent data use the approach that i s much closer to the international definition ("LO" inthe table) but still make one assumption that i s not strict. This assumption concerns the treatment of those respondents who claimed that they were looking for jobs, but also reported having "additional" work during the survey month, although they were not at that work during the survey week. In other countries, such a group i s often counted as employed, and their re-classification ("Strict") puts the unemployment rate at a more comparable level. 3.53 Thus, the survey-based unemployment rate coming from the Labor Force Surveys can be made broadly consistent with L O and OECD specifications. Table3.1: SAMEmployment,UnemploymentandUnemploymentRates Total (1) 6,699 6,646 6,653 6740 Active-total(3)+(4)+(5)+(6)+(7)+(8) (2) 3,583 3,805 3,810 3790 Wage and Self- Employed (Non Agricultural) (3) 2,299 2,332 2,320 2224 Farmers (4) 530 498 542 544 Helping Family Members (5) 171 166 180 186 Other (6) 43 40 33 49 Unemployed seeking employment, even ifperiodically working (7) 528 481 490 539 Currently not employed and not seeking but periodically working (8) 282 288 244 246 Memo: unemployed seeking employment without periodic work (9) 368 356 375 401 Non-active (10) 2,846 2,841 2,843 2950 Unemployment rateFRY definition =(7)/((3)+(7)+(8)) 17.0% 15.5% 16.0% 17.9% Unemployment rate ILO definition = (7)/((3)+(4)+(5)+(6)+(7)+(8)) 14.8% 13.7% 13.7% 15.2% Unemployment rate strict definition = (9)/((3)+(4)+(5)+(6)+(8)+(9)) 10.0% 9.7% 10.2% 11.0% Memo: Employment based on registry (11) 2,298 2,238 2,014 1964 Memo: Unemployment based on registry (12) 811 806 850 923 Memo: Unemployment rate based on registiy (11)/((11)+(12)) 26.1% 26.5% 29.7% 32.0% Sources: FSO Statistical Year Book, RSO, MONSTAT and LFS for 2002. 3.54 The choice of the proper unemployment rate is not a minor issue since the difference between the two rates i s substantial. The number of registered unemployed persons averaged some 920,000 persons, while the number of unemployed persons as per the LFS was roughly 40,000. This difference between the registry and the LFS in the number o f unemployed is among the largest in Central Europe, and i s second only to Bosnia and Herzegovina (in percentage terms). 72 3.55 For the reasons stated above, the LFS-based strict measure o f unemployment i s less subject to distortions and i s consequently more reliable, as it i s free of incentives to cover up the true labor force status. In addition, the ILO indicator o f unemployment, unlike the registered unemployment rate, i s internationally comparable, as it i s not subject to the international variations in registration procedures and regulations. Therefore, being more comparable across countries, the LFS-based unemployment rate brings us closer to answering the question of whether unemployment in SAM i s high. 3.56 Nevertheless, it i s worth noting that the ILO unemployment rate i s not exempt from limitations either. A person may be willing and able to work and available for taking ajob but on thbasis of his or her past experience may consider the effort to continue searching for ajob to be futile, as he or she understands that there are no jobs available for a person of his or her personal characteristics. This i s the case with the so-called discouraged workers. The current design of LFS does not allow it to measure the size of this group. Nevertheless, the inter-country comparability of the ILO unemployment rate i s not questionable. The registered unemployment rate, on the other hand, i s hardly comparable across countries. 3.57 According to the LFS-based unemployment rate, and compared with other countries in the region, unemployment in S A M seems high (Table 3.2). Although unemployment using restrictive measures may appear moderate with respect to other transition countries, and may not be extremely highby EUstandards, the very fact that as many as 4 percent of workers are in the gray area between true unemployment and seasonal, casual periodic work, is alarming. As important as the level of unemployment, however, i s its trend. Inthis case, by whatever measure i s used, unemployment has not been growing throughout the past years. Table 3.2: Unemployment inSelectedTransition Countries, 1993-2002 Serbia and Montenegro* 14.7 14.7 15.1 15.1 14.8 13.7 13.7 15.2 Serbia and Montenegro** 11.0 10.7 10.7 10.4 10.0 9.7 10.2 11.1 Bulgaria 21.4 20.0 15.7 13.5 13.7 12.2 14.0 16.3 19.4 Croatia 10.0 9.9 11.4 13.5 16.1 15.8 Czech Republic 4.3 4.3 4.0 3.9 4.8 6.5 8.7 8.8 8.1 Estonia 6.6 7.6 9.7 9.9 9.6 9.8 12.2 13.6 12.6 Hungary 12.1 10.7 10.2 9.9 8.7 7.8 7.0 6.4 5.7 Latvia 18.9 18.3 14.4 13.8 14.5 14.6 12.8 Lithuania 17.4 17.1 16.4 14.1 13.3 14.1 15.4 17.0 Macedonia, FYR 31.9 36 34.5 34.2 32.2 30.5 Poland 14.0 14.4 13.3 12.4 11.2 10.7 12.5 16.1 18.2 Romania 8.2 8.0 6.6 6.4 6.8 6.2 7.1 6.6 Russian Federation 5.5 7.4 8.8 9.3 11.2 12.4 13.0 10.5 8.7 7.1 Slovak Republic 12.2 13.7 13.1 11.3 11.9 12.6 16.4 18.8 19.3 Slovenia 9.1 9.0 7.4 7.3 7.1 7.7 7.4 7.2 5.9 Ukraine 5.6 7.6 8.9 11.3 11.9 11.7 11.1 EU-15 10.7 11.1 10.7 10.8 10.6 9.9 9.1 8.2 Note*: Broad LFS measure,** Strictdefinition. Source: OECDdata, basedon the LFS of each country, staffestimates for Serbia 73 3.58 The analysis of the trends in 1990-2002 in the first chapter highlighted the mounting problems and the extremely slow response regarding employment generation to the positive changes over the last two years. Figure 3.3, which i s based on the most consistent measure of participation from LFS, further underlines the fact that a stable unemployment rate by itself does not mean good labor market performance. Figure 3.3: ParticipationRates, 1995-2002 (by gender and Republic, to population over 15 years; percent) 75% - 70% Male: Serbia 65% 60% 0 Female: Serbia 55% 50% +Male: Montenegro 45% - 40% A .Female: Montenegro 35% 1995 1996 1997 1998 1999 2000 2001 2002 Source: LFS (1995-2002). 3.59 Figure 3.3 suggests two basic facts: (i)a continuous fall in participation rates; and (ii) significantly low participation rates for women, especially in Montenegro. Figure 3.4 makes the case that stable unemployment rates for the country as a whole hide very large differences in partsof SAM andby gender. 3.60 Figure 3.2 also provides a useful framework for capturing the broad picture of labor market reallocation in S A M during the transition and especially the magnitude of the challenge o f employment restructuring. 3.61 Using LFS data, we can roughly estimate total flows (more accurate analysis would require panel data). So far, as data reported in the PIER (Vol. 2, p. 46) suggest, the largest and the least predictable flow in the labor market was new retirement (labeled as E+D on the figure) -betweenover 120,000 and60,000 ayearfor Serbiaalone. Thisflow toinactivity, however, is relatively simple and does not involve much of a strain on the labor market. As the data show, some of the flow i s reversed in subsequent years with some retirees going back to the labor market in search of jobs (A), or going to employment, often agricultural (reverse D) but these are fairly anemic transitions. There are about 100,000 young adults entering the labor market each year, most of whom end up in unemployment for the first several years. What i s more important i s that the usual annual flow in the key transition from employment to unemployment 74 (B, C and F) is about 250,000-300,000 workers (with an increase in 2002), about half of whom are registered at the LMB and gradually are absorbed. Thus, laying off 150,000 active working age participants will considerably increase the pressure on labor market channels. It i s not, however, an unthinkable load. Figure3.4: ILOUnemploymentRatesfor Adults (Populationabove E),by Gender 30% -Serbia: Men Unemployment 25% +Serbia: Women 20% Unemployment 15% * Montenegro: Men Unemployment 10% 0 Montenegro:Wom en Unemployment 5% 1995 1996 1997 1998 1999 2000 2001 2002 Source: LFS (1995-2002) reported in Jovanovic (2003). 3.62 What i s more worrying i s that the growth process i s not associated to date with job creation. Ifthis situation continues, it may fail to reduce poverty. Moreover, a well functioning labor market can be key to a business climate in which new firms are created and private agents find the proper incentives to invest and innovate. Sound labor market conditions are needed to guarantee the success o f structural reforms, to maintain social support for those reforms, and to ensure that the benefits o f such reforms are widely distributed. Achieving these targets i s crucial to the effective implementation o f poverty reduction strategies. 3.5. Unemployment:Its Characteristicsand Dimensions 3.63 Table 3.3 presents poverty according to the labor market status o f the population based on the respondents' self-declaration. The poverty indicators differ a great deal according to this status. Within the labor market participants, the unemployed were confronted with the highest poverty risk (59.4 percent higher than the population average), and with the greatest poverty depth and severity. 75 3.64 Other labor market participants had a below average poverty risk, except for the category o f "other actives" (mostly unpaid family workers). When looked at observing according to the settlement type (as reported in Krstic[2003]), the unemployed in rural areas were the most vulnerable since their poverty risk was twice as high as the population average, or 39.4 percent higher in relation to the rural population average. In contrast, the employed in urban areas enjoyed the most favorable position - their poverty index was more than twice as small as the population average, and was 38.6 percent lower than the urbanpopulation average. Table 3.3: Poverty According to Socioeconomic Status inSerbia, 2002 (self-reported population over 15 years of age) Vulnerable, % of Poor, %of Relative Share inthe Share among Poverty Poverty population population poverty rate population the poor depth severity Employees 12.7% 6.2% -42.0% 30.6% 17.7% 1.2% 0.4% Employers and self-employed 16.3% 9.2% -13.3% 5.1% 4.4% 1.6% 0.5% Farmers 23.3% 10.6% -0.5% 5.3% 5.3% 2.0% 0.7% Other active 23.1% 13.2% 24.6% 1.4% 1.7% 2.7% 0.8% Unemployed 29.2% 16.9% 59.4% 11.9% 19.0% 3.6% 1.2% Pensioners 22.0% 10.9% 2.9% 24.1% 24.8% 2.4% 0.8% Other non-active 23.7% 13.3% 25.2% 21.7% 27.1% 2.9% 0.9% Total 20.2% 10.6% 100% 100% 2.2 70 0.8% Note: The relative poverty risk i s calculatedinrelation to the poverty index of the referencepopulation. Categoriesare not defined on the basis of the ILO standarddefinitions. Source: Krstic (2003). 3.65 Thus, unemployment i s a very strong predictor of poverty. The previous section discussed the issue of defining unemployment and pointed to a large gap between the economic definition of the unemployed and formal status, which is supposedly reported inTable 3.3. 3.66 To address this question, Table 3.4 lists the key characteristics of ILO and registered unemployed as reported in the SLS in 2002 and group-specific poverty rates. In total, fewer people have declared themselves among the registered in SLS than the official data would suggest. This i s partly due to the design of the module, which asked the question on registration only o f those who claimed to be out of work and unemployed (thus those employed and inactive may have been undercounted), and partly due to problems with covering all IDPs, some of whom may be registered at the LMB. 3.67 The Table 3.4 suggests that while economically the ILO definition is superior to registration, from the point of view o f profile and/or poverty correlates, both are equally serious. Those unemployed according to ILO criteria are somewhat more likely to be poor as a whole, and for most of the categories, but the similarities are striking. 3.68 The long duration o f unemployment rates supports the observation that an underclass may be forming. Data available for several Central and South Eastern European countries show that much of the unemployment i s long term in nature. For example, as much as 80 percent of all o f the unemployed inFYR Macedonia are long-term unemployed, and this pattern i s very similar 76 across socioeconomic groups, in contrast to the more typical pattern found in Europe, where the incidence of unemployment increases with age. Inthis respect, unemployment in S A M i s not yet as malignant as it i s inMacedonia, but the current state i s alarming. 3.69 However strong its link i s to material poverty, unemployment, as noted in Chapter 2, i s a poverty dimension on its own. And some particular features of unemployment render it a very worrying phenomenon: its long duration, the concentration among young adults, new entrants, and those with vocational education. 3.70 Regional data suggest that there are large differences across regions in very long-term unemployment rates, (see Table xx in the Annex). Data from the LFS strongly support this finding. In the highest unemployment region (Western Serbia) the ILO unemployment rate is three times as high as in relatively well-off Belgrade. Vojevodina also has a very high unemployment rate despite its low poverty risk, which points to serious problems in this I dimensionof living standards and shows the results of pressure placed on the labor market by the inflow of refugees andIDPsin this region. Table 3.4: Characteristicsof the Unemployed in Serbia: Registeredversus Strictly Defined ILO Unemployed Registered Unemployed Structure of which poor: Structure poor: Total number of ILO 260,000 100 19% Total number 550,000 100 17% Unemployed of Registered Unemployed Educationattainment Educationattainment Elementary 21.65 32% Elementary 23.18 33% Vocational 64.7 17% Vocational 65.14 13% Secondary and post-secondary 13.65 10% Secondary and post-secondr 11.67 8% Age group Age group under 30 59.65 18% under 30 52.5 16% 30-45 28.06 18% 30-45 32.44 16% 46-60 12.29 27% 46-60 15.07 22% Female 49.29 15% Female 50.79 14% Rural 36.78 19% Rural 43.5 17% With previous job exp 42.9 17% With previousjob exp 43 15% Duration of Duration of unemployment unemployment Less than 1year 36.76 12% Less than 1year 36.76 11% Over 1 less than 3 27.74 7% Over 1 less than 3 27.74 11% over 3 35.51 31% over 3 35.51 20% Injobless household 25.47 42% Injoblesshousehold 14.5 35% Memo: womlation r l 7.470.000 10.60% 7,470,000 10.60% I _ Source: Staff estimates based on SLS 2002. 77 3.71 The link between poverty and unemployment would be much stronger if unemployment data were to concentrate on householdheads instead o f on their children, as i s currently the case inSAM. 3.72 In general, the earnings coming from the household head and his or her spouse are the most important source of income for the household. Therefore, in general the unemployment of either of these two household members i s considered to be more detrimental for the family than the unemployment of other members of the household, say the children. In this respect, unemployment among household heads i s much less important in S A M than in other countries (the EU i s taken as a comparator for lack of comparable data). Whereas in the EU a large proportion of the unemployed (40 percent) are heads of household, as opposed to spouses (24 percent) or children of heads of households (32 percent) - in S A M most of the unemployed are children of household heads (33 percent). Only 14 percent o f the unemployed are household heads, while spouses of household heads are another 26 percent. Among men, the differences from the EU average are more dramatic. In SAM, 22 percent of unemployed men are heads of households, while in the EUthey account for 58 percent. It i s interesting to note that the use o f LFS data fully confirms these results. Such a low share of the unemployed among the heads of households is in contrast even with neighboring Croatia, where head of household males account for about one-third of all ~ n e m p l o y e d . ~ ~ Figure 3.5 Change inAge Specific Unemployment Rates, Cumulative 1995-2002 40.0% 30.0% 20.0% 10.0% -all , 0.0% ,*male , A females -10.0% -20.0% -30.0% -40.0% Source: LFS 1995-2002. 3.73 Unemployment in S A M i s highest among the young. While the overall unemployment rate reached 11percent in 2002; the unemployment rate for those between 18 and 24 years old was 36 percent and was increasing over time, as suggested by Figure 3.5. These young 37See World Bank (2000), "Croatia: Economic Vulnerability and Welfare Study," for sources. 78 unemployed persons represent some 16 percent of total unemployment. Although it i s worrying, the concentration of unemployment among the young i s less marked inSAMthan inthe average Central European transition economy, and in Croatia and the EU. (For the EU, 21.2 percent of this age group is unemployed in the group of first round candidate countries for EU accession [16.9 percen] and in the group of second round candidates C21.8 percent], and for Croatia the figure i s about one-quarter.) 3.74 Figure 3.5 shows that it was not only the youth who originally had high unemployment rates that was affected by the unemployment increase, but that older pre-retirement workers have also gradually movedinto a higher risk zone. 3.6. The WorkingPoor: ` % o o ~and~"Bad" Jobs ~ 3.75 As Section 3.2 demonstrated, employment does not guarantee pay in SAM. More detailed analysis available in a thorough study by Krstic (2002) sheds additional light on how labor participation and earning differences per worker are correlated with poverty outcomes. As noted above, the poor have lower labor force participation rates than the non-poor and higher rates of unemployment. Even when the poor work, they work in less well-paid and less secure positions than the non-poor. 3.76 One of the most striking findings of the SLS reported in Krstic (2003) i s a highincidence of poverty among workers employed in the informal sector. Those who have their principaljob inthe informal sector (30 percentof all workers according to the SLS) have a poverty rate higher thanthe averageby one-third. However, as the analysis inthis chapter suggests, the low earnings observed for some workers in the informal sector are often a result of their overall extremely poor standing in the labor market given their low education, their gender, their ethnic background, etc. These marginal workers are the first to be pushed out of formal employment into informal economic activities, where not many earningopportunities are open to them. 3.77 Krstic notes that earnings in SAM are very strongly correlated with observable characteristics, such as education, age, or experience. The poor are also over-represented in the sectors that pay the lowest wages. Other factors, however, add to the earnings differentials between the poor and the non-poor. The poor tend to work shorter hours. Often, this lower pay i s the result of being forced to take involuntary leave or to accept reduced hours. And being a woman i s also a riskfactor (see Box 111.2) 3.78 Poorly educated households are at higher risk of poverty. Although pay scales were compressed and did not necessarily reward higher education, this i s changing. Education and experience are rewarded in much the same way as in OECD countries, though there i s some evidence that the country i s headedtoward the premium earnings peak ineducation that occurred inother transition countries early. Thus, all elsebeingequal, low levels of education would result in less pay and thus correlate with higher levels of poverty. This is indeed what the poverty profiles show: individuals with a secondary education have a lower relative risk of poverty than those with a primaryeducation. 79 BOX 3.2: Gender and Work inSAM The low gender gap found in the earnings regressions (reported earlier) may hide significant unobservable barriers that women face in entering the wage employment altogether. More children may mean less female labor force participation and higher rates o f poverty. Analysis of the Labor Force Survey finds that if other personal and household characteristics are held constant, the number of children below the age of 7 and the number of elderly ina household are significant factors affecting the probability of urban females contributing to the labor force. This finding suggests that one of the causes of females dropping out o f the labor force in urban areas might be the unavailability o f childcare and household responsibilities assumed by women in extended households. Women's responsibilities for care o f the children and elderly may make it more difficult for them to find and keep jobs, even when they want to enter the labor force. Unemployment rates do not exhibit a strong systematic bias against women. Many qualitative poverty studies reported frequent complaints of discrimination against women in the labor market. Women complained that many factors that normally work in favor of men frequently worked against them, even for unskilled jobs. Source: Krstic (2002) andUNDP (2002) andISSP-CEED. 3.79 Other variables that are found relevant to explaining remuneration levels are the regions inwhich the worker lives andthe sector inwhich the worker is employed. Regional effects are strong and persistent even when the full set of controls i s introduced, suggesting the existence of a regionally segmented labor market. 3.7. Private Sector Development: How Can It Serve the Poor? 3.80 Dynamic labor markets must go beyondjob of shedding through enterprise restructuring, and must also include the creation of new jobs. The S M E sector has an important role to play in this dynamic process, although this role is more complex than is sometimes believed. 3.81 Nonetheless, where analysis has been undertaken, it i s clear that much o f the dynamism in labor markets is centered in the SME sector, in terms of creating new jobs and in terms of destroying existing ones. According to the OECD (1994), in some countries the key source of new jobs i s new firms, while in other countries it i s the expansion of existing firms (often in the S M E sector). 3.82 Mobility and entre reneurship issues in transition economies are generating increasing attention in the literature,P8 but they are far less often studied in the context of Serbia and Montenegro. To a large extent, enterpreneurship and self-employment are regarded as only marginally important. A cross-country empirical analysis of household data suggests, however, that entrepreneurship has been a very successful high earnings strategy for individuals able to actively adapt to the t r a n ~ i t i o n . ~ ~ 38Recent examples include Lehmann and Wadsworth (1999) on Poland, Scharle (2000) on Hungary, and Sabrianova (2000) onRussia. 39Mark Dutz, Celine Kauffmann, Serineh Najarian, Peter Sanfey and Ruslan Yemtsov, "Labour Market States, Mobility and Entrepreneurship inTransition Economies,". EBRDWorking Paper, November, 2001. 80 3.83 Krstic (2002) has looked at the S M E sector in Serbia and Montenegro and concludes that there i s a striking absence of new businesses and a lack of small enterprises. Even using the broad definition o f SMEs, this sector still accounts for a small part of total employment. 3.84 Figure 3.6 uses a most internationally comparable indicator based on LFS data that shows the role of micro enterprises and entrepreneurs in creating employment. It shows the share of the non-agricultural self-employed and small family firm employers in total non-agricultural employment. Figure 3.6 i s sobering. It shows that among all countries in CEE and SEE for which comparable data exist, Serbia has the lowest index: hardly above 2 percent. Section 2.10 has noted that the role of self-employment income i s close to nothing in the structure of household incomes in Serbia, and this finding further highlights significant problems with this activity inSerbia. Figure 3.6: Non-agricultural Self Employment as Percent of Total Non-agricultural Employment 3.85 Barriers to the development of small and micro enterprises can be traced to the high level of taxation and to poor access to credits. High rates of taxation in particular (taxes on labor) can work against job creation. In addition to high corporate income tax rates, Serbian and Montenegrin firms face high rates of social security payroll contributions. High payroll taxes increase the cost of labor and create a wedge between labor costs and wages that burdens employers and discourages labor supply. While this problem i s common to many countries, it may be particularly acute in Serbia and especially in SMEs, for whom the costs of complying with payroll taxes and with the necessary administrative procedures are likely to be higher than for large firms. There i s a growing literature to the effect that high payroll taxes may introduce a bias against unskilled labor, since the wedge is more burdensome for lower incomes. For higher-skilled workers, business can more easily separate compensation into a wage and non- wage component; in addition, it i s easier to understate true incomes for workers at the top of the scale. Moreover, high payroll taxes and non-wage costs are more burdensome in proportional terms inpoorer regions. 81 3.86 An analysis provided by the EBRD using the recent BEEPS data focuses on factors outside o f the labor market to explain the slow rate of enterprise creation and the relative paucity of small firms. It includes a number of deterrents includingdiscrimination in business licensing, blocked access to distribution networks, corruption, real estate problems, and regulatory uncertainties. On most of these counts S A M falls below the average transition economy. 3.8. Rural Incomes and Poverty 3.87 Rural poverty appears to be significant, and growing. With over 1.1 million poor living in rural areas, rural poverty i s a significant problem, as suggested by the data from the Federal Office of Statistics. The rural poor accounted for a significant share of the total number of poor in the country - almost 40%, and rural poverty grew from 1995 to 2000 by 28%, while urban poverty grew by 21%. The high rates of poverty in rural areas o f S A M are partially a historical phenomenon, but they also reflect the immense dislocation experienced by the rural sector immediately following the breakup of the SFRY. The production and marketing structures, the incentive framework, the provision of social services, and farm and non-farm employment opportunities were profoundly affected. 3.88 Untilrecently, a widely heldview was that rural areas were faringmuchbetterthan urban areas given their access to land and their ability to produce their own food, fuel, etc. As a result, little i s known about rural poverty, including the impact of growth of recent macro and sectoral reforms on rural livelihoods and well-being. The poverty analysis in Krstic (2003) has examined aggregate rural poverty levels and has discussed trends, but it has not touched on examining the possible determinants, such as access to land and livestock assets, and participation in agricultural product and factor markets. 3.89 Despite the capacity to produce some of its own food-which may keep many rural poor out of extreme poverty-conditions in rural areas are believed to be difficult. Households interested in farming lack the means to expand into more profitable farm activities. They tend to be cash-poor and therefore they rely on barter. Being cash-poor, with few opportunities to obtain credit, these households are unable to branch out into more profitable crops because they cannot afford equipment, fertilizers, pesticides, seeds, or irrigation, and they often lack able-bodied adults. 3.90 Table 3.5 shows key variables reflecting the asset ownership, human capital characteristics and livelihood sources for an average rural household among the poor and the near poor (the next 10 percent o f the population) and by quintiles o f the distribution. Although the household level data are now being analyzed to produce a thorough assessment, some evidence i s so strikingly clear from the Table that leads can already be given at this stage. 3.91 Ownership of agricultural assets is fairly widespread: a large share of rural households own some land and livestock, and a significant number o f households own some heavy agricultural machinery. According to the most recent SLS data, about two-thirds o f all rural households own some land, with an average holding size o f approximately .5 ha. Only 64 percent of land owned i s typically cultivated by the owner. While there is no direct survey 82 information on "land not cultivated," the rest i s probably either rented out, cultivated by relatives, or sits idle. About 70 percent of households own some livestock. Over one-third o f all rural households own a tractor, and almost one-quarter (22 percent ) own a harvester. 3.92 Many households have other income sources besides farming, or are only employed in agriculture part-time Wage income forms the most share of total household incomes in rural areas (34 percent of the total), and off-farm work i s common (over one-half of households contain a wage earner). Census data from 1991 indicates that at that time, a large number o f rural households were part-time farmers. The dominant role o f wage income in total income reported in the most recent household survey, suggests that this pattern may have continued. The next largest income source i s farm income (23 percent of total income) followed by pensions (14). 3.93 Household labor and capital endowments do not vary much across income classes, despite the income gap between the highest and lowest quintiles. We look at the correlates o f low income by examining household characteristics across quintiles. The expenditures o f the highest quintile are higher than those o f the lowest quintile expenditures by almost a factor of two. This income gap i s similar to that in rural areas of Bulgaria and Romania. Despite this gap, household endowments of human and do not vary much across income classes (with the exception that the lowest quintile averages 7 years of schooling while the highest averages 9 years). The percentage of households owning land does not vary much by income class, indicating that for those households lacking land, many are able to compensate with non-farm forms of income generation. Households in the lowest quintile have a slightly higher endowment o f calves, dairy cows, horses, pigs which may indicate that these are a source of savings in the absence of a functioning rural credit market. 3.94 Lower income households own less land, hire labor less, and participate less in product markets. Despite the overall even spread of resources across income classes, the lowest quintile households tend to own less land and a smaller share of their land i s cultivated than i s the case for other income classes (whether or not any economic benefit i s derived from this, e.g. from renting out, is not clear from the data). Lower income households also hire less labor and own less heavy agricultural machinery. They also participate less in commodity market sales, although sales of some figure importantly, e.g. fruits (not grapes), and forest products. 3.95 Poor households tend to have more elderly members, but not more young dependents. Poorer households tend to be older and pension income i s a higher share of their total income than i s the case for the higher quintiles. Pension payments to the lowest income class are evidently insufficient to raise them out o f poverty, something that separates S A M from a number of other countries in the region (e.g. Romania). On the other hand, unlike other countries in the region (and evidently unlike urban households in SAM), having a low income i s not associated with a larger number o f child dependents (children under 18). Finally, lower income households tend to have more IDPs as members or household head, and more female headed households. 83 e3.5: rsh es of the Ruraf P 84 and the non-poor, then, i s having access to wage incomes outside the farm. And the sharp difference between the poor and non-poor in years of schooling explains the key driver of non- farmincomes. 3.97 This simple presentation shows that the issues to be addressedby policy makers to reduce rural poverty are complex and require a sustained long-term effort and cross-sectoral (education, credit, infrastructure etc.) coordination. 3.9. Conclusions 3.98 An understanding of the labor market in S A M is hampered by bad data. Data either provide an extremely distorted picture (such as registration data), or fail to follow the most critical events (such as the absence of data on layoffs). The quality o f the ongoing LFS needs to be considerably improved for it to become a key monitoring tool and source of data for labor market studies focused on policy issues. 3.99 Usingan array of data, the chapter arrives at a number of conclusions. Its key conclusion i s that the overwhelming concern about labor reallocation hides a deep problem with how the labor market functions and how unfriendly the business environment still i s in S A M , especially in Serbia. It is more productive to focus on these forward-looking issues than to become absorbed in the (incorrectly) measured unemployment. Private firms in a dynamic economy are the only response to the inevitable need to dismiss many workers from socially owned firms. 3.100 The share of the labor force that can be measured as unemployed in an ILO/OECD sense does not exceed 11 percent. A further 10 percent were on involuntary leave or were experiencing severe reductions in working hours, with little or no pay but with some immediate and alternative employment options. Many of the registered unemployed hadjobs inthe informal sector. On the other hand, over a third o f the true unemployed were not registered. This testifies to the poor capacity of the LMB to monitor the situation in the labor market and to deliver benefits to those who need protection. 3.101 The most disturbing finding of this analysis which uses the most robust and recent data available i s the high incidence of joblessness, exclusion, or poorly paid jobs among vulnerable groups -- young people, older workers, women, and the unskilled. The most accurate data obtainable also suggest significant regional disparities in employment and unemployment and the low productivity of rural agricultural employment. This i s a warning sign that these vulnerable groups will not benefit fully from improvements in aggregate macroeconomic conditions andthat thus have borne most of the brunt o f deteriorating conditions. 3.102 On the other hand, three facts are crucial for an understanding o f the role o f the labor market as a driver of poverty in Serbia and Montenegro. First, there i s little doubt that informal activities such as subsistence farming and casual jobs have played a crucial role in providing employment and earnings to many individuals in Serbia. Second, while the unemployed are among those who have suffered the largest decline in living standards, it i s important to keep in mind that the transition has also created "winners," with the most obvious beneficiaries being firm-owners and the self-employed. Entrepreneurship has proved to be a very rewarding 85 strategy for some, but the relatively low number of entrepreneurs implies that impediments to investment remain significant. The latter fact i s constantly overlooked in the debate about the role that the labor market can play inpoverty reduction. 3.103 Ingeneral, the labor marketi s the key determinant of household incomes. For this reason, labor market developments are critical for the determination of household welfare and poverty. In summary, the picture that emerges from the recent literature and empirical evidence is summarized below. The current labor market situation has contributed to poverty in the country and may be critically compromising long-termeconomic growth. It i s characterized by low wages, wage arrears and growing wage inequality. Unemployment is also significant-approximating CEE levels. Hiddenunemployment or over-manning also remains aproblem. The labor market i s not "flexible" in terms of high efficiency in re-allocating resources to their optimal use. Self- employment andother active coping strategies are nonexistent or are small. The formal economy is not creating enoughjobs. Studies of wage determination seem to indicate a spatially segmentedlabor market. 3.104 Considerable advances have to be made to implement a growth-oriented poverty reduction strategy. Some of these issues will be picked up in the forthcoming CEM, but they all have to be regarded as parts of the single agenda: how to make the labor market work for the poor. 86 4. EDUCATION:POLICIESTO COMBAT POVERTY The conflict and social unrest o f the last decade had profound effects on the Serbian education system. Shifts in public expenditures over time provide clues about the seriousness o f the situation, and supply side issues (such as deteriorating school infrastructure and maintenance, declining teacher salaries, etc.) have received a fair amount of attention in policy circles. The demand side issues are not well understood, however. In particular, lack of appropriate data has prevented the analysis of the educational environment faced by the poor. Rising unemployment and future economic uncertainty may have had a more devastating impact on the poor, who are less likely to be able to borrow against future income and who may be more risk averse and thus may opt to "rationalize" household expenditures on schooling by discontinuing some children's schooling and/or by making sacrifices in certain schooling inputs (e.g., textbooks, mode of transportation to school etc.). The focus of the analysis in the chapter i s the poor and the policies that might improve the educational standing of the poor. The first section summarizes the Serbian education system and presents selected aggregate statistics on the schooling status of the population. Then the chapter turns to the access to quality of schooling, followed by analyses of school enrollment at the kindergarten, primary and secondary levels. The section 4.4. focuses on public and private education spending, and the final section summarizes several major policy recommendations and reforms that are needed inthe context of reportedfindings. 4.1 Context 4.1 Serbia enjoys nearly universal enrollment in primary school: the enrollment rate estimates for recent years are around 99 percent. The adult literacy rate is reported to be above 95 percent even inrural areas. Official figures based on the 1991census suggest that 9.5 percent of individuals aged 15 and over have no schooling (males 4.5 percent, females 14.2 percent.40 The 2002 Serbia SLS, however, suggests that 18 percent of individuals ages 15 and over are without schooling (males 12.5 percent, females 22.6 percent). Regardless, the SLS data show that 80 percent of the individuals (ages 15 and over) without schooling are over 55 years old. 4.2 Eight years of primary education is compulsory, and the age of primary school enrollment i s 7 (children may enroll in primary school at age 6 as long as they reach age 7 by the end o f the calendar year). Pre-primary education i s provided extensively by public kindergartens in urban areas to children of ages 3 to 7, although the programs tend to be oriented towards 402002 Statistical Yearbook of Yugoslavia. 87 childcare as opposed to early childhood development. Kindergarten attendance i s not compulsory. The final year of kindergarten often includes preparation classes for primary education. About 1,660 public kindergartens provide pre-primary education to roughly 165,000 children (30 percent of the children in the relevant age group). In the 2001/02 school year, 696,374 students were enrolled in primary schools in Serbia. 41 The number of primary schools has been declining steadily over time, from 5,091 in the 1989/90 school year to 3,598 in the 2001/02 school year; in part reflecting population trends and in part due to the need to consolidate schools with small enrollments. 4.3 In the 2001/02 school year, 696,374 students were enrolled in primary schools in Serbia.42 The number of primary schools has been declining steadily over time, from 5,091 in the 1989/90 school year to 3,598 inthe 2001/02 school year: inpart reflecting population trends and inpart owing to the needto consolidate schools with small enrollments. 4.4 Upon completion of primary school, students take an examination that determines the type of secondary school they will attend (general, professional/technical or vocational). Those in the general education path and most professional/technical students are eligible to enter universities and art academies. Vocational school graduates can continue their education in non- university higher schools, but they do not have access to formal university training. In both cases, students need to pass selective examinations. At the secondary level, the gymnasium provides a four-year general education with the purpose of the continuation of schooling upon graduation; art schools provide four years of education inthe fields of fine arts, music and ballet; vocational schools provide schooling for one, two, three or four years (corresponding to different levels of qualification) in the fields of construction, mechanics, agriculture, forestry, medical courses, economics, commerce, transport etc. Only the four-year vocational school students, who are a small minority of vocational students, have the possibility of higher education upon graduation. Inthe 2001/02 school year the number of enrollments in475 secondary schools was 314,556 (down from 393,001 enrollments in 505 schools inthe 1989/90 school year). 4.5 Higher education i s obtained by attending either post-secondary schools (51 in the 2001/02 school year, down from 63 in the 1989/90 school year) or university faculties and academies (76 in the 2001/02 school year, down from 87 in the 1989/90 school year).43 In the 2001/02 school year, 49,251 students were enrolled in post-secondary schools and 133,690 students were enrolled in faculties and art academies. Less than 10 percent of the adult population in SAM has a post-secondary education on a par with Poland, the Czech Republic, andRomania (where), better than Albania (with 5 percent) and FYR Macedonia(with 8 percent), but still worse than Latvia (with 14 percent), Russia (with 15 percent) or Armenia (with 19 percent). The OECD average i s about 21 percent, with most northern European countries exceedingthat rate of post-secondary graduates. 41Data sources for these figures are the 2003 Statistical Pocket Book and the 2002 Statistical Yearbook o f Yugoslavia; both published by the Serbia and Montenegro Federal Statistical Office. 42Data sources for these figures are the 2003 Statistical Pocket Book and the 2002 Statistical Yearbook o f Yugoslavia; both published by the Serbia and Montenegro Federal Statistical Office. 43These legally autonomous institutes are loosely associated with two universities based inBelgrade and Novi Sad. 88 4.6 Adult education is provided in special schools for those who are 15 years and over. The 2002 Serbia SLS suggests that seminar and training (crafts, pre- qualifications) participation remains at less than 1 percent among the adults. About 3.5 percent of the relevant age group takes courses (language, computers, driving, etc.). Less than 1 percent of the poor do so, however. BOX 4.1. Why Education Matters for Poverty Reduction Educational systems are vital to the process of societal change that both underpins economic reform and is needed in its own right. The empirical studies on the rates of return to schooling have consistently shown high private and social rates of return, especially for lower levels of schooling. The return on human capital raises an individual's material living standards, and there are clear benefits to the economy and society as a whole from having a well functioning education system. Other effects of education enrich a person's life and its quality. Some of these effects may be expected to increase during the transition, notably the returns on certain forms o f human capital, as a vast body o f empirical literature suggests. And this raises the question o f who gains the most from the existing education systems in transition countries This is why equity in access to education is very much a live issue in many countries: unequal access to education opportunities i s seen as one of the culprits inthe entrenchment of long-term unemployment, in the slow responses to the changing demand for jobs with different skills, and in the emergence in some countries of higher inequality. The transition economies face these same changes in what is often an even greater degree and, in addition, have typically had to cope with sharply lower national incomes. But equity o f access is not the only thing that matters. What emerges from the empirical literature is the emphasis on the quality of learning actually achieved and its relevance (e.g. UNICEF, 1999). This seems to be a particular problem in many of the former socialist countries. The over-centralized education systems of the planned period typically involved the minute prescription in schools o f curricula, textbooks. timetables and teaching methods. Outcomes were assumed predictable and there was little external verification o f any learning achieved. The educational systems o f the planned economies streamed many children away from a general education at age 14 or 15, some attending schools which provided them with what was in effect merely firm-specific human capital for a local enterprise. The notably higher rates o f unemployment faced by young people in the majority o f transition countries are a sharp reminder of the need to provide relevant education. Source: Micklewright, John (2000). 4.2.Access to and Quality of Schooling 4.7 It is useful to start the access to schooling discussion by summarizing the status in the early 1990s and comparing it to the current situation. UNICEF (1999) provides useful 1990 statistics: about 8 percent of students in Serbia attended four-year feeder schools, the implication being that they had to travel longer distances for schooling after grade 4. A 1992 survey revealedthat at the time of the survey "in Central Serbia 27.8 percent of the students traveled (by whatever means) between 11and 20 kilometers to school, while 27.9 of them traveled more than 21 kilometers." Busing service to schools was commonly offered (free busing for about 13 percent o f students, subsidized or full-price), but no quantitative information i s available on its effectiveness. 4.8 Based on qualitative evidence, "distance to kindergarten" has been flagged as a possible issue by OECD (2001), which suggests operating with "a larger number of smaller pre-school centers rather than a few large ones." This recommendation needs to be analyzed in the light of serious budgetary constraints for expanding pre-primary schools, especially inrural areas. While 89 a full-blown analysis of this issue is out of the scope of this chapter, it i s useful to provide some quantitative evidence on access. Using the 2002 Serbia SLS, Table 4.1 provides statistics on distance to school for Kindergarten, primary school and secondary school by poverty status. The poor in Serbia have to travel 6 (9.2 in rural areas) kilometers to kindergarten on average, while the non-poor have to travel 3.5 kilometers. For primary school the distances are less problematic: 2 kilometers for the poor, 1.4 kilometers for the non-poor. The nearest secondary school for poor families is, on average, 9.9 kilometers away as opposed to 7.1 kilometers for the non-poor. It i s important to note that children may not be eligible or may not want to enroll in the closest secondary school if that school has highentry examination score requirements or if it i s not the desired type of secondary program. Thus, the distance-to-secondary figures reported here are probably biased downwards. Table 4.1: Estimated Distance (inkms) between Household Residenceand Closest School: by Urban/Rural Residence and by Poverty. Medians and Means [in brackets] Urban/Rural Urban Rural Serbia (urban & nual) Access Poor Noo-poor Au P00r All Poor Non-poor All I I I I I INon-pootI I I I Kindergarten 1.o 0.5 0.5 7.0 3.5 4.0 1.5 1.o 1.o f1.41 [1.01 r1.11 r9.21 L7.01 L7.41 l6.01 L3.51 c3.81 Primary School 1.o 0.5 0.5 1.o 1 1.o 1.o 0.8 0.9 L1.01 L0.81 r0.91 L2.81 r2.21 r2.31 L2.01 E1.41 [lS] Secondary School 1.5 1 1 12.0 12.0 12.0 9.0 3.O 3.5 [3.7] [2.2] [2.3] [14.4] [14.1] [14.1] [9.9] I7.11 [7.4] 4.9 Table 4.2 presents regional differences in access to schooling. Not surprisingly, Belgrade and Vojvodina fare better than other regions in terms of access to schooling. West Serbia and East Serbia rank worst. Table 4.2: Estimated distance (inkms) between Household Residence and Closest School: r Kindergarten 0.5 0.9 3.0 1.2 2 H.81 u.51 V.33 E.71 r6.81 Primary School 0.5 0.8 1.o 1.o 0.8 [1.01 r1.11 [2.31 [1.71 [2.31 Secondary School 2.0 5.O 4.0 5.0 6.0 r4.91 [8.21 B.73 V.91 P.91 .ces: Staff estimates based on SLS 2002. 4.10 About 3.5 percent of individuals who participated in the household survey identified themselves as refugees or IDPs. The refugee and internally displaced children can be expected to face an unfavorable environment at home, in terms o f economic conditions and because o f an unstable lifestyle and stress due to the experience of conflict. In terms of proximity to schooling, however, there does not seem to be a significant problem. The mean distances to the closest kindergarten, primary school and secondary school are 1.96, 1.18 and 5.82 kilometers, 90 respectively, for the refugee and IDPhouseholds. For the remaining households these figures are 3,82, 1,48 and 7.5, in the same order. Despite this, however, the refugee and IDP children are somewhat less likely to attend school, as discussed later in this chapter. 4.11 It is difficult to link the householdsurvey data to the learning performance of children in school, as standardized assessments are only now being developed and data on socioeconomic status are not kept for the 8/9 advancement exam. However, an assessment study conducted in December 2000 (described in detail inUNICEF 2001) provides useful insights, since it tabulates test scores of eighth grade students by parental characteristics and urbadrural residence (which are strong correlates of poverty). This evaluation finds that a child whose father has a university education scores 1.44 times higher on language tests than a child whose father has a primary school education. For mathematics and sciences, this ratio i s 1.66 and 1.47, respectively. Urban versus rural residence also makes a difference: children in urban areas scored 1.16, 1.25, and 1.14 times better in language, mathematics and science areas, respectively. Thus, there i s reason to believe that children from poor households lag behind significantly interms of performance in advancement examinations. 4.3. Children's school enrollment and completion 4.12 Empirical evidence, mostly relying on data from the United States, testifies to the beneficial effects of kindergarten attendance throughout life (see, for example, studies analyzing the impact of the Perry preschool experiment). 44 In Serbia, too, there is some evidence supporting this finding. As described by OECD (2001), which in turn refers to Landers (2001), 97.3 percent of Roma children who attended an NGO pre-schooling program were competent in the Serbian language (as opposed to a 33.3 percent competency rate for those who did not attend); and almost all attendees completed the first year o f primary schooling, as opposed to a 40 percent first year completion rate for others.45 4.13 Former Yugoslavia had the lowest kindergarten enrollment rates prior to transition than any other socialist country (22 percent in 1989, as reported by Micklewright). The SLS reveals that the enrollment rates have not changed very much.46 Figure 4.1 presents public kindergarten attendance patterns by age group, urbadrural residence status, and poverty. Very few children attend kindergartenbefore age 3, and thus the emphasis i s on ages 3 to 7. About 26 percent o f 3- year olds attend kindergarten, and this percentage steadily increases by age to 62 percent at age 7. Enrollment patterns vary significantly between urban and rural areas, especially in the early 44Among others, see W.S. Barnett, (1996). "Lives in Balance: Age-27 Benefit-Cost Analysis of the HigWScope Perry Preschool Program (Monographs of the HigWScope Educational Research Foundation, 1l),. Ypsilanti, MI: HigWScope Press; L.J. Schweinhart, Barnes, H.V., Weikart, D.P., Barnett, W. S. and Epstein, A.S. (1993). Educational ResearchFoundation, lo), Ypsilanti, MI: HigWScope Educational ResearchFoundation. "Significant Benefits: The HigWScope Perry Preschool Study Through Age 27". (Monographs o f the HigWScope 45OECD. "Thematic Review o f National Policies for Education: Serbia,". 2001; Cassie Landers, Early Childhood Development in the Federal Republic of Yugoslavia: SuggestedStrategiesfor UNICEF,.New York,. 2001. 46The following discussion focuses on public kindergartens. Private and religious kindergarten attendance is rare in Serbia: in the 2002 LSMS sample, only 10 children were reported to attend private kindergartens and 4 children were reportedto attend religious kindergartens. 91 ages where only 5 percent of 3-year olds are enrolled in rural areas, as opposed to 40 percent in urban areas. At age 7, the urbadrural gap closes somewhat. But in fact this i s not necessarily a good indication, because at age 7 children are supposed to be enrolled in primary school, not in kindergarten.47 Figure 4.1: Public KindergartenAttendance by Age Group, Povertyand UrbadRural Residence +Allchildren Urban -A- -+-.Poor Rural -A- .Non-poor Sources: Staff estimates based on SLS 2002. 4.14 Differences in kindergarten attendance by poverty are even more revealing. None of the 3- and 4-year old poor children in the SLS sample attended kindergartens, and older poor children's attendance rates vary between 10 and 30 percent. It should be noted also that refugee and IDP children are less likely to attend kindergarten (34 percent versus 41 percent for the remaining population, not shown in Figure 4.1). Enrollment statistics for higher levels o f schooling are not available for refugee and IDP children because of the small sample size. 4.15 The section on Education in Annex I1provides further insights into public kindergarten attendance trends. Multivariate regressions aim at capturing the key factors influencing kindergarten enrollment. Specification 1considers attendance as a function of children's gender and of the schooling, occupation and marital status o f the household head. Specification 2 adds urbadrural residence to that list, and specification 3 further expands it by considering distance to the closest kindergarten as an additional explanatory variable. A Probit model i s estimated where the dependent variable takes the value 1if the child attends kindergarten, or 0 otherwise. The key findings from this exercise are that children's gender does not have an influence on attendance; children are more likely to attend kindergarten if the household head has completed secondary school or higher; and, while the children o f married household heads are less likely to 47The magnitude of the difference in age-of-enrollment between the poor and the non-poor may be influenced by the measurement error on age reporting. It should be noted that such a measurement error would have an impact on these trends only if it varies significantly betweenthe poor and the non-poor. 92 attend kindergarten, the effect i s not statistically significant. But the more interesting findings are the impact of the urbadrural and distance-to-kindergarten variables. When both of these variablesare included (i.e., specification 3), parental characteristicsdo not have a statistically significant impact on kindergarten enrollment. In other words, in areas where kindergartens exist, children with uneducated parents, etc., are as likely to attend kindergarten as other children. Figure4.2: Timingof PrimarySchool Enrollment a Poor Non-Poor Rural Urban Total Sources: Staff estimates based on SLS 2002. 4.16 A survey question also inquired about the reasons for not attending kindergarten. The poor are more likely to mention distance as an obstacle (14 percent of the children from poor families are not enrolled because the kindergarten i s too far away), and they are more than twice as likely to give costs as the reason for not sending children to kindergarten (20 percent among the poor versus 8 percent among the non-poor). According to the Social Child Care Act, the kindergarten expenses of children from low-income families are covered by municipal authorities, But since cost was mentioned for 20 percent o f poor children as the reason for non- enrollment, either these households are not eligible for free or sufficiently subsidized pre- primary education, or they are not aware of the presence of a subsidy scheme, or legally established subsidies are not available inpractice. 4.17 Since enrollment rates at the primary level are close to 100 percent, the key issue at the primary level is not whether a child enrolls but the timing of enrollment. The household survey data suggest that almost all children enroll in primary school by age 8, so it i s sufficient to present the percentage of children enrolled by age 7 (see Figure 4.2). Less than 40 percent of poor children enroll in primary school on time, while 60 percent o f the non-poor do so. Urban versus rural residence also matters somewhat, although differences inresidence are less visible. 4.18 Repetitions and dropouts are believed to be negligible in primary school for the general population, although not for the Roma: UNICEF (2001) suggests that only one-third of the Roma 93 children in Serbia complete primary scho01.~'At this stage, data restrictions do not allow one to investigate this issue further. But a survey i s being fielded as part of the Poverty Assessment initiativethat focuses exclusively on the Roma. BOX 4.2: Equity ofAccess to Education: The HumanRights Perspective The value o f education to the individual can be so high that access to education is recognized as a human right in international law. This might appear to take discussion o f public policy beyond the realm usually inhabited by economists. However, a closer look at the notion of rights related to education in the UN Convention on the Rights of the Child - the most widely ratified international instrument of all - serves as a useful reminder of several issues that public policy has to confront, whatever one's disciplinary standpoint. Countries ratifying the Convention on the Rights o f the Child (which includes all the transition countries, and in fact all but two o f the world's sovereign states) recognize the right to education "on the basis of equal opportunity." Any reasonable interpretation o f this principle goes far beyond the provision of buildings and teachers inall parts of a country. Other general principles inthe Convention underline how the right to education should be promoted. These include (i) an absence o f discrimination, and (ii) primary consideration in policy making o f "the a best interests of the child". An absence o f discrimination on the basis o f income does not, in fact, prohibit the charging under the Convention for school or college places in the public educational system (other than at the primary level where it i s explicitly ruled out). Nor does it preclude private schooling (the right to establish private schools i s written into the Convention). But it does imply that low income families should be assisted with any fees in the public system. The clearest case o f the principle o f "the best interests o f the child" being accepted as paramount is the requirement throughout the industrialized world that schooling should be compulsory over a certain age range. This i s a counter-example to the argument that achieving greater parental choice should be a dominant principle in educational policy making. An analogous situation arises with the decisions made by communities or local governments ina decentralized educational system, where again the argument i s often made that greater choice i s always important to achieve - the case for genuine decentralization i s in part that it allows policy to be more closely aligned with voter preferences. But the choices made by a majority may in some cases be so much to the detriment of a minority that the "best interests" (or similar) principle needs to be invoked by central government to override local decision-making. Source: Micklewright(2000). 4.4. Secondary School Enrollments and the Poor. 4.19 Ministry of Education statistics suggest that in the 2000/01 school year, 20 percent of secondary school entrants enrolled in a gymnasium, 78 percent enrolled in professional schools (less than 1percent in two-year schools, 40 percent in three-year schools, 38.5 percent in four- year schools) and 1.5 percent enrolled in art schools. These statistics are roughly in line with the findings from the household survey data. Since the sample size of the household survey does not allow one to focus on children who entered secondaryinthe previous year, Figure4.3 focuses on the status of children aged 15 to 17 instead (which also allows one to depict the percentage of children who are not enrolled in secondary school). 4.20 About 23 percent of poor children aged 15 to 17 have discontinued schooling without enrolling at the secondary level. Less than 6 percent of non-poor children did so. Also, only 3.3 48 UNICEF.(2001). Comprehensive Analysis o f Primary Education inthe Federal Republic of Yugoslavia. 94 percent o f poor children in this age group are enrolled in a gymnasium. The same statistic i s 13.4 percent of non-poor children. This i s no doubt in large part due to the scores received from the secondary school entry examinations - the success rate in these examinations has varied between 75 percent and 85 percent in recent years, which i s reasonably high, but those without high scores cannot attend gymnasium. The ruralhrban residence differences in enrollment are also sizable (14 percent and 3 percent non-enrollment, respectively), but gymnasium attendance i s less unequal: about 10 percent in rural areas versus 14 percent in urban areas. The share of students outside of the general schooling path (gymnasium) is unusually high. This is a trend that cannot be justified by referring to the demands of the labor market, since 90 percent o f graduates o f vocational schools (including secondary-for-skilled-workers) have been unemployed for from one to five years following graduation (OECD 2001). Figure 4.3. Enrollment at the Secondary Level, Children Aged 15 to 17 70 1 60 -B - 50 40 W -E 30 B 20 10 0 Poor Non-Poor Rural Urban Total HDiscontinuedSchooling&Never EnrolledinSecondary Currently enrolledinelementary UlVocational School Secondaryfor SkilledWorkers 0Gymnasium Sources: Staff estimates basedon SLS 2002. 4.21 A multivariate model of predictors of secondary school attendance is presented inAnnex 11. The dependent variable distinguishes among non-enrollment, vocational secondary school, secondary school for professional/technical workers, and gymnasium. Explanatory variables are identical to those that were used to explain kindergarten attendance, although this time the "distance to school" variable i s not included because the household survey did not inquire about distance by secondary school type (instead there i s a general question on distance to secondary school). The key finding from this estimation i s that, unlike the kindergarten attendance model, urban/rural residence i s not a statistically significant determinant of non-enrollment. Instead, parental characteristics play a very important role: the level o f schooling achieved by the household head matters more than his or her occupation. Finally, if we control for other explanatory variables, female children are more likely to be enrolled in (any type) o f secondary school. 95 4.22 Reasons given for not attending secondary school (for those between ages 15 and 17) are mostly reported not being interested (10 percent); not having the financial means (12.1 percent) and planning for further education in the future (68.1 percent). For the poor, the corresponding percentages are similar with financial means mentions twice as often (22.4 percent).49 4.23 Since entry examinations play an important role in determining who enrolls in what type of secondary school, and since children coming from poor households fall behind right after the completion of the primary level, it i s useful to document the correlation between poverty and taking private classes. The survey question on private classes inquired "Does the child attend organized classes (private classes) 4 or more classes a week." Slightly more than 20 percent of primary school children reported taking private classes, and almost all of these children were non-poor (only 1 percent of poor children took private classes). At the secondary level, 13 percent of children took private classes; similarly, almost all of these children are non-poor. 4.5. Public and Private Education Spending 4.24 Between 1990 and 1999, GDP per capita decreased from US$3,229 to US$1,793. During the same period, public spending on education as a percentage o f GDP decreased from 3.49 percent to 3.22 percent (peaking at 4.5 percent in 1996 and 1997). As a result, public per student spending on primary education decreased from US$554 in 1990 to US$263 in 1999. Over 2001- 02 these spending levels remained stagnant. 4.25 As emphasized by UNICEF (2001), the reduction in per student spending in the 1990s has been most severe at the secondary level (with a 57 percent cut), followed by the primary level (with a 52 percent cut) and the post-secondary and higher levels (with a 50 percent cut). In the year 2002, as the data from the 2002 Republican budget show, higher education accounted for a hefty 23 percent of public spending on education, while secondary education had a very similar 26 per~ent.~' 4.26 This pattern of reduction in spendingis likely to have a more severe impact on the poor. This is becausethe non-poor are more likely to continue schoolingbeyond secondary school, and tertiary schooling experienced the lowest percentage decrease. Also, in Serbia fees paid inhigher education depend in part on the students' performance in the entrance exam (with a weight of 0.6); in addition to school grade average (with a weight o f 0.4). To the extent that household wealth i s correlated with examination performance, the public subsidies to higher education benefit wealthier groups not only because of selection of higher education but also because the wealthier students are more likely to be accepted free into higher education and have a lower 49Inthe 2002 LSMS, there are two questions that ask for reasons for not being enrolled inschool: one as part of the household roster the other as part o f the education section (with slightly different response categories, in particular not having a `planning further schooling' alternative). There were few responses to the latter question (only 42 cases) and thus the statistics reported here rely on the question asked as part o f the household roster -focusing attention on 15 to 17 year olds who were not enrolled at the secondary level. 50The total Ministry of Education budget in 2002 accounted for 2.83 percent o f GDP, but it did not included some of the spending (capita and repairs) that is being financed at the local level. No data on municipal budgets were available. as the PIER noted. 96 tendency to pay fees (fees may also be assigned to those who fail and wish to repeat courses in higher education). Figure 4.4: Monthly HouseholdEducation Expenditure (mean amount by consumption decile) 1200 11 1000 / 800 -I 200 4 0 1 2 3 4 5 6 7 8 9 10 ConsumptionDeciles +Total educationexpenditure (includingtransportation) *Transportation expenditure 1 Sources: Staff estimates basedon SLS 2002. 4.27 Education expenditures increase steadily by consumption until after households are at aroundthe median, remain stable thereafter, only to increase substantially for the wealthiest 10 percent of households (Figure 4.4). The increasing expenditure trend i s in part due to the fact that children from wealthier households are more likely to attend school. At this stage, with one single cross-sectional data set, the observed trend provides few insights. If the wealthier and poor households alike spent about the same amount on education, this would be a red flag that would require immediate attention. But this is not the case; the wealthier spend more on education. Thus, in this context, the relevant questions include "How much more should the wealthy spend compared to the poor?" "Do the wealthy children receive a better quality of schooling in return?" etc. Some insights might arise from an investigation of the types of education expenditures with special attention to discretionary spending. Also, the trend displayed in Figure 4.4 can serve as a base scenario for future reference: it may be difficult to speculate on "the" correct spending pattern, but it would be useful to know whether the changes over time favored the poor or not. 4.28 For reporting household expenses on education, it i s desirable to make the distinction between households that do not pay anything at all (either because these are households with no children or the children do not attend school) and those that report education expenditure^.^^ 51This is useful for presentationpurposes -if households with zero expenditures are included, expenditures will be downward biasedand possibly the bias will differ between poor and non-poor households -and because it allows flexibility in modeling. For example, at an extreme, this separation allows for the possibility of a household 97 Annex 11provides household level estimates of the probability of reporting non-zero education expenditures, and, conditional on non-zero expenditures the amount spent on education. The explanatory variables are: consumption per adult equivalent (and consumption squared), number of household members, and indicators of gender and age distribution o f household members. Predictions based on these models are plotted in Figures listed subsequently in the Annex 11; assuming a household with five members (three females and two males), two adults over age 30 and one child at each of the remaining age groups that correspond to different education levels. The wealthier households are significantly more likely to report non-zero spending, and they are more likely to spend more. In fact, conditional on non-zero spending, if such a five-member household i s among the wealthiest 10 percent, then education spending can be expected to be twice as much as that of a similar householdthat i s among the poorest 10percent. 4.5. Conclusionsand Questionsfor Research 4.29 Poor children do not experience a favorable learning environment, at home or in school. This comment is stating the obvious; it applies to developed countries as much as it applies to developing countries. But poor children in Serbia are further disadvantaged because of the economic downturn in the last decade and because of a selective education system in which poor children do not have much of a chance of staying in the general education path after primary school: they have a greater tendency to drop out after primary school or to be streamed into vocational schools and to discontinue schooling upon graduation. In such an environment, low social mobility should not be a surprise: those who are poor will remainpoor across generations. 4.30 The information and analysis provided above suggest a number o f policy recommendations for education, as well as the need for further research and monitoring in key policy areas. Recommendations are summarized as follows: 4.3 1 Kindergartedpre-school education: as in many countries, knowledge and skill disparities at the time of primary enrollment - which are highly correlated with the level of education of the parents -- lead to divergingeducational outcomes in Serbia. The analysis above shows that the poor, who start out behind their peers, are more likely to drop out early or to be tracked into occupational streams which do not offer the skills needed to cope with a dynamic labor market. InSerbia, differences between the poor and non-poor are further reinforced by the provision of publicly subsidized kindergarten services to only about 30% o f the age group, mostly the non-poor and those who live in urban areas. However, the easy recommendation of rapidly expanding access to publicly-subsidized pre-school or kindergarten to the remaining 70% of the age group i s not likely to be a viable option for Serbia in the short to medium term due to fiscal constraints. To the extent that budgetary increases at the republic or municipal level can be justified, or efficiency gains can be found in the system, an expansion of publicly-subsidized pre-school programs should be targeted on the basis of both poverty and poor learning outcomes in primary education. Further analysis is needed on this issue to determine the fiscal feasibility of pre-school expansion during the PRSP period and beyond, and to analyze the cost- characteristic to decrease the probability of education spending but increase the level of education spending conditional on non-zero spending. 98 effectiveness o f alternative approaches. For example, in lieu of expanding formal pre-school in highly dispersed rural areas, which presents serious efficiency problems, emphasis could be placed on organizing remedial programs in core learning skills such as reading and math in the first year of primary education for targeted groups of students. Given the fact that the poor do not benefit significantly from public expenditure on pre-school, increasing cost recovery should also be considered for working parents, particularly inurban areas. 4.32 Secondary education: the high share of children in specialized occupational secondary programs in Serbia (40% in 3-year programs) and the low share of children in general programs (20%, of which only 3.3% come from poor families) reflects an educational system that was designed to support a socialist, command economy that no longer exists. The evidence o f this broken link with the economy i s the higher probability of unemployment for graduates of those occupational programs. Moving children into new broad-based secondary programs which increase a graduate's flexibility to respond to a modern labor market would help redress poverty concerns, as well as promote economic growth. Because occupational courses are more expensive in terms o f student/ teacher ratios and specialized equipment needs, such a reform would have the added benefit of allowing a more efficient delivery o f public education, and thereby free resources for poverty-targetedinterventions. Further analysis i s needed on the costs of delivering various types o f secondary training programs prior to developing a model which could demonstrate the potential efficiency gains of such a reform. More research would also be useful on the tendency of graduates not to be employed in the occupational fields in which they were trained - anecdotal evidence suggests that such wastage has risen rapidly in recent years - and on income variations for those trained in various types of secondary programs. 4.33 Tertiary education: As only 3.6 percent of young people currently enrolled in higher education in Serbia were from poor households, according to the SLS, it i s true that the large public expenditures and subsidies to higher education in Serbia provide no benefits for the poor. These facts make it an inevitable recommendation o f this poverty assessment that private contributions to higher education should be significantly increased in terms of both increased tuition fees and reduced student support subsidies to non-poor students and their parents. In addition to increasing cost recovery for those who can pay, a number o f actions need to be taken to increase higher education enrollments, particularly among the poor. The reform of secondary education described above should help to increase access, but will also increase pressure for higher public expenditure as demand for higher education increases. Fortunately, it i s likely that increased demand can be financed through both increased private contributions and efficiency improvements in the sector. Although little hard information is currently available, it i s commonly accepted that the fragmentation o f university management across faculties and high repeater and drop out rates related to outdated, professor-centered teaching methods contribute to a cost per graduate in most programs that i s far beyond relative EUnorms. Baselines should be established immediately for graduation rates and costs per graduate, and public funding allocations to universities should be tied to improvements in these rates. While universities should be autonomous when it comes to the content of programs, government in Serbia must demand accountability in terms of the efficient use of public resources in order to improve access to higher education. 4.34 A focus on learning outcomes: While the 2002 Serbia SLS provides some new insights for education policy makers, it i s not designed to address all relevant issues for education 99 planning or policy reforms. As pointed out in this chapter, Serbia i s only now beginning to collect standardized information on student performance in the system, which will permit "inputs" provided to the system (facilities, teachers, training, curricula, textbooks, etc.) to be linked to and measured against system outcomes. For purposes of poverty targeting, student performance data should be produced in a manner consistent with household survey data that contain consumption and other socioeconomic information. To the extent that countrywide educational performance problems are identified in Serbia in the coming years, the most pro- poor policy action would be to consider increasing the number of hours o f schooling per school year in primary education. Such a policy has been advocated by several policy papers in recent years, by indicating that time spent in primary school classes in Serbia i s less than ?4 of OECD average. Since average class size in primary tends to be low by EUstandards (excluding certain urban areas), implementing such a policy could potentially be done over time without an increase in the teaching force, although teacher remuneration would inevitably be expected to increase over time. 52 Such a reform would require both further analysis of fiscal implications, as well as significant stakeholder consultation. To the extent that learning performance problems are identified in specific geographic areas or among specific groups such as the Roma, targeted funding programs aimed at providing remedial support at the point of school entrance (see above) will need to be developed and funded. 4.35 Adult education and life long learning: This chapter provides some evidence that secondary and tertiary systems in Serbia have not offered students the types of flexible skills and competencies that are required in Serbia's new labor market, and that this system failure has a disproportionately negative impact on the poor. These can be addressed by a two-prong strategy: (i)increasing the relevance of secondary school programs, and by (ii) of tertiary reform education. As more secondary students demand higher education and more adults seek re- training opportunities in the new economy, tertiary education in Serbia - both public and private providers - will have to diversify to meet this demand. Priority should be given to the development of modular and short-term degree programs that are recognized across institutions, as well as in the broader EUcontext. Fundingfor these public private tertiary programs should be a mix of employer and student contributions, with public financing heavily targeted on individuals who can prove need, as well as selected disadvantaged groups such as Roma. 4.36 Knowledge gaps: While the 2002 Serbia SLS provides key new insights for education policymakers, it i s not designed to address all relevant issues for education planning. Certain policies seem plausible and may not need elaborate quantitative analysis (such as increasing teaching hours per school year, especially at the primary level where the poor would also benefit) but little is known about how to increase the effectiveness of schools because "inputs" cannot be linked to "outputs." It is especially important to collect such data in a manner that can be merged with household survey data sets that contain consumption information, among other things. Such an initiative should be institutionalized so that data collection and monitoring/evaluation activities produce timely feedback to policymakers. ~~ 52ThematicReview of National Policiesfor Education:Serbia. OECD. June 2001. 100 5. HEALTHCARE REFORMAND POVERTY It is often recognized that health is closely linked to poverty. The way to interpret the linkage between bad health and poverty i s twofold. Bad health can lead to exclusion from the labor market and can thus cause poverty. Poverty can also generate bad health, owing to poor living and housing conditions or difficulties in access to health care. A sound health care system i s one of the pre-conditions for sustainable human development. This chapter summarizes and draws preliminary conclusions from the health-related questions of the 2002 SLS (Poverty Survey) in Serbia. Some other background information and data are included to set the context for these results, but the chapter does not attempt a full summary of all of the available data and background literature relevant to an analysis of the relationship between health and poverty in Serbia. Significantly less information i s available for Montenegro and further efforts are needed to reproduce in that country a basic analysis of poverty and health. 5.1 Introduction: Overview of the Health Sector 5.1 Serbia, like other parts of former Yugoslavia, has inherited a health system that undertakes to provide easy access to comprehensive health services for all of the population. Since the late 1980s, however, the economic decline facing Serbia has resulted in a substantial real reduction in resources for the health system. The health system has not adjusted to this reduction in resources. Capital investment has been curtailed for over a decade, staff salaries are unsustainably low, and the gap between expenditures and revenues has been met through increased out-of-pocket payments by patients, and by the accumulation o f arrears to suppliers. These trends might be expected to have a more negative impact on the poor. 5.2 The health system is financed via a combination of public finance and private contributions. The cornerstone o f the public financing system i s the Republican Health Insurance Fund(HIF)for Serbia. The Yugoslav health care system was unique inEastern Europe because it was historically financed by compulsory social insurance contributions and not directly from the budget. Inthe past, however, the system was much more decentralized. There i s a separate Federal Health Insurance Fund for Military Personnel and their families (FMHIF). The Republican HIF receives earmarked payroll contributions from employees, employers, the self- employed, farmers and the Pension and Labor Market Funds. Transfers from the Serbian budget are intended to cover health care provision for the "vulnerable groups" and refugees. Vulnerable groups include the long-term unemployed and other recipients of social assistance, and the \ elderly (via transfers from the PensionFund). Inline with for poverty outcome. 101 5.3 The health care delivery system i s characterized by an extensive network of public facilities: ambulantas (small primary health care stations that are scattered throughout the country), public pharmacies, dom zdravlje (larger primary health care centers with diagnostic services and primary care pediatricians and obsttrician-gynecologists), hospitals, and clinical centers - tertiary university hospitals located in Belgrade, Nis, and Novi Sad. Overall, there are approximately 58,500 beds, or an average of 5.9 beds per 1,000 population. The level o f service inputs (staff numbers, infrastructure) is almost identical to that which was operating in 1990, but the financial resources flowing into the sector have significantly declined. The existing infrastructure in Serbia i s in disrepair and needs basic repairs and re-equipping to restore it to the point where it can provide a level of minimally acceptable health services. Only one-third of the hospitals in Serbia have functioning sterilization systems. Seventy-five percent o f the medical equipment in the health facilities i s more than 10 years old, which most of the producers consider the upper time limit for the manufacturing and stoclung of spare parts. (EAR, Assessment of EquipmentNeeds inHospitals andHealth Centers inSerbia, January 2002). 5.4 Approximately 115,000 people work in the health sector in Serbia (excluding Kosovo). There are reportedly large imbalances by specialty and by region. Physicians have dominated the system, with less emphasis on nursing and other paramedical specialties. Today, 1,400 doctors are reported to be unemployed in Serbia while 1,000 more graduate each year. The average monthly salary (excluding private practice or informal payments) o f health professionals in 2002 stood at 130 for doctors and 90 for nurses, as opposed to 176 for the national average gross salary. As wages have fallen in real terms and the basic means for delivering health services have deteriorated, the morale and motivation of the work force have deteriorated. 5.5 Private practice i s permitted, and since the transition there has been a rapid development in private practices in some parts of the system, such as dentistry, pharmacy and specialist consultative services. There are reportedly 3,000 registered private institutions, doctors and services, employing over 6,000 workers full time, with 12,000 part-time consultants. It i s a parallel systemthat i s serving a small portion o f the population: those that can pay for services in cash. Many doctors from public services work within the private sector as consultants. 5.2. PopulationHealth Status 5.6 Despite all of the difficulties in Serbia during the 1990s (the economic crisis, war, sanctions, bombings), all of the vital indicators in the population (except for Kosovo) improved duringthat time period according to data based on household surveys conductedby UNICEFin 2000. The under age 5 mortality rate decreased by 5.5 percent per year from 1990 to 1999, while the infant mortality rate decreased from 22.8 to 13.6 deaths per 1,000 live births between 1990 and 2000 (TransMonee Data). Serbia i s on track to meet the Millenium Development Goal (MDG) for under-5 mortality, except that there is a risk arising from some signs of recent slowing in the rate of improvement. Almost 100 percent of births in Serbia are attended by skilled health professionals, and Serbia has a low rate o f maternal mortality (5.6 per 100,000 live births), and as a result the associated MDG target for Serbia is too low to be relevant (TransMonee Data). Today, life expectancy at birth i s estimated to be 69.8 years for males and 102 74.5 years for females. The population's access to improved drinking water sources and to sanitary means of sewage disposal i s almost universal, and vaccine preventable diseases are under control. Regarding the causes o f death, the picture i s clearly one of a developed and transitional country with high levels o f heart disease, strokes, and cancer. Smoking i s estimated to cause 30 percent of the mortality in Serbia. Poor nutrition i s another major risk factor. 5.7 Some minor declines in the health status have been reported recently however, and although they not well documented, together with the recent leveling off o f the trend towards improvement in infant and maternal mortality, they are of concern given the other conditions in the health sector and the experiences in other countries in the region where health status has deteriorated significantly. A high annual incidence of tuberculosis (39 per 100,000 population) indicates a needto continue to be vigilant about infectious diseases as well, particularly given the living situation of the most vulnerable population, such as IDPs and refugees, and the affordability of drugs. The government's view i s that there has been a deterioration in the health status (Government of Serbia, Interim Poverty Reduction Strategy, June 2002), but that this has not yet been documented by reliable data. Of the Millenium Development Goals related to health, the most challenging for Serbia are those concerning poverty, hunger, and HIV/AIDS. Although there i s a low prevalence country, Serbia i s very much at risk for future outbreaks of HIV/AIDS, given the existing transmission patterns in the region (intravenous drug use, commercial sex activity). 5.8 Inthe absenceof comparable objective health data (such as data on premature mortality) by socio-economic groups, the self-defined health status allows the first useful investigation of a particularly important aspect of poverty. The subjective nature of this indicator i s to be kept in minddueto the problems of interpretation. The SLS (Poverty Survey) collected data on the self- reported health status among respondents. The pattern of self-reported chronic illness and acute symptoms (in the previous month) in Serbia shows the typical variation by age and gender seen in most EU and CEE countries (see Annex II). Over 60 percent of those over the age o f 60 report that they suffer from some form of chronic illness, and over 30 percent report having experienced acute symptoms in the previous month. There i s little difference in the rates of reported chronic illness between the poor and the non-poor, though some more specific questions on chronic illness, discussed below, point to a higher burden o f some forms o f chronic illness and invalidity among the working age poor. Fewer of the poor than the non-poor report experiencing acute symptoms in the previous month, and this difference i s marked for infants and young children (as reportedby parents). 5.9 These differences need to be interpreted with caution. Other studies have demonstrated that self-reported measures of health differ significantly from medical measures of diagnosed illness. Potential biases can occur inboth self-reported and medical measures. Inthis case, there i s no evidence that we are aware of that the poor suffer less illness than the non-poor in Serbia, and there i s some contrary evidence. When survey participants who didnot use health care inthe past month were asked the main reason for this, 83 percent of the non-poor said that they did not need health care, compared to 78 percent of the poor. Evidence from other countries tends to find higher rates of infant mortality and higher rates of a range of other illnesses among lower socioeconomic groups. A number o f factors may lead to higher rates of reporting among the non-poor: better access to health care services for the non-poor may mean that asymptomatic illnesses are more likely to be diagnosed among the non-poor; health professionals may face 103 financial incentives to diagnose and treat minor self-limiting conditions among patients who are able to pay; and higher income groups may be more likely to experience anxiety about and seek medical advice for minor conditions or screenindcheck-ups for themselves and their children than the poor - a phenomenon sometimes referred to as the "worried well," which has been observed in other countries to rise over time with rising affluence. Women report higher rates of chronic illness and acute symptoms at all ages except in the 0-4 years age band. This pattern i s observed in many EUand CEE countries. 5.10 Multivariate analyses shed more light on the predictors for reporting chronic and acute illness (estimates are presented inAnnex 11). These (probit) models focus on individuals aged 40 and over. The dependent variable for the chronic illness model takes the value 1if the surveyed individual reports having a chronic illness, 0, if otherwise. Similarly, the dependent variable for the acute illness model takes the value 1if the surveyed individual had an acute symptom in the last month, 0, if otherwise. Both models use the same set of explanatory variables: age, age squared, gender, marital status, schooling attainment, occupation and urbadrural residence. 5.11 The age variables capture the expected trend: the elderly are more likely to report illnesses. If we control for other variables (and when these variables are at their means), females are 6 percent more likely to report a chronic illness and they are 7 percent more likely to report an acute illness. Marital status does not have a statistically significant effect on reporting chronic or acute illnesses. More schooling leads to a smaller probability o f reporting chronic as well as acute illnesses, although both the magnitude and statistical significance of the schooling effect i s much more visible for chronic illnesses: for example, an individual with tertiary schooling i s 8 percent less likely to report a chronic illness compared to one who does not have any schooling. The occupation of an individual i s also a better predictor of reporting a chronic illness rather than an acute illness. Only those who listed their occupation as "agriculture, fishing, mining" stand out as being 4 percent more likely to report acute illnesses. As for chronic illness, at one extreme there are those who are not working (with the highest probability o f reporting chronic illnesses) and at the other extreme there are those whose occupation i s "tradeshervices" (who are almost 15 percent less likely to report a chronic illness compared to those who do not work). 53 Finally, urban residence increases the likelihood o f reporting chronic and acute illnesses, by 7 and 4 percent, respectively. The residence effect i s likely to capture, at least in part, the increased awareness of those who have better access to information and health care. 5.12 The 2002 Serbia SLS also asked respondents about a range o f specific chronic illnesses and acute symptoms. High rates o f missing responses (related to hypertension, cardiovascular and cerebrovascular disease, asthma and chronic obstructive pulmonary disease, cancer, diabetes and ulcers), and very low numbers o f positive responses (related to acute respiratory disease, diarrhea, headache, chest pain, low back pain, injury) have limited the significance of the survey results for many o f these questions. However, relatively high response rates were received for adult age groups for questions related to mental health and disability. A significantly higher percentage of poor than non-poor adults and older people report suffering from chronic illness in this category (28 percent of poor comparedto 12 percent of the non-poor inthe 15-59 age band; 53 The occupation dummies need to be interpreted with caution. While it i s possible that not being employed has a negative impact on health, reverse causality -that is, the possibility that those who are inpoor health do not work -cannotberuledout. 104 14 percent of the poor compared to 9 percent of the non-poor in the over 60 age band). Poor adults report significantly higher rates of insomnia than non-poor. Among working age adults (15-59 years), the poor report higher rates of diseases that lead to permanent bodily impairmenthnvalidity (15.8 percent of the poor compared to 12.0 percent o f the non-poor) and higher rates o f diseases that restrict the individual in performing daily activities (30.4 percent of the poor compared to 25.4 of the non-poor). 5.13 As noted above, the direction of causation is ambiguous in relation to the linkage between poverty and chronic illness or invalidity. Chronic mental illness may cause poverty. But there is also international evidence that one major cause of poverty unemployment may lead to depression. Invalidity and functional disability may also cause - - chronic poverty. The planned analysis of the forthcoming panel might produce more insights into these complex relationships, since panel data on individuals would allow one to estimate dependent variables of interest (e.g., health and labor outcomes at the time of the wave-2 survey) conditional on health and employment status (among other individual and household characteristics) at the time of the wave-1 survey. 5.3. Healthcare Provision and Access to Health Care 5.14 The Serbian health system has an extensive network of health care facilities that provides most citizens with access to an ambulanta (primary health care station) within 1-2 km o f where they live. Infact, the government has identified an excess capacity in the health care network in parts of the country. 5.15 In Serbia, both the hospital occupancy rate (68.7 percent) and the average caseload per physician (133) are low by international comparison, and while the official number of hospital beds per 1,000 population (5-6) i s lower than in many transition and highincome economies, one very preliminary estimate suggests that there may be 17,000 more beds than necessary in Serbia. This would imply an excess capacity of 30 percent. These numbers must be used with caution, however, because they do not account for the social protection that many hospitals take on inthe region, caring for the poor, the elderly, and the mentally ill.Nor do they take account of the fact that hospital utilization i s low in Serbia by comparison with EU and other CEE countries, and this may reflect barriersto accessto care and an unmet need for medicalcare. 5.16 The SLS (Poverty Survey) 2002 illustrates these issues. In urban areas, the median distance from a pharmacy i s only 0.5 km, and the median distance from a hospital i s only 3.0 km. However, as Table 5.1 illustrates, the pattern o f access to primary health care facilities shows disparities between the poor and the non-poor. Part of this difference i s due to the fact that populations in rural areas, where health facilities are inevitably more dispersed, are poorer than urban populations. In urban areas the differences in access to health care between the poor and the non-poor are minor. However, even in rural areas, the poor are more likely to live further from health care facilities. Inthe case of ambulanta, this difference i s small (the median distance from an ambulanta for the rural poor i s 700 m greater than for the non-poor). But currently, ambulanta are poorly equipped and supplied, and are unable to provide access to diagnostic services or pharmacies. The median distance from a dom zdravlje or pharmacy i s 4 km greater 105 for the rural poor than for the non-poor. Five percent of the poor live 25 kmor more away from the nearest dom zdravlje. Table 5.1: DistanceinkmfromHealthCare Facilities.MediansandMeans(inbrackets) Serbia (urban& rural) Poor INon-poor I AU 1.5 1.o 1.o [2.7] [1.9] [2.0] 5.O 2.0 2.0 [7.9] [4.9] [5.2] 12.5 6.5 7.0 L16.61 r12.61 [13.0] 2.0 1.o 1.o r6.01 [3.4] [3.7] Sources: Staff estimates based on SLS 2002. 5.17 Access to primary health care facilities i s closer in Belgrade and Vojvodina. Distances from hospitals by region are harder to interpret: a lower median distance from a hospital may be associated with excess capacity. And patients frequently cross regional boundaries to access hospital services (for example, in a specialist hospital or clinical center outside of their own region). Thus, the closest health care facility reported by some individuals in the household survey may not be located in their region o f residence. Respondents from West Serbia, East Serbia and Central Serbia reported the largest median distances for access to primary care facilities (distances o f up to 13-15.5 kmfrom a dom zdravlje, for example. 5.4. Health Services Utilization 5.18 The percentages of the population, broken down by age, gender and poverty, using outpatient services in a state institutioninthe previous month are shown in Figure 5.1. The poor have a utilization rate of 18.5 percent in aggregate, compared to 23.7 percent for the non-poor. Table 2 shows hospitalization rates by rurality and poverty. (Numbers o f respondents visiting private institutions are too small for this analysis to be significant.) In line with other EU and CEE countries, and mirroring the pattern for the reporting o f illness, utilization i s higher among the very young and the elderly, and higher among women at all ages except the 0-4 age group. 5.19 The most significant finding of this analysis, however, is that utilization by the poor is lower than by the non-poor for all age groups, and the difference is largest in infants and young children from poor families. While this pattern of poor-non-poor difference mirrors the pattern of reported acute illness, the disparity in poor versus non-poor utilization i s greater than the disparity inreported acute illness. This gives a reason for concern about whether the poor are able to access primaryhealth care and ambulatory specialist services as readily as the non-poor. 106 Figure5.1: OutpatientVisitsto DoctorsinStateInstitutionsby GenderandPoverty 50 1 45 i :PX 40 - 58 -8 3 5 - 30- .-3 .kj 3 2520 - $: 1 5 1 10 05 0-4 5-14 15-59 >=60 Age Group Sources: Staff estimates based on SLS 2002. Note: visits over 30 days preceding the survey 5.20 Less than half of 1percent of poor respondents, and no children of the poor, reportedly visited a doctor in apn'vate institution in the previous month, compared to some 2.2 percent of the non-poor (3 percent of the non-poor in the 0-4 and over 60 age bands, 1percent in the 5-14 age band, and 2 percent in the 15-59 age band). The use of private doctor services i s highest in Central Serbia (2.87 percent), Belgrade, andVojvodina, and lowest inEast and South East Serbia (1.13 percent). 5.21 Relatively few people report using hospital services: 5.0 percent o f those surveyed had been hospitalized during the12 months preceding the survey, of whom only 1.5 percent stayed in private hospitals or both private and state hospitals. Hospitalization rates were almost identical for the poor and the non-poor, and were similar for rural and urban populations (5 percent). However, the children of the poor (0-14 years) have a lower rates of hospitalization than the children of the non-poor, though there i s no evidence to suggest that poor children should have lower need for hospital care. Males are more likely to have been hospitalizedthan females in the past 12 months, except in the 15-59 year age band, in which childbirth is likely to account for higher rates o f female hospitalization. 5.22 The survey data indicate a rate of hospital admissions o f only 8.3 per 100 population per year in Serbia, compared to an average o f 18.3 per 100 in CEE and EU countries that have a similar population age structure to Serbia. This comparison should be treated with caution: the survey may under-report hospital admissions compared with administrative data (on which the CEE and EUfigure i s based). Nonetheless, this raises the question o f whether many people who need hospital care in Serbia are going without it. 5.23 The regional pattern of hospitalization i s strikingly different from the regional pattern of primary care use: it is highest inCentral Serbia (5.2 percent), West Serbia and South East Serbia, 107 and lowest in Belgrade (4.5 percent). The pattern of hospitalization appears to reflect differences in the age composition of the regions. It is noteworthy that Belgrade, with the lowest hospitalization rate among resident citizens, has the highest concentration o f hospital beds o f any region in the country. In part this reflects Belgrade's role as a national referral center for tertiary (complex) health care. Further analysis i s needed to assess the extent to which this reflects excess capacity. 5.24 Utilization of dental cure shows the widest disparities between the poor and the non- poor. In the previous month, overall 4.9 percent visited a state dentist; but only 2.6 percent of the poor visited a dentist compared to 5 percent of the non-poor. Given the evidence that patients are paying significant out-of-pocket payments for dental care, this disparity i s consistent with the international evidence that the poor respond to user payments with greater reductions in utilization in the case of non-urgent services (deferrable services or preventive services, such as dental checks) than in the care of urgent services (such as hospitalization), and greater reductions in use for a given level of user payments than in the care with the non-poor. Visits to dentists showed a similar regional pattern as visits to doctors in state institutions, except that West Serbia showedthe lowest rate (3.9 percent). 5.25 Survey findings bear out the extensive emergence of private dentistry. Inthe past month, 3.3 percent o f those surveyed visited a private dentist: that is, 40.6 percent of all dental visits were to a private dentist. However, only 0.5 percent of the poor visited a private dentist, compared to 3.5 percent of the non-poor. In urban areas, 4.2 percent visited a private dentist compared to 2.2 percent in rural areas. Use o f private dental care i s highest in Belgrade and Vojvodina and lowest inEast and Southeast Serbia. 5.26 In the previous year, only 31 people among almost 20,000 surveyed used medical treatment abroad, of whom 2 were poor. The average payment made by patients for this treatment was 171,675 Serbian denars, plus a further 20,850 denars on average for transport costs. 5.5. Health Care Financing and Expenditure 5.27 According to the recent Public Expenditure and Institutions Review (PEIR, 23689-YU), public spending on health care in Serbia was close to 7 percent of GDP in 2001, and has apparently been slowly decreasing over the past few years. When estimates of private expenditure are added, total health expenditure would range between 9 and 11percent o f GDP - among the highest in the region and close to the levels registered by high income countries. These rather high ratios primarily reflect low GDP numbers, however, as Serbia's per capita health expenditure (approximately US$62 per person per year in 2001, (though planned to be US$82 per person in 2002) i s one of the lowest in the region. 5.28 The amount of private expenditure on health care i s unknown, although one survey by UNICEF estimates it to be 40 percent and a small household survey conducted in the Krajlevo region for ICRC found a similar percentage. Private, out of pocket, spending i s considered one o f the major issues by the government. It has attempted to capture some of this expenditure through co-payments, but with limited success. The co-payment system has extensive 108 exemptions, to the point that fewer than one-third percent o f users are required to pay, according to Serbia Ministryo f Health estimates. 5.29 The financial performance o f the HIF over the past five years has been poor, and achieving fiscal sustainability in the HIF i s one of the main sectoral issues that the government plans to address. The net accumulated arrears o f the Serbian HIF by the end of 2001 were 6.7 billion dinars (1.0 percent of GDP). The Serbian HIF has met its deficit by (i)taking out commercial loans; (ii) delaying payments to suppliers, especially pharmaceuticalcompanies; (iii) delaying payments to providers; and (iv) artificially maintaining low reimbursement prices or setting contractual revenues at levels that do not cover all of the costs of services provided to insurees. Sustainability requires that the gap between HIFrevenues and expenditures be bridged, which in turn calls for either an increase inrevenue or a reduction in expenditure or, preferably, a combination of the two. It i s important, as well, that in bridging the gap, costs are not simply pushed on to patients in the form of higher out-of-pocket payments for pharmaceuticals and medical and other supplies that are necessary for their treatment under the HIFbenefits package. 5.30 On the revenue side, the main issues are evasion of contributions andinformal payments. In a system that was designed to provide universal coverage, and where the link between contributions and entitlement to services has grown increasingly weak, the incentives to pay the required contributions for the self-employed and the farmers are minimal and, as a result, they are rarely paid. According to the PER, these two categories o f workers contribute only 3 percent and 1 percent of total contributions, respectively, while the share of GDP derived from the private sector and from non-public agriculture are 40 percent and 20 percent, respectively. 5.31 In addition, as the economic situation deteriorated, an increasing number of large enterprises experiencing financial problems were granted "exceptions" to the statutory payment of contributions to all social funds. The exemptions from contribution payments granted to employers appears to have been eliminated recently. Similarly, the Pension Fund and other social funds have begun to pay their contributions more regularly, and the Republic MoF has budgeted for transfers to the HIF for 2002 an amount sufficient to cover the contributions o f IDPs, refugees and vulnerable groups, through a combination of increased budget transfers for these groups and a general subsidyto finance the deficit inthe HIF. 5.32 The survey confirmed that patients are paying substantially more out of pocket for accessing services in state institutionsthan the small officiul eo-payments,and those who use private services are paying substantial amounts out of pocket. Figure 5.2 shows household out- of-pocket spending on health care by consumption decile. For the poorest 70 percent o f households, out of pocket expenditure rises gradually with income from around 300 denars per month to around 800 dinars per month. Household spending on health rises steeply with income for the top three consumption deciles, reaching almost 4,500 denars per month for the richest decile, reflecting the relatively high use of private health care by this group. 109 Figure 5.2: HouseholdHealth Expenditure by Consumption Decile 5000 4500 1 4000 .? 3500 2 3000 - i WB2 2500 - 2000 - $ 1500 r" 1000 5000 1 1 2 3 4 5 6 I a 9 10 ConsumptionDeciles Sources: Staff estimates based on SLS 2002. 5.33 Those who visit the doctor as an outpatient in state institutions, including primary care clinics, on average pay 617 YUD per month for the cost of the doctor's visit, prescriptions and diagnostic tests. Those who visit private doctors, on average pay 2,158 YUD per month. Those admitted to hospitals on average paid 9,752 YUD over the previous year for hospital care, including drugs and diagnostic tests and procedures. These sums are much higher than the official co-payments (which amount to 1.3 percent of revenue by the HIF, according to budget estimates for 2003). Official co-payments are set at 20 YUD for a medical examination, or a laboratory test, 30 YUD for an x-ray, and 50 YUD for an ultrasound. 5.34 The HIF pays state health care providers for the health care costs of the insurees, apart from official co-payments. However, if insurees find that they have to pay out o f pocket for health care services that they believe they are entitled to receive at HIF expense, they are able to apply to the HIFfor reimbursement of the expenditure. Thirteen percent of people asked the HIF for reimbursement for the costs of doctor visits in state institutions, but only 1percent received it; 18 percent asked for reimbursement for private medical costs, but less than 1percent received it; 24 percent asked for reimbursement for the costs of hospitalization, but only 6 percent received it. 5.35 The pattern o f health expenditure by consumption decile may have both benign and negative causes. The system of official co-payments has extensive exemptions, aiming to benefit older people and households with children. Although this system is poorly targeted, it helps to protect the poor from out-of-pocket costs for health care. On the other hand, there is evidence from the survey that the lower health expenditure on the part of poor households reflects in part the fact that some are going without health care when they need it because they cannot afford the out-of-pocket costs. 5.36 Furtherinsights come from descriptive household level regressions aiming to clarify the answers to "Which households had a non-zero health expenditure in the previous month?" and "How much did these households spend on health?". Annex IIprovides estimation results. The 110 explanatory variables are: consumption per adult equivalent (and consumption squared), number o f household members, and indicators o f gender and age distribution of household members. The directions of the estimated coefficients provide little surprise: high-consumption households are more likely to report non-zero health spending and the amount they spend i s higher; household size is positively correlated with both dependent variables; so i s the share of the elderly (defined here to be ages 60 and over). An interesting finding i s that an increased share of females in the household increases the likelihood of non-zero household expenditure on health. Annex I1 Figures display predictions from these descriptive models, focusing on the age distribution of household members: households with a higher share of the54 elderly incur significantly more out-of-pocket health expenditures -- poor and non-poor alike. 5.37 Some people receive financial help in paying health care costs from others. Of those surveyed, 249 (1.3 percent) received help from a friend or relative in Serbia and Montenegro; 109 (0.6 percent) received help from a friend or relative abroad, 4 received help from a humanitarian organization and 22 from a social care center. The amounts received averaged around 14,500 denars. The poor received somewhat lower amounts of help on average. 5.6.Bam'ersto Access 5.38 The findings reported above provide cause for concern that the lower health care utilization by poor households may be due to their going without care because they are aware that they will face higher costs than the official co-payments and are unable to afford these costs. The survey asked respondents who did not use health services what their main reasons were for not seeking health care in the previous month (Table 5.2). 5.39 Among those who said that they needed health care but did not receive it (other than for minor conditions that can be self-treated), 45 percent say that this was because it i s too expensive. B y comparison with the non-poor, more than twice as many of the poor households reported that the expense of using services, lack of health insurance, or distance from services were reasons for not usingservices. Table 5.2: Main Reasons for Not UsingHealth Services (Percentage) Minor illness, treated myself 45.36 69.86 66.33 Minor illness, not treated 7.77 7.08 7.18 Too expensive 22.5 1 10.19 11.96 No health insurance 9.72 2.14 3.23 Too far 4.33 2.10 2.42 Poor service 1.43 3.22 2.96 Other 8.88 5.41 5.91 Sources: Staff estimates basedon SLS 2002. 54This is not an artifact of restrictions imposed by the empirical methodology (Le., estimating one single parameter for the elderly from poor and non-poor households). Ifone estimates separate models for poor households and non- poor households, the estimated probabilities and health expenditure figures are similar to those that are reported by Figures 6 and 7. 111 5.7. Conclusions 5.40 Mental health problems and disability emerge in the survey as a problem for significantly more of the poor than the non-poor. Otherwise, the pattern of self-perceived health status i s typical o f European countries. Those with higher levels of education report better health, and those working in agriculture and fishing report worse health than other occupational groups and the non-working population, if on control for other variables. 5.41 The survey provides clear evidence that people in Serbia are paying significantly more out of pocket for health care that i s supposed to be covered by Health Insurance than the official co-payments. 5.42 There i s evidence that the costs of health care are deterring the poor, more than the non- poor from usinghealth care when they need it. This i s reflected in the lower utilization rates for the poor, for primary care and the strikingly lower use by the poor o f dental care. Rural people, and the rural poor in particular, have poorer access than others to health care facilities, and this appears to be another factor responsible for the lower utilization of primary health care by the poor. But the low hospital utilization rates found in the survey may suggest that even for the non-poor (who, as a category, include significant numbers of people at the margins o f poverty) there may be barriers to access. Private health care i s not reimbursed by the HIF in most cases, and entails higher out-of-pocket costs for patients. As a result, private health services are overwhelmingly serving the non-poor, and particularly those inthe top consumption decile. 5.43 The co-payment system was introduced to formalize widespread informal payments i s rife with exemptions to the point that an fewer than 20 percent of users are required to pay. On the other hand, private health care provision remains rudimentary and serves almost exclusively the non-poor. The direction of reform i s to redefine health care providers' roles and responsibilities, tighten eligibility of exemptions, define of a package of basic health care services provided on an equitable basis, and increase cost recovery in the provision of health services outside the core package. This recommendation i s in line with the strategic vision of the Government represented in several strategy documents. 5.44 A follow-up survey to be conducted in 2003 will enable the next stage of work on the Serbia Poverty Assessment to examine some issues in more detail. Inparticular, additional data from some vulnerable groups will be available, and this i s expected to enable an analysis to be made of health and poverty as it affects the Roma, IDPs and refugees. As well, a follow-up re- survey of some of the households included inthe 2002 Survey will provide panel data, to enable an exploration to be made of some of the dynamics of the relationship between health and poverty over time, andto shed greater light on the causal mechanisms at work. 5.45 More detailed survey data are required to assess the link between poverty and health in Montenegro. 112 6. PROTECTINGTHE POOR Social protection system i s a crucial component of any strategy to address poverty and the sources of risk and vulnerability. Raising the incomes of those at the bottom end of income distribution and preventing others to fall bellow a certain income threshold require sustained economic growth that creates jobs and increase wages, adequate pensions but also safety net for special groups who are unable to derive income from the labor market or who may need aid even if the overall poverty were low. This chapter investigates how well the social protection system i s fulfilling its role to assist individuals and households to manage risk better and provide support for the critically poor. More specifically, this chapter: i)discusses the overall composition of social protection (SP) expenditure and address the question of whether the mix of programs and their size i s adequate or sensible given constraints; ii) assesses how effective are the main social assistance programs in alleviating poverty; andiii) based on the analysis, and given financial, capacity and political economy constraints, identifies what are the possible policy options conducive to poverty reduction in the future. Due to data limitation the chapter focuses on Serbia and refers mostly to information from 2002; this i s important to bear in mind, a during the last year the Ministry of Social Affairs of Serbia has undertaken a number of significant reforms that are described in the text, but could not be assessedat the time of writing. 6.1Social Protection System in Serbia and Montenegro 6.1 Social protection has been and remains an important policy priority for the Governments of SAM. The system o f social protection i s a combination of programs inherited from the socialist period in which protection was available through a large-scale pension system, employment guarantees and price subsidies, and new programs meant to help households to manage new risks (war, displacement, unemployment) and provide support to the extremely vulnerable. The coverage of this system i s broad. In 2002 almost 63% of the households in Serbia received at least one type o f social cash transfer and a small fraction received benefits in kind or social care in residential institution^^^. In Montenegro about one half of households were recipients of public transfers incash and in-kind. 6.2 The social protection system includes two main pillars: (i) Social insurance: programs that channel benefits to contributors and their dependents contingent of the occurrence of selected social risks, such as loss of working capacity due to old-age, disability, work-related injury or 55In2002, almost 15000frail elderly and disabled adults and children, children who lost one or both parents, and children born into difficult circumstances (to poor teen or single mothers, for example) were placed into residential care 113 occupational disease; or job loss. The main programs included here are pensions (old-age, disability, and survivor's), short-term insurance benefits (sickness benefit), and unemployment benefits. (ii) Social assistance: programs that transfer resources to poor or other vulnerable groups, based on the principle of social solidarity (without contributions). The main programs are Material Support to Families (MOP), and Family Material Support (FMS) in Montenegro, caregiver's allowances, veterans and disabled allowances, child allowance, parental allowances, birth grants, compensation of child care in preschool institutions, assistance to refugee mothers caring for children under one year of age, subsidized vacations in recreational centers for children, educational program for children before entering the first grade, various targeted programs with electricity and heating subsidies, protection for civil victims of war, one time municipal social assistance, soup kitchen etc. 6.3 In2002, consolidated social protection spending represented about 18 percent of GDP in Serbia and almost a half of the consolidated public expenditure (see Table 1.3 in Chapter 1). In Montenegro transfers to households also amounted to 18 percent o f GDP for 2002. By international standards Serbia and Montenegro stand as countries with the high expenditures on social protection (Table 6.1). Table 6.1 Social protection expenditure as a share of GDP, selected ECA countries Romania 2002 10 Bulgaria 2000 13 Croatia a 1998 17 Poland 2001 20 EU accession-7 1995-2000 14 Montenegro 2002 18 Serbia 2002 18 Notes Does not include local budgets. Simple average o f Czech Republic, Hungary, Slovak Republic, a ' Slovenia, Estonia, Latvia and Lithuania. Source: Serbia: Ministry o f Finance and Economy, Memorandum on the budget and the economic and fiscal policies for the budget year 2004; Bulgaria: World Bank (2002) Poverty Assessment, report 245 16; Croatia: World Bank (2001), Economic Vulnerability and Welfare Study, report 2207; Romania: World Bank Poverty Assessment, forthcoming; Poland and EU Accession-7 from World Bank (2003) Poland: Toward a Fiscal Framework for Growth. A Public Expenditure and Institutional Review., report 25033. 6.4 Pension expenditures in SAM, which account for the bulk of social transfers, increased rapidly at the same time as the social contribution tax base eroded, leading to a deficit of around 6% of GDP in 2002 covered from general Government transfer^^^. Today the share of pensions in GDPis among the highest in ECA countries (Table 6.2). In Serbia pensions channeled close 56 Ministry o f Finance and Economy, Memorandum on the budget and the economic and fiscal policies for the budget year 2004 114 to a third o f the consolidated public expenditure in 2002 and almost two thirds of the consolidated social protection expenditure. According to the household surveys these expenditures represent almost 90 percent of the total amount received by households as social transfers in 2002. More than half of the population in Serbia benefits directly or indirectly (through their family members) from pensions. In Montenegro about 45 percent of the population benefitedfrom pensions. Table 6.2 Public expenditures on pensions in selected countries to GDP Slovak Republic 1999 9.5 Hungary 1996 9.7 Bulgaria 2000 10.0 Macedonia, FYR 2001 10.8 Serbia a b 2001 12.7 Croatia 1999 13.5 Slovenia 1996 13.6 Poland 2000 13.6 Montenegro a 2001 14.9 Note: a On accrual basis. Consolidated expenditures for pension and disability funds. Source: SAM: PIER, Poland- Bulgaria-Other countries: The WorldBank,World DevelopmentIndicators,2001 6.5 The pension system was not financially sustainable - the expenditure on accrual basis reported in Table 6.2 in practice were not fully met, and on a cash basis pensions represented a little over 10 percent of GDP in Serbia (and 13 in Montenegro). The gap between obligations and actual payments was accommodated through accumulating arrears and debt to pensioners. But even with pilingup debt, the transfers from the general Government revenues were required to balance cash revenues and spending: in the amount of about 4 percent of GDP in Serbia and 6 percent inMontenegro (2001). 6.6 The Serbian pension system has three separate funds: Workers' Fund; Farmers' Fund; andSelf-Employed Fund,the first one being the largest interms of beneficiaries and contributors (Table 6.3). The Workers' Fundhas experienced a sharp decline in the number o f insured during the 1990' as the number of formal sector employees has declined. This has created major financial problems for the system. The pension fund revenues have dropped because o f tax evasion and a sharp decline in contributors as many workers became unemployed, migrated to the informal sector or abroad or, took advantage of the early retirement. By 2002 the ratio o f employees to pensioners dropped to 1.16 from almost 6 in 1960. In 2002, the pension system was able to meet current legal obligations and reduction of arrears only with the help of budgetary transfers amounting 6 percent o f GDP.57 6.7 Payment arrears, although have been reduced, remain a serious fiscal and social concern, most notably in the case of farmer pensioners. The collapse of the revenue base caused delays and arrears in pension payment and left the state unable to fully compensate pensioners for 57 Ministryof Finance andEconomy, Memorandumon the budgetandthe economicand fiscal policiesfor the budgetyear 2004 115 inflation. InJuly 2003 the farmers have received the pension due for the month of January 2002 which translates into a payment delay of 18 month??`. Adding inflation means that the Government meets pension obligation through lower benefits. Source: Based on data provided by the Pension and Disability Insurance Fund 6.8 Current coverage of farmers stands at 60 percent and does not provide adequate insurance, particularly for survivors and in case of disability. The average pension, according to Table 6.3, i s only about 2,500 YUD, that i s a little over one half of the poverty line for a single person. Contribution compliance among the farmers i s low. The poor state of the Farmers' Fund has lead to the belief that farmers should be allowed to opt out of the system. However, opt-out could result in social and fiscal problems through increased pressure on social assistance expenditures. Mandatory participation of the farmers in the pension system has been retained throughthe 2003 Republican Law on Pension and Disability In~urance.~~ 6.9 Aiming to contain spending while improving both incentives and equity in the pension system, in December 2001, the Federal Parliament passed amendments to the Federal Pension Insurance Law endorsing important parametric changes to the system, including increasing the retirement age from 60 to 63 for men and from 55 to 58 for women in a single step, changing indexation from 100 percent wage to 50 percent wage and 50 percent price indexation, and unifying the minimum pension. Earlier (June 2001), the pension contribution rate was reduced from 32 percent to 19.6 percent of gross wages, while the definition o f wage income was expanded to include a range o f hitherto untaxed supplements and allowances. The change was not revenue neutral, with higher transfers from the budget needed to compensate for the overall loss in contribution revenues. Overall, the reforms in 2001 represented a significant initial adjustment to the pension system, though left a systemic overhaul for later. In 2003 the Parliament approved a new Law on pension and disability insurance tightening eligibility to disability pension and to early retirement, introducing a new pension formula based on individual points, lengthening work history to entire career used to calculate pension in the new point formula. 6.10 These measures have already reduced the deficit, but even in 2003 the pension system in Serbia required a monthly transfer o f around 4 bln. YUD (over 3 percent o f GDP and twice the total social assistance envelope). As a result of all measures planned the expenditure will be reduced from 11.5 percent of GDP in 2001 to 10.9percent o f GDP in 2005, providing savings of 0.7 percent of GDP in 2002 which rise to 1.8 percent of GDP in 2005 relative to the no change '*In July 2003 the pensioners inthe workers fund have received the pension due for the month o f March 2003. The Self Employed Fundpaid the pension intime. 59 The Law stipulates that the head o f an agricultural household shall be covered by compulsory insurance or at least one member of the household. 116 scenario. The deficit of the pension system falls. While these measures reverse the deteriorating trend in fiscal balance, they do not eliminate it entirely. 6.11 Social assistance programs ina broad sense are numerous and heterogeneous, interms of their objectives, target group, targeting mechanisms, source of financing, or implementing agencies. There are several agencies in S A M responsible for social policy. The Ministry of Social Affairs in Serbia and the Ministry of Labor and Social Welfare in Montenegro, are the main agencies with role in defining the social protection policy and administrates the main programs to support the poor.60They are spending directly on their programs 1.7 percent of GDP in Serbia and 2 percent of GDP in Montenegro. An estimated 5 of GDP in Serbia represents social expenditures of the Unemployment Insurance Fund, the local municipalities61and of other social welfare institutions. 6.12 Social assistance benefits including child benefits reached one in fifth households in Serbia and one in ten in Montenegro (SLS 2002 and ISSP H H S 2002). Some programs have national scope, other are implemented by local administrations exercising their right to local autonomy (Table 6.4). There are as many as 26 different publicly financed social assistance programs in Serbia (Blagoevic [2003]. Among these programs, however, there are only two that engage most o f the public resources and involve direct cash transfers: child allowances and MOP in Serbia and FMS in Montenegro. 6.13 Child allowance i s the social assistance benefit in Serbia with the highest coverage. Though it is managed by the MOSA, local administrative authorities (municipalities) are responsible for runningthe program. A new Law on Financial Support to Families with Children came into effect in June 2002 with the overall objective to improve the targeting of the child protection system, aiming for adequate and sustainable benefits for poor households with children. The child allowance amounted 1017 YUD in June 2003 per child or almost 9% o f the average wage (see Table 6.5).62 The right to child allowance can be exercised only for the first four children. The Law aims to strengthen support to children through promoting uniform benefit throughout Serbia, tightening targeting of child allowances through the inclusion of assets in the means testing, indexing benefits to cost of living, and reducing provision o f some duplicative benefits. The main changes incorporated inthe Law include: 0 Adoption of a uniform eligibility threshold for child allowances across the Republic (3054 YUD per capita in May 2003), significantly increasing the equity of benefit incidence; 0 Strengthening of non-income means-testing to address asset ownership by applicant households more effectively; 0 Indexing benefit levels to the cost of living in order to preserve their real value and break inappropriate linkages to wages; 6o In Serbia the Ministry for Social Affairs i s in charge o f a system o f state support to the poor regulated by three Republican laws: (1) The Law on Social Protection and Provision o f Social Security for Citizens (2) The Law on Social Protection of Children and (3) The Law on Financial Support to Families with Children " Local administrative authorities (municipality, or city) can provide additional funding for state support to the poor established by law and for various types o f assistance funded from local budgets - assistance incash, in kindor inservices, one-time assistance, subsidizing of utilities, assistanceinpaying the rent and electric power, establishing soup kitchens, and others. 62 The child allowance amount is increased by 20% for single parents, foster parents, guardians and welfare beneficiaries, and by 50% for children with special needs who are not placed ininstitutions 117 0 Inclusion of a provision allowing households to apply for assistance as soon as required, without having to wait for a specific date inthe year; 0 Introduction of means-testing (including income and assets) for child allowances for families with 3 or more children, as it existed for families with less than three children. Elimination of child allowance for children beyond the fourth child. Material Support to Families-MOP IChildallowance Caregiver'sAllowance Wage compensation during matemity leave, absence from work for taking care of child andabsencefrom work for taking care of special needs children Accommodationininstitution Parentalallowance for second, third andfourth child Family accommodation Compensationof cost for pre-schoolinstitutions for special needs children Job Training Compensationof cost for pre-schoolinstitutions for low incomefamilies Compensationof cost for pre-schoolinstitutions for childrenwithout parentalcare Educationalprogramone year beforebeginningof elementary schoolin durationof 3 hrs throughout schoolyear and for childem hospitalized for long time (under the I competenceof Ministry of EducationandSport) Assistance to refugeemotherswith childrenup to 1y. of age I Financedout of of local administrativeauthoritiesbudget (municipality, or city) Materials for beneficiaries in residential Accommodation, pre-school education, health care of children of pre-school age and institutions accommodation of children up to 10 One-offassistance ISubsidised vacatiodrecreationof children up to 15 yrs of age inrecreationalcenters for children Soup kitchen Subsidizingcost for children inpre-schoolinstitutions andvacatiodrecreation Subsidizedutilitiesheating Assistance inkind Subsidizingcost of pre-schoolinstitutionsfor children from low income families Providingheatingfuel Compnesationfor funeral expenses I Healthcare I Solutionof housingproblems Source: Based on Biljana Bogicevic (2003) and Law on financial assistanceof families with children 6.14 Other child benefits such as parental allowance may seem more generous, but they are providedonly once at the birthof a second, third or fourth 63 The parental allowance amount 5 average wages for the second child, 8 monthly wages for the 3Idand 11 monthly wages for the 4th child 118 6.15 The 2002 Law was expected to reduce the total amount of expenditures from 0.7 percent of GDP in 2002 to around 0.6 percent of GDP by 2003. Already in May 2003 there were 27 percent fewer beneficiaries of child allowance than in May 200264. Anecdotic evidence suggests that the reduction in the number of beneficiaries happened among relatively well-off families. However, further scrutiny i s needed to see if the administrative complications resulting from strengthening the means testing did not hampered also some poor households from claiming the benefits. Table 6.5 Main child benefitsinSerbia, 2003 Amounts as of June 2003 Child allowance 1,017 9 Supplementarychildallowance 1,323 12 Parentalallowance: 2nd child 56,75 1 520 3rd child 1 0 2 31 935 4th child 136,201 1,247 Compensationfor job loss during maternity 989 9 Income threshold: Regular child allowance 3,113 29 supplementary child allowance 3,726 34 Memo: Average wage 2003 10,920 100 Source: basedon data provided by the Ministry of Social Affairs; average wage i s basedon the 2003 plannedbudget 6.16 Material Support to Families (Materijalno Obezbedenje Porodice, MOP) is the main social assistance benefit in Serbia, provided to around 40,000 households in 2002, and accounting for approximately 0.1 percent of GDP. Unlike the child allowance, the program i s fully run by MOSA. MOP i s aimed at individuals and families whose income is lower than the guaranteed "social security level". The program fills the gap between the beneficiaries income andthe threshold. This threshold is fixed by law at a percentage of the average municipal wage, adjusted for the number of family members, and in practice constituted around 2,477 YUD in March 2003 per a single member beneficiary h o ~ s e h o l d .The adjustment for the size of family ~ ~ i s very stringent, and goes beyond any reasonable assumptions about economies of scale (see Mijatovic {20031 for in-depth discussion). The second family member i s considered to increase the equivalent size by only 38 percent, and the forth and fifth has a weight of only 0.13 (and anyone beyond fifth i s simply not counted). As a result for the family o f 5 or 7 members under the current rules the income threshold i s only double the threshold for a single person. This excludes many larger families from the beneficiaries o f MOP. 6.17 Benefit amounts are indexed to wages and eligibility criteria also include an assets test. Decentralized eligibility thresholds for social assistance may conduce to a vicious circle where households in relatively well-off communities end up receiving more generous assistance whereas those living in poor communities with low paid jobs and few opportunities of employment get lower benefits. Decentralized eligibility thresholds for social assistance, based 64 InMay 2003, 31% of the children under 18receivedchild allowance comparative with 42% in2002. The reduction was more important inVojvodina (35% ) than inSerbia (23%). Inthe Municipality of Belgrade the number of beneficiariesdecreasedby 45%. 65 The social security levelrepresented 16%of the average wage for a single member family. 119 upon the average municipal wage, have contributed to low coverage rates inpoorer areas. MOSA has attempted to address this through temporary measures which provide additional benefits for households which fall between the level o f the republican and municipalaverage wage levels. In August, 2001, MOSA created two additional groups of beneficiaries, MOPl and MOP2, by providing supplementary benefits based upon the threshold of the average republic wage. MOPl beneficiaries are entitled to a benefit supplement based on the difference between the municipal and republican average wages. MOP2 benefits are paid to individuals and families who are not eligible under the municipality threshold but whose income levels are below the republican wage threshold. MOPl and MOP2 benefits are administered separately, as they are not included in the Law on Welfare and are financed separately by donor funds. 6.18 The current proposal for reforming the MOP, as reflected in the PRSP, includes both the introduction o f a single national eligibility threshold. The work i s also on-going to prepare the decision amending the rules regardingthe equivalence scale. 6.19 Child benefits and Material Support to Families in Serbia were the largest expenditure items in the social protection budget, 1.7% o f GDP being devoted to them in 2002 (Table 6.6). Slightly less than half percent of GDP was devoted to the direct programs to fight poverty, among which MOP, a very well targeted means-tested program providing financial assistance to poor families absorbed les than 0.1% of GDP. Table 6.6 Serbia: Public Expenditures on social assistance 2000 2001 2002 2003" Total child benefits and assistancefor low 0.97 1.65 1.69 1.70 income Childbenefits, ow: 0.73 1.38 1.31 1.26 Wage compensation during maternity 0.20 0.23 0.33 Child allowance 0.37 0.89 0.70 0.6 Parental allowance 0.10 Educational program before first grade 0.09 0.09 0.07 Others 0.08 0.16 0.12 Assistance for low income, ow: 0.23 0.27 0.38 0.44 MOP 0.02 0.09 0.09 Caregiver's allowance 0.03 0.03 0.04 Social care residential institutions 0.10 0.06 0.08 Others 0.09 0.09 0.16 Source: calculations based on data provided by the Ministry of Social Affairs and the 2003 Budget;* estimated 6.20 In addition to direct income transfers, the central Govemment and local municipalities in Serbia use price subsidies and tax exemptions to meet social protection objectives. The prices of public monopolies and the prices of utilities are under administrative control. Municipalities run special assistance programs to help the poor with utilities bill payment. The EPS (electric energy company of Serbia) also runs a special program to partially compensate the beneficiaries of M O P for the increase in energy tariffs. 6.21 The tax exemption list included at the beginning of 2003 basic food products, utilities, basic medicine, etc. While such subsidies succeed to increase consumption of a target commodity in keeping with policy objectives, they usually distort production incentives. The 120 incidence o f benefits from a general price subsidy are proportional to purchases or consumption, and wealthier households receive larger transfers in absolute terms, yet the amount of transfer a poor household receives will be a larger share o f its budget (Alderman, 2002). There is little evidence and data to analyze the extent and impact of price subsides or tax exemptions in Serbia but the experience elsewhere with housing andutility subsidies shows that these benefit more the upper-income dwellers since they have larger dwellings, and tend to benefit urban residents connected to public utilities. The rural poor, relying on wood, coal, and other unsubsidized heating fuels, do not benefit. The impact of price subsidies will need further scrutiny. 6.22 New forms of social protection emerged when big state companies started the restructuring (severance payments) and when the population increasingly faced problems to pay the utility bill (heating subsidies). Severance Payments - `Transition Fund': to mitigate the social impact o f restructuring, the Government of Serbia approved a Social Program for Employees Dismissed during Restructuring and Privatization in March 2002 (these were analyzed in section 3.1 and 3.2). The redundant employees are entitled to receive a lump-sumor several installments equivalent to Euro 100 for each year of service. In 2002, 57000 employees were paid off and a budget of 7 billion YUD was spent (Mijatovic 2003). The 2003 budget has allocated around 7.5 bln. YUD for payments under the Program to an estimated 150,000 dismissed workers. There are several concerns about the impact of the severance payments. These are significant funds and, as the analysis in Chapter 3 has shown, they are not used for small businesses start-ups, as expected. Redundant workers receiving severance payments get only a temporary improvement in their living standard, but as soon as this money goes they will face problems with incomes. 6.23 Compensation for heating. The Government has introduced in the winter of 2001 emergency measures to mitigate the negative impact on the very poor and vulnerable households of rising utility prices. This includes new cash benefits during the heating season to households with disabled members and selected poor households with low energy consumption. The new program i s coordinated with social welfare beneficiary identification at municipal level to avoid duplication of administration. 6.24 A detailed description of the main types of state support to the poor (institutional arrangements, financing, eligibility, beneficiaries) i s given in a study by Biljana Bogicevic (See Bogicevic et al. Chapter 3 [2003]) developed simultaneously with the PRSP and need not be repeated here. 6.25 Montenegro has instituted sweeping reforms of the social protection system in an effort to improve targeting and ensure fiscal sustainability of social assistance programs. Currently, spending on five main social protection programs - Unemployment Benefits, Family Material Support, Child Allowance, One-off Support and Other Person Care - represents almost 2 percent of GDP. In addition to holding the total budget constant in nominal terms since 2000, the reforms have included major shifts in budget allocation across the major social protection programs. Legislation enacted in 2001 broadened the scope of entitlement for Family Material Support (FMS). Although not all entitled families received FMS, the increase in the number of eligible families i s mirrored inthe increased share o f F M S in the total social protection, from 18 to 37 percent of the total budget. While F M S has expanded, amendments to Child Allowance (CA) regulations made in 2001 restricted the definition of eligibility for CA to include only households that already receive FMS, or households with developmentally disabled children regardless of income. 121 6.26 In comparison, the structure of social assistance in Montenegro is quite different for Serbia. Specifically, family assistance (MOP) in Serbia has much smaller coverage than in Montenegro, while the C A program in Serbia reaches a significantly larger share of the population. Recent estimates show that 37 percent of child population up to 19 years old receive C A in Serbia (Bogicevic [2003]). Ministry of Labor and Social Welfare reports that CA was received by 7 percent of children in Montenegro in August 2002. While the program has been sharply re-targeted, the amounts remained small and its relevance as social policy tool has diminished. 6.27 Preliminary conclusions and issues. Despite very tight budget constraints both Republics run quite extensive safety nets. The resources are predominantly absorbed by the pensions system with very little left over and above basic obligations of social insurance. The pattern of social protection spending, with high social insurance spending and small budget social assistance programs, has been much debated and criticized elsewhere. Implicit in this criticism i s that, if there has been less spending on pensions and other untargeted assistance programs, more would have been available for targeted poverty relief (Fox, 2003). Inprinciple, if the poverty risks are well addressedby the social insurance, relatively limited role of the social assistance may not be a problem, especially if they provide focused and adequate support to the needy. In SAM, however, despite lack of resources, there are multiple social assistance programs run by various agents with different rules and eligibility criteria. Further progress in reforming the system requires careful examination of its costs andbenefits. 6.28 Incidence analysis o f transfer system i s therefore required to assess the adequacy and efficiency of the system. Such analysis will be presented in the Section 6.3. Table 6.7 gives a snapshot of findings for Serbia andMontenegro. Table 6.7 Serbia and Montenegro: Coverage of the Population and the Poor by Social Assistance Programs, 2002 u Percentage of population and percentage among the poor receiving a transfer, 2002 Serbia Montenegro Population a Poor Population a Poor Total Social Assistance 11 30 9 11 Childbenefits, olw: Child allowance 8 20 2 4 Others 3 1 2 1 Assistancefor low income, olw: MOPEMS 2 4 4 10 Others 0 6 2 1 Notes and sources: a Dataonpopulationcoveragefrom administrativesources and Bogicevic et al(2003) for Serbiaand Radovic(2003) for Montenegro; Dataon coverageof the poor are staffestimatesbasedon householddata (not comparableto populationcoveragefigures) and are from SLS 2002 and HHS ISSP 2002, poor are definedaccordingto this reportpoverty line with aconsumptionpre-transfer(see Tesluk [2003] and BeegleandRadevic[2003]for Details of methodology); programs overlap, and same householdmayreceiveas multiple benefits, thus the sumof all administrativedataon beneficiaries of programsis different than actual combinedcoverage; Dataon child allowance inSerbiarefer to the periodprior thereform(June2003) thatimprovedthe targetingandreducedthe coverage. 6.29 It is clear from the Table 6.7 that the two tier system of social assistance covers only a small fraction of those in need in both republics. At the same time, the existence of multiple 122 programs may create duplication, administrative complication and confusion for the applicants. The eligibility criteria are different for MOP and child allowances in Serbia, and the F M S and other forms of social assistance are not well harmonized in Montenegro. The institutions receiving the applications are not the same, and sometimes there are territorial differences in the implementation of the same type of benefit (as i s the case of MOP). Notwithstanding these problems that will be analyzed further in this chapter, the key reason behind low coverage i s inadequate finding. This i s why a careful analysis of social safety net as a whole i s needed to identify the patterns of resource use, main beneficiaries and aggregate outcomes. 6.2. TheRole of NGOs. 6.30 Social capital-broadly defined as the networks, noms, and values that enable people to act collectively to produce social benefits-has been recognized as an essential element of post- conflict reconstruction. 66 O f the three types of social capital (bonding, bridging and linking67), bonding social capital can be seen most frequently in the Yugoslav successor states.68Bonding social capital and social networks can serve as a collective coping and risk-management mechanism in the absence of other resource^.^^ This has taken the forms of individuals receiving money from relatives inthe diaspora, or refugees and DPs staying with family or friends or with membersof the same nationallethnic group. 6.31 SLS and ISSP surveys highlights the importance of private transfers as a safety net. In the month preceding the survey as many as 40 percent of households have reported receiving gifts in kindfrom their relatives, neighbors and friends, while 14 percent report to have received private cash transfers. Even higher share i s reported in Montenegro, where 19 percent of households are reporting to have received money transfer. These transfers had significant impact on household welfare, as illustrated by Beegle and Radevic (2003). 6.32 Civil society organizations (CSOs) have proliferated during the last years of regime, even though human rights organizations experienced politically-motivated accusations o f being traitors. There are varying estimates of how many civil society organizations exist. One estimate places the number at 600 NGOs registered by 1999.70 Another source, based on Ministry of Justice sources, doubles the figure for 1999 and has 2,800 registered by 2001. Taking a wider definition and includin all citizens' associations and social organization yields a dramatically higher number, 19,129.$1 Both sources agree, however, that the increase began in the aftermath of the 1996-1997 protests and as a response to the increased crisis. ~ 66 C. Grootaert, , Social Capital: The Missing Link? 1998 and Nat Colletta and Michelle Cullen, (eds.), Violent Conjlict and the Transformation of Social Capital, 2000. 67 Bonding social capital refers to intra-group networks or formal associations.Bridging social capital refers to those networks or formal associations linking individuals and groups beyond major social categories and cleavages, and linking to the linkspeoplehave with higher levels of decision-making and resourceallocation. See Poggi et al., Bosnia-HerzegovinaLocal Level Institutions and Social Capital Study. 69 Other coping strategies, not dependent on networks, include: selling of possessions or land or delay in bill '%?;:de Centrefor Human Rights, Human Rights in Yugoslavia 1999. 71NGOPolicy Group, Third Sector in Serbia: Status and Prospects. 123 6.33 Institutions focused on the needs of ethnic minorities reinforce bonding social capital. The Roma Information Center lists 56 Roma associations and CSOs. Many of the active associations focus on the needs o f women and children (particularly on health and education) and on preservation o f Roma culture. The other main areas are provision of humanitarian aid and infrastructure within Roma settlement^.^^ In Vojvodina, minority nationalities identified cultural institutions as key institutions, rather than those provided services or benefits.73 6.34 According to data compiled for "Breaking with the Past..." (World Bank 2001b) programs of various local and international donors and NGOs cover about 8 percent of the total population, but are targeted mostly to refugees and IDPs (close to 90 percent of whom receive different forms o f humanitarian assistance). In the general population the share of households receiving such assistance i s much smaller (7 percent in Montenegro - UNOCHA - and around 1 percent in Serbia). Humanitarian relief played an important role in adverting the severe deprivation during the worst period of 1999 (there were more than 1 mln. applications for Red Cross food assistance), but have not significantly (and had not been intended to) changed the mainfactors determining the distribution of incomes. In 2002 international donors remained an important source in providing the safety nets for the vulnerable in Serbia, as the subsequent analysis of humanitarian assistance will show. 6.35 Humanitarian aid provided by international donors to refugees and IDPs usedto be as important as the spending on poverty programs but gradually arrived at an end during 2003. As of January 2003, there were 337,066 refugees in Serbia and 13,345 in Montenegro (UNHCR- WFP, 2003). A substantial number of refugees have opted for Serbian citizenship, most frequently in the case of able-bodied active households who have identified options for local integration and employment. There i s greater reluctance to exercise this option among older refugees, who continue to be uncertain about the possibility o f reclaiming property in Croatia or BiH once they obtain Serbian ~itizenship.In~the second half of 2003 the WFP proposed a ~ reduction of the caseload to an average of 63,000 limited to those refugees unable to support themselves (elderly, children, expectant and nursing mothers, residents o f collective centers). The phase out of WFP food aid i s no way an indication that there i s no vulnerability among refugees in Serbia. On the contrary, if "grey economy" activities continue to be reduced and legal employment opportunities are not created, a large number of refugee families who have been dependent on these activities for their basic income over the last 10 years will join the ranks o f the poor (UNHCR-WFP Joint assessment, 2003). . 6.36 The withdrawal o f ICRC food parcels in September 2003 added to the problems of more than 59,000 IDPs. Supplying food parcels to the IDPs more than four years after the conflict ended was not considered a pragmatic solution and does not meet the long-term basic need of IDPs (ICRC Vulnerability Assessment, 2003). 72 Catholic Relief Services, Integration withoutAssimilation 73Helsinki Committee for HumanRights inSerbia, Minorities and Refugees in Vojvodina, p. 64. 74The majority of refugees living inSerbia are from Croatia and are ethnic Serbs, mainly concentrated inVojvodina and Central Serbia around Belgrade. 124 6.3.Incidence Analysis of Key Transfers 6.37 Given how much the Government i s spending, the concern i s how these resources improve the distribution of income. Though most of the social protection expenditures contribute to redistribution, they do so with different intensity. 6.38 From the large array o f social transfers only few of them can be monitored using household survey data to asses their coverage, adequacy and incidence in order to understand the poverty impact o f these transfers. This section offers an analysis of programs that are captured inthe householdsurvey andtheir impact. 6.39 There are two approaches to assess how well the welfare regime works. Administratively, one evaluates how strictly the benefits are granted according to the internal standards set in program's legislative. Within the scope of this report it is impossibleto conduct any detailed assessment of whether or not particular claimants formally qualify for a benefit. Instead, we assess the program interms of the extent to which benefits go to the poor. Of course, none of our results should be taken as conclusive evidence that program fails or succeeds in its own terms. As we have already noted, the eligibility criteria for two major social assistance programs in Serbia,- MOP and child allowances - differ quite substantially (and are much lower) from the definition of poverty adopted in this report. Our perspective is based on the view that an ultimate aim of welfare regime i s to alleviate poverty. 6.40 Accordingto the Poverty Household Survey (2002), social protection (SP) transfers reach 63% of the Serbian households, or 60.5% of individuals in these households (Table 6.8). Among these households, 47.6% of the households receive social insurance benefits and 20.7% receive social assistance benefits (including here child benefits). Almost 40 percent o f population does not receive any SP transfer while there are 8% of households that receive both types of benefits. Social insurance benefits - that provide replacement incomes contingent of socially insuredrisks -haveboththe largest coverage andthe highestbenefit level. Themostwidespreadis the old- age pensions that reaches 30% of individuals in recipient households, and channels about 57% of the SP transfers. The other two important transfers are disability and survivorship ("family") pensions that reach 10.3% and 8.4% of individuals respectively, and account for 15.1% and 10.6% of the SP transfers. Foreign pensions7' appear as the most generous transfer but their outreach i s very limited: less than one percent o f Serbians receiving this type of benefit. 6.41 Since pensions account for the bulk of consolidated social protection spending, the overall impact of the welfare system in Serbia and Montenegro as a whole on poverty could be seen as almost completely determined by the features of the pension system. It may seem redundant to conduct a distribution analysis o f the pension system, as its outcome i s pre- determined by the past working histories and current legislation. Evidently, social insurance system has objectives other than poverty reduction, but the deficit in the pension system covered from general Government revenues i s diverting budget resources that could otherwise be redirected towards the poor. In addition the claims for social assistance transfers are often determined by the performance and adequacy of the insurance programs, especially pensions. 1 5 Foreign pension (mostly for rights accrued inSlovenia) are not part of the public system 125 Therefore it i s important to conduct the incidence analysis of pensions and to assess whether the existing system adequately protects elderly anddisabled from falling into poverty. Table 6.8: Coverage of MainSocial Protection Programs inSerbia, 2002 recipient HHs* Index Social Insurance, ow: 47.6% 7543 1.17 92.7% Old age Pension 30.3% 7519 0.48 57.5% Disability Pension 10.3% 5990 0.35 15.1% Family Pension 8.4% 4399 0.34 10.6% Unemployment Benefit 1.6% 4158 0.34 1.3% Social Assistance, ow: 20.7% 1980 2.77 7.3% Caregiver's allowance 1.7% 2514 0.34 1.O% Veterans and disabled protection 0.2% 3206 1.10 0.2% Material Support to Families (MOP) 1.2% 3785 7.39 1.O% Humanitarian Aid 3.4% 1039 2.17 0.8% One time municipal assistance 0.4% 1462 1.34 0.2% Child allowance 15.2% 1450 0.27 3.4% Birthallowance 0.7% 2501 0.31 0.2% Monthly allowance for mothers 0.8% 1303 0.30 0.1% Alimony 0.6% 3655 2.16 0.1% Total Social Protection 60.5 % 6897 1.33 100.0% * BeneficiaryHouseholdsweightedby householdsize, totalalso includesforeignpensions, not shown separately Source:Tesliuc basedon SerbiaPovertyHouseholdSurvey, (SLS) 2002 6.42 Pensions and family allowances represent a bulk of all social transfers. However, since pensions and family allowances are "special" social transfers in the sense that they have an income-smoothing function and may be regarded as deferred wages, our interest in how the social transfers system performs will be focused mainly on the smaller transfers whose primary function, in principle, should be to help the poor households. Out of the numerous social assistance programs only four have significant coverage: the child allowance (that reaches 15.2% of households and amount 3.4% o f total household SP receipts in 2002); the humanitarian aid (3.4% and 0.8% respectively); the caregiver allowance (1.7% and 1% respectively) and the MOP (1.2% and 1% respectively). The level o f benefits per households varies significantly across these programs. The child allowance transferred at the time of the survey a flat amount of 900 YUD per child, or 1,450 YUD per recipient household. Humanitarian aid transfers about 1,000 YUD per households inaverage. Humanitarian aid program is included for comparison purposes and should not be used as a reference since has been phased out in 2003. The benefits from MOP andthe caregiver allowances rangedbetween 2,500 and 3,700 YUD per household. 6.43 The outreach of the overall Social Assistance System (including here child benefits) i s low compared to other countries in the region (Table 6.9). This may come from the fact that the Serbia main means-tested social assistance programs (MOP and child allowance for families with 1-2 children in 2002) were under severed budget constraint (and their implementation is doing a goodjob in helping only those in need o f assistance while reducing the budgetarycost to a minimum) and the other social assistance programs are covering very narrow risks (disability, 126 victims of war, internally displaced and refugees) and specific events (birth) while in other countries the highcoverage i s determined by the universality of child or family allowance. Table 6.9: Share of households receiving: social transfers - Share of total households benefitingfrom social transfers Country Year Any social transfer, ow: Pensions Social assistance ChiWfamily (non-contributory) a'1owance Bulgaria* 2001 80.4 54.0 19.1 40.5 Croatia 1998 67.0 57.0 16.0 7.0 Romania 2002 87.0 51.0 16.8 55.5 Serbia* 2002 60.5 47.6 5.9 15.2 * Households weighted by householdsize; Source: Serbia: SLS 2002, Bulgaria:World Bank (2002) Poverty Assessment, report; Croatia: World Bank (2001), Economic Vulnerability andWelfare Study, report2207; Romania: World BankPoverty Assessment, forthcoming 6.44 Program participation rate i s different across economic groups (Table 6.10). Both social insurance and social assistance programs have greater coverage for the poor compared to non- poor (or for the poorest quintiles compared to the richer ones). The largest differences in program participation rate between poor and non-poor are found for the MOP (5.9 times), caregivers' allowance (3.2 times), humanitarian aid (2.9), and unemployment benefit (1.8 times). 6.45 Social assistance programs have low coverage o f the poor. The humanitarian aid, the program with the largest coverage, merely covers 8% o f the without-transfer poor, while MOP, the main anti-poverty program covers only 4% of the poor. 6.46 Child allowances have higher coverage of the poor (20%), but at the expense of higher leakage to the non-poor (14.5% of the non-poor households also receive this transfer). The survey data were collected at the time when child allowance was means tested for families with one or two children. Immediately after, the means testing was expanded to families with 3 children and the assets testing has been strengthened. This may result in the future in lower coverage of non-poor but also of the poor, given the low take-up ratio in 2002. Administrative data already confirmed that the number of beneficiaries in May 2003 was 29% less than in May 2002 but there i s no information whether the reduction occurred also among poor families. In addition to current action to strengthen the mean testing, the Government may consider stimulating the demand for benefits through providing clear information about eligibility, flexible schedule and reducing the transactions costs for applicants as many households indicated that they did not applied for child allowance because they did not know about it. Milanovic (2003) makes the same point with reference to MOP after a detailed analysis that underscore the lack of information as a strong explanation for the low participation rate. 6.47 There are differences in program participation rate by area of residence. Child allowances are more widespread among rural households consistent with lower income and relatively higher adequacy o f the benefit for the rural households with children. Pensions have a higher coverage inurban area where a higher share o f pensioners lives. 6.48 Female-headed households benefit from most o f the survivorship pensions, have higher odds of receiving MOP, humanitarian aid or caregivers' allowances; but lower odds of getting 127 child allowances (as most of these households are comprised o f either younghnmarried singles, or elderly widows). 6.49 The wide range of benefits, many covering very narrow risks, creates duplication while other vulnerable groups remain uncovered. About 8.2% of total population (up to 20.5% of the without-transfer poor) benefit from more than one social transfer. Most frequent combination i s a low-benefit pension (survivorship or invalidity), coupled with one of the SA programs (MOP, child allowance or humanitarian aid). I Table 6.10 Serbia: coverage of programs R e orted statist 'otal Poverty Status ConsumptionQuintile Notpoor Poor 1-Poorest 2 3 4 5- Richest, Social Insurance 47.6% 45.6% 60.9% 60.4% 54.5% 47.8% 42.2% 33.0% Old age Pension 30.3% 29.3% 37.0% 36.4% 36.6% 32.0% 26.6% 20.0% Disability Pension 10.3% 9.8% 13.8% 13.9% 9.7% 10.5% 10.3% 7.3% Family Pension 8.4% 7.9% 11.4% 11.7% 9.1% 6.8% 7.8% 6.4% UnemploymentBenefit 1.6% 1.5% 2.7% 2.6% 2.2% 1.4% 1.4% 0.6% Social Assistance 20.7% 19.2% 30.6% 29.8% 23.7% 19.8% 16.3% 13.9% Caregiverallowance 1.7% 1.3% 4.2% 3.6% 1.4% 1.0% 1.O% 1.5% Veterans anddisabledprotection 0.2% 0.2% 0.1% 0.1% 0.0% 0.2% 0.1% 0.5% MOP 1.2% 0.7% 4.1% 3.5% 0.8% 0.5% 0.6% 0.3% HumanitarianAid 3.4% 2.7% 7.8% 7.2% 4.2% 3.1% 1.4% 1.2% One time municipal assistance 0.4% 0.3% 1.3% 1.0% 0.5% 0.3% 0.1% 0.4% Child allowance 15.2% 14.5% 20.0% 20.5% 17.7% 15.5% 12.7% 9.4% Birthallowance 0.7% 0.8% 0.7% 0.6% 0.5% 1.0% 0.7% 0.8% Monthly allowance for mothers 0.8% 0.9% 0.2% 0.6% 0.3% 0.4% 1.4% 1.2% Alimony 0.6% 0.7% 0.0% 0.3% 0.8% 0.9% 0.4% 0.5% Total Social Protection 60.5% 57.9% 77.7% 75.9% 69.0% 60.8% 53.5% 43.3% Memo: Share of population in each 100.0% 86.8% 13.2% 20.0% 20.0% 20.0% 20.0% 20.0% Note: Poverty and consumption quintiles are estimated inthe absence of the social protection transfers Source: SLS, 2002 6.50 Finally, there are substantial gaps in the coverage o f the poor with SP programs. About 22 % of the (without-transfer) poor are not covered by any SP transfers. The poor who are not coveredby any of the main social benefits are more likely to live in larger households, headed by relatively younger males, working as an employee and concentrated inrural areas. 6.51 The ability of government action to influence inequality without major tradeoffs depends critically on the specific policies, institutions and social arrangements that underpin the distribution of income (World Bank 2000, Making transition work for everyone). Distributional impact of the main social protection programs i s conducted using concentration curves to detect ifprogramspending is progressive or regressive. 128 6.52 Benefits incidence results can be readily be portrayed in graphic form. Tracking the cumulative distribution of total household expenditures against the cumulative population ranked by per adult equivalent expenditures gives the expenditure Lorenz cuwe. This provides a point of comparison with which to judge the distribution of social protection transfers (Demery 2000). Figure 1 presents concentration curves for the main social insurance and social assistance programs. First,compare the concentration curves with the 45' diagonal. If the curve lies above the diagonal, it means that the poorest quintile gains more than 20% of the total transfer. Such a distribution i s progressive in absolute terms. Second, comparisons should be made with the Lorenz curve. Concentration curves lying above the Lorenz curve (and bellow the 45' diagonal) are mildly progressive relative to consumption expenditures. Concentration curves lying bellow the Lorenz distribution indicate regressive transfers. From Figure 6.1 it i s clear that public social transfers are progressive in absolute terms (the concentration curves lying above the diagonal), though with different intensity. 6.53 Child allowances i s a progressive transfer but not far from that that would have occurred with an universal, untargeted program, despite the means testing effort for families with one or two children. The expanding of means testing for child allowance for families with 3 children (starting June 2002) coupled with the strengthening of the non-income means testing, to address asset ownership by applicant households, will eventually improve the targeting of the benefits. However, the Government should be aware of the application costs for families; some families might find the cost of providing evidence for eligibility to be higher relative to the child allowance benefit. The dramatic fall in the number of beneficiaries in 2003 compared to 2002 needs further investigation to see if the changes determined poor households to give up child allowance because the increase inthe application cost. 6.54 From an inequality reduction perspective, the overdll message i s positive. The system of social transfers in Serbia achieves a sizeable redistribution in incomes, through most programs and, even more important, through the most important programs in terms of spending. 6.55 SP transfers are important for their beneficiaries, especially for the poor. Overall SP i s equivalent to 14.7% of the estimated consumption without the transfers (see Box VI 1 for a discussion regarding the welfare ranking in the absence o f social benefits). However, there are wide discrepancies between poor and non-poor. SP transfers represent about 13% of the consumption of the non-poor, but more than half of the consumption of the (without-transfer) poor. For the group of non-poor, only pensions represent a significant share of their consumption (12%). All social assistance transfers account for only 0.8%. For the group of poor, both social insurance and social assistance transfers are important. Social insurance (notably pensions) accounts for 45% of their consumption, while social assistance accounts for another 6%. By consumption decile, the importance of SP transfers relative to the consumption of the group declines from 42.5% for the poorest quintile to 6.7% for the richest one. 129 8 0 i? 130 6.56 Without social protection programs, poverty rate would increase substantially among benefi~iaries.~~ The reduction in the poverty gap and the severity of poverty triggered by the social spending i s larger than the reduction in the headcount. Without social protection or under the assumption of substantial reduction in its level, the poverty gap may increase by 50% and severity will double. Under the scenario of lower welfare in the absence o f social transfers (for example, a net gain of social transfers of 50%), the withdrawal of social protection will cause an increase of 100% of the headcount, and a three times increase in the poverty gap and five times inpoverty severity. This highelasticity of the poverty statistics to a withdrawal of SPtransfers is largely driven by the low level of poverty. I BOX 6.1. Introduction to evaluation of socialpolicies: the levelofwelfare inthe absenceof SP I To estimate the impact of social protection programs in reducing poverty and inequality; assess the program coverage, duplications and gaps, and what programs are more cost-effective in covering the poverty gap, one needs to rank individuals according to their level of welfare in the absence of the SP transfers, to avoid the spurious correlation that exist between these transfers and the economic status o f the household, correlation that would lead to wrong conclusions insimple benefit incidence studies. A simple example illustrates this. Say, in the absence of any SP transfers, a household is poor. Then, the household receive a transfer equal to the poverty gap, and it is lifted out of poverty. Such a transfer is the most effective: the household receives only enough to be lifted out of poverty, but no extra Dinar. Given our interest in poverty reduction, we use a poverty-related social-welfare-function that gives 100% weights to transfers to the poor, and 0 weights to transfers to the non-poor. If we rank the households according to the pre-transfer income (or consumption), this household will be classified as poor, and the transfer will be counted as pro-poor spending. However, if we rank the household based on the post-transfer income (or consumption), then the household will be classified as non poor, and the transfer will be counted as leakage to the non poor. This will be a mistake. To avoid it, we need to rank households according to their pre-transfer level of welfare. However, this i s not simple, as the pre-transfer level of welfare is not observed. To estimate the counterfactual consumption for all SP recipient households the analysis proceeds by (i) estimating the marginal propensity to consume out o f transfers (following van de Walle 2001, 2002); and then (ii) estimating "consumption net of SP transfers" by deducting the actual transfer multiplied by propensity to consume from the observed consumption. The counterfactual consumption in the absence o f the SP transfers is estimated as observed consumption minus 13% o f the SP transfers in urban areas, or 28% o f the transfers inrural areas. We use this counterfactual consumption to re-rank households into poor and non-poor. Source: Tesliuc (2003) 6.57 Different programs contribute toward the poverty reduction goal proportionally with the amount of resources channeled through the program (transferred to the beneficiaries) and the degree of targeting. One way to focus on the targeting performance o f various programs i s to estimate the amount of resources spent in order to reduce the poverty gap of the program beneficiaries by 1Dinar. A cost-benefit ratio i s thus computed for each SP program, where (i) the program benefits were measured as the reduction in the poverty gap due to the program; and (ii) programcostsarethetotalreceiptsundertheprogram.Theestimationproceedsinthree the steps: 76 Tesliuk (2003) estimates that poverty among beneficiaries of social protection would rise by 34% (from 13% after all transfers to 17% inthe absence o f transfers). I t is important to note that this is a rough estimate obtained with certain assumptions (see Box 6.1.) on how the withdrawal of benefits would affect the consumption o f beneficiaries. Tesliuk (2003) adopts a rather conservative measure of this effect with consumption falling only by a fracti,ono f the benefit amount when it i s withdrawn; this therefore provides a lower bound estimate o f the effectiveness o f transfers. 131 0 Estimate current and counterfactual poverty gap. 0 Determine the poverty-reduction benefit of the program: contribution of the program to the reduction in the poverty gap (difference between the observed and counterfactual poverty gap). 0 Compute the cost-benefit ratio by dividing program benefits by its costs (total benefit receipts under the program) 6.58 The analysis points once again that the MOP has the best targeting performance, followed by the humanitarian aid program which anyhow has been phased out during last part of 2003. In the case of MOP, every 2 Dinars program spending reduces the poverty gap by 1Dinar. Most other programs, including all types of domestic pensions and the caregivers' allowance, have higher and similar cost-benefit ratios. From poverty reduction perspective, the most costly SP intervention i s the child allowances. The analysis did not take into consideration the administrative cost of providing the benefits. 6.59 Within social assistance programs, the Government may achieve a higher impact on poverty reduction if would transfer resources from the child allowance andcaregivers' allowance programs towards MOP. Mijatovic (2003) discusses the political economy of consolidating child allowances and MOP. 6.60 But what are current accomplishments of the social protection administration in helping the poor? We draw upon three related concepts to capture the capacity of the administration in channeling funds to the poor: (i) coverage of the poor: what share of the poor are covered by the program; (ii) targeting of the benefit to the poor: what share o f transfers are captured by the poor; and (iii) benefit adequacy: how important is the benefit transferred by a program in the consumption of the poor beneficiaries. 6.61 From poverty reduction perspective, more efficient programs would have higher coverage of the poor, higher targeting performance and greater adequacy. Figure 6. 2 summarizes the three related concepts of coverage, targeting and adequacy, for social insurance and social assistance programs altogether. Social insurance programs (pensions and unemployment benefit) cover two thirds of the without-transfer poor and channels 16% o f benefits to the poor. All social assistance programs including child allowance cover one third of the poor and transfer three quarters of the benefits to the poor. Social assistance have lower coverage of the poor but better targeting reflecting the targeting feature of the programs' design. A perfect program would be located in the upper-right quadrant: it will have 100% coverage of the poor, and 100% targeting. As for program adequacy, to be just right a program should provide benefits equal with the household consumption deficit (poverty gap) before the transfer (accounting thus for incentive effects). 132 Figure6.2 Coverage, Targeting and Adequacy of Consolidated Social Insurance and Social Assistance 0% I&)%overage 20% ofthe poor:% o poor receiMng 30% 40F 50%. .60& bene 70g 80% Note:'he size ofthe bubblesis proportionalto the relative jmportance ofthe transfer mthe consunptionofthe poor Source: Serbia Househo'ldPoverty Survey (SLS), 2002 6.62 None of the social protection programs implementedin Serbia in 2002 are close to the "perfect program" benchmark in terms of coverage or targeting. This i s not a deficiency of the Serbian social protection system. Such programs are difficult to design and implement. That right quadrant of the graph is empty for other countries too. However, the more programs are close to that quadrant, the better the social protection system i s at reducing poverty efficiently, that i s minimizing the amount of budgetary resources spent for one Dinar reduction in the poverty gap. 6.63 Based on figure 6.3 from a poverty alleviation perspective, we distinguish the MOP program as a valid instrument the Government could use to further fight poverty. MOP exhibits good targeting and sufficiently adequate benefits for the beneficiaries but very low coverage. Increased funding, better information and lower costs for the applicants may bring many more poor households to participate in the program. The data seems to suggest that, among the poor household not covered by SP transfers, able- bodied household are over represented. For households with able bodied adults, making MOP conditional on work requirement,may prove to better protect these households while increasing the support for this type of program and developing a system of self-targeting. Child allowance i s the best program in terms of coverage of poor households, but more than 80% of the funds go to non-poor households despite program's design (2002 data!). 133 Figure6.3 Coverage, Targeting and Adequacy of the Main Social Protection Programs 0% 10% 20% 30% 40% 50% 60% 706 80% 906 lOO% Coverage of the poor: % of poor receiving the benefit Note: The size ofthe bubbles is proportional to the relative hqortance of the transfer inthe consurrption of the poor Source: Serbia Household Poverty Survey, 2002 6.64 Low coverageof individualprograms may not necessarilybe a problem. If the social protection system i s built on many programs that address narrow, specialized social risks, then the system may provide adequate coverage despite the low coverage of the individual components. The Serbian social protection system i s not close to this situation since all social assistance cover only one third of the without-benefit poor. In addition, as shown previously, there i s a considerable overlap between the existing programs. Consequently, there i s ample room for both increase in coverage, rationalization of overlapping programs, and better means testing to reduce the leakage of funds toward the without-benefit non-poor. 6.4. Pension reform: approaches and issues 6.65 At present the pension system has an important role in helping recipients to smooth consumption beyond the working age. Nevertheless, over the next decades the change in the dependency ratio will impose a hardship on the pension system to provide pensions to an increasingly larger number of elderly. 6.66 The ratio of employees to pensioners has dropped continuously since 1960 and quite dramatically in the last 12 years from 2.6 in 1990 to 1.16 in 2002 (Figure 6.4). inrecent years on the background of changes in the demographic composition and decreasing formal sector employees. 134 Figure 6.4 Changes inthe demographic composition and dependency Demographic composition Ratio of employees/pensioners - .. 1948 1953 1961 1971 1981 1991 2002.' Source: The Republic Statistics Office and RepublicanPensionand Disability Fund; * excluding Kosovo and Metohija. 6.67 Poverty among pensioners with the exemption of pensioners-farmers was still less prevalent than among the rest of the population in 2002 (see Table 6.11). This was because the income status of the large number of workers who lost their job or get low wages deteriorated even more than that o f the pensioners, who in 2002 received their pension in time. On the contrary, pensioners-farmers were considerably more vulnerable, especially in rural area where the risk of poverty was two times higher than for the general population (see Table 6.11). Table 6.11 Poverty among Pensioners inSerbia in2002 (in %) %of the Relutive Pensw. Structure Poverty Poverty poor poverty structure ofthe depth severity risk* poor pension. Total pensioners 10.9 2.9 24.1 24.8 2.4 0.8 Pensioners-farmers 19.6 84.4 2.1 3.9 5.0 1.8 Other pensioners 10.1 -5.O 22.0 20.9 2.1 0.7 Total population 10.6 2.2 0.7 Total pensioners 7.9 1.3 26.0 26.3 1.5 0.5 Pensioners-farmers 12.5 61.0 0.6 1.0 3.5 1.1 Other pensioners 7.7 -0.1 25.4 25.3 1.5 0.5 Total urbanpopul. 7.8 1.4 0.4 Total pensioners 15.7 9.6 21.6 23.7 3.6 1.4 Pensioners-farmers 21.0 46.5 4.1 5.9 5.3 2.0 Other pensioners 14.5 1.1 17.5 17.7 3.3 1.1 Total ruralpopul. 14.3 3.3 1.2 Source: Krstic (2003) 6.68 There are not significant differences inthe risk o f poverty by gender, within the groups o f pensioners or elderly without pensions. Among all pension recipients in 2002 as many as one third received pensions below the poverty line. The group of elderly without pensions was also quite vulnerable to poverty. Overall about 10 percent of all elderly lived in households not receiving any pension or social assistance transfer. Out of elderly not receiving pensions 18 percent were poor. A disproportional share among not covered pensioners was represented by women (See Box VI.2). 135 BOX 6.2. Pensionsystem: who is out? I InSerbia there are 1.75 millionpeople of retirement age (men 63 and women 58), and one quarter million does not get a pension (old-age, survivorship, disability, or foreign pension). Some o f them live together with other pensioners or active persons. Poverty rate for the elderly are higher than for adults or children (13.90% versus 10.53 and 9.27). The poverty risk for elderly without pensions i s 18.22% compared with 11.90% for those with pension. However, there are about 100,000 poor elderly without pensions compared to 142,000 poor pensioners, thus a majority o f poor elderly are covered by pensions. Both groups merit attention. The probability of being a poor elderly (of pensionable age) without a pension increases with being single, female, IDP-refugee, from rural area and low education; and decreases with being widowed, or being located in westkentralteast Serbia compared to the rest of the country. Children living on survivorship pensions have much higher poverty risk (20% versus 10.5 for the rest of the children). Source: SLS (2002) 6.5Medium-termframework for publicfinances 6.69 The Government of Serbia foresees that social protection expenditures will decrease over the next years in an effort to reduce the government's public deficit. Both pensions and social welfare expenditures are expected to fall as a share of GDP. These expectations however, are hard to reconcile with the Government's intentions to maintain the purchasing power of pensions, to pay the old pension debts to almost 1 million retirees and their heirs77, to secure funds for regular payment of child allowance and for increasing the number of families eligible for Material Support. The main challenge for Serbia's public finances i s to reduce the public deficit concomitant with a planned drop in the share of public revenues in the GDP (Figure 6.5). Public revenues are expected to increase on average at a lower rate than the GDP, in an attempt to relax the fiscal pressure and improve competitiveness and the business e n ~ i r o n m e n t . ~ ~ 6.70 The projected reduction in social protection expenditure if true must be accompanied by an increase in the targeting and rationalization of expenditures. Overspending on pension and financing the deficit using budgetary revenues i s an important issue and the Government should be commended for the efforts to contain the expenditures. However, one would expect that the savings could supplement the limited resources devoted so far for poverty programs (MOP) and child benefits. 6.71 Ministry of Social Affairs in the PRSP process has done a very detailed analysis of the social assistance program and simulated several reform scenarios to expand its coverage o f MOP while keeping the cost manageable and maintaining the good targeting (Mijatovic, 2003). But even a conservative estimate (revision o f equivalence scale and moderate increase in the eligibility threshold which will remain well below the poverty line), should result in at least doubling of MOP budget. These options need to be incorporated in the action during the implementation of PRSP. 77 The old debts to retirees refer to unpaid pensions due for the period 1994-1995. The number o f persons entitled to compensations amount to 694000 living retirees and the heirs o f 257000 retirees who have meanwhile deceased. The liability of the state, of almost 20 billion dinars in2002, represents almost 2% o f GDP. 78 InDecember 2002, the profit tax was reducedfrom 20% to 14% 136 Figure 6.5 Trend inconsolidatedpublic revenues and expenditure, and selected public sectors, Serbia 2001 2002 2003 2004 2005 2006 --tDefense -Economic affairs Health care x Education -x- Pensions --eSocialwelfare Source: Ministry of Finance and Economy, Memorandum on the budget and the economic and fiscal policies for the budget year 2003 and 2004 Note: Consolidated public revenues include all revenues at central level (the budget of Serbia, Pension and Disability Insurance Fund, Healthcare Fundand Unemployment Insurance Fund) and at the local level (autonomous provinces, cities and municipalities) 6.6. Protectingthe vulnerableand improvingthe targeting The analysis undertaken inthis paper points out to a number of policy recommendations: 6.72 The household data allow in-depth policy analysis of various options for the reform of social benefits, as demonstrated by Bogicevic et a1 (2003) and by this Chapter. Due to data limitations, such analysis could be conducted only for Serbia. The Government of Montenegro and donors should seriously consider the need for the comprehensive assessment of the welfare system in Montenegro and finance the collection o f household data that are needed for such an assessment. 6.73 There i s no doubt that the combination of high social insurance spending combined with child allowance and small family assistance programs worked quite well to reduce poverty in Serbia. It i s estimated that poverty would have been 34% higher without social transfers and poverty gap would have been 50% higher. Therefore i s important that the spending on social welfare i s protected. At the same time the leakage i s important, especially for social insurance programs, which by their nature are not targeted and therefore are designed to be "leaky". But even in the case of child allowance in 2002 only 30% o f child allowance went to the bottom quintile, compared to almost 80% of the MOP benefits. 6.74 Main problem of social assistance, however, was not targeting, but exclusion errors. In fact all child benefits and social assistance programs taken together in 2002 cover not more than one third of the poor, at the expense o f spending three quarters o f the funds on non-poor. To address the issue of exclusion adequate financing for safety net programs is key. MOP, the only 137 well-targeted program with good administrative capacity has a very small coverage o f the poor owing to a tiny budget. Thus, there i s ample room for both increase inMOP coverage, andbetter means o f targeting of other programs to reduce the leakage o f funds toward non-poor. 6.75 Regional disparities in the social assistance will need further scrutiny to check anecdotic evidence that relatively well-off municipalities can offer more generous assistance while the poor municipalities are less openhanded. The proposed reform of MOP design that departs from average municipalwage and introduces single national eligibility threshold will address some of regional equity issues. So far donor funds helped the equalization of the threshold across the country but public resources have not yet been secured to avoid regional disparities in the MOP implementation nor has the framework to establish a uniform income threshold been legislated. 6.76 There are many social assistance programs in operation under different levels of government or administrative bodies, with different objectives and eligibility criteria that fragmentize the system of social assistance. Harmonizing programs consistent with poverty alleviation objective may be a first step to avoid overlap, expand coverage and simplify access. The Government may even consider the option of consolidating the child allowance and other child benefits with the MOP as further step in an effort to rationalize expenditures and improve targeting. This step, however, needs to be carefully assessedbased on newly collected SLS data (2003), which reflect the results of the child allowance reform. 6.77 The exclusion of the children beyond the forth child both from the child allowance and the parental allowance reduces the level of assistance for the most needy families. In Serbia less than 0.2 percent of population leaves in households with more than four children but the risk of poverty among them i s almost 6 times higher than for the general population. The cost of expanding the scheme for this group i s very small and its effectiveness for poverty alleviation i s undisputable. 6.78 The Government within the PRSP process developed the expertise to derive a poverty line based on the cost of basic needs to compute the needed assistance for poor families. Recent simulation done within the PRSP process to set the income threshold at different levels of the poverty line and testing different coefficients o f economies o f scale show that this will lead to the increase in the number of beneficiary households eligible for social assistance. Thus they will be helpful in reducing the exclusion errors. Again, the implementation of these proposals will require considerable increase inthe findingo f targeted social programs. 6.79 Inaddition to current action to strengthen the means testing for the child allowance andto index benefits to cost of living, the Government may consider stepping up its current efforts in stimulating the demand for benefits by providing clear information about eligibility, flexible schedule and reducing the application costs. The household survey shows that many families do not apply for child allowance because they do not know about the program. 6.80 Further, the option o f unifying eligibility criteria for all means tested benefits and of placing the responsibility of delivering the services within one agency devoted to fight poverty might increase transparency and responsibility. As a first step Social work centers, the municipal child allowance services and the EPS agencies may consolidate expertise and databases and make savings while improving the services for the beneficiaries through simplification, staff availability, getting geographically closer to the beneficiary and increasing accountability. 138 6.81 With better performing labor market the Government may reconsider the open-ended welfare entitlements with a system that placed more emphasis on getting people off welfare and into the workplace. 6.82 Pension system insolvency i s one of the key risks for making adequate resources available for direct targeted programs. The 2003 Republican Law on pension and Disability Insurance creates an environment to gradually eliminate the deficit and thus free up resources for poverty alleviation programs. It i s therefore important to sustain efforts in implementing the new law and regulate the voluntary pension schemes. 139 140 7. MONITORINGAND EVALUATIONFOR POVERTY REDUCTION A monitoring system is essential for the success in reducing poverty: it is on the basis of tracking over time, and the discovery of the changes that have taken place, that the progress of the strategy can be assessed. The objective of this chapter i s to review the state of the existing system to assist both countries in developing a system to monitor and evaluate whether a poverty reduction strategy i s effective in reducing poverty. S A M already has poverty monitoring systems in place, so the task at hand i s to assess their adequacy and strengthen them as necessary. The chapter discusses the strengths and weaknesses of the existing system, drawing particular attention to data access issues, and focuses on how to build capacity and in particular to strengthen the processes that provide policymakers and others with feedback on the impact of policies and programs. The chapter also discusses the role of non-governmental actors -- research institutions, civil society organizations, special- interest and advocacy groups, or others -- in the design of the monitoring and evaluation system, in the actual carrying out o f the monitoring and evaluation activities, and in the use of the results. 7.1 Components of a Monitoring Sy~tem'~ 7.1 Monitoring i s the systematic collection and analysis of information on the implementation of interventions and on the changes in outcomes indicators. Monitoring i s essential to the PRSP process as it helps to achieve improved interventions through the detailed formulation of objectives, activities and expected results and through the fine-tuning o f future activities, and it provides greater accountability through explicit commitment and means of control and monitoring. 7.2 The PRSPs for Serbia and Montenegro contain a thorough discussion o f the indicators and their types (input output, outcome), and therefore this section will focus on the monitoring of a single issue without which PRSP i s unthinkable: that of poverty. Poverty i s a multidimensional phenomenon, as described in this report, and therefore its monitoring requires cross-sectoral efforts. The main participants in collecting data on poverty are the Statistical Offices, the Health Care System, the Social Welfare System, and other line ministries. A number of polling agencies and research institutes are engaged in carrying out regular or occasional surveys on the poverty and economic status of the population (i.e., particular groups of the population, economic trends or other parameters relevant to poverty monitoring). A number of international humanitarian and development agencies are also engaged in the field collecting data or in hiring local research agencies for the purpose o f monitoring poverty and the 79This chapter relies heavily on the discussion of monitoring inthe PRSP source book by Giovanna Prennushi, Gloria Rubio, and Kalanidhi Subbarao. 141 state o f the population in Serbia, and more often the poverty and state of particular vulnerable groups. 7.3 The chapter will review the roles of these actors, which are specific to each constituent state of SAM. When monitoring poverty, the following activities are undertaken: the collection and processing of information; data analysis; dissemination and feedback. 7.2. Serbia 7.4 The Poverty Surveys (SLS survey for Serbia produced, for the first time in both Republics, a set of reliable and disaggregated poverty indicators. Very importantly, the initiative and leadership for these efforts in Serbia came from the leading PRSP agency - MOSA. The data were analyzed by local researchers and combined with a variety o f other data sources. These efforts are fully incorporated in the draft PRSP. Thus, the first task of creating baseline data on poverty i s successfully completed. 7.5 The first round or data collection was being followed by the panel survey in Serbia. The analysis of new data and feedback to policymakers will represent the next step in creating national monitoring and evaluation systems for the implementation o f PRSP. 7.6 The next step will consists of institutionalizing poverty monitoring in the regular operations of data collection by state agencies. The poverty data used in this report and in the draft PRSP come mostly from externally funded surveys implemented by NGOs. Household Budget Survey (HBS) carried previously by the Federal Statistical Office (FSO) and now by the Republic Statistical and Information Office of Serbia (RSO) gives a detailed picture of household income and expenditures and provides general data on the demographic structure o f households, with the possibility of quarterly monitoring. 7.7 The ongoing official Household Budget Survey in Serbia are being reformed with the support of SIDA and Statistics Sweden to play a key role as the PRSP monitoring tool. Starting inlate 2003 these data will become available for the analysis andmonitoring of poverty. 7.8 The key problem with the HBS i s the "no access to primary data" policy pursued by the official data collection agencies before. 7.9 As the data from SLS are being analyzed and widely used in Serbia and elsewhere, the focus now has to shift to lessons from this project for the statistical system. Some components of the survey have to be studied by statisticians from FSO and RSO to draw lessons and reform the ongoing HBS. 7.10 The issue of the revision o f the system of statistical surveys has been addressed through the elaboration of the new Master Plan for Statistical Surveys and the development of a concrete action plan for making a revision of the surveys (proposed by FSO and RSO of both republics). This plan recognizes most of the criticisms made to date and foresees the revision of quite a number of surveys. In addition, the plan also foresees general personnel and technological improvement of the statistical system. 142 7.11 The capacity to use household data for policymaking in Serbia needs to be consolidated, and the first round of the Poverty Survey will serve as a catalyst in this process. The Poverty Survey established a baseline poverty profile usingthe reformed household survey data. 7.3. Montenegro 7.12 The statistics from this report are drawn mainly from the Household Survey (HHS) data collected by ISSP-CEED in July and October of 2002. The Poverty Surveys (CEED-ISSP H H S in Montenegro) produced, for the first time, a set of reliable and disaggregated poverty indicators. The data were analyzed by local researchers and combined with a variety of other data sources. These efforts are fully incorporated in the PRSP. Thus, the first task of creating baseline data on poverty i s successfully completed. 7.13 The next step will consists of institutionalizing poverty monitoring in the regular operations of data collection by state agencies. The Household Consumption Survey (APD) i s conducted b y Federal Statistical Office (FSO) of Serbia and Montenegro. This quarterly survey included 280 households from 12 municipalities in Montenegro, about 11 percent of the total household sample, all part of the permanent population. Not only this sample would be inadequate for assessing the dynamics o f poverty, the survey was stopped in 2003 and no data are collected at the moment. 7.14 The reform of official survey in Montenegro considerably lags behind. After the completion of Population Census in November 2003 efforts need to be stepped up to restart the HouseholdBudget Survey. 7.15 The new ISSP surveys i s an important step in unifying these approaches and reaching a consensus over the basic poverty diagnostics. However, this consensus remains to be reached. The discussions of the ISSP H H S results will be an important step inbuildingsuch a consensus. 7.4. Challengesin Strengthening Feedback Mechanisms. 7.16 Monitoring should not be a stand-alone, technical activity. it should be closely linked to the decision-malung processes at all levels and should provide feedback to project managers, policymakers, and civil society on, among other things, the performance of existing policies and programs. 7.17 Thus, a crucial element of the monitoring and evaluation system is the existence of a feedback process. A feedback process i s a mechanism by which monitoring and evaluation results are disseminated and used to decide on future courses o f action. Monitoring and evaluation systems that provide results to only a select group o f users (central ministries, for example) risk being underused and losing financial and political support. Wide dissemination of results reinforces the system by strengthening an outcome-based culture. 7.18 The dissemination strategy for SLS (Serbia) and H H S (Montenegro) accommodated the diverse information needs o f different groups, including policymakers, program managers, 143 program beneficiaries, the general public, the media, and academics. For example, several conferences and workshops for academic groups have taken place. At the same time, the media were briefed even on preliminary results early on. That created a demand for the data not only in the research community, but also among the general public. 7.19 Reports that include main findings and emphasize implications for policy and program design have been prepared and now are distributed among government officials in central and line ministries as well as local administrations. Detailed reports were also produced for program administrators and researchers. 7.20 The posting of survey information on the PRSP web site made it available to interested audiences within and outside of the country. It i s important, however to ensure that findings and recommendations are accessible to community councils, local women's organizations, and ethnic, religious, environmental, and other groups representing the communities to which programs are targeted. Most of these groups in Serbia, however, have access to information technology andconventional dissemination mechanisms. 7.21 The active participation of NGOs and other local organizations may be crucial to ensure that all sectors of the community are reached. 7.22 In addition to different results compared to HBS, the key difference of new surveys is that actual data and careful documentation of methods of analysis will be made available to the public. Reluctance to release unit record data can give rise to suspicion, while open access and discussion of data, methods, andresults foster transparency andbroad acceptance of the findings. 7.23 Open access to unit record data also enables NGOs to carry out independent analysis and increases the demand for data, which helps ensure the sustainability o f the monitoring and evaluation system. 7.24 S A M has no record of providing public access to unit record household level data top all interested parties. A study i s needed to establish whether there are legal impediments to the dissemination of raw data related to the protection of privacy. It i s important to note that such impediments could be overcome with technical solutions that make it very hard to identify respondents and changes in the legal framework; many countries now grant open data access, and lessons have been gained from their experience. 144 Aghion, P. and 0.Blanchard (1994), "On the speed of transition in Central Europe", National Bureaufor Economic ResearchMacroeconomic Annual, pp. 283-320. Alderman, Harold (2002), Subsidies as a Social Safety Net: Effectiveness and Challenges, SP Discussion Paper No. 0224 Atkinson, A. (1998), Poverty inEurope, Blackwell Atkinson andF.Bourguignon (eds), Handbook of Income Distribution, Elsevier Atkinson T., Cantillon B., Marlier E. and Nolan B. (2002) Social Indicators: The EU and Social Inclusion, Oxford UniversityPress, Oxford Barnett, W.S. 1996. Lives in balance: Age-27 benefit-cost analysis of the High/Scope Perry Preschool Program (Monographs of the HigWScope Educational Research Foundation, 11). Ypsilanti, MI:HigWScope Press Boeri, T. (1999), "Optimal speed of transition 10 Years After," IGIER Worlung Paper No. 154, 1999. BogiCeviC B., G. KrstiC, B. MijatoviC. (2001) "Siromagtvo u Srbiji iReforma Driavne PomoCi SiromaSnima"( The Poverty in Serbia and Reform of the State Assistance to the Poor). Beograd : Centar za liberalno-demokratske studije Belgrade, 2001 BogiCeviC B., G. KrstiC, B. MijatoviC and B. MilanoviC. (2003) Siromagtvo i Reforma Finansijske Podrske Siromagnima (Poverty and Reform of the Financial Assistance to the Poor). Beograd : Centar za liberalno-demokratske studije : Ministarstvo za socijalna pitania, 2003 (Beograd : Goragraf). BogiCeviC, Biljana (2003) System of State Support to the Poor andLaw on Financial Assistance of Families with ChildreninB.BogiCeviCet al. (2003). Bourgignon, FranCois (1999) Absolute poverty, relative deprivation and social exclusion. Paper prepared for the Workshop on the World Development Report organized by the DES, Berlin, villa Borsig, February 2-3, 1999. CPNCPS (Center for Policy Studies) Cvejic S. and Babovic M. The Survival strategies of householdsin Serbia inthe year 2000. Belgrade, 2000. Deaton, Angus (1997) The Analysis of Household Surveys: A Microeconomic Approach to Development Policy. Washington, D.C.: World Bank. 145 Deaton, Angus and Zaidi Salman `` Guidelines for Constructing Consumption Aggregate", SLS Working Paper Series, World Bank, 2002. Demery, Lionel (2000), Benefit Incidence: A Practitioner's Guide, World Bank mimeo Dollar, D. and Kraay, A. (2000) "Growth i s Good for the Poor", World Bank, Development Research Group Dutz Mark, Celine Kauffmann, Serineh Najarian, Peter Sanfey and Ruslan Yemtsov Labour Market States, Mobility and Entrepreneurship in Transition Economies. EBRD Working Paper, Nov, 2001. Earle, J. and Z. Sakova (2000), "Entrepreneurship from scratch: lessons on the entry decision into self-employment fromtransition economies", IZA Discussion Paper, No. 79. EBRD(2000), TransitionReport 2000, London. Eurostat (2003), Population and Social Conditions, Poverty and social exclusion in the EU after Laeken. Report no N o 8/2003. Luxembourg, 2003. Farber, H. (1997), "Job creation in the United States: Good jobs or bad?" Industrial Relations Section Working Paper 385, July, PrincetonUniversity. Federal Republic of Yugoslavia: InterimPoverty Reduction Strategy andJoint Staff Assessment, Report No. 24490-YU, July 2002 Figini Paolo (1998) Inequality Measures, Equivalence Scales and Adjustment for Household Size and Composition, Maxwell School of Citizenship and Public Affairs Syracuse University WorkingPaper No. 185 Syracuse, New York 1998 FSO (Serbia and Montenegro Federal Statistical Office): Statistical Yearbook of Yugoslavia 2000,2001 and 2002 2003 Statistical Pocket Book Belgrade. Foster, Michael F., andIstvan Gyorgy Toth. 1999. "Trends in Child Poverty and Social Transfers in the Czech Republic, Hungary, and Poland: Experiences from the Years after Transition." Paper presented at the conference "Child Well-Being in Rich and Transition Countries: Are Children in GrowingDanger of Social Exclusion?" September 30 - October 2, 1999, Luxemborg. Fox, Louise (2003), Safety Nets in Transition Economies: A primer, SP Discussion Paper No. 0306 Galor, 0. and J. Zeira (1993), Income Distribution and Macroeconomics, Review o f Economic Studies, 60,35-52 Gavin, M. (1996), "Unemployment and the economics o f gradualist policy reform", Joumal of Policy Reform, 1(3), pp. 239-258. 146 Government o f the Republic of Serbia (2002) Interim Poverty Reduction Strategy Paper for the Republic of Serbia. Belgrade, June 18,2002. Government o f the Republic o f Serbia (2003) Poverty Reduction Strategy Paper InThe Republic of Serbia. FirstDraft. Belgrade April 24,2003. Government of the Republic of Montenegro (2002) Interim Poverty Reduction Strategy Paper. Podgorica, June, 2002 Grootaert, C. (1998) Social Capital: The Missing Link?Nat Colletta and Michelle Cullen, (eds.), Violent Conflict and the Transformation o f Social Capital, 2000 IMF/IDA: The International Monetary Fund and The International Development Association (2002) FRY: Assessment of the InterimPoverty Reduction Strategy Paper (PRSP). Washington, July 23, 2002. Institute for Strategic Studies and Prognoses (ISSP). 2002b. "Household Survey Report #6." Podgorica, Montenegro. Institute for Strategic Studies and Prognoses (ISSP). 2002a. "Household Survey Report #5." Podgorica, Montenegro. Institute for Strategic Studies andPrognoses (ISSP). 2003. "Montenegro Economic Trends (MONET)." Podgorica, Montenegro. International Committee of the Red Cross (2003), The vulnerability Assessment of Internally Displaced Persons in Serbia and Montenegro, mimeo Kakwani, Nanak, Shahidur Khandker and HyunH.Son, Poverty Equivalent Growth Rate: With Applications to Korea and Thailand. 2003 Klugman, Jeni, John Micklewright and Gerry Redmond. (2002) Fighting Poverty in the Transition: Social Expenditures and the Working-Age Poor. Paper Presented at EBRD 10th anniversary conference, December 2001;CEPR DiscussionPaper DP3389,2002 KrstiC G. (2003) Profil siromagtva u Srbiji u 2002 (Poverty Profile in Serbia) - in B. BogiCeviC et al. (2003). KrstiC G. (2002) An empirical Analysis o f the Formal and Informal Labour Markets in FR Yugoslavia (1995-2000). University o f Sussex, PhDDissertation. KrstiC G., Reilly B. The Gender Pay Gap in the FRY, 2001. JoviEiC M. , NojcoviC A., Paranos A. Labor force survey data on wages and earnings in the private and social sectors inFR Yugoslavia, 2000. Landers, Cassie (2001) Early Childhood Development in the Federal Republic o f Yugoslavia: Suggested Strategies for UNICEF. New York. 2001. 147 Lanjouw, Peter, Branko Milanovic and Stefan0Paternostro, Poverty andEconomic Transition: How Do Changes inEconomiesof Scale Affect Poverty Rates of Different Households?//World BankPolicy ResearchWorking paperno 2009, November 1998. Lehmann, H. and J. Wadsworth (1999), "Tenures that shook the world: worker turnover in Russia, Poland and Britain", IZA DiscussionPaper, No. 90. Lokshin M., Jovanovic B. (2002) State-Private Sector Employment Choice And Wage Differentials InYugoslavia. Working Papers -- Labor & Employment. Labor market policies and institutions.2959. World Bank. Micklewright, John (2000). Education, Inequality and Transition. Innocenti Worlung Papers. Economic and Social Policy Series no.74 January 2000 Mistaen Johan A. and Martin Ravallion. Survey Compliance and the Distribution of Income The World Bank Development ResearchGroup Working Paper, 2002 MijatoviC, BoSko (2003), Proposal of Reforms of Financial Assistance to the Poor - in B. BogiCeviCet al. (2003). MilanoviC, Branko (2003), Distribucija (raspodela) socijalnih transfera. Nejednakost (Distribution of social transfers. Inequality.) -inB.BogiCeviCet al. (2003). Milanovic B. (1990) Poverty inPoland, Hungary andYugoslavia inthe Years of Crisis, 1978-87. 1990World DevelopmentReport Working paper, WPS 507, 1990,World Bank. Milanovic B. and B. Jovanovic (1999). Change In The Perception Of The Poverty Line During The Times Of Depression: Russia 1993-96. The World Bank. Transition Economies Working Papers 2077. Ministry of Finance and Economy (2003), Memorandum on the budget and the economic and fiscal policies for the budget year 2003 and 2004 Ministry of Finance andEconomy (2002), Report on the activities of the Ministry of Finance and Economy andthe Public RevenueAgency of the Republic of Serbia, Jan 2001- Sept 2002 OCHA 2000. UNOCHA sub-office in Podgorica and Economics Institute of Belgrade. 2000. "Income andExpendituresinMontenegro." OECD (2001) Thematic Review of National Policies for Education: Serbia.Paris, 2001. PoSarac, A. (1991). Poverty inYugoslavia, 1978-89. Sudosteuropa, 1991, Vol. 40, No 2. PoSarac, A. Unemployment, Labor Market and Social Sector in FRY 1990-1996. World bank, 1997Mimeo. 148 PoSarac, A., BogosavljeviC S. Soup-Kitchen Program Beneficiaries in FR Yugoslavia and Identification of the Most Vulnerable Groups in Need of the Humanitarian Assistance, International Federation of Red Cross and Red Crescent Societies and Yugoslav Red Cross, Belgrade, 1996 Radevic D. and Beegle K. (2003) Living Standards and Poverty in Montenegro. Background paperfor Serbia andMontenegro: Poverty Assessment. ISSP andThe World Bank,June, 2003. Radovic, Milivoje (2003) The Effects of Social Contribution on Reduction of Poverty in Montenegro. Mimeo. Universityof Montenegro. Ravallion M. and Chen S. What Can New Survey Data Tell Us about Recent Changes in Distribution andPoverty?World BankEconomic Review, 1997, vol. 11,No.2 RavallionM.and Chen S. Measuring Pro-Poor Growth. The World Bank Development Research Group Working Paper, 2002. Ravallion, Martin, andMichael Lokshin. 1999. "Subjective Economic Welfare." Policy Research Working Paper 2106. Washington, D.C.: World Bank. Ravallion, M.(1992), Poverty Comparisons: a Guide to Concepts andMethods, Fundamentals of Pure and AppliedEconomics, 56, HarwoodAcademic Publishers Sen, A. (1983), Poor, relatively speaking, OxfordEconomic Papers, 35, 153-69 Scharle, A. (2000), Self-employment in Hungary, unpublishedPhD dissertation. Schneider, Friedrich and Dominik Enste (2000): Shadow Economies: Size, Causes, and Consequences,The Journal of Economic Literature, 38/1, pp. 77-114. Schweinhart, L.J., Barnes, H.V., Weikart, D.P., Barnett, w. S. & Epstein, AS. 1993. Significant benefits: The HigMScope Perry Preschool Study Through Age 27. Monographs of the HigMScope Educational Research Foundation, 10, Ypsilanti, MI: HigWScope Educational ResearchFoundation. Tesliuk, C. (2003) Social Protection and Poverty In Serbia. Background paper for Serbia and Montenegro: Poverty Assessment. The World Bank,mimeo, May, 2003. UNICEF. 2001. Comprehensive Analysis of Primary Education in the Federal Republic of Yugoslavia. UNDP, Human Development Report: Yugoslavia 1997. UNDP. 2001. "Employment, Labour Market and Standard of Living in Montenegro." Podgorica 2001. UNHCR-WFP (2002), Joint Assessment Mission Report, Serbia andMontenegro, mimeo 149 van de Walle, Dominique, (2001) "Viet N a " s Safety Net: Protection and Promotion from Poverty", mimeo,DevelopmentResearchGroup, WorldBank, WashingtonDC van de Walle, Dominique, (2002) "The Static and Dynamic Incidence of Vietnam's Public SafetyNet", mimeo,DevelopmentResearchGroup, World Bank,WashingtonDC World Bank (2000a), Making Transition Workfor Everyone: Poverty and Znequality in Europe and Central Asia, TheWorldBank,Washington,D.C. World Bank (2000b) Croatia: Economic Vulnerability and Welfare Study. Poverty Reduction andEconomic ManagementSectorUnit,EuropeandCentralAsia Region.Washington, D.C. WorldBank (2001a)FRY: Transitional Support Strategy,World Bank, 2001. ReportNo. 22090- YU. World Bank (2001b). Breakingwith the Past: The Path to Stability and Growth. World Bank, 2001. World Bank (2002a) Kosovo: Poverty Assessment. Poverty Reduction and Economic Management Sector Unit, Europe and Central Asia Region. Washington, D.C. Report No. 23390-KOS. World Bank (2002b) Transition: The First Ten Years. Analysis and Lessonsfor EasternEurope andthe FormerSoviet Union,WorldBank,2002. World Bank (2002~)FRY: Public Expenditure And InstitutionalReview, World Bank, 2002. ReportNo. 23689-YU. WorldBank (2003a) SAM: TransitionalSupportStrategyUpdatefor FY 04, World Bank,2003.. WorldBank(2003b)Albania: Poverty Assessment.Washington,D.C. World Food Programme (WFP 2001) , The Survey on Poverty and Living Conditions of the RefugeeBeneficiariesof WFP Aid inSerbia.Belgrade,2001 Case3~11 3a~oa3a cTaTacTaKy. IIoTpouama Kopna, ueHe IInpocewa HeTo sapaaa, Eeorpaa, 1999-2002. 150 ANNEX I MEASUREMENTOF POVERTY This Annex summarizes key methodological principles for poverty measurement developed in Krstic (2003) and Beegle and Radevic (2003). It also includes a section on the newly available results for Montenegro (2003 data, ISSP-UNDP [2003]). This Annex provides only a short overview andfor detailed references the reader should address theprimary sources. Survey Data This report is based on special integrated surveys conducted by an NGOs in 2002. The key objective of these surveys was to establish a baseline poverty profile. The survey was conceived as a "pilot" activity to be conducted by an independent research agency that would (i) serve as a pilot for the reform of the official HBS, addressing its key shortcomings, and (ii) provide the data and analytical inputs needed for reforming the social protection system and improving its targeting. In Serbia The Ministry of Social Affairs saw the survey as a key tool in the reform of social policy during the year 2002. In addition to social policy, the survey was also seen as key to reform measures in other areas, notably energy price reform, owing to the social impact of rising prices. Implementation arrangements for the survey included the Steering Committee, the survey team, the implementing agency and coordination between different stakeholders (relevant ministries, academic participants, NGOs, international agencies andthe World Bank). The survey was designed to contribute to buildingthe capacity o f Statistical offices in reforming the existing the HBS. The staff members from the Statistical Offices (Federal and Republican) were involved in the planning and implementation of the survey, and they provided critical inputs from the April 2002 Census to the sampling of the survey during May 2002. The Strategic Marketing and Media Research Institute (SMMRI), a local research firm with an outstanding record in survey data collection, implemented the Poverty Survey and prepared its full dataset for the analysis. Its responsibility was to consult with the Steering Committee and other relevant stakeholders in the process. The Steering Committee was be responsible for monitoring and for following up the progress and dissemination of survey findings and data. Serbian "SLS" (living standards measurement study) questionnaire design was based on the principles o f combining critical elements o f the Bank's model questionnaire of the same name (focus on the multi-dimensional aspects o f living standards, focus on consumption measurement) with the elements of the standard HBS (Household Budget Survey: daily records of purchases, 151 coding o f items, standard classification for employment, elements o f income modules). The decision to adopt formats as close as possible to HBS design (without compromising on quality) was taken to ensure that the results and lessons would be applicable to the reform of the official household survey. The new Survey conducted inSerbia inMay-June 2002 is the largest and most comprehensive study of living standards in the last decade. It was implemented as the Government of Serbia's own study, supported by the World Bank advisory team and the Dutch Trust Fund. It i s called in Serbia "SLS" (living standards measurement study) to highlight its close links to the Bank's program of surveys which developed a model questionnaire with the same name. Though many principles o f SLS influenced the design of the study, it was not the only source, and the SLS group was not directly involved in the survey. W e therefore keep the original title of the survey as "Household Poverty Survey", highlighting one of its main objectives. The Strategic Marketing and Media Research Institute - SMMRI, local research firm with an outstanding record in survey data collection, implemented the Poverty Survey and prepared its full dataset for the analysis. Its responsibility was to consult with the Steering Committee and other relevant stakeholders in the process. The Steering Committee was to be responsible for monitoring and following up the progress and dissemination of the survey findings and data. The activity started with the meeting of the key stakeholders to identify the broad objectives of the activity in December 2002. The process o f consultations involved all relevant stakeholders in the process. The progress was followed by the establishment of the Steering Committee of the Survey, and the subcontracting of the survey implementation to SMMRI. The questionnaire design was ajoint Ministry of Social Affairs-SMMRI-World Bank activity with the participation of key stakeholders that and was completed by April 2002. At this time, the survey team conducted the pilot test, adjusted the questionnaire, and built a random stratified sample using the maps and listings from the April 2002 Census. The survey was conducted in June 2002. Duringthe field work the data entry started and was completed by the end of August 2002. In September-October 2002 the survey team worked together with Steering Committee and the World Bank experts to finalize key definitions of the variables, and carried out a full set of consistency checks. This work was completed by mid-November 2002. Key findings o f the joint analysis by the Ministry of Social Affairs o f Serbia and the World Bank team were produced during the workshop in mid-October at the World Bank and the follow-up data analysis by the SMMRI staff completed jointly on November, 9. These basic findings were reviewed and finalized as agreed upon as the decision on methodological principles and jointly presented duringthe PRSP conference inDecember 2002 inBelgrade. The survey was designed to be representative for six macro regions (Begrade, Vojvodina, Western Serbia, Central Serbia, Eastern Serbia, and South-East Serbia) and within each region for urban and rural areas. The sampling procedure relied on a two-stage sampling. At the first stage, a random selection of a given number of enumeration districts as PSUs was based on probability proportional to size (PPS) in each of 12 strata. The enumeration district i s the smallest area defined in the 2002 Census data and on average in Serbia consists of about 70 households. In total, 618 primary sampling units (enumeration districts) were selected for the sample. 152 At the second stage households were randomly selected from the Control Book (the listing of households with addresses) for the Census 2002 in each enumeration district (PSU). A list o f 11 randomly drawn households with addresses was prepared for the primary sample, in each enumeration area, with two additional addresses as reserves incase some households on the main l i s t would decline to participate. Thus, the estimated target size of the sample was about 6,800 households. The realized sample size was 6,386 households. Weights or expansion factors have been developed at the PSU level to bring the structure of the survey by strata to the 2002 Census estimates and to correct for the non-responses. The variation of such weights across PSUs i s relatively moderate (the coefficient of variation i s around 30 percent), malung overall estimates of means from the survey rather precise. The information base for the survey i s therefore the most updated possible and produces overall estimates generalizable to the actual population of Serbia. It included refugees and IDPs privately accommodated, and minorities to the extent that they were covered by the 2002 Census. Some special issues of coverage o f an important ethnic minority, Roma, was addressed in the follow-up round 2 special survey of Roma (500 Households), results of which are now being analyzed. In Montenegro, the survey HHS have been developed and fielded by the Institute for Strategic Studies and Prognoses (ISSP). This work has been supported by the European Commission Food Security Programme (with significant contribution coming from Mr. Vasilis Panoutsopoulos), USAID Montenegro, and Chesapeake Associates. In addition, for H H S Rounds 4, 5 & 6, ISSP&CEED have received technical assistance and feedback from World Bank staff. In an effort to enhance the usefulness of the H H S to policymakers and researchers studying poverty and living standards inMontenegro, support from the World Bank for Round6 was used for two main purposes: expand the sample size, and undertake collaborative poverty analysis. This support greatly improved the data source in terms of level o f representation and quality, and, therefore, its usefulness as a source for analysis and fact-based policy-making. The sample size of 500 households inRounds 4 and 5 was consideredto bejust large enough to generate national statistics, but i s too small to generate statistics for regional estimates (North, Center, South) and too small for most subgroups of particular interest (such as vulnerable populations, includingthe poor, elderly, children, and unemployed). Using the existing sampling frame (the listing o f individuals from the Mass Voucher Privatization program, MVP8') the sample has been expanded appropriately, to include a total o f 800 households from the MVP listing. The emphasis in the first three rounds of the ISSP H H S was the basic demographic characteristics of the household, individual income earnings, and non-earned sources of income for the household. InRounds 1-3, thus, income i s the main indicator o f well-being. Pursuant to field experiences and discussions with World Bank staff, representatives from other donors, and government agencies, the questionnaire has been subsequently revised since Round 3. Rounds 4, 5 and 6 have introduced detailed consumption and expenditure modules as well as greater range of topics. The main topical areas of the integrated surveys inRounds 4, 5, and 6 are: 0 Householdroster including migration information Housing conditions, including information on utilities The HouseholdSurvey Project gratefully acknowledgesMr.Zeljko Brkovic from ZOP (Clearing house) in providing the MVP and assistingthe project inits use. 153 Durable assets owned by the household Foodconsumption values for about 87 food categories (see Annex 3) Non-food expenditures Receipt of assistance from social programs: pensions (old age and disability), scholarships for school, unemployment benefits, Family Material Support (FMS), Child Allowance, one-off support, and other person care Other non-labor income sources Employment characteristics intended to measure labor force participation in the formal and informal sector Health status Subjective well-being (perceived quality of life) Subsequent support for the survey has allowed the sample size to be increased to 800 households in Round 6 (with the support of the World Bank, as mentioned above). The questionnaires of Rounds 4, 5 and 6 have the appropriate set o f questions for the purposes o f measuring poverty usingconsumption and expenditures. All rounds collected food consumption for about 87 food items. However, the data are not comparable in the way in which food consumption i s measured and in the recall period for non-food items. Specifically, in Round 4, households report only quantities of food consumed across three sources: purchases, self-produced and received as gift from others. Thus, the value of food consumed must be imputed from reported quantities and the regional price averages (as collected b y the Statistics office). The data were collected in face-to-face interviewers by a core set of interviewers, about 28, who are in most cases from the municipality in which they work. The training o f interviewers took about 25 hours for each survey round. Given that ISSP retain a large share o f interviewers, the majority of interviewers now have considerable experience in fielding the questionnaire. The field work lasts about 3 weeks. The data are entered and processed at ISSP offices in Podgorica. The samples for Round 5 (500 households) and 6 (800 households) are the basis for the poverty statistics inthis assessment. Data were collected from each of the three main regions o f Montenegro, which constitute the north (mountainous, less populated and least developed), the central region (most populated and industrialized), and the south (coastal, most developed with tourism as main economic sector). The last census in Montenegro was done in 1991. A new census was planned for April 2002 but was postponed and i s now planned for November 2003. Given the lack of current Census data for Montenegro with which to draw a sampling list for the HHS, the ISSP team had to look elsewhere for household or individual listings. They identified two possible sources for developing the sample frame. The first i s the Voting Registration list. The second source i s the Mass Voucher Privatization (MVP) listing of all people compiled in order to distribute vouchers among the population of citizens over 18 years in the summer of 2001. Both lists exclude IDPs (which includes the Roma population inits definition). At the time when sampling was done, the MVP list was newer than the voting registration list. ISSP concluded that these two lists were fairly comparable. The MVP list was usedto identify the target sample ineach o f the 21 municipalities included in Rounds 4, 5 and 6. The number o f households targeted for each municipality was based on the population share o f the municipality inthe total population 18 and over. Households were 154 interviewed based on the sampling list for the municipality, with no clustering design inthe sample within municipality, thereby reducing survey design effects which increase standard errors. The new round of data collection addressed the issue of coverage inthe sample for some marginalized groups (see special section on page 159). Survey data collected in H H S are not comparable to other sources. The sampling frame for the H H S was based on an adult population listing from 2002, while previous studies drew samples based on the census from 1991. Moreover, the sample size o f the FSO survey (280) will be insufficient for regional statistics and perhaps national statistics also. The questionnaire of the H H S includes detailed consumption and expenditure modules as well as covering a range of other topics (such as migration information, housing conditions, durable asset ownership, labor and non-labor income including assistance from social programs and private transfers, employment characteristics, health care utilization, and a subjective assessment of well-being). In comparison, the APD questionnaire was developed and implemented without changes ten years ago and covers fewer food items. The OCHA survey used a very limited questionnaire containing a simple basket of food (15 items) plus personal and household hygiene items, while UNDP surveys asked cumulative questions about households' expenditures for a one-month recall period. The differences in the specific list o f items as well as the recall period make intra-survey comparisons difficult. A third difference in approaches is the poverty indicators used in different studies. These indicators differ according to the welfare measure (consumptiodexpenditure or income) and the construction of a poverty line. Usingthe HHS, our welfare measure i s consumptiodexpenditure (see Radevic and Beegle [2003]). The poverty line in this study identifies the cost o f basic needs as the cost of a food basket to reach nutrition (food poverty line) and the cost o f non-food expenditures of households whose food consumption i s near the food poverty line. The FSO poverty statistics compare household income (without imputed housing value and with savings) against the official "minimal food basket" which i s a food poverty line. The OCHA Survey uses both the official FSO food poverty line and a poverty line based on the "OCHA Podgorica Shopping Basket'' based on expenditures for only 15 food items, expenditures on hygiene items, and limitedexpenditures for electricity, heating, education, and medicine. The UNDP study uses multiple poverty indicators, including ratio of food expenditures, income-to-expenditures ratio, income per capita below 100 DEM, and household expenditures below 150DEM; the 100DEM cutoff was chosen based on the OCHA report. Definition of welfare. Household material welfare i s defined in broad terms. Consumption i s our preferred measure, as we believe it i s better declared and i s less subject to short-term fluctuations. I t s composition and link to the questionnaire is given in the table at the end of this annex. We also experimented with income in combination with consumption to check and validate consumption based results. Consumption i s a comprehensive aggregate o f current consumption expenditures (investment type and productive expenditures are excluded), in-kindconsumption of own production, value of gifts and transfers received in kind, imputed value of owner occupied housing, and value o f flow of services from durables owned by the household. To get as accurate a ranking o f 155 households as possible, we correct for differences inneeds by age groups, for economies of scale inthe household, andfor regionalprice differences (details are givenbelow). Definition of poverty. We define poverty in absolute terms. The value of the poverty line is the minimum cost of food and non-food goods and services at which basic nutritional requirements are fully met. It i s anchored in the minimumfood basket and in the actual consumption structure of the population (details are given below). As an alternative and a cross check we also apply commonly used absolute poverty lines, such as those of the Republican Statistical Office, the Federal Statistical Office (planned), and the WFP poverty line, as well as a subjective poverty line based on the survey. We also use a simple definition of food poor, and compare it to the value of food consumption only. The poor belong to households whose current consumption (or income) per equivalent unit i s below the value of the poverty line. Minimum Food Basket The absolute poverty line i s based on the minimum food basket. The survey allowed the development of the minimumfood basket that meets the basic nutritional requirements at 100(or more) percent and has the minimumcost. In Serbia the new basket reflects the actual consumption structure in Serbia, as its composition by items was set to be at least equal to the consumption inthe lowest decile. The comparison to other minimumfood baskets in use shows that it has adequate variety and specificity, and has the advantage of being very up-to-date. Its costs for a standard family of four (one male of working age, one female of working age, one male child below 6 and one female child 11-13) i s 7,605 dinars per month inthe average Serbian prices for May-June 2002. InMontenegroThe minimumliving standard inEuros is based on the value of.a minimumfood basket. This minimum consumption basket of food i s itself based on a standard of supplying 2,2B8' calories per person per day. The value o f this food basket i s then calculated based on price data. This i s the FoodPoverty Line (FPL). The FPLrelies on three data sources: 1. data on consumption patterns in order to allocate calories across different food items based on a chosen reference population 2. data on the caloric content of each food item 3. data on prices in order to establish a monetary cost o f the minimum food basket (the FPL) We base the minimum food basket on the consumption patterns o f the households in the survey data in the lowest 15%82o f per capita consumption and expenditure. For this population, we compute one-half the mean intake (quantity, typically in kilograms) of each food item.83 The caloric content o f the 86 food items (excluding the category "other misc foods") in the survey i s 81In order to have data and results comparable with the results in the region, for Montenegro we use the same standard that was used inSerbia. 82The FPL amounts to about 41 EUR. This is a rough average between the three FPL estimated: 42.26 EUR for 20% reference population, 41.0 for 15%, and40.15 EURfor 10%reference population.) 83When the actual amount o f intake was taken, the level o f consumption was 100%above the minimumfor all items and the optimization procedure does not work. 156 based on U S D A (2002). Usingthese two sets o f data, an optimization procedure i s applied that establishes the a minimum quantity for each food item that will reach the standard of 2,288 calories per person per day and maintain the consumption patterns in the reference population's food basket. Finally, the price data from the Statistics department are applied to establish the cost o f this basket. This i s the MinimumFoodBasket. there are alternative food baskets that have been used to identify the minimum amount of Euros needed to buy a minimum standard of calories. These include the FSO food basket and the OCHA food basket. The differences in the food poverty lines used by different groups will depend on the contents of the food basket (specifically, the distribution across foods that have different prices per-calorie), the minimumstandard the basket aims to reach (interms of calories and other nutritional criterion) which impacts the quantity of each food item in the basket, and the prices used to cost the basket. Taking one o f these factors, the distribution o f total basket calories, the FSO basket and the Serbia poverty profile basket can be compared with the basket usedinthis study. (Details for the OCHA basket are not available). Note that this comparison i s only a partial evaluation of differences inthe food poverty lines, since it ignores the quantities in the basket (set to reach a calorie standard that varies) and the prices used. Table A.1 shows that the FSO basket allocates a higher share of calories in the minimum food basket to more expensive foods (where "expensive" refers to a per-calorie cost): meat and dairy. On the other hand, the Serbia food basket used by Krstic (2003) looks similar to the basket Beelge and Radevic (2003) calculate for Montenegro. This i s consistent with the development of the food baskets. Both of the Montenegro and Serbian baskets are determined based on the consumption patterns of the lower-income population as reflected in the recent household surveys used. The FSO basket, on the other hand, i s not developed from recent food consumption patterns of the population. Table A.l: Share of Food Basket Calories by Food Groups Montenegro Serbia SLS Foodgroup (ISSPKEEDHHS5 & 6) FSO (Krstic, 2003) Flour, bread, pasta 44.3 30.2 43.5 Vegetables 3.3 5.6 6.3 Fruit 3.5 4.2 4.0 Meat 4.8 8.5 3.9 Fish 0.1 0.1 0.0 Dairy and eggs 23.7 35.7 16.0 Butter and fats 12.5 8.7 14.4 Beverages 1.5 0.5 5.0 Other (other condiments, spices, 6.4 6.5 7.0 baby food, etc..) Source: Beegle and Radevic (2003) Correctingfor Regional Price Differences Actual survey price (median) were used to get the local value of the food basket. We assume that the rest of the prices are the same across the country, and we take their share among the poor to weigh all price components and to arrive at the overall regional price index given in the last row of the table. This price index i s used to compare the actual consumption to the poverty line. 157 Derivation and Composition of the Full Absolute Poverty Line To get the value of the full absolute poverty line per equivalent unit, we estimate the level of consumption at which the minimum food requirements can be met. In order to do this, we estimate the Engel curve for food and we find the value of consumption at whish a household i s expected to spend as much on food as i s necessary to meet its basic food needs (cost of the minimumbasket). The Figure A1 give the actual derivation for the Serbian equivalence scale of the poverty line per equivalent adult. We also experiment with other scales, and get the full consistency of results (see below). FigureA.l: Engel curve and the Poverty Line Poverty line pfsr :4654.097164043827 4-1 -4 1 6I Ln8I EquivalentConsumption '9I 12I The vertical axis gives the actual food consumption of a household relative to the cost of the minimumbasket: it is zero when the HHspends exactly as much as needed. Horizontal axis is the consumption per equivalent unit. Each dot is a HHin SLS. The intersections of three lines (Engel curve, minimum food requirements, and level o f consumption) gives the poverty line. It i s important to see whether this line implies enough to spend on non-food, as it i s not explicitly based on fixed allowances. Table A.2 below for a family of four and a single person show such check. It would appear that the amounts left for basic non-food needs are adequate. Table A.2: Poverty line by regions Food Imputedrent Everything else Total line per family of 4 City Other City Other City Other City Other Belgrade 8500 8500 3000 800 6300 6100 17800 15400 Vojvodin 7200 7000 1500 600 6200 6100 14900 13700 West Ser 7000 6600 1700 900 6100 6000 14800 13500 Central 7200 6600 1900 800 6100 6100 15200 13500 East Ser 7100 7100 1300 700 6200 6100 14600 13900 South-Ea 6800 6400 1600 700 6200 6000 14600 13100 Single person Food rent everythingelse City Other City Other City Other City Other Belgrade 2200 2200 1500 500 1800 2100 5500 4800 Vojvodin 2000 2000 700 300 1900 1900 4600 4200 West Ser 2000 1900 900 400 1700 1900 4600 4200 Central 2000 1900 1000 400 1700 1900 4700 4200 East Ser 2000 2000 600 300 1900 2000 4500 4300 South-Ea 2000 1900 800 300 1700 1900 4500 4100 158 In Montenegro order to establish a total poverty line, Beegle and Radevic (2003) defined a minimum basket for non-food item by imputing minimum needs from food and total expenditures in the survey data. Based on our F'PLo f 41 Euros per person per month, a total poverty line (food and non-food) was estimated by evaluating the median non-food consumption for those whose food consumption lies within a range of the FPL. In this analysis, we will use the poverty line of 107 Euros per person per months (41 for food and 66 for non-food goods and services). Equivalence Scale Krtsic (2003) and Beegle and Radevic (2003) present evaluations of the equivalent scale using the Engel method. Such an equivalent scale for Serbia gets the OECD-type form of the following shape: Serbian Scale = (1 0.81" (adults - 1) 0.24*children06 + + +0.75*children718). This equivalent scale was used to bringhouseholdspendingdown to the spendingper consumer unit. This new equivalence scale is significantly different from FSO scale which accounts for differences in sex, age and occupation. The consumer unit value for men varies from 0.33 (for up to 1-year-old children) to 1.50 (for miners and industrialworkers performing heavy-duty labor). For women, it varies from 0.33 (for up to 1-year-old children) to 1 (for industrial workers and farmers). Tests for Montenegro discarded any equivalence scale The standard error of the estimate i s large enough that we cannot reject the null hypothesis that there are no economies of scale. The food share i s monotonically falling as welfare increases (negative sign on bl). The explanatory power of the regression i s very weak, indicating significant "noise" in the data that makes it difficult to calibrate the exact relationship. Beegle and Radevic (2003) concluded that the estimates suggest usingthe simplest linear per capita equivalence scale as our preferredestimate. New Findings of the 2003 Survey in Montenegro and Integrated Poverty Profile The 2002 HHS data produced by ISSP with the assistance from the World Bank, as noted above, hadsome serious limitations interms of coverage of marginalized groups of society: Roma (Roma, Ashkaelia and Egyptians), IDPs and refugees. In 2003 UNDP, having both the commitment to alleviation of poverty and with the strong support o f UNHCR, UNICEF, UNOHCHR, ICRC and key RAENGO partners has been able to overcome these limitation in technical partnership with ISSP. The Report on Household Survey o f Roma, Ashkaelia and Egyptians, Refugees and Internally Displaced Persons based inthe survey conducted by the Institute for Strategic Studies and Prognoses (ISSP), was intended to complement the existing sources o f householddata for the regular Montenegrin population (ISSP-UNDP [2003]). This report usedthe best practice approach to sample selected groups of the population and internationally recognized methodology to assess their poverty, comparable to the previous assessments (Beegle and Radevic [2003]) and provided a quality input for the final PRSP. The 159 National Development and Poverty Reduction Strategy posted on the PRSP web site in November, 2003 (http://www .mnmsp.v1ada.w.vu/en/en.html) incorporates these new findings. This new rich document will be discussed inthe upcomingJSA andby the civil society in Montenegro. Here only a short technical summary o f the poverty profile is reported. The data for the Roma sub sample was collected incooperation with the informal network o f the non-government Roma organization "Romski hug". According to this source, there are approximately 19,500 Roma in more than 3,160 households living in Montenegro at the present time. Montenegro i s home to domicile Roma people, Roma who are displaced persons, as well as Roma refugees from ex Yugoslav republics. The ISSP-UNDP survey engaged a sample of 269 Roma households inwhich live 1,540 Roma people. Information for reguees from other former republics of SFRY came from the refugee database in Montenegro, provided by UNHCR. After all data was collected, a total o f 166 refugee households were included in the survey. Intotal, representing 650 individuals living in those households in 11different municipalities. This sample represents approximately 4.5% of the estimated refugee population inMontenegro and it i s representative on the republic and the regional level. Sampling information on IDPs came from complete official registration of IDPs inMontenegro that was conducted at the end of 1999, the total number of IDPsregistered was 19,132.84 In total, 263 displaced households with 929 individuals were included in this research. Survey also covered the regular population with a sub sample based on the same principles as in round 6 of ISSP HHS described above; intotal, 838 valid questionnaires were collected. The sample for the regular population had a control function, which allowed for other sub-samples to be representative at the national and regional level and enabled valid comparisons with 2002. The questionnaire was designedjointly with UNorganizations, Government agencies and NGOs. It followed basic structure of the round 5 and 6 ISSP HHS with its detailedconsumption module and questions on a variety of aspects o f well-being and access to services. It therefore allowed to reproduce basic information on poverty as reported inBeegle and Radevic (2003) with the new data. The resultingintegratedprofile i s reproduced below Several important issues stand out. The national level poverty incidence for domicile population has not changed in a statistically significant sense compared to 2002. Therefore, key conclusions of this report remain valid for the country in general. But the gain inaccuracy and specifically in understanding conditions of the most vulnerable groups i s very impressive. The survey result show that relatively limited poverty estimate at the national level may hide staggering deprivation for specific groups in a society and require serious attention. 84A new registration of IDPs was recently completed, showing that sincethe prior registration, approximately 10,000 IDPs haveleft Montenegro. According to international organizations that deal with these issues, the number of registeredIDPs who came back to Kosovo is considerably less; therefore, it is possible that the IDPs have lost interestinregisteringbecause humanitarian aid is lower andbecause IDPshave socially and economically integrated themselves into the localenvironment. Additional researchthat the Commissariat for Displaced Persons in cooperationwith UNHCR will conduct should determine whether those personsreally left Montenegro. 160 TableA3. Indicatorsof PovertyandPopulationVulnerability inMontenegro(2003) 2,7% 1,9% 3,1% (19.534) (13.308) (22.105) Expenditure/expenses under 52.3% 38.8% 38.6% the absolutepoverty line (10.216) (5.164) (8.532) I I 11.7% 5.94 9.9% Number of total poor per reeions(%)*** Regionalpoverty rates N 193 I ( 1I 1108 8,( 14 9I 6 4 6 8 13 t Economic vulnerability and absolute poverty: Expenditure/expensesbelow the 34.4 31.1 75.6 68.9 13.2 absolutepoverty line +SO% (173.4 Der month Der Derson) Poverty gap 2.7 23.2 12.1 10.2 36 Severity of poverty 1.5 I1 1.o 13.8 1 5.5 1 3.8 44.4 31.2 26.4 percentageof the poverty line Poverty with respectto education: 16 to 24 years old that 17.2 70.0 29.3 8.0 are notinschoolanddidnot 4.7 attendsecondarv school Poverty andhealth: 6.1 6.2 9.3 3.3 3.2 Any illness/injury inthe last 30 days which preventedor disabled standardactivity ~ Poverty with respectto employment 16 to 65 years old 17.4 17.0 43.3 32.5 that are not working,but ready to 30.4 work if employmentopportunities are identified Poverty with respectto living conditions The source of drinking 18.6 16.0 74.7 28.5 39.9 water inthe flathouse is not from the water network or the flathouse does not havea bathroom Less than 10m2per personin the 11.3 8.2 85.8 54.5 50.1 flathouse Notes: *RAE in a term used to describe Roma, Ashkaelia and Egyptians; Total RAE population consists of both domestic RAE (14 856) and intemally displacedRAE (about4.680). **Without RAEfrom Kosovo ***Regions: N-Northern, C-Central,S-Southem) Source: The ISSP survey on household expenditure of Roma, refugees and IDPs, October 2003 and ISSP&WB Researchon Poverty and Living Standards in Montenegro, June 2003; National Development and Poverty Reduction Strategy. November, 2003 (http://www.mnprsp.vlada.cg.ytUeden.htm1) 161 ANNEX I1 STATISTICAL TABLES AND FIGURES. Table A.l.l: Serbia and Montenegro: Overview of Poverty Lines in2000 and 2002, (all values, including derived for equivalent adult are expressed as per month per capita Dinars and US$ at the current market exchange rate). 2000 2002 YUD USD YUD USD New minimumfood basketpoverty line (baseline) na na 2,158 $32 Full absolute poverty line (baseline) na na 3,704 $56 The cost o f the minimumWFP poverty basket 947 $19 3,392 $51 The cost of the food basket by FSO 1,114 $24 2,985 $45 Minimumwage* 258 $6 Minimumpension** 974 $21 Eligibility threshold for social assistance** 1,123 $24 2447 Relative poverty line (60% of the equivalent income) na na 4557 $68 Subjective** * 2,150 $44 6,600 $99 Notes: data represent weightedaverages for Serbia and Montenegro *40 percent of average wage **50 percent of municipality's average wage, here assumedmeannationalwage *** Only for Serbia Sources: Staff estimates, for exact values and detailsof derivation anddefinition see Krstic (2003), Radevic andBeegle (2003) TableA.1.2: Regional Variation of MultidimensionalPoverty Indicators (percent of population below corresponding poverty threshold) Non-income Non-income Region Consumpti Extreme Rights Housing Health Education Employm onPoor poor poor Poor poor poor ent poor poor (at least extreme poor 2) (3 and more) Belgrade 7.9 3.3 8.4 16.4 2.7 7.5 2.1 5.1 0.6 Vojvodina 8.8 2.2 9.7 9.6 3.2 16.1 3.5 6.9 0.9 West Serbia 13.5 1.7 7.0 23.6 4.6 23.1 4.3 12.3 1.8 Central 10.2 2.0 8.9 18.4 4.3 22.2 3.7 10.8 2.0 East Serbia 10.1 1.6 7.8 13.9 3.5 21.2 2.4 8.8 1.o South-East 16.6 2.6 9.2 22.1 3.8 21.4 3.7 10.9 1.7 City 7.8 2.5 8.7 11.9 3.1 7.9 3.3 4.8 0.6 Rural 14.2 2.1 8.7 21.8 4.1 29.4 3.2 13.4 2.2 Total 10.6 2.3 8.7 16.2 3.6 17.3 3.3 8.5 1.3 Source: staff estimates basedon SLS 162 Table A.1.3: Determinates of Poverty inSerbia in2002. Dependent variable: In (spending per consumer unit) Poor Vulnerable Characteristics of the household head or household Coeff s.e. Coeff s.e. Share o f children 0-6 in household 0.541 (0.074) ** 0.507 (0.079)"" Share o f childr. 7-14 inhousehold -0.149 (0.059) * -0.156 (0.060) ** Share o f childr. 15-18 inhousehold 0.039 (0.068) -0.004 (0.069) Share of adults 19-25 in household 0.243 (0.053) ** 0.244 (0.052)** Share of adults 26-45 inhousehold 0.180 (0.035) ** 0.157 (0.035)"" Share o f elderly 65+ in household -0.149 (0.026) ** -0.175 (0.025)** Ln(no. of household members) -0.220 (0.020) ** -0.193 (0.019)"" Age of household head -0.003 (0.001) ** -0.003 (0.001)** Woman head of household 0.011 (0.018) 0.001 (0.018) Incomplete elementary education -0.267 (0.022) ** ** -0.262 (0.021)"" Elementary education -0.213 (0.017) -0.184 (0.018)** Jr. college education 0.124 (0.025) ** 0.128 (0.027)** University education 0.287 (0.024) ** 0.299 (0.026)** Employer or self-employed 0.096 (0.040) * 0.134 (0.040)** Farmer 0.028 (0.051) 0.061 (0.047) Other active 0.038 (0.117) 0.082 (0.092) Unemployed -0.060 (0.031) -0.041 (0.034) Retired -0.018 (0.021) -0.012 (0.021) Other non-active -0.145 (0.038) ** -0.139 (0.033)** % employers or self-employed inhousehold. -0.026 (0.044) -0.100 (0.043)* % farmers in household 0.011 (0.055) -0.049 (0.052) % other active inhousehold -0.074 (0.081) -0.143 (0.069)* % unemployed inhousehold -0.264 (0.027) ** -0.279 (0.028)** % employed in a private reg. firm 0.168 (0.019) ** 0.165 (0.019)** % employed ina private unreg. firm 0.034 (0.024) 0.058 (0.022)** % employed inother sectors 0.019 (0.051) -0.072 (0.050) Belgrade -rural -0.136 (0.040) ** -0.170 (0.040)"" Vojvodina - urban 0.037 (0.021) 0.011 (0.023) Vojvodina - rural 0.005 (0.025) 0.008 (0.025) W. Serbia - urban 0.009 (0.032) 0.022 (0.034) W. Serbia -rural -0.030 (0.030) -0.001 (0.030) Central Serbia - urban 0.021 (0.025) 0.020 (0.026) Central Serbia -rural 0.040 (0.028) 0.042 (0.028) E.Serbia-urban -0.001 (0.032) -0.032 (0.033) E.Serbia -rural 0.093 (0.038) * 0.102 (0.031)** SE. Serbia - urban -0.094 (0.026) ** -0.111 (0.028)** SE. Serbia -rural -0.154 (0.029) ** -0.197 (0.029)** Constant 9.471 (0.056) ** 9.464 (0.054)** F(37,6348) 50.80 [O.ooO] MSD 2338.1 Adj R squared 0.2481 RSD 2712.2 RootMSE 0.4745 PseudoR2 0.1379 Note: denominates 5% of significance *** denominates 1% of significance Source: K r s t i c (2003) 163 S o u r c e I ss df MS Number o f obs = 6379 -------------+------------------------------ F ( 15, 6363) = 1 0 5 . 6 9 M o d e lI 378.261954 1 5 25.2174636 P r o b > F = 0.0000 R e s i d u a l 1 1518.19738 6363 .238597734 R - s q u a r e d = 0.1995 -------------+------------------------------ Ad] R - s q u a r e d = 0.1976 T o t a l 1 1896.45933 6378 .29734389 R o o t MSE = .48846 lny I C o e f . S t d . E r r . t P > / t [95% C o n f . I n t e r v a l ] -------------+---------------------------------------------------------------- o l d .0241559 .0283977 0 . 8 5 0.395 -.0315131 .079825 c h l -.0238347 .0373396 - 0 . 6 4 0.523 -.0970328 .0493634 u -.0656445 .0096352 - 6 . 8 1 0.000 -.OB45327 -.0467564 e .2395975 .028345 8 . 4 5 0.000 .1840319 .2951632 farmer .0383411 .0249849 1 . 5 3 0.125 -.0106377 .0873198 indpolj .0442638 .016539 2 . 6 8 0.007 .0118418 .0766859 transf .0695026 .0240602 2 . 8 9 0.004 .0223365 .1166686 y r s c h o l .0686694 .002603 2 6 . 3 8 0.000 .0635666 .0737722 rural -.0234819 .0168655 -1.39 0.164 -.0565439 .0095802 fhh .0509886 .015352 3.32 0 . 0 0 1 .0208935 .0810837 -Istratm-2 .OB68756 .0190647 4.56 0.000 .0495023 .1242489 -Istratm-3 1 1IIIIII11I11 .0415842 .0243889 1 . 7 1 0.088 -.0062262 .OB93946 -Istratm-4 .0721077 .0211491 3 . 4 1 0 . 0 0 1 .0306485 .113567 -Istratum-5 .0950997 .0245295 3.88 0.000 .0470136 .1431858 -Istra-m-6t -.0879291 .022042 - 3 . 9 9 0.000 -.1311389 -.0447193 cons 1II 8.275437 .0402262 205.72 0.000 8.19658 8.354294 Source: own e s t i m a t e s based o n SLS (2003) 164 EducationAppendix Table A.1.4: Probability of spending on education (household level). The dependent variable takes the value 1for non-zero expenditures on education. 0 otherwise. The Probit model is estimated (marginal effects reported), Itl- statistics inparantheses. Spendingon Spendingon Spendingon Total spending Means& kindergarten elementary & higher on education Standard secondary education Deviations Consumptionper adult equivalent 1.3%-03 9.79e-03 5.67e-03 3.30e-02 10049 (thousandDinars) (4.39) (4.20) (8.21) (10.9) [6814] Consumption (thousandDinars) -1.89e-05 -1.74e-04 -9.61e-05 -5.56e-04 squared (2.56) (2.98) (5.93) (7.77) Number of householdmembers ,0019 ,0596 ,0181 ,0872 3.09 (3.48) (14.0) (13.1) (15.2) [1.62] Share of females in the household -.0048 ,0479 ,0352 ,0891 0.54 (1.14) (1.56) (4.18) (2.49) [0.26] Share of household members aged ,1160 -.0995 -.3730 ,1931 0.04 6 or less (21.1) (1.89) (11.2) (3.22) [0.11] 7-18 ,0323 1.68 -.0395 2.41 0.10 (6.20) (39.3) (2.90) (39.1) [0.17] 19-30 -.0065 ,0575 ,1761 ,592 0.13 (1.48) (1.77) (21.4) (17.0) [0.21] 31 or more -___- __-_- ____- 0.73 [0.27] ~ ~ LogLikelihood -638.5 -971.5 -1239.4 -1811.8 Number of households 6386 6386 6386 6386 6386 165 7Table A 1.5: Conditional on non-zero household expenditure on education, predictors of amount spent. OLS re ession estimates are re orte Jtl-statisticsinp entheses. Spendingon Spendingon Spendingon higher Total spendingon kindergarten elementary& Education education I secondary Consumptionper adult equivalent 96.9 89.9 117.6 120.1 (thousandDinars) (4.06) (10.8) (4.83) (11.7) Consumption (thousandDinars) -1.46 -1.47 -1.48 -1.74 squared (2.96) (6.64) (2.90) (6.83) Number of householdmembers 152.9 131.6 370.3 158.8 (2.88) (7.85) (5.73) (7.68) Share of females in the household 186.8 -19.7 -106.3 35.9 (0.50) (0.18) (0.34) (0.27) Share of household members aged 965.8 -1562.7 -4531.5 -973.1 6 or less (1.43) (6.30) (2.99) (4.03) 7-18 -187.6 881.1 -813.4 -214.9 (0.31) (4.87) (1.30) (1.10) 19-30 -277.3 70.6 566.8 764.8 (0.86) (0.41) (1.42) (4.43) 31 or more IR Constant -645.0 -589.2 -958.6 -470.9 (1.25) (3.96) (2.07) (2.67) 0.11 0.14 0.11 0.13 Number of households 1635 528 2185 FigureA.2: Probability of non-zero householdexpenditure on education (last month, basedon a Probit model) 1 1 0.65 0.61 10th 25th 50th 75th 90th Consumption Percentiles 166 Figure A.3: Household expenditure on education OLS predictions based on a sample of households with non-zero monthly educ. Expenditures 1600 1400 - 1200- 2 8B1000- 800 - 600 - 400 - 200 - 10th 25th 50th 75th 90th ConsumptionPercentiles 167 Health Appendix Table A1.6. Predictors of reporting chronic diseaseand acute symptoms for ages 40 and older. Probit model is estimatedwhere the dependentvariable for the first model takes the value 1if the individual reporteda chronic disease, 0 otherwiseandthe dependent variable for the secondmodeltakes the value 1if the individualhadan acute symptomin the previousmonth, 0otherwise. The repc :d figures are marginalz ects, andthe Itl-statisticsare in parentheses. Meansand Chronic Disease 4cute Symptoms IIStandardDeviations ~ .048 .009 ( 1 0 . 7 2 ) ( 2 . 5 1 ) Age squared I 100 -.030 - .004 ( 8 . 1 1 ) ( 1 . 3 8 ) Gender (1if male, 0 otherwise) -.061 -.072 ( 5 . 4 8 ) (7.45) [ . 4 9 9 ] Marital status of household -.013 -.009 .743 ( 0 . 9 9 ) ( 0 . 8 3 ) [ ,4371 head (1if married, 0 otherwise) Schooling attainment ____- ---_- .282 N o schooling [ . 4 5 0 ] -.049 -. 005 .238 Elementary ( 3 . 2 5 ) ( 0 . 3 8 ) [ .426] -.038 -.015 .365 ( 2 . 3 3 ) ( 1 . 0 7 ) [ . 4 8 1 ] Secondary -.082 -.032 .116 (3.79) ( 1 . 7 0 ) [ ,3201 Tertiary Occupation _ _ _ _ _ -___- .494 Not working -.1 [ . 4 9 9 ] 0 2 .037 .195 Agriculture/fishing/mining -( .120 6 . 9 4 ) ( 2 . 7 3 ) [ . 3 9 6 ] . 0 0 1 .112 -.1 ( 6 . 1 6 ) ( 0 . 0 3 ) [ .3161 Utilitieslmanufacturing 4 7 -.009 . l o 6 ( 7 . 5 2 ) -( .009 0 . 4 9 ) [ . 3 0 8 ] Tradelservices -.112 .056 -.094 ( 4 . 5 0 ) ( 0 . 4 0 ) [ . 2 3 0 ] .017 .036 Public-admideducatiodhealth ( 3 . 2 4 ) (0.64) L.1861 Otherlunknown Urbadrural residence .068 .038 ( 5 . 4 9 ) ( 3 . 5 1 ) L.5001 (1if urban, 0 otherwise) Log-likelihood -6335.9 -6093.7 1 0 3 6 2 10362 10362 Number o f observations 168 Table A1.7. Model 1 focuses on the probability of spending on health during last month (household level, marginal effects are reported). The dependent variable takes the value 1 for non-zero expenditures on health, 0 otherwise. Model 2 investigates the predictors of spending on health conditional on non-zero spending. It1statistics are reported inparantheses. Model1 (probitestimates) Model2 (OLS estimates) Meansand Meansand Coefficient I ndardDeviations Coefficient ! ndardDeviations Consumption per adult equivalent ,0204 10049 326.5 10586 (thousandDinars) (12.3) [6814] :19.67) [7027] Consumption (thousand Dinars) - 1.95*l o 4 -2.082 squared (7.30) (7.47) Number of householdmembers ,056 3.08 558.0 3.17 (11.4) [1.621 (11.4) [1.641 Shareof females inthe household ,076 ,535 -221.8 ,542 (3.22) [.259] (0.87) [.246] Shareof householdmembersaged 4 or less -.117 ,027 -3102.5 ,028 (1.56) [.OS61 (4.12) [.OS81 5-14 -.153 ,073 -1698.6 ,070 (3.35) [.148] (3.51) [.143] 15-59 ,543 ,532 [.374] [.370] 60 or more ,168 ,356 1034.1 ,369 (9.23) [.412] (5.51) [.409] LogLikelihood -4006.8 R-squared 0.15 Number of households 6386 6386 Figure A.4: Chronic and Acute Illness by Poverty 70 1 -Chronic Illness --- Poor --t Chronic Illness --- ---Non-Poor Acute Symptom --- Poor - - -x --Acute Symptom --- Non-Poor 0-4 5-14 15-59 >=60 Age Group 169 ANNEX I11 GLOSSARY OF TERMS Administrative costs Costs associated with the identification of target groups and the delivery of program benefits Adult illiteracy rate The proportion of the population over age fifteen who cannot, with understanding, read and write a simple statement about their everyday life anddo simple mathematical calculations. Access to safe water Measuredby the number of people who have a reasonable means of getting and adequate amount of clean water, expressed as a percentage of the total population. It reflects the health of a country's people and the country's ability to collect, clean, and distribute water. Inurban areas "reasonable" access means there i s a public fountain or water spigot located within 200 meters of the household. In rural areas, it implies that members of the household do not have to spend excessive time each day fetching water. Water i s safe or unsafe depending on the amount of bacteria init. An adequate amount o f water i s enough to satisfy metabolic, hygienic, and domestic requirements, usually about 20 liters per person per day. Access to sanitation Refers to the share of the population with at least adequate excreta disposal facilities that can effectively prevent human,animal, and insect contact with excreta. Suitable facilities range from simple but protectedpit latrines to flush toilets with sewerage. To be effective, all facilities must be correctly constructed and maintained. Births attendedby skilledhealthstaff Percentage of deliveries attended by personnel trained to give the necessary supervision, care, and advice to women duringpregnancy, labor, andthe postpartum period, to conduct deliveries on their own, and to care for the newborns. Chronic poverty A situation where an individual is poor as a result of long-term structural factors. 170 Consumption(income)smoothing The reduction of fluctuations in a person's (or household's) consumption (income) over time DependencyRatio Ratio of non-income earning (or dependent) to income-earning members in the household Employed Anyone who worked at all inthe last seven days preceding the survey and anyone with a permanent job who has not worked for the following reasons: own illness, maternity leave, a household member sick, holidays, education or training, temporary work load reduction and strike or suspension. Employmentinthe nonagriculturalsector Workers in the nonagricultural sector (industry and services), expressed as a percentage of total. Industryincludes miningand quarrying (including oil production), manufacturing, construction, electricity, gas, and water, corresponding to divisions 2-5 (ISIC revision 2) or tabulation categories C-F (ISIC revision 3). Services include wholesale and retail trade and restaurants and hotels; transport, storage, and communications; financing, insurance, real estate, andbusiness services; and community, social, and personal services- correspondingto divisions 6-9 (ISIC revision 2) or tabulation categories G-P (ISIC revision 3). Employmentrate Employed divided by Working Age Population Exclusionerrors Errors intargeting where intended beneficiaries are excluded from program benefits Familykhildassistance(allowance) Public cash transfer based on the number of children in a household Full-timeemployment Employment i s considered full-time if an individual works 35+ hours per week. Giniindex Inequality statistic for income or consumption, showing how unequally these are distributed inthe population. It ranges between 0 (perfect equality) and 1 (complete inequality). In-kindtransfers Transfers in the form o f goods or services, as opposed to cash Immunizationrate (i.e. againstmeasles) Percentage o f children under one year o f age who receivedmeasles vaccine. A child i s considered adequately immunized against measles after receiving one dose of vaccine. (World Health Organization and UnitedNations Children's Fund) Incentivecosts 171 Costs associated with adverse behaviour triggered by targeted programs, e.g. reduction of labor activity for the purpose o f reducing income to qualify for an income-based transfer program Inclusionerrors Errors intargeting arising out of inclusion of unintended(non-poor) beneficiaries inthe program Incomegeneratingprograms Programs designedto generate income that require some form of contribution (labor, or time, or repayment of loan). Inthis study it includes public works and credit-based livelihood programs Infant mortality rate The number of infants, out of every 1,000 babies born in a given year, who die before reaching age 1.The lower the rate, the fewer the infant deaths, and generally the greater the level of health care available in a country. Informaleconomy The exchange of goods and services not accurately recordedin government figures and accounting. The informal economy, which i s generally untaxed, commonly includes goods and services including day care, tutoring, or black market exchanges. Indicator of living standards A numerical measure of quality of life ina country. Indicators are usedto illustrate progress of a country inmeeting a range of economic, social, and environmental goals. Since indicators represent data that have been collected by a variety o f agencies using different collection methods, there may be inconsistencies amongthem. Human capital People and their ability to be economically productive. Education, training, and health care can help increase human capital. Labor force Sum of employed and unemployed Labor force participation rate (LFP) Labor Force divided by Working Age Population Life expectancy at birth The average number of years newborn babies can be expected to live based on current health conditions. This indicator reflects environmental conditions in a country, the health o f its people, the quality of care they receive when they are sick, andtheir living conditions. Malnutrition Not having enoughnourishingfood with the adequate amounts of protein, vitamins, minerals, calories, etc. to support growth and development. Maternal mortality ratio 172 The number of women who die during pregnancy and childbirth, per 100,000 live births. (Demographic andHealth Surveys and other WHO sources, the United Nations Children's Fund) Meanstest Tests designed to identify and separate the poor and usually done on the basis of income MillenniumDevelopmentGoals Internationally agreed goals for development, derived from the World Summits andconferences of the 1990s, adopted by 189nations inthe Millennium Declaration in September 2000. Provide benchmarks for measuring progress in promoting human development andpoverty reduction untilthe year 2015 and include eight goals: Eradicate extreme poverty and hunger; Achieve universal primary education; Promote gender equality and empower women; Reduce child mortality; Improve maternal health; Combat HIV/AJDS, TB, Malaria and other diseases; Ensure environmental sustainability; Develop a global partnership for development. Moralhazard A situation where information asymmetries causedby the unobservability of the actions by the beneficiariesresults inthem gainingmore program benefits than they would otherwise -i.e. under full information Netprimary enrollmentratio The ratio of the number of children of official school age (as defined by the national education system) who are enrolled in school to the population o f the corresponding official school age. Primaryeducation provides children with basic reading, writing, and mathematics skills along with an elementary understanding of such subjects as history, geography, natural science, social science, art, and music. Based on the International Standard Classification of Education, 1997 (ISCED97). Out of labor force (inactive) An individual is considered out of labor force is she is not employed andnot looking for ajob. Part-timeemployment Employment i s considered part-time if an individual works less than 35 hours per week. Percentageof cohortreachinggrade 5 The share of children enrolled inprimary school who eventually reach grade 5. The estimate i s based on the reconstructedcohort method. Poverty(material) The percentage of the population livingwith consumption below the poverty line; Poverty incidence at one dollar a day -percentage of the populationliving on less than $1.08 a day at 1993 international prices (equivalent to $1 in 1985 prices, adjusted for purchasing power parity). Poverty rates are comparable across countries, but as a result o f use o f special PPP exchange rates, they cannot be 173 compared with poverty rates usingcurrent exchange rates or national poverty lines for individual countries. (World Bank) Poverty gap The mean shortfall from the poverty line (counting the nonpoor as having zero shortfall), expressed as a percentage of the poverty line. This measure reflects the depth of poverty as well as its incidence. Prevalence of child malnutrition The percentage of children under five whose weight for age i s less than minustwo standard deviations from the median for the international reference population ages 0 to 59 months. The reference population adopted by the WHO in 1983, i s based on children from the United States, who are assumed to be well nourished. (World Health Organization) Primary health care Health services, including family planning, clean water supply, sanitation, immunization, and nutrition education, that are designed to be affordable for both the poor people who receive the services and the governments that provide them; the emphasis i s on preventing disease as well as curing it. Private transfers Informaltransfers made by individuals (households) without government intervention to other individuals (households). Such transfers could be in cash or kind Public transfers Transfers made by the government or its agents to individuals or households Public works Government-funded projects to develop or maintain physical infrastructure, largely labor-intensive innature Purchasing power parity (PPP) A method of measuringthe relative purchasingpower of different countries' currencies over the same types of goods and services. Because goods and services may cost more inone country than in another, PPP allows us to make more accurate comparisons of standards of living across countries. PPP estimates use price comparisons o f comparable items but since not all items can be matched exactly across countries and time, the estimates are not always "robust." Regressiveprogram A transfer program that provides increasing benefits with income Screening (or targeting) The process of sorting out those inneed of a program from those that are not in need Self-selection Targeting through design features that ensure that only the target population makes use o f the program, e.g. through setting a wage rate inpublic works that i s lower than market rate - creating a disincentive for those who have employment options in the market 174 Social assistance A rangeof benefits incash or kindto provide protection for the most vulnerable persons in society. These programs are usually financed from government revenues Social insurance A range of programs (usually incash) designed to protect individuals inthe event of a decline inincome (due to unemployment, retirement, illness). Benefits are financed through contributions that are usually earnings-related or collected through payroll taxes Stigma Negative social labels attached to beneficiaries who are participating in targeted programs Transient poverty A situation where an individual is poor because of some temporary shock which could be reversed over time Unemployed An individualwho has not worked inthe last seven days precedingthe survey, but has looked for ajob inthe last four weeks, and i s available to work. Unemployment rate Unemployed divided by Labor Force Under 5 child mortality rate The probability that a newborn baby will die before reaching age five, if subject to current age-specific mortality rates. The probability i s expressed as a rate per 1,000. (United Nations Statistics Division's Population and Vital Statistics Report; country statistical offices; Demographic andHealth Surveys and the United Nations Children's Fund's (UNICEF) State of the World's Children 2000) Working age population InCEE countries, population of ages 15to 64years oldis considered as working age population. Official statistics defines the working age as 15-54 years for women and 15-59 years for men. Youth literacy rate The percentage of people ages 15-24 who can, with understanding, read and write a short, simple statement on their everyday life. 175