Report No. 29619-TU Turkey Joint Poverty Assessment Report (In Two Volumes) Volume I: Main Report August 8, 2005 Human Development Sector Unit State Institute of Statistics Europe and Central Asia Region Turkey Document of the World Bank and the State Institute of Statistics, Turkey HBS Household BudgetSurvey HCIS Household Consumptionand Income Survey HDR HumanDevelopmentReport HH Household ILO International Labor Organization IMF International Monetary Fund JPAR Joint Poverty Assessment Report Kcallday Kilocalories per day Kg Kilograms LFS Labor Force Surveys M F C MinimumFoodCost MOH MinistryofHealth NA National Accounts n.e.c. Not Elsewhere Classified NUTS Nomenclatureof Statistical Territorial Units OECD Organisation for Economic Cooperation and Development P A Poverty Assessment PEC Per Equivalent Consumption PEI Per Equivalent Income PGI Poverty Gap Index PPP Purchasing Power Parity PSBR Public Sector Borrowing Requirement PTT Postal Telephone & Telegraph REER Real Effective Exchange Rate SHCEK Social Services and Children ProtectionOrganization, Sosyal Hizmetler ve Cocuk Esirgeme Kurumu SME Small andMediumEnterprises SOEs State-owned Enterprises SRMP Social RiskMitigation Project SSK Social Insurance Organization, Sosyal Sigortalar Kurumu SYDTF Social Assistance and Solidarity Encouragement Fund, Sosyal Yardimlasma ve Dayanisma Tesvik Fonu SYDVs Social Assistance and Solidarity Foundations, Sosyal Yardimlasma ve Dayanisma Vakiflar TIMSS Trends inInternational Mathematics and Science Study TL TurkishLira UNDP United Nations Development Programme USD US.Dollars USDA United StatesDepartment o fAgriculture UNESCO UnitedNations Educational, Scientific and Cultural Organization WB World Bank W D I World Development Indicators WHO World Health Organization TURKEY JOINT POVERTYASSESSMENT REPORT Table ofContents PREFACE ACKNOWLEDGEMENTS EXECUTIVE SUMMARY ............................................................................................................. i A. Data Comparability............................................................................................................................. i B. Recent Economic Developments........................................................................................................ ii C. Poverty Profile................................................................................................................................... ... 111 D. Education........................................................................................................................................... F. Labor Market inTurkey................................................................................................................... viii E. Health................................................................................................................................................. iv vi ... G. SocialProtectioninTurkey............................................................................................................... ix CHAPTER I:DATA COMPARABILITY, 1994-2002................................................................. 1 A. RealIncome and ConsumptionofHouseholdsBetween 1994and 2002............................. 1 B. Inequality inConsumption ofHouseholdsBetween1994and2002: HBS........................................ 5 C. Poverty Lines: 1994 and 2002 Methodology ..................................................................................... D. Poverty in 1994and 2002: HBS......................................................................................................... 5 7 E. Usingthe 1994HBSand 2002 HBS to Decompose ChangesinPoverty ....................................... 10 F. Conclusions....................................................................................................................................... 13 CHAPTER 11: MACROECONOMIC CONTEXT...................................................................... 14 A. Macroeconomic Instability Has HelpedKeep GrowthBelow Potential.......................................... 15 B. Fiscal Imbalance Has Beenthe Root ofChronic Macroeconomic Instability inTurkey ................. 15 D. Exchange Rate-BasedDisinflationPrograms Are Vulnerable inaGlobalizedWorld.................... C. Turkey's Exchange Rate-BasedDisinflation Program..................................................................... 16 17 F. Turkey's Crisis ResponseProgramIncorporatedthe Experience ofOther EmergingMarkets ....... 18 E. The Government RespondedQuickly to the 2001 Crisis ................................................................. 18 G. Strong Recovery Has BeenUnderWay Since Early 2002............................................................... 20 H. Macroeconomic Outcomes in2003 Were Favorable....................................................................... 21 ITurkeyHasAchievedSizableFiscalAdjustment............................................................................. J. DeteriorationinExternalBalance Has CreatedVolatility inDomestic FinancialIndicators........... 21 . 22 K. OngoingStructural Reforms Are Expected To Stimulate Growth................................................... 23 CHAPTER 111: POVERTY PROFILE......................................................................................... 24 A. HouseholdSize and Composition .................................................................................................... 24 B. HouseholdHeadCharacteristics....................................................................................................... 29 C. Spatial Characteristics ...................................................................................................................... 31 D. Non-IncomeAspects ofPoverty...................................................................................................... 34 E. InequalityandRegionalDifferences................................................................................................ 37 F. Multivariate Analysis........................................................................................................................ 40 CHAPTER IV: EDUCATION..................................................................................................... 43 A. The Context...................................................................................................................................... 43 B. Public Spendingon Education.......................................................................................................... 45 C. Traces of Inefficient Resource Allocation inthe Education Sector ................................................. 47 D. DistributionalandEducation-Quality ImplicationsofPublic Policies on Education...................... 50 E. Economic Growth andLabor MarketImplications o fa Public Policythat Undervalues Education62 CHAPTER V: HEALTH............................................................................................................. 70 A. The HealthCare System................................................................................................................... B. HealthInsurance............................................................................................................................... 72 70 D. Self-Reported Morbidity.................................................................................................................. C. Health Outcomes .............................................................................................................................. 74 75 E. Access To andUse ofHealthCare................................................................................................... 76 F. Health Care Expenditure................................................................................................................... 84 CHAPTER VI: LABOR .............................................................................................................. 90 A. Official Unemployment andLabor ForceParticipationRatesfrom the LFS................................... B. Unemployment andInactivity .......................................................................................................... 90 C. Quality o fEmployment.................................................................................................................... 92 94 CHAPTERVII: SOCIAL PROTECTION................................................................................ D. Sector ofEmployment...................................................................................................................... 98 100 A. TurkishPensionSystem................................................................................................................. 100 B. Social Solidarity Fund.................................................................................................................... C. Conditional Cash Transfers ............................................................................................................ 107 112 113 REFERENCES........................................................................................................................... D. LocalInitiatives.............................................................................................................................. 200 Box VII.l. Parameters ofthe Current Pension System ........................................................................... 101 List of Figures FigureI1. Consumption Per Capita with Various Deflators. 1990 .100................................................... . 4 Figure1.2. PriceIndexes inTurkey. 1990.2002. Annual PercentageChanges........................................... 4 Figure 1.3. Turkey: Poverty and GDP Growth ........................................................................................... 8 Figure 1.4. Poverty Changes...................................................................................................................... 10 Figure11.1. Per Capita GDP Comparedto EUAverage............................................................................ 14 Figure11.2. GDP and GDP Per Capita GrowthRates inEmergingEconomies (1965-2001) ................... 14 Figure 11.3. Turkey: Volatility of REERand GDP Growth, 1990-2000 .................................................. 15 Figure 11.4. AdjustedPublic Sector Borrowing Requirement (% of GNP) ............................................... 16 Figure11.5. Total Net Public Debt (%of GNP)......................................................................................... 19 Figure11.6. GNP Growth........................................................................................................................... 20 Figure111.1. Turkey: Poverty andHouseholdSize................................................................................... 25 Figure 111.2. Turkey: Poverty and Number of Children............................................................................ 26 Figure 111.3. Turkey: Poverty and Number of Elderly.............................................................................. 27 Figure111.4. Turkey: Share ofPoor inPercentby HouseholdConsumption........................................... 28 Figure 111.5. Turkey: Poverty Rate and Age............................................................................................. 29 Figure 111.6. Turkey: Poverty and Gender of Head .................................................................................. 30 Figure 111.7. Turkey: Poverty and Educationof HouseholdHead............................................................ 30 Figure 111.8. Turkey: Poverty and Location.............................................................................................. 31 Figure111.9. Turkey: HouseholdSize and Location................................................................................. 32 Figure 111.10. Rural Employment by Sector.............................................................................................. 33 Figure111.11. UrbanEmployment by Sector............................................................................................. 33 Figure 111.12. Rural Employmentby Sector.............................................................................................. 34 Figure 111.13. Turkey: Poverty and Dwelling........................................................................................... 35 Figure 111.14. Turkey: Poverty and Use of Dung..................................................................................... 35 Figure 111.15. Turkey: Poverty and Discretionary Activities .................................................................... 37 Figure 111.16. Turkey: Poverty and Shopping........................................................................................... 37 Figure 111.17. Per Capita GDP Index......................................................................................................... 41 Figure111.18. Socioeconomic Development Levels (2003) ...................................................................... 41 Figure 111.19. Female Literacy Rates (2000) ............................................................................................. 42 FigureIV.1. Public Spendingon Educationas Percentageo fGross DomesticProduct........................... 45 FigureIV.2. Number of Students by School Type.................................................................................... 48 Figure IV.3. Number of Classrooms Over Time ....................................................................................... 48 FigureIV.4. Numberof StudentsPer Classroom by School Type............................................................ 49 Figure IV.5. Student-Teacher Ratios inthe 1997-1998 and 2002-2003 School Years ........................... 50 Figure IV.6. 1999 TIMSS Mathematics Achievement of Eighth-Grade Students.................................... 57 Figure IV.7. 1999 TIMSS ScienceAchievement of Eighth-Grade Students ............................................ 58 59 Figure IV.9. Problems with Schools, Urban, and Rural............................................................................ Figure IV.8. Problems with Schools, Public and Private .......................................................................... 60 Figure IV.10. Satisfaction with Quality of Education ............................................................................... 61 Figure IV.11. Satisfaction with Quality of Education............................................................................... 61 FigureIV.12. UnemploymentRatesOver Time ....................................................................................... 65 66 Figure IV.14. Literate Without Diploma Unemployment Rates Over Time............................................. Figure IV.13. Illiterate Unemployment RatesOver Time......................................................................... 66 67 FigureIV.16. Junior SecondaryGraduates' UnemploymentRatesOver Time ........................................ Figure IV.15. Primary School Graduates' Unemployment RatesOver Time........................................... 67 FigureIV.17. Vocational Junior SecondaryGraduates' UnemploymentRatesOver Time...................... 68 FigureIV.18. SecondarySchool Graduates' UnemploymentRatesOver Time ....................................... 68 FigureIV 19. Vocational SecondarySchool Graduates' Unemployment RatesOver Time..................... . 69 FigureIV.20. Higher EducationGraduates' Unemployment RatesOver Time........................................ 69 Figure V.l. Share of Populationwith Health Insurance, by Quintile........................................................ 73 Figure V.2. RegionalVariation inInfant and Child Mortality Rate and Prenatal Care and Home Births 74 Figure V.3. Incidence o f Illness Income Quintiles and Location.............................................................. 75 FigureV.4. Incidence ofMorbidityby HealthInsurance Status............................................................... 76 FigureV.5. Shareo fThose ReportingIllnessWho Sought Care by Quintile and Location..................... 77 FigureV.6. Propensity to SeekTreatment ................................................................................................ 77 FigureV.7. Propensity to SeekCare When Illby Insurance Status.......................................................... 78 79 Figure V.9. Evolution o f Public Sector Spending on Health..................................................................... Figure V.8. HospitalizationAmong Those RequiringHospitalization, by Quintile ................................. 87 FigureV.10. Compositionof Public Sector SpendingonHealth, 2003.................................................... 87 FigureV.11. Compositiono fTotal Health Sector Spending (2000) ......................................................... 88 FigureV.12. Destinationof Central Government Fundingfor HealthCare ............................................. 89 Figure VI.1. Turkey: Unemployment Rate............................................................................................... 91 Figure VI.2. Turkey: Labor ForceParticipationRate............................................................................... 91 Figure VI.3. Turkey: Adults Employed.................................................................................................... 92 Figure VI.4. Turkey: Adults Not Employed............................................................................................. 92 Figure VI.5. Turkey: Poverty and Inactivity ............................................................................................ 94 FigureVI.6. Turkey: Poverty and Employment Situation........................................................................ 94 FigureVI.7. Turkey: Poverty and Employment Status............................................................................ 95 FigureVI.8. Turkey: Povertyand Social Security ................................................................................... 96 Figure VI.9. Turkey: Poverty and Type o f Workplace............................................................................. 96 Figure VI.10. Turkey: Poverty and Size o f Firm...................................................................................... 97 Figure VII.l. Budgetary Transfers to Social Security Institutions .......................................................... 105 List of Tables Table 1.1. Macroeconomic and Household Survey-Based Data on Living Standards. 1994-2002 .............2 Table 1.2. Indexes for DeflatingNominal Values: 1994 and 2002. 1994 = 1.00......................................... 3 Table 1.3. Change inMainIndicators for Real Incomes and Consumption. 1994.2002 ............................. 3 Table 1.4. Inequality .................................................................................................................................... 5 Table 1.5. Food Basket for Equivalent Adult............................................................................................... 6 Table 1.6. Poverty Lines, 1994-2002 .......................................................................................................... 7 Table 1.7. Comparison o f Poverty inTurkey Between 1994and 2002 ....................................................... 7 Table 1.8. Comparable Poverty Incidence ................................................................................................... 8 Table 1.9. Absolute Poverty Rates o f Europe and Central Asia .................................................................. Table I.10. Decomposition o fPoverty Change Between 1994and 2002.................................................. 9 12 Table 11.1. K e y Economic Indicators......................................................................................................... 19 Table 111.1. Turkey: Poverty and Household Size .................................................................................... 24 Table III.2. Turkey: Poverty and Number o f Children inHousehold ...................................................... 25 Table 111.3. Turkey: Age Structure o f HBS and Census........................................................................... 26 Table 111.4. Turkey: Poverty and Number o f Elderly ............................................................................... 27 Table 111.5. Turkey: Poverty and Household Composition ...................................................................... 28 Table 111.6. Turkey: Poverty Rate of Age................................................................................................. 28 Table 111.7. Turkey: Poverty Rate o f Consumer Durables........................................................................ 36 Table 111.8. Comparable Inequality ........................................................................................................... 38 Table 111.9. New Methodology Inequality Measures................................................................................. 38 Table 111.10. Per Capita Gross Domestic Product, by Statistical Regions, 2001....................................... 39 Table 111.11. Turkey: Social and Human Development Indicators 1997 ................................................. 40 Table 111.12. Probit Estimates.................................................................................................................... 40 Table IV.2. Incidence o fPublic Spending on Educationin 1994and in2001.......................................... Table IV.1. Number o f Schools, Students, and Teachers inthe 2001-2002 School Year ........................ 44 46 Table IV.3, Incidence o fHousehold Expenditures on Education in 1994 and in2001............................. 47 Table IV.4. Characteristics o f Children Aged 6 to 14, by Primary School Attendance Status ................. 51 Table IV.5. Primary School Attendance by Parental Schooling ............................................................... Table IV.6. EstimatedDistance Conditional on School Availability inthe Residential Area ..................52 Table IV.7. Predictors o f Continued Education after Completion o f 8-Year Primary School.................. 53 54 Table IV.8. Conditional on School Availability inthe Residential Area .................................................. 56 Table IV.9. Problems with School ............................................................................................................ Table IV.10. Satisfaction with Quality o f Education ................................................................................ 60 62 Table N.11. Ordinary Least Squares Estimates o f Log-Hourly-Earnings ................................................ 64 Table V.1, Distribution o f Primary Care Health Staff, by Region, 2002 .................................................. 71 Table V.2. MOH Health Centers and Health Posts Lacking Key Staff, by Region, 2002 ........................ 71 Table V.3. Regional Distribution o fHospital Beds, 2002......................................................................... 72 Table V.4. Health Insurance Coverage and Poverty.................................................................................. 73 Table V.5. Health Status Comparisons...................................................................................................... 74 Table V.6. Vaccination Coverage o f Infants and Pregnant Women (TT-2T), by Region, 1998-2002 ..... 75 Table V.7. Self-Reported Morbidity by Income Quintile.......................................................................... 76 Table V.8. Health Care Utilization by Severity o f Illness......................................................................... 78 Table V.9. Reasons for Not Seeking Outpatient Care When I11................................................................ 79 Table V.10. Reasons for Not Being Hospitalized When Needed.............................................................. 80 Table V.11. Problems Encountered When Seeking Outpatient Care, by Quintile .................................... 80 Table V.12. Average Amount Paid DuringLast Outpatient Treatment by Those Who Paid.................... 81 Table V.13. Share and Average Amount of Payments for Last Hospital Admission................................ 81 Table V.14. Determinants o f Health Care-Seelung Behavior, Probit Estimates ....................................... 81 Table V.15. Location o f Outpatient Care by Quintile ............................................................................... 83 Table V.16. Average Amount Paid for Outpatient Treatment by Facility Type ....................................... Table V.17. Utilization o fPreventive Care ............................................................................................... 83 84 Table V.18. Composition o f Health Care Expenditure, by Expenditure Quintiles ................................... 84 Table V.19. Proportion o f People with Catastrophic Health Care Expenditure........................................ 85 Table V.20. Impoverishing Effects o f Health Care Expenditures ............................................................. 86 Table V.21. Distribution o f Public Sector Spending on Health Across Income Quintiles........................ 88 90 Table VI.2. Turkey: PovertyRateofReasonfor Not SeekingJob .......................................................... Table VI.1. Turkey: UnemploymentRateandLaborForceParticipationRate....................................... 93 Table VI.3. Turkey: PovertyRateof StatusofWorkplace....................................................................... 97 TableVI.4. Turkey: PovertyRateofLegal Status ofWorkplace ............................................................ 98 Table VI.5. Turkey: PovertyRateofBasicCodeofMainActivity ofWorkplace .................................. 99 TableVII.1. ContributionRatesto SSK.................................................................................................. 102 Table VII.3. SYDTFEmployment-GeneratingProjectAssistance, July 1, 1997-March26, 1999.........108 Table VII.2. Types ofSYDTFAssistance, July 1, 1997-March 26. 1999.............................................. 109 TableVII.4. SYDTFRevenues............................................................................................................... Table VIIS. SYDTFExpenditures.......................................................................................................... 110 TableVII.6. QuarterlyFiguresfor CCT.................................................................................................. 110 Table VII.7. Turkey: IncidenceofNomina1Transfers........................................................................... 113 114 List of Annexes ANNEX I:METHODOLOGY................................................................................................................ 115 ANNEX 11: POVERTYINTURKEY: A LITERATUREREVIEW..................................................... 185 List of Annex Figures FigureA.I.l. CumulativeDistributionofPEC........................................................................................ 123 List of Annex Tables Table A.I.1. HouseholdSize bythe Type of Settlement......................................................................... 119 120 Table A.I.3. The Structure ofConsumption............................................................................................ Table A.I.2. The Structure ofCashIncome............................................................................................. 121 Table A.I.4. AverageMonthlyConsumptionbyDeciles........................................................................ 122 Table A.1.5. InequalityMeasures............................................................................................................ 124 Table A.I.6. The Structure ofaMinimumFoodBasket EstimatedBasedon SurveyData .................... 125 Table A.I.7. Average Value ofMonthly,CompletePovertyLine,byHouseholdSize .......................... 126 Table A.I.8. PovertyHeadcount, Gap, andSeverityIndex..................................................................... 127 Table A.I.9. SensitivityofPoverty Statisticsto CPIandHousingImputedRent................................... 128 Table A.I.lO. FoodConsumptionandExpenditures, 1994and2002...................................................... 129 Table A.I.ll. ComparablePovertyMeasures, 1994and2002................................................................ 130 Table A.I.12. FoodBasketfor EquivalentAdult..................................................................................... 130 Table A.II.1. DifferentPovertyMeasuresandResults............................................................................ 186 Table A.II.2. PovertybyRegion............................................................................................................. 186 Table A.II.3. UrbanandRuralPovertyRates.......................................................................................... 187 Table A.II.4. CalorieRequirementsbyAge andGender ........................................................................ 187 Table A.II.5. Structure ofPovertybyAge andGender (FoodPoverty).................................................. 188 Table A.II.6. Structure ofPovertybyAge andGender(Basic Needs)................................................... 189 Table A.II.7. Structure ofPovertybyEmploymentStatus(FoodPoverty)............................................ 189 Table A.II.8. Structure ofPovertybyEmploymentStatus (Basic Needs).............................................. 190 Table A.II.9. Structure ofPovertybyEconomic Activity (FoodPoverty).............................................. 190 Table A.II.l 1. RegionalPoverty.............................................................................................................. Table A.II.10. Structure ofPovertybyEconomic Activity (Basic Needs)............................................. 191 191 Table A.II.12. Poverty andGenderofHouseholdHead......................................................................... 192 Table A.11. 13. PovertyandAgriculturalActivity ofthe Household...................................................... 192 Table A.II.14. PovertyandAgriculturalLandSize................................................................................. 193 Table A.II.15. PovertyandAgriculturalLandSize Across Regions....................................................... 194 Table A.II.16. Poverty and Number o f Tractors Owned......................................................................... 194 Table A.II.17. Structure o fPoverty According to Number of Tractors Owned...................................... 195 Table A.II.18. Poverty LinePer Capita (Annual) .................................................................................... 195 Table A.II.19. Comparisono f Incomeand PovertyLines by 5 Percent o f Population........................... 195 Table A.II.20. Poverty inTurkey, 1987-1994 (Minimum-Food-Cost Approach ................................... 196 Table A.II.21. Poverty inTurkey, 1987-1994, (Cost-of-Basic-Needs Approach) .................................. 197 List of Appendices Appendix 1: Income and Consumption by Poverty Status and UrbadRural Dimention ........................ 132 Appendix 2: Poverty Figuresfor Various Poverty Lines ........................................................................ 142 Appendix 3: Sensitivity of Poverty ......................................................................................................... 157 Appendix 4: Regression Analysis ........................................................................................................... 161 Appendix 5: FoodBasket........................................................................................................................ 165 Appendix 6: EstimationofAverage HouseholdAdult Equivalent Size.................................................. 171 Appendix 7: General Descriptive Statistics............................................................................................. 175 This volume is aproduct o fthe staffofthe StateInstitute of Statistics. Turkey and the InternationalBank for Reconstruction and Development/The World Bank. The findings. interpretations. and conclusions expressedinthis paper do not necessarily reflect the views of the ExecutiveDirectors of the World Bank, or the govemments they represent. The State Institute of Statistics and the World Bank do not guaranteethe accuracy of the data includedin this work. The boundaries. colors. denominations. and other information shown on any map inthis paper do not imply anyjudgment on the part of the State Institute o f Statistics and the World Bank concerning the legal status of any territoty or the endorsement or acceptance o f suchboundaries. PREFACE This report represents the first time that the State Institute of Statistics (DIE) of Turkey and the World Bank have issued ajoint report on poverty inTurkey. The findings o f the Volume One Report have been subject to thorough technical review by both institutions, and are considered to bejoint findings. While this report presents a comprehensive analysis of poverty inTurkey, considerable researchinto the subject has already beendone by the World Bank (World Bank 2000,2003), the TurkishGovernment (2003), the UnitedNations Development Programme (UNDP 2001), and academicians and scholars (see Annex I1for a review of the literature). Under the institutional development component o f the Social RiskMitigation Project (SRMP), the World Bank has delivered technical assistanceto DIE on poverty measurement and analysis, leadingto thisreport. The basic data used inthis report are from the official 2002 HouseholdBudget Survey (HBS), conducted by DIE. Comparisons over time are made to the 1987 and 1994 official DIE HBS surveys. Additional qualitative information was gathered from a variety of primary and secondary sources. Limited quantitative data from an unofficial survey of 2001 (which was not conducted by DIE) are used as secondary sources insome o f the chapters. Administrative data from other TurkishGovernment agencies are also used. The 2002 HBS sample was designed to be representative o f the population o f Turkey, and to provide reliable information needed for an urban-rural breakdown o f data. It was not designed to be regionally representative. Thus, only qualitative data and secondary-source data are used herein to discuss the regional dimensions of poverty in Turkey. The World Bank and DIE anticipate a further joint report on regional aspects of poverty in Turkey, based on data from the 2003 HBS, which has a larger and regionally representative sample. This report sets out a new poverty line' methodology for Turkey as the basic measure of poverty in the country. However, severalpoverty lines are calculated for the purpose o f international comparability, and comparability to the World Bank's poverty measures using the 1987 and 1994 data. The basic findings for Turkey for 2002 were publishedin a press release by DIE (April 13, 2004), and this report provides the underlyinganalysis and methodology for these figures. In 2002, 27 percent of the Turkish population was poor, based on the new poverty line methodology detailed in Annex One (food and non-food consumption). However, very few (nearly zero) consumed under the food line or under the $1 per person per day line used in international comparisons (Table A.I.8). The analysis inthis report refers generally to the new poverty line methodology that results in27 percent poor. This line i s called "complete" poverty line, and i s referred to as "Total poverty" in statistical tables. Additional poverty lines and rates can be found inAnnex One. This volume begins with a comparison of poverty trends from 1994 to 2002 (Chapter One), reviews macroeconomic developments and poverty in Chapter Two, draws a portrait or profile of the poor in Chapter Three, and considers education (Chapter Four), health (Chapter Five), labor (Chapter Six) and Social Protection (Chapter Seven). ' Thepovertyline i s the minimumamount o f consumptionneeded for anindividualor householdto cover its basic needsfor foodandnon-foodgoods. ACKNOWLEDGEMENTS This task was managed by Jeanine Braithwaite (ECSHD) for the World Bank and by Ozlem Sarica for State Institute o f Statistics (DIE). The first volume is a joint report o f the World Bank and DIE. The second volume i s a World Bankreport. The team i s grateful for the support of Andrew Vorkink, Country Director, World Bank, and Omer Demir, President, DIE. The team benefited from comments o f peer reviewers Jeni Klugman (AFTP2), Peter Lanjouw (DECRG), andEdmundo Murmgara (ECSHD). Volume One: Managed by Jeanine Braithwaite and Ozlem Sarica. The World Bank team included: Ruslan Yemtsov, Ismail Arslan, Kamer Ozdemir (ECSPE), Cem Mete, Monika Huppi, Anita Schwarz (ECSHD) and Ahu Gemici (Consultant). The DIE team included Sema Alici, Zuhal Daskiran, Guzin Erdogan, Muzzeyyen Pamuk, Gullu Calik, Murat Karakas and Enver Tasti, and Sasun Tsirunyan (Consultant). Special thanks to Carmen Laurente, Jennifer Manghinang, Gizella Diaz and Selma Karaman (ECSHD) for document finalization and logistics and to Shruti Kapoor (Consultant) for research assistance. Volume Two: Managed by Jeanine Braithwaite (ECSHD) with contributions o f the World Bank team members noted above and the ECSPE team working inTurkey. Bothvolumes were edited by Diane Stamm (Consultant). EXECUTIVE SUMMARY The basic data used in Volume One are from the official 2002 Household Budget Survey (HBS), conducted by DIE. Comparisons over time are made to the 1987 and 1994 official DIE HBS surveys. Additional qualitative information was gathered from a variety o f primary and secondary sources. Limited quantitative data from an unofficial survey o f 2001 are used as secondary sources insome of the chapters. Administrative data from other TurkishGovernment agenciesare also used. The 2002 HBS sample was designed to be representative o f the population of Turkey, and to provide reliable information needed for an urban-rural breakdown o f data. It was not designed to be regionally representative. Thus, only qualitative data and secondary source data are used herein to discuss the regional dimensions of poverty inTurkey. The World Bank and DIE anticipate a further joint report on regional aspects o f poverty in Turkey, based on data from the 2003 HBS, which has a larger and regionally representative sample. This report sets out a new poverty line2 methodology for Turkey as the basic measure o f poverty inthe country. However, severalpoverty lines are calculated for the purpose o f international comparability, and comparability to the World Bank's poverty measures using the 1987 and 1994 data. The basic findings for Turkey for 2002 were publishedin a press release by DIE (April 13, 2004), and this report provides the underlyinganalysis andmethodology for these figures. In2002, 27 percent of the Turlushpopulation was poor, based on the new poverty line methodology detailed in Annex One (food and non-food consumption). However, very few (nearly zero) consumed under the food line or under the $1 per personper day line used ininternationalcomparisons (Table A.I.8 found inVolume One). The analysis inthis report refers generally to the new poverty line methodology that results in 27 percent poor. This line i s called "complete" poverty line, and i s referred to as "Total poverty" in statistical tables. Additional poverty lines and rates can be found in Annex One of Volume One. A. DATA COMPARABILITY An in-depth analysis o f the 2002 Household Budget Survey (HBS) compared to that from 1994 shows that living standards inTurkey remained almost unchanged. Poverty based on the previous methodology declined gradually from 1987 to 2002, from 38.5 percent to 34.5 percent. Poverty based on the updated methodology3declinedfrom 28.3 percent to 27 percent from 1994 to 2002. On the other hand, inequality marginally increased. Extreme poverty, already low, further declined from 1994 to 2002. Food poverty declined from 2.9 to 1.4 percent, while $1 per person per day poverty, depending on purchasing power parity (PPP) used, was 2-3 percent or even negligible (0.2 percent). A poverty-growth decomposition demonstrated that while economic growth was a main driving force in poverty reduction, much of the gains from growth were offset by inequality, which slightly worsened from 1994 to 2002. These conclusions should be treated with caution in that both 1994 and 2002 were either crisis years, or immediately following crises, and so there was no measurement o f the effect o f The poverty line is the minimumamount o f consumption needed for an individual or household to cover its basic needs for food andnon-food goods. This is the updated poverty line for 1994, but the consumption aggregate for 1994 is the old version. For 2002, both the consumption aggregate and the poverty line were the new methodology. Additional details are found in Chapter One (Data Comparisons) and Annex One (Methodology) o fVolume One. sustained growth on poverty. DIE is now conducting surveys to measure poverty annually, so the measurement problem should abate inthe future. B. RECENT ECONOMICDEVELOPMENTS Despite many advantages ranging fkom its strategic location to its dynamic population, Turkey has not achieved the stable high growth o f leading emerging market economies. On average, the Turlush economy grew slightly under 3 percent per year over the past decade, well below the best-performing emergingeconomies. Turkey has suffered from an exceptional degree of macroeconomic instability characterized by high inflation and sharp swings inbusiness cycles. Inflation was higher and growth lower inthe 1990sthan in the 1980s. Inthis period, unsustainable fiscal policy has repeatedly put pressure on the TurkishLira (TL), fueled inflation, and undermined financial stability. Open capital accounts and a poorly regulated banking system have magnified the impact o f unsustainable fiscal policy on macroeconomic stability. Short-term capital flows have fluctuated widely as investors responded to the boom-bust cycle driven by unstable conditions. In2000, a crawling peg exchange rate regime was launched to rid the economy of inflation. Key structural reforms in social security, infkastructure, agriculture, privatization, and banking were introduced. However, these achievements were insufficient to avoid a crisis, given the extent o f Turkey's underlying fiscal and financial sector weaknesses built up over decades o f instability and delayedreform. By 2001, persistent doubts about the peg, and underlying fiscal instability, ledto a full-blown speculative attack against the currency. Interest rates shot up to several thousand percent, forcing the Government to abandon the crawling peg and float the TL. The TL immediately lost 40 percent o f its value in a single day. In response to the crisis, following the collapse of the crawling peg and subsequent devaluation, the Government announced a strengthened economic program in M a y 2001. The key structural and social elements o f the program were: (a) a macroeconomic fkamework designed to restore financial stability and ensure public debt sustainability; (b) a rapid restructuring o f the banking sector; (c) a public sector reform program; (d) renewedprivatization; and (e) enhanced social assistance. The Turkish economy started to grow at a fast pace in 2002. Economic growth reached 5.9 percent in 2003, following 7.9 percent growth in 2002. While the inventory build-up led the recovery in 2002, private consumption and investment was behind the strong growth performance in 2003. The current account deficit widened to almost 3 percent of GNP in2003, but was easily financed by short-term capital inflows, public sector borrowing abroad and reverse currency substitution. Inflation fell to 18.4 percent in 2003, and the latest data suggest that in mid-2004, inflation declined to the important single digit level for the first time since the 1970s. Aggregate unemployment remained stable at around 10 percent but this was helped by a temporary shrinkage in the labor force. Fiscal gains were significant in 2003, and the primary surplus rose from 4 percent o f gross national product (GNP) in 2002 to over 6 percent o f GNP in 2003, close to the programmed 6.5 percent target. Monetary policy followed a policy o f implicit inflation targeting, with the Central Bank o f Turkey (CBT) occasionally intervening in the foreign exchange market to dampen what was deemed to be excessive fluctuation in the exchange rate. The decline in inflation, which was aided by the strength o f the TL, led to a commensurate decline in interest rates from a nominal 60 percent inthe first quarter o f 2003, to about 25 percent early in2004. .. 11 However, the deterioration inglobal financial indicators in early M a y 2004 combined with the higher than expected current account deficit figures have led to a sharp weakening in domestic financial indicators. The excess volatility in the foreign exchange market was curbed to some extent by the Central Bank's intervention. It appears that some relative stability has been achieved in domestic financial markets. These recent deteriorations have underlined Turkey's exposure to shocks from the extemal environment. While the depreciation o f the TL i s an adjusting factor to the deteriorating current account balance, it i s also likely to affect inflationary expectations and domestic interest rates. Higher domestic interest rates, intum, together with the impact of the TL's depreciation, would influence the overall fiscal deficit and economic activity. The spillover from global liquidity tightening and rising spreads are likely to increase the cost o f extemal borrowing. Such developments, ifpersistent, could disrupt the virtuous cycle that the economy has experienced over the last year and a half. C. POVERTYPROFILE Household Size and Composition. Poverty in Turkey i s strongly associated with age and household composition; children and families with children are poorer than average. Poverty increases monotonically with additional household members, starting at three members. Larger households are poorer than smaller households, and this is primarily due to the fact that the additional household members are more likely to be children, who have a higher poverty rate. Households with no children or only one child had poverty rates below the average. There i s also a correlation between having elderly members and household poverty, though this correlation i s not as marked as with having additional children. Having one more elderly member did not appreciably increase the risk o f poverty, and having two or more slightly elevated the risk. With respect to correlation with age, younger children are poorer, active-aged adults are not as poor on average, and the elderly are poorer than adults, but not as poor as children. Household Head Characteristics. The household head has a substantial impact on the poverty status of his or her household, through the employmenb`inactivity nexus and the amount o f income she or he can contribute to the household. The poverty rate for households with unemployed heads i s 35.4 percent, compared to 27 percent for the total sample. Besides this, other demographic features that are associated with poverty include whether the household head is a female or not (32 percent poor versus 26.6 percent poor for female and male heads respectively), and the education o f the household head. Illiterate heads and those who had not completed primary school heads had poverty rates nearly twice the average, while those few with masters or other advanced degrees had a poverty rate o f zero. Spatial Characteristics. There i s a sharp difference in the poverty rates between rural and urban households; the poverty rate is nearly 35 percent for the rural population, but only 22 percent for the urbanpopulation. The factors for highrural poverty are the same as for poverty overall. Rural areas are characterized by limited employment opportunities, and rural households where the head is unemployed face a substantial risk o f poverty. Other kinds o f inactivity have different implications on rural and urban areas. The major driver for rural and urban employment findings appears to be sector of employment, where rural location i s dominated by agriculture, which offers less lucrative options compared to formal employment found in urban areas. Education has identical effects on both rural and urban areas, whereby those who are illiterate or whose education i s limited to primary school have higher poverty rates than average, and graduates o f higher education are much less likely to be poor. In both areas, poverty rates steadily decrease as years of education increase. ... 111 Non-Income Aspects of Poverty. Poverty restricts the poor fiom accessing many goods and services. Non-income here refers to material items, assets, or services that are ultimately obtained through income. InTurkey, there isamarkeddifference inthekindofdwelling. Abouthalfthe samplepopulationlivesin a house, and another 27 percent lives inapartments. While individual houses are primarily inrural areas, apartments are almost exclusively an urban phenomenon. Only 6.5 percent o f the apartment dwellers are poor, while 36 percent of those who live inhousesare poor. In terms of land ownership, 27 percent of the sample population reported they had a field, but these households were poorer than average. The mean size o f the fields for poor households was 75 percent of the non-poor fields. More than one-fifth of the population possessed a car, but only 6 percent o f these were poor. The poor are less able to afford discretionary expenditures. The poverty rate of those who smoke, drink, take computer courses, and have access to public transportation to school, for example, i s lower for all than the overall poverty rate of 27 percent poor. Shopping patterns also vary betweenpoor and non-poor; the latter are more likely to shop at markets or bazaars. Inequality and Regional Dzferences. Turkey i s a middle-income country, and its inequality is high. Both consumption and income indexes indicate that inequality i s higher in urban areas than inrural areas, but not much. Other data confirm overwhelmingly that there is a sharp East-West divide inTurkey where the Southeastern and Eastem Anatolia regions are much poorer and have sharply lower human development indicatorsthan Westem Turkey. D. EDUCATION In1923,theyear inwhichtheRepublicofTurkeywas founded, theadultliteracyratewas approximately 10 percent. Such a low starting point (not uncommon for that era) gave the Republic a major cause to introduce a series of ambitious reforms that included a move toward secular education and the adaptation of the Latin alphabet. While the schooling environment in Turkey gradually improved over time, the 1997-1998 school year marked another major leap forward inthat compulsory schooling increased from 5 years to 8 years for children aged 6 to 14. Enrollment rates increased soon after this reform, not only for the %year basic education cycle but also for secondary education. Public Spending on Education Turkey's public spending on education increased significantly after 1998, both in real terms and as a percentage of gross domestic product (GDP). Consequently, as o f 2000, Turkey's public spending on education as a percentage of GDP has become comparable to the spending patterns observed in countries of similar levels o f economic development. The expansion of compulsory schooling to 8 years had an extremely positive impact on the distribution o f public spending on education across rich and poor households. In2001, 21.7 percent of public spending on basic education reached the poorest 20 percent of households (as opposedto 15.8 percent in 1994). But there i s significant room for improvement when it comes to secondary education, since only 13 percent o f public spending reached the poorest 20 percent of the population in 2001. Household expenditures on education strongly reinforce the disadvantageous situation o f poor children: in 2002, the wealthiest 20 percent of households spent 6.4 times more on education than did the poorest 20 percent of households. The current situation represents a significant improvement compared to 1994, when the wealthiest households spent 28.8 times more on education than did the poorest households. iV Inefficient ResourceAllocation in the Education Sector After the expansion o f compulsory schooling, the number o f primary school students and students enrolled in general secondary schools has significantly increased. The number o f students enrolled in vocational secondary schools has remained roughly the same. Classroom construction at the basic education level has been impressive: students per classroom at this level declined between the 1997-1998 school year and 2002-2003 school year despite increasing enrollments. At the secondary level, however: (i) the number of vocational school classrooms increased significantly in2002-2003 (reflecting previous investments) even though enrollment invocational schools didnot increase inrecent years; and (ii) number o f general secondary classrooms remained stagnant, the even though general secondary school enrollment increased visibly inrecent years. As a result, as o f the 2002-2003 school year, average classroom size was 18 in vocational secondary schools and 45 in general secondary schools. Enrollment in Early Levels of Schooling According to the 2002 HBS data, 97 percent o f children aged 6 to 14 are either enrolled in school or have completed the basic education cycle. Parental schooling is a very good predictor o f children's enrollment level: about 70 percent o f the children who are not attending school (but who are of school age) have at least one parent who has not completed primary school. An econometric model ofenrollment insecondary schools reveals that males are 7 to 8 percentage points more likely to continue their education beyond compulsory schooling. Secondary school availability in the residential area has a strong predictive power, boosting the probability o f enrollment by 10percentage points-only 64 percent o f households reported that a secondary school is available in their residential area. Other important correlates o f secondary school enrollment are household wealth and presence of a mother inthe household. Universitiesand ThePoor In Turkey, entrance to universities is primarily based on a student's performance in a centrally administered examination. Grades in secondary school (and a specialized field in secondary school) are other factors that influence the overall score. An analysis o f the 1997 University Student Survey found that students from high-income families are much more likely to be enrolled in private universities and well-established institutions. Thus, (the few) students from poor households who are enrolled in universities do not enroll inuniversitieso f the same quality as those o f wealthier students. Some o f the other findings were that private tutoring plays a key role in determining who attends what type o f university. As the main reason for not receiving private tutoring, 57 percent o f surveyed undergraduate students mentioned lack o f economic resources. Parental ViewsAbout Quality of Education Analysis o f 2001 household survey data reveals that household members are more likely to report problems with public schools (36.7 percent) compared to private schools (25.8 percent). The lack of books and supplies emerges as the leading problem-reported as a problem for 15 percent of children enrolled inpublic schools and 10 percent of children enrolled inprivate schools. The next major problem i s poor teaching (in both public and private schools), reported in roughly 10 percent o f cases. The V urbadrural differences are pronounced-only 45 percent o f rural responses indicated "no problem with school" as opposed to 68 percent inurban areas. The wealthier are much less likely to report problems with school. Complaints about lack o f books and supplies decline rapidly with increases inhousehold wealth. Complaints about the condition o f facilities also decline with wealth. While complaints about poor teaching remain constant across all wealth groups, complaints about lack o f teachers decline with wealth. Schooling Attainment and SelectedLabor Market Outcomes Private returns to schooling are very high inTurkey. The gender gap inearnings is visible-on average-- males earn 45 percent more per hour than females with similar characteristics. Schooling has a robust, positive, and large impact on earnings for both genders. Ifone estimates separate earning regressions for males and females, the impact o f schooling on earnings i s found to be more pronounced for females. In other words, while females earn less on average, the variation inearnings by schooling attainment is more significant for females when compared with males. These findings, jointly with the findingthat only 15 percent o f those who report non-zero wages are females, suggest that by under-investing in girls' schooling and by operating with extremely low female labor force participation rates, Turkey foregoes a vast potential human capital resource that can fuel the economy. Finally, contrary to common perceptions, the unemployment rate among vocational secondary graduates tends to be at about the same level as the unemploymentrate among general secondary school graduates. E. HEALTH TheHealth CareSystem Turkey's health care delivery and financing system i s fiagmented. Both public and private providers supply health services. The Ministry of Health (MOH), the Social Insurance Organization (SSK), and university hospitals are the main providers. While the public health care sector primarily predominates, private sector provisions are gaining importance inwestern and urban parts o fthe country. Despite significant efforts, the service delivery network remains highly uneven, with major concentrations in urban areas, particularly in the western part of the country. This skewed distribution has led to significant regional differences inaccess to and use o f health care and, concomitantly, health outcomes. Health Insurance Several public health insurance schemes currently provide financial protection to various target groups. SSK's health insurance i s the most important one, catering to those employed in the formal sector. The green card system, introduced in 1992, i s intended to provide coverage to low income groups who are not cove4red otherwise. The Government soon plans to shift to universal health insurance, which would operate on the principles o f solidarity and risk pooling, and provide coverage to the entire population. Both DIE HBS and Household Consumption and Income Survey (HCIS) suggest that over one-third (36 to 37 percent) o f the population in Turkey does not have access to health insurance, and almost half the population in rural areas remains without any coverage. The green card program fails to provide broad coverage to all those living inpoverty. Thus an important share o f the lowest-income households remains without access to health insurance coverage even ifthey may be eligible for it, and makes significant out- HCIS is an unofficial survey which was not conductedby DIE. vi of-pocket payments when seeking health care. The lack o f insurance coverage leads many low income households to forego healthcare, which inturnnegatively affects their healthoutcomes. Health Outcomes Despite considerable progress achieved in the recent past, Turkey continues to rank far behind most middle-income and European Union (EU) accession countries on key health indicators. Health outcomes vary significantly across regions, reflecting the uneven supply o f and access to health care in various parts. The Turkish health system faces a dual challenge. Significant parts of the population continue to be afflicted with a highburden o f disease fi-om preventable infectious diseases, and highmaternal and infant mortality rates as i s typical o f developing countries. At the same time, a growing share of the population is affectedby non-communicable diseases prevalent indeveloped countries. Access To and Useof Health Care People with health insurance, including a green card, are more likely to seek health care when illthan those without insurance. The likelihood o f seeking care when illi s lower among the bottomtwo quintiles than among the upper quintiles, with the difference being particularly marked in rural areas. Low-income groups suffer fi-omsignificant access problems. Determinants of CareSeeking The prime reason for not seeking care when sick, or not seeking hospital admission when required, was lack o f affordability. One out o f five people fi-om the lowest income quintile who sought outpatient care reported that the main problem with the care was that it was too expensive. Lack o f a facility nearby did not appear to be a prime determinant o fnot seeking care when ill. The share o f the population that had to pay for outpatient treatment, drugs, and hospitalization is higher among the lowest income quintile than among the upper-income groups. Furthermore, a multivariate analysis o f the determinants o f health care-seeking behavior confirms that income, insurance coverage, household size, gender o f household head, and severity o f illness are the most important determinants o f an individual seeking health care. Health CareExpenditure Out-of-Pocket Payments. Households in Turkey allocate a modest share o f their total expenditure to health care in the form o f out-of-pocket payments. Income is a major deciding factor on the amount spent. The top quintile spends about twice as high a share o f total household expenditure on health care as the lowest quintile. The largest share o f out-of-pocket expenditure on health is allocated to the purchase o f drugs, with payments for outpatient consultation ranlung second across all income levels. Public Expenditure on Health Care Public expenditure on health care grew at an average annual rate o f 7.3 percent between 1999 and 2003. The main public financiers are the Central Government and the social security institutions. According to the recent National HealthAccounts study estimates inTurkey, Central Government fundingaccounts for over one-third o f Turkey's health care expenditure, employer contributions account for less than one-fifth, and households pay over two-fifths through out-of-pocket payments and social security contributions. vii Public sector spending on health care i s skewed in favor of the upper income groups, particularly spending on outpatient care. Budgetary funds are not well targeted toward assuring equitable access by the entire population. Thus, overall, the relatively important public subsidies to health care are benefiting the middle- and upper-income households, while the poor continue to face significant access barriers to health. F. LABORMARKETTURKEY IN InTurkey, as in most countries, poverty is closely related to employment status and the type ofjob, whether formal or informal. Informally employed and casual workers have a higher rate of poverty. Educationplays a keyrole inexplaining employment and poverty outcomes. Unemploymentand Labor Force Participation Unemployment in Turkey was 10.3 percent in 2002. Labor market outcomes are mostly driven by low levels of labor force participation. Those who do not work drop out o f the labor force, and are thus not captured in the unemployment rate figures. The poverty rate o f the non-participants in the labor force reflects strongly the situation o f children: inactive household members younger than 15 years of age had a poverty rate o f 35 percent, but older inactive household members had a poverty rate o f only 22 percent, under the total poverty rate. There are very sharp differences inlabor market participationrates between men and women, with extremely low rates of female labor market participation in Turkey and even decreases in the rate for females since the 1999 level of 30 percent, down to 27.9 percent in 2002. The male labor force participation rate was 72 percent in 2002, or more than twice that of women's. Female unemployment rates have typically been slightly lower than the male unemployment rate. This i s primarily becauseso few women are inthe labor force Unemploymentand Inactivity In2002, 35 percentofthoseaged 12andabovereportedthat theyhadworked inapaidjob inthemonth of the survey. The poverty rate o f these 35 percent was 25 percent. Another 43 percent of those aged 12 and above who reported that they did not work had almost the same poverty rate (24 percent poor) as the employed. Intermsofunemployment, only 7.2 percent ofthoseaged 12andabove reportedthat they were looking for a job. Households with employed heads had a poverty rate of 26 percent compared to 35 percent where the household headwas unemployed. Reasons for not seeking a job varied fiom factors relating to age or family structure (student, housewife, elderly) to disability or seasonalemployment. Quality of Employment InTurkey, there exists a strong association betweenthe type of employment and the poverty statusofthe individual or household. Poverty rates o f those who had permanent employment were lower compared to those with casual or temporary jobs. The relative risk of poverty for casual workers was 3.7 times greater than for the permanently employed. Poverty rates for self-employed and unpaid family workers were higher. Poverty was also found to be sharply associated with a lack o f registration at a social security institution. Of the 35 percent who reported being employed, 32 percent were enrolled insocial security. The poverty Vlll ... riskfor peoplewith formal jobs was the lowest for those employedbythe government and instate-owned enterprises. However, it should be emphasizedthat very few people have such employment. Poverty is associated with the size of the enterprise. People employed in larger firms were less likely to be poor compared to people fkom firms with 1 to 9 people. Almost 70 percent of the respondents (aged 12 and over) reportedthat they worked ina firm o f 1to 9 people, and thus had ahigherpoverty rate. The mean number o f hours worked by the poor was 43.4 per week, whereas by the non-poor, it was 46.3 hours. Also, the mean duration of employment of the poor inthe same low-payingjob was longer than of the non-poor. Sector of Employment The largest sector o f employment inTurkey is agriculture. Agriculture is also the sector with the highest poverty rate o f those employed in it. Of the 35 percent aged 12 and above, 40 percent are engaged in agriculture. The next-highest poverty rate i s that of construction. The poverty rate is lowest for mining and quarrying. After agriculture, the other significant sectors interms of employment are manufacturing, and wholesale and retail trade. G. SOCIALPROTECTIONINTURKEY Social protection inTurkey consists primarily o f limited formal systems inpensions and social assistance, supplemented greatly by informal mechanisms. Formal elements of social protection are the pension (social security) system, and the Social Assistance and Solidarity Encouragement Fund (SYDTF) and its 931 affiliated Social Solidarity Foundations (SYDVs). Turkish PensionSystem Turkey's social security system is highly fragmented. Benefits and contributions depend on one's occupation. Sosyal Sigortalar Kurumu(SSK) covers the bulk o f the labor force, especially private sector employees, and those public sector employees who do not qualify as civil servants. Civil servants are covered by Emekli Sandigi (ES), and the self-employed and farmers are coveredby Bag-Kur (BK). Also, there are separate occupational schemes that cover various other groups. Overall, 42 percent (11 million people) of the labor force i s contributing to one or the other scheme. On the beneficiary side, only 29 percent (1.2 million people) of the population over age 65 is receiving an old-age pension. However, almost 3 million individuals below the age of 65 are receiving pensions, primarily old age pensions, with many of the recipients considerably younger than 65. As a result, the pensionsystem i s showing large fiscal losses eachyear and i s inneed o f transfers from the government to cover those losses expected to be around 4 percent o f GDP in2004. SSK is by far the largest system, and covers mostly private sector employees. Employers contribute 11 percent of wages for pension, and employees contribute 9 percent o f wages. In 1999, the Social Security Law changed most of the SSK benefit parameters. Pre-reform and post-reform benefits are discussed in detail inthe social protection chapter. Bag-Kur primarily covers the self-employed and some farmers. Contribution rates are 20 percent for pensions and 20 percent for health coverage. To combat the perennial problem o f evaluating income eamed, Turkey uses the system o f minimumearnings steps, which are attributedto individuals regardless of what they actually eam. ix Bag-Kur has a very low collection rate for its contribution revenue. Workers pay very little during their working years, and just prior to retirement, they pay Bag-Kur a lump-sum equivalent to the past-due contributions with interest, and thenreceive their retirement. Emekli Sandigi covers civil servants, including military personnel. Financing of Emekli Sandigi is somewhat different from the other plans in that health insurance during working years is not covered by the pension fund. Instead, it is covered directly by the line ministries in which the civil servants are employed. Another difference between Emekli Sandigi and other schemes i s that the basis for contributions and the basis for benefits are different. Contributions are paid on the basis of basic salary. On the other hand, when pension benefits are paid, they are paid on full remuneration. Thus, there i s both a financing gap and an equity issue, where lower-grade workers pay contributions on a larger share o f their salary than higher-grade workers. Noncontributory Pension Benejh. Turkey also provides a small noncontributory benefit to those over age 65 who earn below the level of the benefit. Social Assistanceand Solidarity Encouragement Fund The Social Assistance and Solidarity Encouragement Fund (SYDTF) is an extra budgetary fund financed by earmarked taxes and administered by a Cabinet Minister. The SYDTF, together with its local affiliates, i s the largest programo f social assistanceinTurkey interms o fnumber o f beneficiaries. Conditional Cash Transfers Conditional Cash Transfers (CCTs) are a national programrecently introduced in Turkey, supported by a loan from the World Bank, the Social RiskMitigation Project (SRMP). CCTs are payments made to the mothers of poor children, provided they attend school or visit health clinics. CCTs are important tools that are targeted to the poorest o f the poor, many o f whom are not able to afford the out-of-pocket expenses o f sending their children to school. In May 2005, 1.6 million children and 7,000 pregnant women benefited from the program. Local Initiatives Supported under the SRMP, the Government of Turkey has undertaken a significant expansion o f the microprojects traditionally handled by the SYDVs with approval from the SYDTF, along with a tightening of procedures. At the end o f 2004, 94,490 people benefited from income generation, employment, and social service opportunities under the SRMP Local Initiatives component, which seeks to provide these people with sustainable livelihoods, thereby lifting them permanently out of poverty. X CHAPTER I:DATA COMPARABILITY, 1994-2002 The State Institute o f StatisticsiWorld Bank (DIEM) team analyzed the data from the 1994 and the 2002 Household Budget Survey (HBS) to (a) assess comparability of consumption measures of welfare between the two main surveys datasets, (b) establish comparable poverty lines, (c) and discem trends in poverty and inequality inTurkey between 1994 and 2002. The 1994 and 2002 HBS are the largest and the only nationally representative household surveys inTurkey for 1987-2002. The in-depth analysis has shown that the results o fthese surveys are broadly comparable, and offer a solid base for analyzing dynamics o f living standards and poverty inTurkey. Comparisons o f living standards in Turkey between 1994 and 2002 are extremely sensitive to the price indexes used to convert current nominal figures into real values. Between 1994 and 2002, prices increased between 53 and 67 times, depending on the indicator o f inflation. The cost of living in Turkey increased more rapidly than the GDP-based deflator would suggest. The use of appropriate deflators and consumer baskets reflecting the consumption pattems of the populationreveals that living standards between 1994 and 2002 remainedalmost unchanged. Inequality, on the other hand, increased only marginally, remaining highby regional standards. The combination o f unchanged inequality and zero growth in real consumption produced the outcome whereby absolute poverty remained unchanged between 1994 and 2002. Intemationally-basedmeasures o f extreme poverty show a deterioration, while national poverty lines, anchored exclusively on food, indicate some improvement. This discrepancy reflects significant changes in relative prices over time, with food becomingcheaper, and non-food goods, mainly services, becoming relatively more expensive. The DIEiWB team concluded that these results are consistent with macroeconomic data, and are robust withrespectto measurementassumptions. A. REALINCOMEAND CONSUMPTIONOFHOUSEHOLDS BETWEEN1994AND 2002 The timing of the two major surveys inTurkey-1994 and 20024ictates the frame for the comparisons. Table 1.1 shows main macroeconomic indicators of living standards based on national accounts and survey-based estimates o f consumption. Key macro-based measures o f living standards are gross domestic product (GDP) per capita and personal consumption per capita. Per capita GDP i s not a particularly good measure o f standard o f living; it i s a measure of the total output inthe economy, which includes many items such as investment goods that do not make a contribution to the household welfare. Moreover, it has no direct relevance for poverty measurement. That i s why another measure-per capita personal consumption-will be usedinthe table. Table 1.1 presents two measures of personal (household) consumption based on survey data. The first measure i s constructed according to the principles o f National Accounts (NA) methodology ; the second measure intends to capture current consumption components, which i s more appropriate for the analysis o f poverty and inequality. It i s also more appropriate to take into account economies o f scale in the household, which are influencedbythe household size. The table shows that there is a gap (as in all countries) between consumption measured in NA and consumption measured in the surveys (this is due inpart to different definitions, and to underreporting). However, while this gap existed in both 1994 and 2002, it has narrowed over time: in 1994, the total consumption estimate based on the survey was around 55 percent of the NA estimate; in 2002, it increased to 66 percent. That i s why the comparison of the two surveys gives higher growth rates o f consumption than suggested by macro indicators. A slight fall in household size contributed to marginally slower growth o f the main welfare indicator per equivalent adult compared to changes in the per capita measure. Table 1.1. Macroeconomicand Household Survey-Based Data on Living Standards, 1994-2002 1994 I 2002 I Change, GDP, current prices, million TL* 4,027,176,295 277,163,385,986 69 Total personal consumption (fromNA), million TL* 2,706,262,470 184,036,488,295 68 Per capita GDP, current TL 67,729,991 4,052,509,121 60 Per capita personal consumption, current TL 453 14,554 2,690,866,056 59 - .~ Per caoita annual consumation INA method). current TL 71 I, 1.500.338.527 I 116.349.713.174 I 24.917.311 1 1.762.864.991 I ,I I , I , I Total householdconsumation (PA 1994method). million TL 78 ~~~~~ ~~ I ! , , , , , , Per capita consumption (PA 1994method), current TL 25,233,044 1,701,192,501 67 Consumption per equiv. adult (PA 1994 method), current TL** 40,427,172 2,667,781,670 66 Memo: Population 59,459,277 68,393,032 1.15 Numberofhouseholds 13,342,055 16,446,644 1.23 Average householdsize 4.456 4.15848 0.93 Average equivalent size o f household** 2.71 2.58 0.95 The main factor affecting direct comparability o fpoverty fkom various years inTurkey is inflation. Table 1.1 shows that consumptionincreased 66 times as measured by the survey. Of course, this has to be seen against the background o f high inflation. Table 1.2 lists main price indexes for 1994-2002. There are clear differences between different price indexes-note especially the Consumer Price Index (CPI) and GDP deflator. It i s not unusual, o f course, for fixed-weight price indexes (Laspeyres) such as the CPI to exceed indexes with variable weights (Paasche), such as GDP deflators, especially duringhigh inflation. The geometric average o f the two indexes i s often used to obtain the best estimate o f the true change in the cost o f living. The GDP private consumption deflator would be close to such an estimate. It i s our preferred measure o f changes in real cost o f living in Turkey. Also worth noting is the difference between the increase in food prices and total CPI, suggesting changes inrelative prices. Exchange rate- based indexes show much slower change in prices than proper price indexes. This calls for extreme cautioninusingvarious measures o fprices for making comparisons over time. 2 Table 1.2. Indexesfor Deflating NominalValues: 1994 and 2002,1994 = 1.00 1994 2002 Change, Times CPI all 1 66.5 66.5 CPI food 1 58.2 58.2 GDP deflator 1 53.0 53.0 GDP private consumptiondeflator 1 59.6 59.6 Current exchange rate, to USD 29,826 1,504,119 50.0 Current PPP exchange rate, to USD 12,096 663,575 55.0 deflator data are from World Bank 2003a. Based on different price indexes, the changes in mean real consumption are depicted in Table 1.3. The table clearly shows that when measured with the proper cost-of-living index (the personal consumption deflator from the national accounts), per capita indicators o f living standards have not changed much. CPI-based figures show a decline, while exchange rate-based indexes and total GDP deflators increase in real consumption. The table also shows that the comparison is extremely sensitive to the price indexes used, and that the preferred measures based on macroeconomic data and based on surveys consistently show practically zero growth o freal consumptionbetween 1994 and 2002. Table 1.3. Change inMain Indicators for Real Incomes and Consumption, 1994,2002 1994 2002 Change, Times ** 1994equivalence scale (nutrition based). Sources: SIMA unified survey 2002, the World Bank for real GDP and gross national income (GNI) per capita data, other data providedby DIE. The baseline survey-based measure-per capita consumption (highlightedhincreased by only 1percent between 1994 and 2002 (while the per equivalent adult measure fell by 1 percent)-in line with the corresponding macro variable. The picture o f the relative stability o f consumption inreal terms i s qualified when annual indexes are used instead o f point-to-point comparisons. The extreme volatility becomes evident. The NA measure of the population's living standard-private consumption per capita deflated with the GDP personal consumption deflator-shows modest growth between 1994 and 2002. But once the CPI price index is 3 used, the result is the opposite: real consumption fell between 1994 and 2002. The current dollar consumption shows even greater volatility. Overall, in 2002 it appears to have increased only when compared with the through figures of 1994 and 2001 (Figure 1.1). Figure 1.1. ConsumptionPer Capita with Various Deflators, 1990 100 - -capita, Private consumption per constant prices, private constant deflator Private consumption per capita, current USD Private consumption per capita, constant prices, CPI It is evident that such an outcome is drivenby different trends inprice indexes. Figure 1.1illustrates this point. For the entire period, the CPIindex was above both GDP deflators. Figure 1.2. PriceIndexesinTurkey, 1990-2002, Annual PercentageChanges 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 -Inflation, CPI to GDP private consumption ev pr deflator ye ar =I nn -GDP Deflator Exchange rate chain index 4 B. INEQUALITY INCONSUMPTIONOFHOUSEHOLDS BETWEEN 1994AND 2002: HBS Inequality as measured by consumption from HBS has not changed much once the same standards in constructing the consumption aggregate and equivalence scales are applied to the 2002 data, as were used by the World Bank with the 1994 data (see Table 1.4 and Annex I). Table1.4. Inequality 1994 2002 Change, Times Inequality, Ginifor 1994 consumption per equiv. 0.385 0.390 1.01 adult* Inequality, Ginifor 1994 consumptionper capita 0.408 0.413 1.01 c. POVERTYLINES: 1994AND 2002 METHODOLOGY There are important differences in defining the poverty line according to 1994 and 2002 methodologies. They arise mostly from the composition o f the minimum food basket. The 1994 methodology used a minimumfood basket developed by Hacettepe University. The new methodology proposes to use actual (HBS) survey data to obtain the composition o f the minimum food basket. The HBS provide us with expenditures and quantities of each food item consumed by the households. The 2002 HBS analysis team determined the average expenditure and average quantity o f each food item in each population decile. Based on the U.S. Department of Agnculture (USDA) database (available on the World Wide Web), the team also obtained the total calorie content of these baskets. The calorie requirements (needs) of individuals may be used as a starting point for constructing food poverty lines. These requirements depend on several factors such as age, sex, body weight, climatic conditions, and activity levels. The new approach takes actual consumption o f 80 main food items in the third and fourth deciles o f total consumption, calculates total caloric value, and scales each quantity proportionately so that the total calorie intake from the basket i s 2,100 Kcal on average per person (or 2,450 Kcal per adult). The quantities obtained are quite different from the Hacettepe University basket, previously used in Turkey. Table 1.5 demonstrates that the Hacattepe basket i s quite different from actual food consumption patterns. The new 2002 methodological approach also follows "best practice" by defining products very narrowly (down to 10-digit codes), whereas the Hacattepe basket i s defined in broad product groups, which again has implications for pricing the basket properly. Based on the value o f a minimumfood basket, the non-food component was estimated based on the actual consumption structure o f the poor. This part o f the methodology is very similar to the approach developed before for the analysis of the 1994 data in the previous Poverty Assessment (World Bank 2000), but has some differences in details, producing slightly different results. To obtain the real value o f a 2002 basket o f food and non-food goods for 1994, components of the CPI (food and non-food) were used. With the 1994 data, to update the poverty line to 2002, direct survey-based prices were used to obtain the value o f the minimum food part, and the fixed markup was updated using the non-food CPI index (regionally differentiated). This makes the 1994baseline comparisons more accurate. 5 Table 1.5. Food Basketfor EquivalentAdult: 2002 Survey-Basedand HacettepeUniversity (kilograms per day per adult) n.e.c. = Not elsewhere classified, To complement these national poverty lines, the team also developed a set o f international poverty lines. The 1990 World Development Report presented for the first time the global estimates of poverty on the basis of a US$l-a-day poverty line estimated using the 1985 purchasing power parity (PPP) exchange rate. Since, inthe late 1990s, 1993 PPP exchange rates were available for a large number of developing countries, the World Bank re-estimated the international poverty line. Local poverty lines were used for 33 countries, convertedto 1993 PPP dollars. The medianpoverty line o f 10 countries with lowest poverty lines was calculated to be equal to US$l.OS per person per day, and it was adopted as the new international poverty line. It should be noted that this new poverty line i s not strictly comparable to the US$l-a-day poverty line in 1985 PPP dollars. The US$1-a-day poverty line is appropriate only for low-income countries that are situated in tropical regions. People living in the Europe and Central Asia (ECA) region have to face a harsh, cold climate, which requires them to heat their houses, wear warm clothing, and ingest even greater energy in kilocalories (Kcal) to maintainthe same metabolism. Their basic needs are obviously very different from other countries, so their poverty lines should also be different. That i s why ECA countries are currently using PPP US$2.15 and PPP US$4.30 lines for international comparisons. In addition, the 1993 PPP conversion rates are not available for the Commonwealth o f Independent States (CIS). However, the 1996 PPP conversion rates were available for the CIS countries. The recommended US$2.15-a-day PPP uses the same methodology as the standard US$1-a-day methodology by taking the median of countries with the lowest national poverty lines converted with the 1996 PPP. Turkey has all PPP estimates needed for such calculations: 1985, 1993, and 1996. In addition, PPP rates are available for each year. Thus (although it is not a correct application of the World Bank methodology), one could use each year's estimate o f PPP. It must be emphasized, however, that PPP exchange rates were not designed for making international poverty comparisons; they were mainly designed for comparing aggregates from national accounts. The PPP exchange rates are based on prices of commodities that are not representative o f the consumption baskets o f the poor. More important, weights inthe PPP baskets of goods and services do not adequately represent the consumption basket of the poor. That i s why using CPI indexes to update the base year estimate i s required. Consistent estimates are obtained usingeither (a) the 1996 PPP, as done by ECAPOV, converted to current Turkish Lira (TL) usingthe CPI to 1994 and 2002 prices; (b) or the 1993 PPP, converted to current TL usingthe CPI. 6 Table 1.6 presents the main variants o f setting the poverty line. The two top rows present two variants o f setting the poverty lines based on methodology proposed injoint DIE/WBwork on the 1994 HBS data, and the new methodology used in this report to set the poverty line (2003). Lower rows give values for the international poverty line (US$1-a- day per capita at PPP) and the relative poverty line. Table 1.6. PovertyLines, 1994-2002 II I I I Foodbasket, 1994(Hacateppe), TLlday* 38,794 2,159,424 56 Poverty line, 1994methodology, TL/day** 69,142 4,712,879 68 Foodbasket, 2002 methodology. TLlday**** 18,571 1,082,2 11 58 Povertyline, 2002 methodology, TLlday*** 38,928 2,510,930 65 USDcurrent PPPUS$1a day, TLlday**** 12,096 663,575 55 - USD 1985PPPUS$1a day, CPIto update, TLlday** ** 14,803 984,375 67 USD 1993PPPUS$1.08 a day, CPIbased, TLlday*** * 13,221 879,195 67 USD 1996PPPUS$1.08 a day, CPIbased, TLlday**** 15,266 1,015,205 67 60% of the median 1994methodologycons, TL/day** 50,246 3,3 19,177 66 D. POVERTY IN1994AND 2002: HBS It i s impossible to apply all elements o f the methodology in the same way to 1994 as to 2002. The consumption index developed for the 2002 HBS i s more accurate, but it is impossible to replicate this methodology with 1994 HBS data. The decision was therefore to recalculate the consumption indicator in the 2002 dataset to make it completely identical to the 1994 indicator. With this indicator, a variety o f poverty lines are used: 1994 and 2002, and the international PPP-based lines. The comparison undertaken is presentedinTable 1.7 by "X". Table1.7. ComparisonofPoverty inTurkey Between1994 and 2002, with NationalAbsolute PovertyLine POVERTY LINEMETHODOLOGY Poverty Line 1994 Est. 1994HBS+ Est. Poverty Line 2002 Consumption 1994 PA (2000): 1994HBS (baseline)+ Est. 2002 HBS 2002 HBS X X Consumption2002 New report, only 2002 HRS 7 Based on comparable consumption measures (see Annex I)and comparable poverty lines, Table 1.8 presents results. Table 1.8. PovertyIncidencewith Comparable ConsumptionAggregate, Equivalence Scale and PovertyLines 1 1994 1 2002 1 I Times National poverty lines Economic vulnerability, 1994Complete PL 36.3% 34.5% 0.95 r Poverty, 1994FoodPL 7.3% 4.9% 0.67 t I t I Poverty, 2002 complete PL 28.3% 27.0% 0.95 Extreme poverty, 2002 food PL 2.9% 1.4% 0.48 World Bank international lines WDR 1990method: 1985 PPP US$l aday 2.5% 3.2% 1.28 Updated WB: US$1.08 at 1993 PPP a day 1.7% 2.0% 1.18 UpdatedWB: US$2.15 at 1993 PPP a day 15.9% 16.8% 1.06 ECAPOV US$2.15 at 1996PPP a day 22.1% 22.6% 1.02 US4.30 at 1996PPP a day 61.O% 60.6% 0.99 I I Current PPP exchange rate I I I Poor, US$l per day 1.1% 0.2% 0.18 Poor, US$2.15 per day 15.3% 9.2% 0.60 Poor, US$4.30 per day 51.7% 38.9% 0.75 Usingthe 1994 methodologywith the 1987 data leads to the resultthat 38.5 percent o fthe population was poor in 1987, suggesting that there has been a gradual decline in poverty over the last two decades, but that the many shocks experienced in the macroeconomy meant that poverty did not decline significantly over the period. Figure 1.3 presents this information graphically, combining the "stock" measurement of poverty in 1987, 1994, and 2002 with the "flow'' data on GPD growth over the time period. Figure1.3. Turkey: Povertyand GDP Growth 8 The relative poverty line can be used to make some comparisons of consumption poverty. The relative line i s fixed as a constant share of the median consumption. However, note that such figures will not be entirely comparable, because the methodology requires recalculation o f the poverty line for each year. Using60 percent o f the currentyear per equivalence adult (usingthe 1994 scale) consumption as a cutoff, 20.1 percent o f the population was below the poverty line in 1994, which increased slightly to 21.5 percent by 2002. The use o f international poverty lines allows cross-country comparisons. The decision to compare Turkey with other countries in ECA dictates the choice o f poverty lines (Table 1.9). Because countries differ in the quality of their survey data, the table ranks all countries by the ratio o f survey-based consumption mean to the estimates of personal consumption from national accounts. Turkey seems to occupy a unique position with respect to this indicator. The closest comparator countries, interms o fboth national income per capita and o f survey coverage o f personal consumption, are Lithuania and Latvia. Both countries have dramatically lower poverty. Such comparison highlights the significant role o f inequality inTurkey as a driver ofpoverty. Highinequality inTurkey worsens poverty considerably. Table 1.9. Absolute Poverty Ratesof Europeand CentralAsia (ECAPOV methodology) I I Survey IHeadcount IHeadcount I Ratioof I 1998GNP I 1998GNP 1 Date- I Survey 1 US$2.15/Day US$4.30/Day To N A Private Atlas Per Capita Cons. Method TURKEY 1994 22.1 60.9 0.55 2,640** 5,873" TURKEY 2002 22.6 60.6 0.63 2,500** 5,285" Belarus 1999 1 10.4 0.99 2,180 6,318 Tajikistan 1999 68.3 95.8 1.02 370 1,040 9 IAzerbaijan 23.5 I 64.2 I 1.39 I 480 I 2,168 ***ForTurkey, GDP data for correspondingyears are used. GNI. The survey for Albania did not cover the capital city Tirana. Note: Recent household survey data are not available for Bosnia and Herzegovina and Uzbekistan. Private consumptiondata are not available for Tajikistan, Turkmenistan, and Kazakhstan. GDP per capita in current prices is used instead. Private consumption data for Azerbaijan are for 1998; GDP per capita (first half 1999) are used for Ukraine. The poverty headcount numbers are based on the international poverty lines of US$2.15 and US4.30 per person per day; PPP estimates are for 1996. Sources: "Making Transition Work for Everyone." Turkey-estimates based on 1994and 2002 HBS. E. USINGTHE 1994HBSAND2002HBS TODECOMPOSE CHANGESINPOVERTY Methodology Poverty can change over time depending on two factors. The first factor is the magnitude o f economic growth rate; the larger the growth rate, the greater the poverty reduction. The second factor is the distribution o f benefits o f growth; ifthe benefits o f growth go more to the poor than to the non-poor, then the poverty reduction will be larger. Poverty levels can then fall for two reasons: either growth increases the consumption o f all members of society, or the share o f the poor intotal consumption increases due to shifts in the distribution o f total welfare. On the other hand, poverty may rise if either consumption falls or the distribution shifts against the poor. Often the two processes-growth and redistribution-operate simultaneously, and they reinforce each other ifthey work inthe same direction, or weaken each other. Decomposition of poverty changes. Changes in poverty can be decomposed into changes due to growth and changes due to an increase ininequality. This decompositioni s crucial to proper identification of the link between poverty and growth. Figure 1.4 illustrates the principles o f such decomposition (see Bourguignon 2002). Figure 1.4. Poverty Changes -initial distr. 0.5 -new distr. ----only mean shift 0.4 0.3 0.2 0.1 0 10 Initially the poverty rate i s equal to A+B+C, that is, to the area below the initial distribution to the left of the poverty line. When the distribution changes to the new one, the new poverty rate equals A. The difference, area B+C, can be split into a growth component, area C, which shows how the poverty would have changed if only growth effect had been in place (that is, without a change in the shape of the distribution-only shift in mean). Area B is the additional poverty reduction achieved due to falling inequality. The move from the initial to the new distribution can be regarded as the combination o f two effects: a pure proportionate growth effect captured by the rightward shift o f the distribution function; and a pure redistribution effect. This allows the total change in poverty to be decomposed in a similar fashion, taking into account the contribution of income growth and redistribution. In the situation portrayed in Figure 1.4, the two effects reinforce eachother to produce a significant reductioninthe headcount poverty rate, but the same analysis can also be applied inless-favorable circumstances. Formally, we calibrate the distribution of well-being relative to the poverty line.5 Poverty (P) is then a function o f mean consumption and the Lorenz curve, or P(z/p,L), where z is poverty line, p is mean consumption, and L is Lorenz curve, showing the distribution o f consumption among individuals. This procedure allows the change inpoverty for Turkey between, say, 1994 and 2002 to be expressedas: p2002-p1994= g -k d + where the contributions o f growth (g)and redistribution (d) are definedas: and r i s aresidual due to interactionbetweengrowth and distribution. The problem with this specification is that g indicates the marginal effect of the change inmean income with the distribution held constant at the initial configuration (L1994) while d computes the marginal impact of redistribution holding mean income constant at the final level (pLzOo2). One can equally well generate a decomposition with the other conditions interchanged, and since there is no logical reason for preferring one configuration over the other, symmetry arguments suggest that the two effects should be averagedto yield the growth effect: g = ?4[P(z/p2002,L1994)-P(~~p1994,L1994)1 + ?4 [~(~p2002,~2002)-~(z~p1994,L2002)], and the redistribution effect These expressionsturn out to be the contributions associatedwith the level and distribution of income ina two-way Shapley decomposition o f the change in poverty.6 We apply this, as well as a traditional decomposition, inwhat follows. Every poverty measure can be decomposed in a simple way to quantify the relative importance o f growth and changes inthe distribution, as shown inDatt and Ravallion (1992). The Shapley decomposition is inspiredby the classic cooperative game theory problem o f dividing a pie fairly, the Shapley solution to which assigns to each player his or her marginal contribution averaged over all possible coalitions o f agents. The reinterpretation described inShorrocks (1999) considers the various factors (n intotal, say) that together determine an indicator such as the overall level o f poverty, and assigns to each factor the average marginal contribution taken over all the n! possible ways in which the factors may be "removed" in sequence, The 11 Results: Turkeybetween 1994and 2002 The annual results o f such decomposition for Turkey are presented in Table 1.10. It is evident that economic growth was a main driving force in poverty reduction, while inequality was acting in the opposite direction to reduce the effect of growth. Table 1.10 shows that poverty incidence decreasedover the period by 1.77 percentage points, from 36.2 percent of the population to 34.5 per~ent.~This corresponds to an increase in the consumption (per equivalent unit defined according to 1994 methodology) by about 5 percent between these two years. If only mean consumption had changedbetween 1994and 2004 and its distributionbetween the rich and the poor had not, the decrease in poverty would have been significantly bigger than the observed decrease (poverty according to 1994 methodology would fall to about 32 percent o f the population). If only changes in the distribution o f consumption had affected poverty, as a result o f the increase in inequality (the Gini index edged up by about 1 percent) we would instead have observed an increase inpoverty to 38.7 percent. Table1.10. Decompositionof Poverty ChangeBetween1994 and2002 Into Growthand RedistributionComponents 1994 2002 200211994 Percent Poor 36.2% 34.5% -0.018 Growth Component -0.046% Redistribution Component +0.025% Residual +0.003% Full(Shapley Decomposition) Growth Component -0.044% Redistribution Component +0.026% Source: Calculated from DIE2002 HBS. This simple simulation shows boththe dominance of growth as a key factor inpoverty changes, and the importance of distribution, especially inthe periods o f low growth that Turkey experienced between 1994 and 2002, when it becomes a significant explanatory factor o f poverty dynamics. Unfortunately, overall changes inmean consumption and distribution between 1994 and 2002 were quite marginal, and thus the difference is only marginally significant. It should be kept inmindthat 1994 and 2001 were crisis years, and thus, little growth was measured in the period. Sustained growth could have had a much greater impact on poverty, had it occurred. particular attractions o f this technique are that the decomposition is always exact, and the factors are treated symmetrically. As shown inChapter Ion the comparison o fthe 1994 HBS, baseline comparability canbe assured only when 1994 poverty lines and methodology are applied to the 2002 data. 12 F. CONCLUSIONS The timing of two major surveys in Turkey-1994 and 20024ictates the frame for the comparisons. Unfortunately, due to macroeconomic instability, living standards between these two years have not improved. The conclusion that stems from this analysis i s that growth between 1994 and 2002 was not sufficiently strong to produce any sizable reduction in poverty, and the impact o f the little growth there was, was dampened by an increase ininequality. 13 CHAPTER 11: MACROECONOMIC CONTEXT Despite many advantages ranging from its strategic location to its dynamic population, Turkey has not achieved the stable high growth o f leadingemerging market economies.' Nor has it matched the growth rate o f European Union (EU) accession countries such as Hungary and Poland, or the fast-growing cohesion counties such as Spain and Portugal (Figure 11.1). Turkey's per capita income level declined from 26 percent o f the EU average in 1991 to 22 percent in 2002.9 Duringthe same period, Poland and Hungary made significant progress in reducing the per capita income differences with the EU. On average, the Turkisheconomy grew slightly under 3 percent per year over the past decade-respectable, but well below the best-performing emerging economies (Figure 11.2). Figure11.1. Per Capita GDP at PPS: 1991-2002 Compared to EUAverage 9 0 T- Spain v1 8 0 a a .-c 7 0 P o r t u g a l z An _ _ 0- Hungary II 5 0 3 W 4 0 P o l a n d 3 0 2 0 Turkey Source: World Bank Figure 11.2. GDPand GDP Per Capita Growth Rates inEmerging Economies (1965-2001) 10 8 s-6 - 4 I 2 0 ... Source: World Bank 8 This chapter draws on "Turkey: Country Economic Memorandum-Towards Macroeconomic Stability and Sustained Growth," July 28, 2003, World Bank Report No. 26301-TU. These figures do not reflect the informal economy, which is likely to be substantially larger as a percentage of GNP inTurkey relative to the EUaverage. 14 A. MACROECONOMIC INSTABILITY HASHELPED KEEP GROWTH BELOWPOTENTIAL Analysis suggests that macroeconomic instability-among many factors-has played an important role in Turkey's inability to realize its full growth potential. Cross-country comparisons and analytical work suggest that countries that grew faster than Turkey did so in part because they achieved a greater degree of macroeconomic stability, accumulated physical capital faster, invested more inhuman capital, and did more to improve government effectiveness and the business climate. Of all these factors, the contrast in the degree of macroeconomic stability stands out. Turkey has suffered from an exceptional degree of macroeconomic instability characterized by chronically high inflation and sharp swings in the business cycle. Many emerging market countries have experienced large fluctuations ineither growth or the real effective exchange rate (REER), but Turkey experienced instability in both (Figure 11.3). Repeated attempts to stabilize the economy fell short, and highgrowth was never sustained for long. Inflation was higher and growth lower, on average, in the 1990s than in the 1980s. Income volatility doubled during the 1980s and 1990s as the standard deviation o f real GDP growth increased from 2.7 percent to 5.5 percent. The boom-bust cycle has continued into the new decade, with a record contraction of real gross national product (GNP)o f over 9.5 percent in2001, followed by a strong recovery, with estimated growth o f 7.9 percent in2002. Figure11.3. Turkey: REERand GDPGrowth,1990-2000 1 2 - + 10 2 5 s + T u r k e y S 6 U > 4 T h a i l a n d 2 C h * 0 0.5 0 . 1 5 1 1 . 2 5 1.5 C V : G D P ( C h % ) Sources: W o r l d B a n k , JPM organ; volatility measured b y coefficient o f variation applied to growth rates o f t h e variables variation applied to B. FISCAL IMBALANCEHAS BEENTHE ROOT CHRONIC MACROECONOMIC OF INSTABILITY IN TURKEY The 2000 Country Economic Memorandum (World Bank 2000a)" demonstrated that fiscal imbalances are key to understanding Turkey's macroeconomic instability. Unsustainable fiscal policy has repeatedly putpressure onthe TL, fueled inflation, andunderminedfinancial stability. Fiscalpolicy has been unable to act as a smoothing influence on the business cycle. When crises have hit, contractionary fiscal and monetary policies have been required to restore a semblance o f financial stability, worsening the real impacts o f internal and external shocks. The impact o f unsustainable fiscal policy on macroeconomic stability has been magnified by Turkey's open capital account and, until recently, its poorly regulated lo "Turkey: Country Economic Memorandu-Structural Reforms for Sustainable Growth," September 15, 2000, ReportNo. 20657-TU, World Bank. 15 banking system. Short-term capital flows have fluctuated widely as investors responded to the boom- bust cycle driven by unstable macroeconomic conditions. The causal linkages between fiscal imbalances and instability in Turkey, as in many other emerging markets, suggest that the key to macroeconomic stability lies in sustained fiscal adjustment underpinned by credible structural reforms (Figure 11.4, data from World Bank 2003a). Figure11.4. AdjustedPublicSector BorrowingRequirement(YOof GNP) I 25 20 15 ~ 10 I 5 0 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 C. TURKEY'SEXCHANGERATE-BASEDDISINFLATIONPROGRAM In2000, an exchange rate-based disinflationprogramwas launched inTurkey ina boldattempt to ridthe economy o f inflation. The centerpiece o f the program was a crawling peg exchange rate regime to act as the nominal anchor. The peg was supported by front-loaded fiscal adjustment. K e y structural reforms in social security, infrastructure, agriculture, privatization, and banhng were initiated. Infact, fiscal policy was significantly tightened in 2000, and inflation began to fall, dropping to 39 percent by the end o f the year. Turkey also carried out significant structural reforms under the program. These included establishment o f an independent banking authority; passage of legslation for an electricity market; reform of the public pension system; a constitutional amendment to allow intemational arbitration; launch o f an ambitious agriculture reform; establishment o f a telecommunications regulator; and a serious, although short-lived, acceleration o f privatization. However, these impressive achievements were insufficient to avoid a crisis, given the extent o f Turkey's underlying fiscal and financial sector weaknesses builtup over decades o f instability and delayed reform. Internal factors combinedwith unfavorable external developments started to undermine the exchange rate pegby mid-2000. Onthe external side, rising oil prices and a prolonged slide inthe Euro contributedto a softening o f Turkey's extemal accounts. On the internal front, the disinflation program was confronted with deep-rooted structural fiscal problems and a fragile banking sector burdened by huge contingent liabilities. A sharp drop in interest rates following the onset o f the crawling p e g 4 r i v e n in part by a resurgence in short-term capital inflows-fueled a surge in demand. The economy soon began to overheat. While falling, inflation did not come down as quickly as anticipated, generating a significant appreciation o f the real exchange rate under the peg. Imports increased sharply as consumption boomed, contributing to a deterioration in the current account, which recorded a deficit o f 4.9 percent o f GNP in 2000. Domestic banks took advantage o f the peg to borrow cheap foreign exchange in order to finance their expanding domestic operations, including growing purchases o f government securities and consumer 16 lending. The expansion indomestic credit contributed to the consumptionboom as banks quickly builtup large open foreign exchange positions and aggravated maturity mismatches in their portfolios. Bank restructuring got off to a slow start, and the state banks continued to be burdened with the costs o f large "duty losses" from government-mandated subsidized lending to agriculture and small and medium enterprises (SMEs). The average maturity o f Turkish Lira (TL) deposits remained extremely short because confidence inthe TL remained fragile. A first bout of financial instability hit Turkey in November 2000, presaging the full-fledged currency crisis o f early 2001, which short-circuited the exchange rate-based disinflation program. As banks came under increasing pressure from shrinking profit margins on government securities and growing liquidity needs, isolated speculative attacks emerged in November 2000, which soon plunged the banking system into a struggle for survival. Desperate for liquidity, certain banks engaged in fire sales o f government paper, causing interest rates to skyrocket and international investors to exit the market. The result was a liquidity crunch, aggravated by the Central Bank's inability to inject additional liquidity into the system under the quasi-currency boardrules added to the disinflation programjust prior to launcho f the crawling Peg. The situation stabilized temporarily in December 2000 when the International Monetary Fund (IMF) acted to prevent a financial meltdown by announcing an additional US$lO billion in financial assistance. This additional financing was conditionedon the Government's commitment to strengthen the p r o g r a m inparticular, to accelerate financial sector restructuring andprivatization. The Government introducedan explicit blanket guarantee effectively covering all banking liabilities, excluding capital. While designed to contain systemic risks in the banking system, the blanket guarantee highlighted the potentially enormous fiscal costs in case o f a systemic failure, and made explicit this contingent fiscal liability. In the wake o fthese events, interest rates declined and a precarious degree of financial stability returned, but this provedto be a temporary respite. Inearly 2001, persistent doubts about the peg and underlying fiscal sustainability led to a full-blown speculative attack against the currency. Interest rates shot up to several thousand percent, forcing the Government to abandon the crawling peg and float the Lira on February 21, 2001. The Lira immediately lost 40 percent o f its value ina single day. D. EXCHANGE RATE-BASED DISINFLATIONPROGRAMSARE VULNERABLE INA GLOBALIZED WORLD The reliance on a pegged or fixed exchange rate in an environment o f free capital flows and an unreformed banlung system entails risks. Exchange rate anchors-while generally successful in setting inflation on a downward long-term trend in chronic-inflation countries-have often been associated with currency crises. Many of the exchange rate-based stabilization programs in the 1980s encountered currency crises at some stage, and the economic crises that broke out in the second half o f the 1990s occurred in countries with fixed or managed exchange rate regimes. Currency attacks have often been accompanied by banking crises (for example, Chile, Mexico, and East Asia). The Asian experience shows that, with limited capital mobility, even a weak banking system can function reasonably well and support economic growth. However, such a system may not be able to handle massive entry and exit o f short-term capital induced by capital account liberalization in the context of globalization. Exchange rate-based stabilization also generally results insizable real exchange rate appreciation and a deterioration o f the current account, which can undermine investor confidence. International experience has shown that early moves to introduce exchange rate flexibility can minimize the extent o f subsequent currency crises, as in Israel. Turkey's program featured a predetermined transition to a widening exchange rate band 18 months after the launch of the peg. However, this preannounced exit-unique among pegged exchange rate systems-did not prevent the collapse o f the peg after only one year. 17 E. THEGOVERNMENT RESPONDEDQUICKLY TO THE 2001CRISIS The Government announced a strengthened economic program in M a y 2001 in response to the crisis following the collapse o f the crawling peg and subsequent devaluation. The key structural and social elements o f the program were: (a) a macroeconomic framework designed to restore financial stability and ensure public debt sustainability-principally through a further tightening o f fiscal policy; (b) rapid restructuring o f the banking sector-especially o f state banks and insolvent private banks intervened by the regulatory authority, the Banlung Regulation and Supervision Agency (BRSA)-based on large resource transfers from the budget; (c) a more ambitious program o f public sector reforms centered on deeper structural and institutional reforms to improve fiscal management and public governance; (d) a renewed privatization drive-in combination with further liberalization measures focused on energy, telecommunications, and agriculture-and strengthening o f independent regulatory bodies to improve the private investment climate; and (e) enhanced social assistance to help low-income groups adversely affected by the crisis. F. TURKEY'S CRISIS RESPONSEPROGRAMINCORPORATEDTHE EXPERIENCE OTHER OF EMERGING MARKETS Turkey's crisis response program benefited from the lessons learned by other emerging market countries facing crisis (Liviatan 2002; Brahmbhatt 2001). Immediate fiscal measures were introduced to shore up the primary surplus and strengthen confidence in the sustainability o f the public debt. A front-loaded program of bank restructuring was launched, backed by extensive fiscal resources. Bank restructuring was complemented by additional structural reforms in the financial sector designed to further strengthen prudential regulation, adopt internationally accepted financial reporting standards and practices, and enforce compliance. Inparallel with accelerated financial sector reform, a comprehensive public sector reform program, including institutional reforms, was introduced to tackle the structural roots o f Turkey's chronic fiscal imbalance. Strengthened financial and public sector reforms were placedwithin a medium- term programmatic framework in an effort to bolster investor confidence by demonstrating the Government's intent to address the core structural causes behind the crisis, and not just the immediate symptoms. Initial outcomes under the crisis response program were mixed as the Government struggled to contain the fallout from the crisis and reestablish its policy credibility. The immediate financial turmoil arising from the crisis was fairly quickly contained, but at the price o f a sharp increase in the public debt as the costs o f bank restructuringwere bome by the budget. The price spike following the initial devaluation in February 2001 was contained, but inflationary pressures persisted, with inflation reaching 69 percent by the end o f the year (Table 11.1). Following the decision to abandon the peg, uncertainty about exchange rate policy persisted for some time because the Government was slow to confirm its commitment to the float, and the Central Bank repeatedly intervened in the foreign exchange market. Interest rates were brought down from their post-crisis peaks, but remained well above the projected program path throughout 2001, mainly due to the needto roll over large amounts of short-term public debt inthe face of a slower-than-expected recovery in investor confidence. The primary surplus target o f 5.5 percent o f GNP for 2001 was met, but doubts continued about the medium-term sustainability o f the fiscal adjustment. The economic recession turned out much deeper than projected as real GNP shrank by an estimated 9.5 percent for the year. A major factor inthe recession was the sharp turnaround inthe current account driven by capital outflows. The current account recorded a surplus o f 2.4 percent o f GNP in 2001. The combination o f highreal interest rates, devaluation, the huge fiscal cost o f bank restructuring, and deep recession caused the stock o f public debt to rise significantly. The ratio o f net public debt to 18 GNP increased from 58 percent of GNP at the end of 2000 to an estimated 94 percent of GNP by the end of 2001 (Figure 11.5). Table 11.1. Key EconomicIndicators Actual 21 Est. 21 Prop. 31 1999 2000 2001 2002 2 0 0 3 2004 iMAIN M A C R O INDICATORS GNP Growth -6.1 6.3 -9.5 7.9 5.9 5.1 CPI Inflation (Dec-Dec) 68.8 39.0 68.5 29.7 18.4 12.1 Nominal Interest Rate 106.2 38.0 99.1 63.5 44.1 29.1 Real ex ante Interest Rate 4/ 32.0 -9.5 35.5 30.3 28.6 11.: Unemployment Rate 7.7 6.5 8.4 10.3 10.5 Unit Wage Index (1997 = 100) 113.7 103.0 73.9 73.1 86.3 PUBLIC SECTOR Primary Balance (% GNP) -0.2 2.7 5.5 4.1 6.1 6.: Overall Deficit (% GNP) 22.3 18.9 21.1 12.1 9.9 6.1 Net Public Debt (% GNP) I / 61.0 58.3 93.9 78.8 70.1 65.' o f which net external debt (% GNP) 20.1 19.0 37.7 32.1 22.2 20.' Privatization ($ bn) 0.1 3.3 2.8 0.5 0.3 3.1 EXTERNALBALANCE Current account balance (% GNP) -0.7 -4.9 2.4 -0.8 -3.4 -3.4 Exports (fob, S bn) 51 28.8 30.7 34.4 40.1 51.2 55. Tourism (%bn) 5.2 7.6 8.1 8.5 13.2 9.8 External Debt (% GNP) 55.0 59.0 79.0 71.3 61.1 48. CBT Foreign Exchange Reserves ($ bn) 24.3 23.2 19.8 28.1 35.2 32. GNP (TL quadrillion) 78.3 125.6 176.5 275.0 356.7 415. I/IMF, includes government securities issuedto recapitalize SDIF and state banks. 2/ Government figures as adjusted by I M F and W B estimates. 2002 and 2003 G S P figures are as announced by SIS in April 2004. 31 Updated as ofthe 7th I M F review. 4/ Average o f monthly nominal interest rate divided by 12-month ahead CPI inflation. 5 / Includes shuttle trade. Sources: Government, IMF and W B estimates. Figure11.5. TotalNetPublicDebt(YOof GNP) 100 I "^"-"I " ^ - ^ Cost o f rertructurmg o f state banks 80 60 40 20 1994 1995 1996 1997 1998 1999 2000 2001 2002~- 2003 -~ ~- _ ~ 19 G. STRONGRECOVERYHAS BEENUNDERWAY SINCEEARLY 2002 Economic activity rebounded strongly in 2002 and the recovery continued into 2003. Real GNP growth reached 7.9 percent in2002, exceeding programprojections by a wide margin (Figure 11.6). The recovery was led by robust export performance and exceptionally large inventory rebuilding inthe first half o f the year. The recovery was further buoyed by a sharp rise inpublic consumption and investment during the second semester, reflecting accelerated government spending ahead of early elections held in November 2002. An increase in agricultural output estimated at 6.9 percent was another factor. Overall, stock building accounted for some 90 percent o f 2002 growth, with private consumption and government spending making significant contributions to offset declines in private investment and net factor income. The impact o f strong export performance inthe 2002 growth accounting was offset by even faster growth in imports as the recovery gained steam. Importantly, private consumption and investment ledthe way for the first time since the crisis, recording increases in 2003 o f 6.6 percent and 20.3 percent, respectively. While stronger-than-expected growth has been due in part to base effects from the recession, the genuinely positive news is that the recovery has been export led, with exports inUS.dollar terms increasing by some 30 percent in2003. Strong export performance, buoyant tourism, and renewed capital inflows have eased the pressure on the balance o f payments, even as imports have expanded rapidly with the economic recovery. The modest current account of about 1percent o f GNP in2002 was easily financed. The extent of the recovery, and its basis in export growth, place Turkey squarely inthe category o f rapid-recovery, post-crisis countries, such as Mexico in 1995 and Korea in 1999. Turkey's recovery began three quarters after the crisis trough was reached, in line with the fastest recoveries worldwide over the past decade. Figure 11.6. GNP Growth In contrast with the real-side recovery, financial outcomes were mixed in 2002. On the positive side, inflation fell sharply. Consumer prices increased 29.7 percent over the course of 2002, well below the program target o f 35 percent. The fall in inflation was helped by the rebound o f the nominal exchange rate from its crisis lows. The strengthening TL did not hurt export performance because it was counterbalanced by a very sharp drop in real wages. More problematic were slippages in the fiscal program duringthe runup to the November elections. A gap o f about 2.5 percent of GNP emerged with respect to the 2002 primary surplus target o f 6.5 percent o f GNP. Contributing factors included: (a) cost overruns in the social security system; (b) pre-election spending (new agriculture support purchases and 20 civil service wage increases); (c) an unexpected drop-off in tax revenues, driven by expectations o f a post-election tax amnesty; and (d) unplanned spending through earmarked accounts left over following closure o f the extrabudgetary funds. A series o f stopgap fiscal measures were identified in late 2002 to close the fiscal gap, but were left largely unimplemented. Despite the fiscal slippage, the stock of net public debt fell to an estimated 80 percent o f GNP by the end o f 2002, helped by the rebound inthe real exchange rate after the 2001 overshooting* and higher-than-expected growth. The new Govemment's grace period with the financial markets was short lived as concems about a slow start on economic reform, hints o f political tension, and the looming threat o f hostilities in neighboring Iraq started to weaken investor confidence in mid-December. Ad hoc increases in pensions in early January and other populistmeasures raised concems about the Govemment's political will to implement tough economic reforms. The average yield o f the benchmark government paper moved up to the 60 percent range, and the Lira came under some pressure. Financial market volatility continued during the first quarter o f 2003. However, the windingdown o f hostilities inIraq and approval by the U.S. Congress in early April of a scaled-down assistance package for Turkey (a grant equivalent of US$l billion, potentially convertible to upto US$8.5 billion inloans) eased some o f the market tension once again. H. MACROECONOMIC OUTCOMESIN2003WERE FAVORABLE The Turkish economy continues to grow at a fast pace. Economic growth reached 5.9 percent in 2003 following 7.9 percent growth in 2002. The major contributing factor to the favorable growth outcome in 2003, because o f its weight inthe national accounts, was private consumption growth. However, showing much faster rates o f growth was private sector capital formation. This augurs well for sustaining growth, with capacity utilization levels reaching historic highs. Exports continuedto play an important role inthe recovery. The current account deficit widened to almost 3 percent o f GNP, but was easily financed by short-term capital inflows, public sector borrowing abroad, and reverse currency substitution. Inflation fell to 18.4 percent in2003, and the latest data suggest that inflation i s falling toward the important single- digit level for the first time since the 1970s. Employment declined in 2003 following public and private sector restructuring, which, together with three years of decline in real wages, helped preserve competitiveness in spite o f strong currency appreciation. Aggregate unemployment remained stable at around 10 percent, but this was helped by a temporary shnkage in the labor force. Inurban areas, the unemployment rate approached 15 percent, and unemployment o f educated youth rose above 30 percent at the end o f 2003. With a trend increase in the labor force o f at least 1.8 percent per year, further reforms are neededto strengthenjob creation. I.TURKEYHAS ACHIEVEDSIZABLE FISCAL ADJUSTMENT Strong fiscal performance has been the comerstone o f the economic program. Fiscal gains were significant in2003, and the primary surplus rose from 4 percent o f GNP in2002 to over 6 percent o f GNP in 2003, close to the programmed 6.5 percent target. Nevertheless, the overall fiscal deficit remained considerable at 9.9 percent o f GNP. Although the 2004 budget passed inDecember was consistent with the 6.5 percent primary surplus target, a sizable fiscal gap quickly emerged. The Govemment announced above-inflation increases in minimum wages, and cut contribution rates for social security to reduce the additional costs to employers. Inaddition, the Govemment increased pensions by 21 percent, well above the inflation target. These initiatives, together with revenue shortfalls relative to the budget, created a " A 10 percent move in the real exchange rate causes a 4 to 5 percentage point adjustment in the public debt-to- GNP ratio. The real exchange rate path is given inTable 11.1, 21 financing gap o f close to 1.7 percent o f GNP. The Government introduced a fiscal package in March to close the fiscal gap. This package has two main components: a supplementary budget and revenue measures. The supplementary budget passed in March cuts discretionary expenditures by 13 percent across all ministries. The Government also introduced measures to increase tax revenues by adjusting excises on petroleum, alcohol, tobacco, and natural gas. While the Government has demonstrated a willingness to undertake action to meet the fiscal target, good public expenditure management and delivery o f services to citizens will require less reliance on ad hoc, short-term measures, and a focus on sustainable fiscal adjustment. Monetary policy followed a policy of implicit inflation targeting, with the Central Bank occasionally intervening inthe foreign exchange market to dampen what was deemed to be excessive fluctuation inthe exchange rate. The decline ininflation, which was aided by the strength o f the TL, led to a commensurate decline ininterest rates from a nominal 60 percent inthe first quarter o f 2003, to about 25 percent early in 2004. Onthe external front, despite appreciation o f the TL, risingproductivity and declining labor costs helped sustain external competitiveness and export growth. Exports grew sharply in 2003. Textile and vehicle exports were best-performing sectors. One encouraging sign i s the growing importance o f new export markets. Iraq has already become a large export market for Turkey, and there was strong growth in exports to China, Russia, and Central and Eastern European countries. Imports also grew rapidly, with oil, increased imports o f machinery and equipment, and rising demand for imported consumption goods being major contributors. Intermediate and capital goods accounted for 86 percent of the increase in imports during 2002 and 2003. Consumption goods also rose sharply in the second half o f 2003, and continued in early 2004. Much o f the increase in consumption goods imports was driven by automobile imports, which benefited from a temporary tax credit on automobile purchases introduced in August 2003. Tourism receipts were maintainedin2003 despite the Istanbulbombings and uncertainty caused by the Iraq war. The current account deficit increased to 3 percent o f GNP in 2003. Continued market confidence has spurred an improvement in capital inflows, although green field investment has remained low. These inflows easily financed the current account deficit and allowed the sharp increase in international foreign exchange reserves to US$33 billion, equivalent to 5 months of goods and services imports. The combination o f strong real and financial market performance had a favorable impact on the public debt burden. The net public debt-to-GNP ratio fell sharply during 2002-2003 from its end-2001 peak o f 93.9 percent o f GNP. Helped by declining real interest rates, strong fiscal performance, the recovery of economic growth and, above all, by the continued appreciation o f the real exchange rate, Turkey's net public sector debt i s estimated to have fallen to about 70 percent o f GNP at end-2003. The decline would have been larger without the issuance o f new debt (of TL 6.8 quadrillion) for the takeover o f the failing Imar Bank in July 2003. With capital flows increasing, and a growing appetite for Turkish Government paper, the Treasury had no problem servicing the debt. Nevertheless, the highrollover rate, 88 percent in 2003, indicates the continued dependence on market sentiment, and thus vulnerability to external developments. J. DETERIORATIONEXTERNALIN BALANCE HAS CREATEDVOLATILITY IN DOMESTIC FINANCIAL INDICATORS The deterioration in global financial indicators in early May 2004, combined with the higher-than- expected current account deficit figures, has led to a sharp weakening in domestic financial indicators. While the TL depreciated about 14percent, taking it to its late-2002 level, the stock market plunged by 18 percent, although from an all-time-high level, and the benchmark Treasury bond rate reached a higho f 30 22 percent by mid-May from 22 percent in early April. Turkey's Eurobond spreads widened by over 200 basis points to 525 basis points during the same period. The excess volatility in the foreign exchange market was curbed to some degree by the Central Bank's intervention. It appears that some relative stability has been achieved in domestic financial markets. These recent developments have underlined Turkey's exposure to shocks from the external environment. While the depreciation o f the TL is an adjusting factor to the deteriorating current account balance, it is also likely to affect inflationary expectations and domestic interest rates. Higher domestic interest rates, together with the impact o f the TL's depreciation, would influence the overall fiscal deficit and economic activity. Spillover from global liquidity tightening, and rising spreads, are likely to increase the cost of external borrowing. Such developments, if persistent, could disrupt the virtuous cycle that the economy has experienced over the last year and a half. K. ONGOINGSTRUCTURAL REFORMS EXPECTED STIMULATE ARE To GROWTH In response to market pressures, the Govemment has shown a renewed commitment to program implementation. Sustaining the momentum o f the ongoing recovery and the confidence o f markets continues to be the challenge facing policymakers. Under stable domestic and international conditions, Turkey could repeat or improve on last year's macroeconomic performance in 2004. Growth should again meet the 5 percent target. Inflation i s already running below projections. Carryover from the strong increase in industrial output in 2003, and a more normal harvest inthe agriculture sector in 2004, should deliver the growth target from the production side. On the demand side, confidence indicators are strengthening, and lower interest rates and easier credit are providing stimuli to private investment and consumption. Despite firm domestic demand, there are strong prospects of meeting the inflation target o f 12percent, which would outperformtargets for the thirdyear insuccession. Medium-term projections demonstrate that sustained implementation o f economic reform is necessary if Turkey i s to attain its macroeconomic stability and growth objectives. Under the structural reform program, the economy is expected to grow about 5 percent during 2005-2006. Specific factors underlying stability and sustained growth include: (a) greater confidence in the policy framework; (b) improved macroeconomic stability and declining real interest rates-which would stimulate private investment and consumption demand; (c) an increase inproductivity resulting from structural reforms; (d) stronger exports performance-which would permit faster import and output growth; and (e) higher external inflows, including sizable foreign direct investment. Under the sustained reform scenario, fiscal adjustment will yield a permanent reduction in the public sector borrowing requirement from about 10 percent o f GNP in2003 to 5 percent o f GNP in2006. 23 CHAPTER111: POVERTYPROFILE In 2002, 27 percent of the Turkish population was poor, as determined by the complete poverty line methodology detailedinAnnex One. This poverty profile is based on that definition of poverty. Poverty inTurkey is strongly associatedwith age and household composition, where children and families with children are poorer than average. This association is robust with equivalence taken into consideration (Ravallion and Lanjouw 1995), because the main poverty line i s one with the adjusted equivalence scale (see Annex I). Other correlates of poverty are standard, but still important-rural location, unemployed household head, female-headed households. Poverty and labor force variables are also interrelated (see chapter on labor). The 2002 Household Budget Survey (HBS) also captured many correlates of the well off, including the ownership of many assets and consumer durables andmany housing attributes. A. HOUSEHOLDAND COMPOSITION SIZE Poverty increasessomewhat monotonically with additional household members (Table 111.1,Figure111.l), starting at three members. There i s a small "blip" whereby households comprised of one member or two membersare slightly poorer than those with three, but there are few such households inTurkey-only 1 percent o f the sample population i s a single-person household, and 8 percent are two-person households. Those households are primarily composed o f elderly, who face a poverty risk higher than average, but still lower than that o f children. Table 111.1. Turkey: Poverty and Household Size 9 47.5 52.5 3.1 2,100,168 3.1 1o+ 55.9 44.1 5.8 4,022,830 5.9 Total 27.0 73.0 100.0 68,393,031 100.0 24 Figure 111.1. Turkey: Poverty and Household Size 1 2 3 4 5 6 7 8 9 1 0 + HouseholdSize Larger households are poorer than smaller households, and this result is driven primarily by the fact that additional household members are more likely to be children, which have a higher poverty rate. This conclusion i s demonstrated by a number o f cross-tabulations (poverty and the number o f children, the number o f elderly, poverty by household composition, and poverty by age, which are presented inTables 111.2, III.3, 111.4, and 111.5 and Figures 111.2, 111.3, 111.4, and 111.5, respectively). Households with no children or only one child had poverty rates sharply below the average, whereas households with three or more children had poverty rates above or substantially above average (Table 111.2, Figure 111.2). The modal household is one o f two parents and two children, and was used inthe poverty line methodology- households with two children have a poverty rate slightlybelow average. Table III.2. Turkey: Poverty and Number of Children inHousehold *The percent ofthose who answeredthe question or for whomwe have data. Notes: Povertyin percentages. PopulationSubtotals: Weighted number ofobservations. Source: DIE 2002 HBS. 25 Figure111.2. Turkey: Povertyand Number of Children Poor +Total Poor 0 1 2 3 4 5 6+ I I HouseholdSize Given the population structure o f Turkey (Table II1.3), it i s not surprising that many fewer households have elderly members (about a quarter) than those with children (79 percent). There i s a correlation between having elderly members and household poverty, but this correlation i s not as marked as with having additional children, and far fewer households have elderly members (Table 111.4, Figure 111.3). Having one elderly member did not appreciably increase the risk o f poverty, and having two or more slightly increased the risko fpoverty, but not as muchas having three or more children. Table 111.3. Turkey: Age Structureof HBS and Census 45-49 2,035,002 1,978,808 3.0 2.9 50-54 1,571,553 1,786,466 2.3 2.6 55-59 495,889 1,8 10,571 0.7 2.6 60-64 401,241 1,615,043 0.6 2.4 65+ 1,280,363 3,533,030 1.9 5.2 Subtotal 33,307,408 35,085,627 48.7 51.3 26 Number of Elderly Percent of Population inHousehold Poor Non-Poor Valid* Subtotals Percent of Total 0 25.9 74.1 73.7 50,394,138 73.7 1 28.0 72.0 17.1 11,684,961 17.1 2+ 33.7 66.3 9.2 6,3 13,935 9.2 Total 27.0 73.0 100.0 68,393,034 100.0 * The percent o f those who answered the question or for whom we havedata. Notes: Povertyin percentages. PopulationSubtotals: Weighted number ofobservations. Source: DIE 2002 HBS. Figure111.3. Turkey: Poverty and Number of Elderly 40.0 __II__~ __llll.__.-lll._l 35.0 -- -; 30.0 - 25.0 -- 3 E I 20.0 - Poor 2 --tTotalPoorII 15.0 -- 10.0 - I 5.0 - 0.0 I 0 1 2+ The finding that having children increases poverty risk more than having elderly members is also demonstrated by looking at the data in terms o f family composition (Table 111.5): whether a family has children, elderly, both, or neither. The latter type o f family i s unusual in Turkey-only about 12 percent of the population lives in such households, while 66 percent o f households have children but no elderly, and only 4 percent have elderly members but no children. Households with no dependents are rarely poor (only a 12 percent poverty rate), while households with both children and elderly are the poorest (38 percent are poor), although there are relatively few such households (under one-fifth). The majority o f the poor is comprised o f families with children (Figure 111.4). 27 Householdswith NoElderlyandNo Percent Population Percentof Children Poor Non-Poor of Valid* Subtotals Total No children, noelderly 11.9 88.1 12.3 7,915,326 11.6 Children. no elderlv 28.5 71.5 65.7 42.478.812 62.1 Elderly, no children 18.2 81.8 4.4 2,865,057 4.2 Bothchildren andelderly 37.6 62.4 17.6 11,345,848 16.6 Total 27.0 73.0 100.0 68,393,033 100.0 * The percent of those who answeredthe question or for whom we havedata. Notes: Povertyinpercentages. PopulationSubtotals: Weighted numberofobservations. Source: DIE2002 HBS. Figure111.4. Turkey: Share ofPoor inPercentby HouseholdConsumption 5.3 No Children, No Bderly Children, No Bderly 2.9 0Elderly,NoChildren [3BothChildren & Bderly InTurkey, poverty is strongly associated with age-younger children are poorer, and active-aged adults are not as poor, and the elderly are poorer than adults, but not as poor as children (Table 111.6, Figure 111.5), except for a peak in the age range 30 to 34, which is the prime age bracket for adults to have children. Table III.6. Turkey: PovertyRateof Age 28 Age Poor Non-Poor Percent of Valid* Population Subtotals Percent of Total 45-49 20.5 79.5 5.9 4,013,810 5.9 50-54 18.1 81.9 4.9 3,358,019 4.9 55-59 21.5 78.5 3.4 2,306,460 3.4 60-64 19.9 80.1 2.9 2,016,284 2.9 65+ 26.6 73.4 7.0 4,813,393 7.0 Total 27.0 73.0 100.0 68,393,035 100.0 Figure 111.5. Turkey: Poverty Rate and Age 40.0 35.0 30.0 25.0 +Poor i 20.0 15.0 l-Total Poor 10.0 5.0 0.0 B. HOUSEHOLDHEAD CHARACTERISTICS The household head has an enormous impact on the poverty status o f her or his household, through the employmenthnactivity nexus and the amount of income she or he can contribute to the household. Employment and labor market variables are covered in the chapter on labor (but it bears mentioning that the poverty rate for unemployed heads was 35.4 percent compared with total (average) poverty of 27 percent), while education i s treated below. Other household demographic characteristics that are associated with poverty include whether the householdi s female headed (Figure 111.6), with a poverty risk o f 32 percent, compared to a male-headed poverty rate o f 26.6 percent. However, few Turhsh households are headed by women-only 6.5 percent o f the sample population lives in a female-headed household. 29 Figure 111.6. Turkey: Povertyand Gender of Head 30 c 25 - I 20 - i =Poor n Z 15 - 10 - i1II +Total Poor __ 5 - 1 0 . ! 1 Male Female The education o f household head is an even more significant influence on household poverty than gender or unemployment for the illiterate and those who didnot receive a primary school diploma (Figure 111.7), with relative risk premia o f nearly 100 percent (meaning that the rate o f poverty o f these kinds o f households i s twice the national average of 27 percent).12 Additionally, nearly 9 percent and more than 7 percent o f the sample population live in households with illiterate or no-diploma heads, respectively. At the opposite end o f the spectrum, n o households with heads with masters or doctorates were poor-but less than 1percent o f the sample population lives in such households, and only 1 percent o f households headed by someone with four years o f university education was poor. Even finishing high school or vocational education was a bulwark against poverty-households with these sorts of heads were only 9 percent and 8 percent poor, respectively. Figure111.7. Turkey: Povertyand Educationof HouseholdHead Poor +Total Poor ''Turkey recently changed the length o f mandatory education from 5 to 8 years, so very few household heads have completed 8 years, because this would make these heads quite young. 30 C. SPATIAL CHARACTERISTICS The 2002 HBS sample was not designed to be regionally representative o f Turkey's seven regions, and therefore no cross-tabulations are presented on the regional level. The 2003 survey, which was recently completed, was designed to be regionally representative. When the 2003 data are cleaned and available, DIEand the WorldBank intendto complete a Poverty Update that will quantify the regional dimensions o f poverty. In the meantime, all indicators, quantitative and qualitative and the 1994 findings, indicate that spatial dimensions o f poverty are highly significant, with Eastem and Southeastem Anatolia being muchpoorer than other regions of Turkey, and Marmara beingthe best-off area. These two areas are also the least urbanized in Turkey, and the 2002 data are representative for rural and urban differences, which are discussed below. There i s a sharp difference inthe poverty rates between rural and urban households, with a poverty rate o f nearly 35 percent for the rural population, but only 22 percent for the urban population (Figure 111.8), while most people live inurban areas (three-fifths o f the sample population). Figure111.8. Turkey: Poverty and Location , 40.0 ~ I 35.0 30.0 Eal 25.0 1 22 20.0 15.0 ~ I 10.0 5.0 0.0 Rural Urban , The drivers for the predominance o f rural poverty are the same as for poverty overall-household composition where larger families with more children predominate in rural areas, limited employment opportunities (agriculture predominates inrural areas and i s the sector with elevated poverty risk, with 70 percent o f the rural employed working in asculture), and the influence o f education. Mean household size in rural areas is 4.30 members, whereas in urban areas it i s 4.07. In both urban and rural areas, household size i s larger for poor than for non-poor families, but the gap i s larger in urban areas (Figure 111.9). 31 Figure 111.9. Turkey: Household Size and Location Total Non-Fbor ILkban Wral Fbor 0 1 2 3 4 5 6 7 I I Number of Members I Rural areas are characterized by limited employment opportunities, and rural households where the head i s unemployed face a substantial risk o f poverty-65 percent o f them are poor. Numerically, however, few households in rural areas have unemployed heads-only 5 percent o f the rural population lives in such households (10 percent o f the urban population have unemployed heads, but their poverty rate is only 26 percent). Other kinds o f inactivity have different implications in rural and urban areas. For example, heads who gave their reason for not working as elderly (over 60) were 55 percent poor in rural areas, but only 34 percent were poor in urban areas (comprising 4.3 and 2.7 percent o f the rural and urban populations, respectively). Housewives were 47 percent poor in rural areas compared to 32 percent poor in urban areas (1.2 and 3.8 percent o f the respective populations). The only kind o f inactivity for representative numbers o f households that had similar poverty rates in rural and urban areas was when the household heads reportedtheir inactivity was due to retirement-1 3 percent poor in rural areas and 11 percent poor in urban areas. However, far fewer household heads were retired inrural areas (only 6 percent o f the population) compared to 14.5 percent o f the urbanpopulation. The major driver for rural and urban employment findings appears to be sector o f employment, where rural location i s dominated by agriculture (Figure 111. lo), which offers few o f the more lucrative formal employment options found in urban areas (Figure 111.11). As demonstrated in the chapter on labor, poverty rates for those employed in agnculture are sharply above those for other, urban sectors. In2001, the Government introduced a Direct Income Support (DIS) program for agriculture, which involved a payment o f about US$90 per hectare to farmers incompensation for the removal o f agricultural subsidies. DIS was estimated to cover about half the income loss to farmers in 2002 for the removal o f subsidies (World Bank, 2004). The HBS did not include any questions about the DIS, so further quantification in this report based on the HBS i s not possible. Interestingly, about 10 percent o f the employed urban population i s active in agriculture, so this sector i s predominately but not exclusively rural. For example, 8 percent o f urban households report that they produce vegetables and livestock for their own consumption, and these households are much poorer (32 percent) than those that do not (21 percent poor). Inrural areas, 69 percent of households produce their own food, but the differences inpoverty rates among those that do (36 percent poor) and those that do not (3 1percent poor) are not as pronounced as for urban households. 32 Figure111.10. RuralEmployment by Sector Agriculture, huntinq, and I Manufacturing I3Construction I Wholesale and retail trade 9.1 W Remainder Figure111.11. UrbanEmploymentby Sector __ I 9.7 Agriculture, hunting, and forestry Manufacturing I 0Construction - 'ElWholesaleand retail I trade I Remainder 22.3 Education appears to have identical effects in both urban and rural settings, whereby those who are illiterate or limited to primary school have sharply higher poverty rates than average, and higher education graduates are much less likely to be poor (Figure 111.12). Inboth areas, poverty rates steadily decrease as years of education increase, and always, at every education level, the poverty risk premium o f rural location persists untilthe very highest levels for which there are essentially no observations inrural areas. 33 Figure111.12. RuralEmployment by Sector 50 0 45 0 40 0 35 0 _. u 300 c ,+Rural Pwr 2 Q 250 p u r a 1Total -- UrbanPwr n Q Z O O "mwU*an Total 15 0 100 5 0 0 0 Nonmaterial aspects of poverty also confirm that poverty is more pervasive in rural areas. In terms o f quality o f diet, the rural poor consume more bread (8.37 units) and therefore less higher-calorie items per week than the urban poor (7.59). D. NON-INCOME ASPECTS POVERTY OF Poverty restricts the poor from accessing many goods and services. The 2002 HBS collected some information on these other aspects o f poverty, which show the typical pattern o f much lower participation o f the poor in the activity, good, or service profiled. Alternatively, there i s information on assets that uniformly shows that assets are distributed to the non-poor, which, o f course, is to be expected since the poor lack the resources necessary to amass assets. This section details the "non-income" aspects o f poverty, whereby non-income i s understood to be shorthand for material items, assets, or services that are ultimately obtained through income. Dwelling place i s typically the most important household asset, followed by land in rural areas and automobiles. InTurkey, following on from the spatial differences, there i s a pronounced difference inthe kind o f dwelling. About half the sample population lives in a house, and another 27 percent lives in apartments. However, while individual houses are primarily inrural areas, they are also still prevalent in urban areas-particularly in gecekondu (slum) areas. However, apartments are almost exclusively an urban phenomenon, and one that pertains to the non-poor. Only 6.5 percent o f apartment dwellers are poor (Figure 111.13), while 36 percent of those who live inhouses are poor (this corresponds to the overall total rural poverty rate). The urban poor are clustered in the gecekondu areas-the poverty rate o f those who report that they live in a gecekondu house i s 35 percent. Virtually all gecekondu dwellings are houses. 34 Figure 111.13. Turkey: Poverty and Dwelling 40 -__ __ I ______l____l___ -___.__ 35 - 30 - 25 + -- E Poor 20-- 2 g 15 - --tTotalPoverty 10 - 5 -- 0 ' I Apartment House ~ J Other housing-related indicators reflect similar findings on the spatial distribution o f the poor, and particularly on the urban poor being locatedingecekondu areas. For example, 7.7 percent of the sample population reported that they had central heating (an urban attribute), but only 2 percent o f these were poor. However, the 86 percent of the population that reported their heating source was a stove had a poverty risk of 31percent, which i s more than the total poverty rate o f 27 percent. Electricity is virtually universal inTurkey; only 0.1 percent of the sample populationreportedthat they livedwithout it-but all of these were poor (too few observations for reliable inference). Only 4 percent of those surveyed reported that they owned an additional home, apartment, or summer resort, or that their home was a "luxury" one, but o f these, very few were poor, with average poverty rates inthe single digits. Conversely, the use ofdung for heatingwas strongly correlated with poverty (Figure 111.14). Figure 111.14. Turkey: Poverty and Use of Dung --! 45 -- ! 40 - 35 - I I $ - , - - 30 e I -Poor 2 20 - I 15 - I 10 - 5 - No Yes 35 Land is an important asset in Turkey, particularly for the poor. Twenty-seven percent of the sample population reportedthat they had a field, but these households were poorer than average, with a poverty rate of 34 percent. The mean size of the fields for poor households was 75 percent the size o f the non- poor fields. More than one-fifth (23 percent) of the sample population reported owning a car, but only 6 percent of these were poor. Other consumer durables showed similar results. Table 111.7. Turkey: PovertyRate of ConsumerDurables Water heater 11.9 88.1 49.1 33,594,478 49.1 Aspirator 5.5 94.5 22.7 15,535,335 22.7 The poor are less able to afford discretionary expenditures-the poverty rates o f those reporting that someone in the household smokes, drinks, takes course in computers or foreign language or driving, attends nursery schoolkindergarten, takes public transportation to school, dines out, reads newspapers, goes to moviedtheatre, habituates cafes, plays the lottery, or takes medicationregularly are all lower, and often substantially lower, than the overall poverty rate of 27 percent (Figure 111.15). However, the participationin each of these activities vanes greatly, from a highof 67 percent of the sample population reporting that someone smokes, to a low of 1.6 percent for nursery schoolkindergarten participation. Fifteenpercent of households reported use o f public transportation, but more non-poor than poor accessed this-the poverty rate for usage was 17percent compared to the total poverty rate o f 27 percent. 36 Figure 111.15. Turkey: Poverty and DiscretionaryActivities Shopping patterns vary between the poor and the non-poor because the latter are more likely to shop at markets or bazaars. The poverty rate o f those who do not shop at markets or bazaars i s higher than average (Figure III.16), while very few poor have access to a credit card-only 21 percent o f households reported usinga credit card, but o f those, the poverty rate was only 5 percent. Figure 111.16. Turkey: Poverty and Shopping 40 -- "_-- . ~ I I----- ! 35-- 30 - 4 7 A c 25 C -- - Q) 2 20 - -Poor 0) I +Total Poor n. 15 I 10 .- 5 - 0 - I Market- Market--no Bazaar- Bazaar-no i Yes Yes E. INEQUALITYAND REGIONAL DIFFERENCES Turkey i s a high-inequality country, and its inequality has not improved since 1994 (Table 111.8). Inequality, as measured by consumption from the HBS, does not change much once the same standards in constructing the consumption aggregate and equivalence scales are applied to the 2002 data, as were used by the World Bank with the 1994 data (see Annex I). 37 Table 111.8. ComparableInequality ~ 1994 2002 Change, Times Inequality, Gini for 1994 consumption per 0.385 0.390 1.01 equiv. adult* Inequality, Gini for 1994 consumption per 0.408 0.413 1.01 * 1994 equivalence ~~ scale (nutrition based) However, in this report, a new methodology for consumption was used, which excluded some o f the variation inconsumption o f the old methodology interms o f handling o f consumer durables (Table 111.9). The new methodology shows no change in income inequality, but a lower level o f consumption inequality. This lower level o f consumption inequality is not due to any improvement in the distribution, butis simply the result o fthe new methodology. Table111.9. New Methodology InequalityMeasures Both indexes indicate that inequality i s higher in urban areas than in rural areas, but not by much. As i s typical, inequality based on income i s greater than that based on consumption. Turkey is a middle-income country, and its inequality i s high. The worst inequality i s in the poorest countries (Milanovic, 200l), the middle-income countries have medium levels o f inequality (but Turkey i s relatively high), and the lowest inequality i s among the best-off countries (except for the outlier of the United States, with highincome inequality). Inprevious World Bank research and the Turkishliterature, the major driver for Turkey's highlevel of inequality i s regional differences, with urban-rural differences playing a secondary role. Eastern and Southeastem Anatolia are markedly poorer than the rest o f the country. This has been the case for Turkey for a very long time, and certainly as documented duringthe past 30 years. This finding is corroborated by every major study on poverty in Turkey, including the Government's latest Preliminary National Development Plan (2003), studies undertaken by the Bank in its previous Poverty Assessment and Poverty Update (World Bank, 2000, 2003), the United Nations Development Programme inits "Human Development Reports," the Turkish academic literature (for example, Ogut and Barbaros undated), reports from the State Planning Organization, and numerous qualitative assessments undertaken for World Bank projects inTurkey over the past decade. The Government's most recent study, the Preliminary National Development Plan (2003) provided several figures-Figures 111.17 through 111.19-ranking Turkey's regions on a variety o f indicators, including GDP per capita and other social and human development indicators. Inall these, much greater poverty inEastern and Southeastern Anatolia i s demonstrated. Table 111.10, on per capita GDP, ranks the 26 regions o f Turkey from highest to lowest, and Table 111.11 provides social and human development 38 indicators by region. Note that inFigure 111.18, the index o f socio-economic development used is lowest for the best-offregions and highest for the worst-off. Table 111.10. Per CapitaGross DomesticProduct,by StatisticalRegions, 2001* Statistical Regions US% Turkey 2,146.00 Source: DIE. *At Current Prices Ranked inDescending Order - 39 Table III.ll. Turkey: Social and Human Development Indicators 1997 ~ Under-weight Adult LifeExpectancy Children Literacy at Birth(years) Under 5 (YO) Rate (YO) Marmara 71.59 6.20 89.29 Aegean 71.53 7.25 85.16 Black Sea 67.22 9.29 81.45 Central Anatolia 65.52 10.54 84.60 IMediterranean I 66.63 I 11.29 I 84.28 I Southeastem Anatolia 65.89 17.00 62.34 EastemAnatolia 62.05 19.00 70.99 Turkey 68.90 10.00 83.20 F. MULTIVARIATE ANALYSIS The multivariate analysis confirms almost all o f the findings o f the bivariate analysis presented above in poverty cross-tabulations. A probit analysis demonstrates that the probability of the household's being poor i s correlated with household demographics, employment status o f the head, education o f the head, and location. In Table 111.10, coefficients with negative signs decrease the probability of the household being poor, and coefficients with positive signs increase the chance of the household being poor. The coefficients themselves represent the percent change in poverty likelihood. For example, rural location increases the chance that the householdis poor by 9 percent, controlling for the other factors. Table 111.12. Probit Estimates Number o f obs = 3,555 LR chi2(15) =1821.66 Prob > chi2 = 0.0000 Log l i k e l i h o o d = -4041.3858 Pseudo R2 = 0.1839 ....._........__________________________~- POcPla I dP/dx Std. Err. z P > l z / x-bar I 95% C . I . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . hunemp*/ ,068221 .0169959 4.37 0 . 0 0 0 ,083203 .03431 ,101532 dhigher*/ -.1346925 ,0125338 -6.53 0.000 ,090215 -.153258 -.110127 dprimar*/ ,0938267 .0089331 3.82 0.000 ,640502 ,076201 ,111453 dsocsec*/ -.1123116 .0086383 -12.52 0.000 ,428362 -.123361 -.095262 dsecjob*/ -.0054283 .0131209 -0.28 0.779 ,045107 -.042905 ,032047 elder65 I ,0077759 ,0078997 0.98 0.325 ,223653 -.007707 ,023259 adult -.0250377 ,00369 -6.78 0 . 0 0 0 2.45313 -.03227 -.017805 c h i l d 11 ,0667864 ,0026853 25.73 0 . 0 0 0 1.57415 ,061523 .072049 unpaid*l ,0049726 .0158795 0.32 0.752 ,106541 -.026151 ,036096 r u r a l * / ,096574 .0143081 7.42 0 . 0 0 0 .153218 ,068531 .124617 femalhd*/ -.0043401 .0133545 -0.32 0.747 . l o 0 5 7 6 -.030514 ,021834 s e l f * / .012003 ,0111668 1 . 0 3 0.277 ,241444 -.009878 .033896 employer*/ - . l o 4 8 3 4 ,0134159 - 5 . 5 8 0.000 ,052747 -.131123 -.078539 casual*/ ,1033531 ,0168735 6 . 8 8 0 . 0 0 0 ,084458 .070282 .136425 temporar*i .0436523 ,0153984 2.90 0.004 ,078074 ,012297 .075003 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . obs. P ~ ,2132915 pred. P 1 ,1648576 ( a t x-bar) .............................................................................. ( * ) dFjdx i s f o r discrete change o f dummy variable from 0 t o 1 z and P > / z i are the t e s t o f the underlying c o e f f i c i e n t being 0 40 The greatest impact on reducing the probability o f poverty i s when the head o f household has higher education (reduces poverty by 13 percent), followed by whether the head has social security (in other words, i s formally employed), which reduces the chance o f poverty by 11 percent. The role o f formal employment i s also demonstrated in the series of dummy variables relating to employment status o f the head (whereby permanent employment i s the omitted variable) because casual or temporary worker status increases the probability o f the household being poor, but the category of employer reduces the risk o f poverty by 10percent. Factors that increase the probability o f poverty, other than rural location, include the number o f children inthe family (each additional childincreases the chance ofpovertybynearly 7 percent). The coefficient estimates for four variables were not statistically significant (elderly, self-employed, female-headed households, and unpaid family members), probably because they are relatively few innumber. Figure 111.17. Per Capita GDPIndex Nomenclature of Statlstlcal Territorial Unltsll Regions ___-___ 01987-1989AW3'.AGES 1990-1994AmGES 019952000A~GES Figure 111.18. Socioeconomic Development Levels (2003) Nomenclature of StatisticalTerritorialUnits II Regions 41 Figure111.19. FemaleLiteracyRates(2000) Nomenclatureof StatisticalTerritorialUnits II Regions 42 CHAPTER IV: EDUCATION A. THECONTEXT In1923,theyear inwhichtheRepublicofTurkeywas founded, the adultliteracyratewasapproximately 10 percent. Such a low starting point interms o f human capital (not uncommon for that era) is certainly undesirable, but this allowed the founders of the Republic to carry out major reforms that might otherwise have been more difficult to implement. Indeed, the key features of the Turkish education system were established inthe early years o f the Republic. The Tevhid-i Tedrisat law of 1924marked a move toward secular education, and in 1928, the Latin alphabet was adopted. In 1961, the duration o f primary schooling was increased from three years (inrural area schools) to five years.13 The idea of 8-year primary education first emergedin 1970, with the establishment of a working group to formulate the specifics of the proposed changes. This led to a 1973 law (1973 sayili Milli Egitim Temel Kanunu), which declared the duration of basic education to be 8 years. Inpractice, however, the current structure of the Turkish education system was established during the 1997-1998 school year, with the increase in compulsory schooling from 5 years to 8 years for children aged 6 to 14. Upon completion of the 8-year primary school cycle, students may enroll ingeneral or vocational secondary schools, which i s three years induration. Tertiary schooling i s providedby universities. A key implication of the 1997-1998 education reform was the phasing out o f lower-secondary (general and vocational) schools that served grades 6 to 8. This reform was controversial, since both lower- secondary parts of vocational schools that offered religious training and other lower-secondary schools had to be closed. The reform, however, also resulted in a renewedpublic and private commitment to the achievement o f universal enrollment levels inbasic education. Onthe public side, extrabudgetary sources were channeled into education to cope with increased enrollments. Furthermore, sizable resources were obtained from the World Bank (through the June 25, 1998 Basic Education Program Loan and other projects that followed), the European Union, and other sources. On the private side, the "build a school" campaign enjoyed a boost, and several relatively large organizations emerged to collect donations with the objective of supporting the schooling ofpoor children. Enrollment rates increased soon after the reform. Duringthe 1995-1996 school year, the primary school enrollment rate was 89.8 percent; by the 1999-2000 school year it had grown to 97.6 percent. For secondary school, the enrollment rate increased from 55 percent to 59.4 percent during the same period. After the start o f the reform inthe 1997-1998 school year, enrollments inprimary schools increased from 9.1 million students to over 10.3 million students inthe 2001-2002 school year. A substantial school and classroom construction effort dominated this period, and the number of classrooms increased from 210,905 to 280,257. Furthermore, over 70,000 new primary school teachers have been recruited since 1997.14 Other developments that took place in 2000s include equipment o f some schools with ICT and improvement of the teachers' competencies especially inthe disadvantaged regions. A support campaing to increase the schooling o f girls i s also under implementation. For the 2001-2002 school year, the numbers of schools, students, and teachers by education level and urbadrural residence are depicted in Table IV.~.'~ l3A detailedhistoricaloverview can be found inTurkiye I s Bankasi (1999). l4A comprehensive review o f the events and issues that characterized the Turkish education system in the 1990s until2004 canbe found inDulger (2004). l5Extracted from TC Milli EgitimBakanligi (2001). 43 Table IV.1. Number of Schools, Students, and Teachers inthe 2001-2002 School Year Public Schools Schooling Number of Schools Number of Students Number of Teachers Level Total Urban Rural Total Urban Rural Total Urban Rural Pre- 10,554 7,361 3,193 256,392 216,625 39,767 14,520 12,579 1,941 Primary Primary 34,993 9,906 25,087 10,310,844 7,500,373 2,810,471 375,511 256,272 119,239 Secondary I I I I I I I I I Total 6,065 5,463 602 2,312,271 2,212,658 99,613 138,785 131,482 7,303 General 2,637 2,276 361 1,490,376 1,427,845 62,531 72,609 68,092 4,517 Vocational 3,428 3,187 241 821,895 784,813 37,082 66,176 63,390 2,786 University 53 1,156,9 15 63,029 Schooling Number of Schools Number of Students Number of Teachers Level Pre- I 799 18,152 1,822 Prima ~ 642 171,623 14,811 Secondary Total 487 73,136 8,229 General 465 72,05 1 8,128 Vocational 22 1,085 101 University t 23 55,022 3,721 The TurkishGovernment originally planned to increase compulsory schooling once again, from 8 to 12 years, fairly quickly. This policy proposal seems to be on hold, although it i s very much a part o f the long-term agenda,16 and thus an improved understanding o f the implications o f the previous reform (of 1997-1 998) for the poor would provide guidance for further modifications to the compulsory schooling law. This report aims to do that by describing the positive developments inrecent years. It also identifies key problem areas that emerge as challenges to the schooling o f poor children. Finally, the findings of this study will serve as inputs into the ongoing Turkey Education Sector Study titled, "Sustainable 16See, for example, TC MilliEgitimBakanligi (2001). 44 Pathways to an Equitable, Effective, and Efficient Education System," which aims to build a technical foundation for policy analysis, and provide an open forum for a technically baseddialogue inTurkey. B. PUBLICSPENDINGONEDUCATION How Does Turkey Compare with Other Countries? Turkey's public spending on education significantly increased after 1998, both in real terms and as a percentage of gross domestic product (GDP) (Figure IV.1). However, even after such a dramatic increase, as o f year 2000, Turkey's public spending on education as a percent o f GDP was slightly less than (but comparable to) that o f lower-middle-income countries (the category into which Turkey falls) (Figure IV.l)." FigureIV.l. PublicSpendingon EducationasPercentageof Gross DomesticProduct 2 - 1990 1991 1992 1993 1994 1995 1998 1999 Yea, -e- Turkey t i o w i w o m s +Lauwmlddbiwome - - X - l J p ~ r m l d d l ~ i n s m s-Hgh i n 4 I Source: World Development Indicators (WDI). Years 1996and 1997are not included since the WDIdata set does not provide spending figures for those two years. Distribution of Public Resources Across Different Levels of Schooling Interms of distribution ofpublic expenditure on education across different levels, the year 2000 figures from the World Bank's EdStats data base suggest an emphasis on primary schooling-about 49 percent o f public funds are allocated to primary schooling, where the public retums are high and from which the poor are most likely to benefit. The same source reveals that secondary schooling receives 20 percent of public expenditure and 31percent goes to tertiary education. The low share o f spending on the secondary level might signal a problem, and this is a topic that will be examined indepth by the forthcoming Turkey Education Sector study which will focus on spending trends inmore recent years. "ThetotalspendingoneducationinTurkeyisestimatedas7.31percentofGDP,bythe"Research onTurkey's Spending on Educationin2002" report publishedby the State Institute o f Statistics in2003. The same document reveals that public spending on education comprised o f 63.5 percent of the total (i.e., 4.64 percent o f GDP in2002). 45 Incidence ofpublic Spendingon Education The expansion of compulsory schooling to 8 years hadan extremely positive impact on the distributionof public education spending across poor and rich households (Table IV.2). In 1994, only 15.8 percent of public spending on basic education reached the poorest 20 percent of households. In 2001, 21.7 percent of public spending on basic education reached the poorest 20 percent of households, because the enrollment rates o f children coming from poor households increased substantially. Table IV.2 also shows that while some pro-poor redistributionoccurred inthe distribution of public secondary education funding between 1994 and 2002, much remains to be done: only 13 percent o f public secondary school spending reached the poorest 20 percent ofpopulation in2001 (up from 8.7 percent in 1994). These estimates of incidence of public spending on education are obtained by imposing certain assumptions. One such assumption i s that the public schools attended by poor children, on average, receive the same amount o f fundingas the public schools attendedby wealthy students. Ifpublic schools with higher poor-student presence receive less fundingon average, then this meansthe figures reportedby Table IV.2 overestimate the amount of resourcesthat reach the poor. The figures for 1994 come from the education section o f the Turkey Public Expenditure and Institutional Review (World Bank, 2001), using student enrollment information from the 1994 Household Expenditures Survey, which does not identify enrollment status in private schools, and ascertaining level of schooling currently attended has to be determined by certain assumptions (described fully in the Turkey Public Expenditure and Institutional Review). These limitations (that is, lack o f information on private school attendance and imprecise school enrollment information) do not apply to the reported 2001 figures, which are constructed using enrollment data from the 2001HouseholdConsumption and Income Survey (HCIS). Table IV.2. Incidence of Public Spending on Education in 1994 (before the expansion of compulsory schooling) and in 2001 (after the expansion of compulsory schooling) 1 2 3 4 5 (poorest) (richest) Basic education(8 years, primary + middle) 15.8% 21.1% 22.2% 20.6% 20.3% Secondary education 8.7% 16.2% 22.3% 25.4% 27.5% Total public expenditures 13.5% 19.5% 22.2% 22.2% 22.7% Primary 25.4 22.4 20.0 18.4 13.7 Secondary 14.0 17.4 21.3 23.3 24.0 Tertiary 4.5 10.0 14.4 25.5 45.5 Basic education(8 years) 21.7% 21.4% 21.0% 22.0% 13.9% Secondary education 13.0% 14.6% 25.4% 22.8% 24.2% Total public expenditures 19.2% 19.4% 22.3% 22.2% 16.9% database, and being categorized as a lower-middle-income country by the 2003 WDI. These are Ecuador, Guyana, Jamaica, Morocco, Peru, Romania, and South Africa. None of these countries, when considered alone, provides an adequate reference point when comparedwith Turkey. As a group, however, they provide some insights into the general circumstances in lower- middle-incomecountries. 46 Inorder to putthe outcome ofthe incidence analysis incontext, the middle segment ofTable IV.2 reports average statistics for other lower-middle-income countries. Such a cross-country comparison suggests that prior to the expansion o f compulsory schooling, Turkey was at an extreme when it comes to distribution o f public resources in a way that benefited the wealthier households. After the implementation o f 8-year compulsory schooling, the distribution of public funding became more in line withthe experience of other countries at about the same level of economic development. Incidence of Household Spending on Education The distribution o f household expenditures on education is even more unequal (Table IV.3). Only 2.2 percent o f total household expenditures on education was by poor households in 1994. After the expansion o f compulsory schooling, as more children from poor households started participating inbasic education, the poor households had to spend more, and thus education expenditures' share in total household spending increased to 6.2 percent. Regardless, household-level spending strongly reinforces the existing differences inthe schooling environment experienced by poor children. Household Income Quintiles 1 2 3 4 5 (poorest) (richest) 1994 2.2% 7.1% 9.4% 18.0% 63.3% znni 6.2% 14.1% 16.6% 23.5% 39.6% c. TRACES OF INEFFICIENT RESOURCEALLOCATIONINTHE EDUCATION SECTOR A full discussion of inefficiencies in the education sector is beyond the scope of this paper-this topic will be covered in detail by the ongoing education sector study titled, "Sustainable Pathways to an Equitable, Effective, and Efficient Education System." Among the topics that are important but left out are ideal classroom size, teacher characteristics, teacher compensation, and school amenities. Also, in Turkey, there would be large payoffs to an elaborate evaluation o f the short-term and longer-term effectiveness o f information technology investments in schools to identify which groups benefit most from computer availability inschools, and under what conditions. ClassroomSize and Student-Per-Teacher Ratios in Primary, General, and VocationalSecondary Schools Inrecent years, the number of primary school students has increased significantly (FigureIV.2). The number of students enrolledingeneral secondary schools increased slightly and enrollments invocational secondary schools has remained about the same. Inresponse to (and to enable) rising primary school enrollments, the number o fprimary school classrooms increased sharply after 1998 (Figure IV.3). 47 FigureIV.2. Number of Studentsby School Type 12000000 ~ 4000000 2000000 -B I b 0 -~ 1997-1998 1998-1999 I9942000 2000-2001 2001-2002 2002-2003 SchoolYear I +Primary -General high school -Vocational high school Source. Ministry of National Education Statistical Yearbooks. FigureIV.3. Number ofClassroomsOver Time 300000 ___- -~ I 0 - - ~- - 1997-1998 1998-1999 1999-2000 2000-2001 2001-2002 2002-2003 School Year +Primary +General _hlSh school -*VocaJonal ~ high s c h L K Source Ministry ofNational Education Statistical Yearbooks. 48 Starting with the positive development that the increase inthe number o f classrooms at the primary cycle has been impressive inrecent years, this has prevented increases in students per class at this level despite increased enrollments due to the expansion o f compulsory schooling. Infact, the number of students per classroom at the primary level has declined from 43.2 inthe 1997-1998 school year to 36.9 inthe 2002- 2003 school year. The cohorts affected the most by the expansion of compulsory schooling are candidates for enrolling in secondary schools. The demand for general secondary schools has increased-enrollment rose from 1,094,610 in the 1998-1999 school year to 1,588,800 in the 2002-2003 school year (excluding "open" general secondary enrollments). As a result, average classroom size increased steadily after 1998, and peaked at 33 in the 2002-2003 school year. Incontrast, average number o f students per classroom i s 27 for vocational secondary schools. Similarly, in vocational secondary schools the student-teacher ratio was only 14 inthe 2002-2003 school year, compared to more than 20 for general secondary schools. As a result, Figures IV.4 and IV.5 reveal misallocation of scarce resources at the secondary level, but they also show very encouraging results at the primary level. Figure IV.4. Number of Students Per Classroomby School Type 1997-1998 iwa1999 19942030 2m2001 2001-2002 2002-2003 _ _-General ~ ~ SchW Year -Pnmaw hylhschoolsvocatoonal highschool Source: Ministry o fNational Education Statistical Yearbooks. 49 Figure IV.5. Student-Teacher Ratios in the 1997-1998 and 2002-2003 School Years 30 25 ; Q 20 c t - 15 2 U 0 10 5 1997-1998 2W2-2W3 SchoolYear ' OPrimaw mGT&fil high s c b l mvocatlona hlghsmoOl Source: Ministry ofNational Education Statistical Yearbooks. In2002, about 32percentofsecondaryschool students were enrolledinvocational schools. Theintention i s to increase the share of vocational school students even more (TC Milli EgitimBakanligi ZOOl), even though there are unresolved questions about the efficiency o f and demand for vocational secondary education. D. DISTRIBUTIONAL AND EDUCATION-QUALITY IMPLICATIONSOF PUBLIC POLICIES ON EDUCATION The question, "who suffers from inadequate public education funding?" is not that meaningful, since a policy that has such large effects on economic growth affects everybody. An alternative approach would be to consider "who suffers from inadequate public education fundingthe most"? Inthe Turkishcontext, two leading contenders are (a) females, and (b) poor children. Despite a consistent decline in the percentage o f illiterate adults, as of 2002, 22 percent o f female adults in Turkey were illiterate (down from 33 percent in 1990). For males, the percentage o f illiterate adults declined from 11percent in 1990 to 6 percent in2002.'' The situation of the poor is described indetail below. Characteristicsof Children Not Enrolled in (Compulsory) Primary Education Although it might have been natural to focus on pre-primary enrollments first, this section starts with an analysis of determinants of enrollment inprimary education. The reason i s because o f extremely low pre- primary enrollment (and lack o f relevant questions inhousehold surveys). UNESCO estimated that only 6 percent of childrenof the relevant age group were enrolled in pre-primary school in 2000. There are 9,225 public and 432 private pre-primary schools in Turkey (1,657 are categorized under "other") (MinistryofNationalEducation, 2003). '* The source for these statistics is the 2003 World Development Indicators database. 50 According to the 2002 HBS, 97 percent of the relevant age group (ages 6 to 14) are either enrolled in school or have already completed the primary cycle. The Turkish Government has attempted to increase compliance with the compulsory schooling law, by providing financial support to the poorest households who cannot afford to send their children to school-under both ad hoc social assistance and as supported by the World Bank under the conditional cash transfer component o f the Social Risk Mitigation Project -and also, in some cases, by fining parents who might be unwilling to send their children to school for one reason or another (the reasons mightbe unwillingness to send girls to school inrural areas, the desire to supplement household income through child labor, and so forth). Table IV.4 displays the characteristics o f the "3 percent" that i s left out o f the schooling system. The distinguishing features o f these children would be useful in targeting efforts to achieve universal enrollment inprimary school. The age distribution o f children who are not currently attending primary school rules out the possibility that many o f these students are simply late enrollers: those age 9 or older make up more than 90 percent of all children who are primary-school age but who are outside the schooling system. About 50 percent of the children who are not attending school reside in Eastern Anatolia or South Eastern Anatolia. However, the 2002 survey is not representative at the regional level, and thus there may be some error in regional estimates. More than 70 percent o f the children who are not enrolled in primary school are female, and more than 55 percent have illiterate mothers. These children tend to reside inrural areas (67 percent) andthey tend to be poor (53 percent). TableIV.4. Characteristicsof ChildrenAged6 to 14, byPrimarySchoolAttendance Status (reportednumbersare percentages) Attending or Already Not Attending and Never Completed Completed (compulsory) (compulsory) Primary Schooling Primary Schooling GENDER Male 51.82 27.79 Female 48.18 72.2 1 MOTHER'S SCHOOLING 51 200,000) Rural 60.02 66.68 HOUSEHOLD POVERTY STATUS 33.69 53.15 Poor Non-poor 66.3 1 46.85 Sample size 6,587 205 It might seem difficult to identify a small proportion o f children, such as 3 percent, but for households with certain characteristics, the percentage o f children who are not enrolled in school is much higher. Parental schooling and gender o f the child alone may not be good enough to narrow down children who are not in school, butjointly, these two indicators are adequate. Table IV.5 shows that if one focuses on fathers with no schooling, in almost 1 out o f 10 cases, a child (aged 6 to 14) who does not attend school would be identified. A systematic intervention that reaches all mothers and fathers without primary schooling would identify roughly 70 percent o f children who should attend primary education but who do not. If these children attended school, the primary enrollment rate would increase from 97 percent to 99 percent. Table IV.5. Primary School Attendance by Parental Schooling Percentageof ChildrenAged 6-14, Who Do Not Attend School (andwho havenot already completedprimary) Both Genders Male Female Father didnot complete primary 9.1% 5% 12.5% (illiterate or literate without diploma) Mother did not complete primary 6.1% 2.7% 9.7% (illiterate or literate without diploma) School availability inthe residential area might be one reason for non-enrollment. This issue seems to be especially relevant in less-developed areas for female children, even if transport to the nearest school is arranged by the local authorities. The 2001 HCIS data reveal that 88 percent o f households have a primary school in their residential area. Primary school availability does not vary much by wealth, as measured by a household possessions index. In fact, if anything, the distribution o f primary school availability is pro-poor (about 91 percent o f the least-wealthy households report the availability o f a primary school; about 85 percent o f wealthiest households report availability). This trend i s primarily driven by urban versus rural residence: 87 percent o f urban households reveal primary school availability in their neighborhoods, while in rural areas this percentage increases to 95 percent. These statistics suggest that the school availability issue at the basic education level has been more or less resolved. In urbanareas, even ifa primary school i s not available inthe immediate residential area, there would be one inan adjacent neighborhood. For small rural villages, on the other hand, access to schooling is through government-subsidized transportation o f children to nearby villages or towns. Table IV.6 presents the average distance to primary school, by residence and household wealth, conditional on school availability inthe residential area. The mean distance to primary school is higher inrural areas and for the poor, but 52 the differences are not large. Having said that, long distances might be more significant in rural areas becauseo f physical barriers such as streams, hills, andpoor road conditions. Table IV.6. Conditionalon SchoolAvailability inthe ResidentialArea, EstimatedDistance Residence Wealth Quintiles Access Urban Rural All (basedon a householdpossessionsindex) Group 1(poorest) Group2 Group 3 Group 4 Primary School 300 500 300 300 300 300 300 [445] [600] [475] [5 141 [4721 ~4281 ~4861 WhatHappens After Eight Yearsof Primary Education? As shown, the demand for (noncompulsory) general secondary education has increased in recent years. This i s likely to be one of the positive side effects of 8-year compulsory schooling, since it increased the number of potential secondary school students, and also increased the amount of time parents interact with teachers and the schoo1,system. A convenient way to present the impact o f the main factors that influence whether a child enrolls in secondary school is through the presentation of marginal effects on enrollment, as estimated by a probit model (Table IV.7). This is a reduced form approach, takmg into account only the following: child's gender, proxies for household wealth, indicators for parental schooling and presence in the household, and dummy variables for urbadrural residence and secondary school availability in the residential area. These models are estimated with and without school availability indicators, and also separately for males and females. The key insightsfrom this exercise are as follows: 0 Taking into account other factors listed above, being a male increases the probability of secondary school enrollment by 7 to 8 percentage points. 0 Compared to the least-wealthy group of households, the next wealth category increases enrollment probability by roughly 7 percentage points. This figure increases to over 11percent for other wealth categories. 0 When other factors are taken into account, mother's and father's schooling coefficients are not statistically significant, except for the category "more than primary," which i s associated with a more than 10 percentage point increase in enrollment probability. (If both the mother and father have more than primary schooling, this translates into an increase o f over 20 percentage points in the probability of enrollment, which is quite substantial initself). 0 Mother's absence from the household severely reduces the chances o f secondary school enrollment-by almost 16 percentage points. A peculiar result i s the statistically insignificant coefficient estimates for father's absence." l 9This might bebecause we focus on the schooling o fhouseholdhead and his or her spouse's children. A selection bias might be responsible for this peculiar finding: it could be that if the father is not a part o f the household (dead, divorced, and so forth), only women with certain unobserved (but pro-child education) characteristics become household heads. Inother words, some women separated from the father o f their children may become household members in an extended household, and the schooling status o f these women's children is not captured in this estimate (because the household survey questionnaire allows only household head's children to be matched to their mother and father). 53 0 Secondary school availability in the residential area i s a very strong predictor o f enrollment, leading to a 10 percent increase in probability o f enrollment. Those who reside in urban areas reap another 5 percent increase.20 Estimating separate models by gender allows evaluation o f the differential effects of certain household characteristics on the enrollment o f boys and girls. Occurrence in the wealthiest group o f households perfectly predicts secondary school enrollment for boys. Having a mother with more than primary schooling is more important for girls' schooling than for boys', but the differences are minor in models that control for school availability. The importance o f father's more-than-primary schooling for girls' enrollment i s quite pronounced though, increasing the probability o f enrollment by 17 to 20 percentage points in alternative specifications. Mother's absence hurts girls' schooling prospects more. The impact of secondary school availability is gender neutral. Urban residence does not have a statistically significant influence on boys' enrollment, but it i s crucial for girls' enrollment (giving a boost o f 21 to 26 percentage points in the probability o f enrollment). The presence o f such a large urbadrural residence effect-even after taking into account parental schooling and secondary school availability-deserves careful consideration. Some possibilities that might explain this trend are demand for girls' time in farm work, and the prevalence o f teenage marriage. Table IV.7. Predictorsof ContinuedEducationafter Completionof %Year Primary School WEALTHGROUPS MOTHER'S SCHOOLING 2oWhen high school availability is excluded from the model, the "urban residence premium" increases to 10percent, partly capturing the fact that high schools are more likely to be available in urban areas. While availability of secondary schools has a clear impact on enrollments, it should also be recognized that in some rural areas with sparse population distribution, alternative approaches may be cost efficient (these include subsidized bus service to schools or the establishment o fboarding schools etc.) 54 Means Both genders Males Females & Model1 Model2 Model3 Model4 Model5 Model6 Std. Dev. More than primary .115 .099 .129 .121 .155 .124 .27 (3.11) (2.74) (2.16) (2.06) (2.22) (1.78) (.44) FATHER'S SCHOOLING ----- ----- ----- ----- ----- No schooling _---- .14 (.35) Some primary .039 .030 .040 .041 ,070 .035 .15 (1.12) (0.87) (0.75) (0.80) (0.94) (0.47) (.35) Primary graduate .049 .036 .005 -.010 .127 .lo4 .29 (1.55) (1.20) (0.10) (0.20) (2.05) (1.77) (.45) More than primary .117 .lo2 .lo1 .086 .196 .169 .42 (3.13) (2.88) (1.81) (1.59) (2.69) (2.44) (.49) Mother absent fromhousehold -.159 -.158 -.152 -.156 -.226 -.238 .07 (3.05) (3.05) (2.15) (2.22) (2.00) (2.02) (.26) Father absent from household .048 .032 .011 -.009 .110 .081 .06 (1.19) (0.78) (0.16) (0.12) (1.51) (1.07) (.24) Secondary school availability inthe residentialarea (1ifavailable, 0 otherwise) ----- .099 ----- .133 ----- .126 .66 (4.15) (3.47) (2.74) (.47) Urbanresidence .loo .052 .011 -.034 .259 .209 .81 (1 ifurban, 0 ifrural) (3.40) (1.91) (0.29) (0.94) (4.25) (3.46) (.39) LogLikelihood -261.3 -244.2 -123.3 -116.4 -128.9 -117.7 Sample Size 792 768 355 346 343 332 792 Since secondary school availability emerged as an important determinant o f enrollment that can be influencedby policymakers inthe short run(as opposed to parental schooling, for example), the next step i s to see the extent to which certain households are disadvantaged when it comes to access to secondary school education. Overall, there is significant room for improvement when it comes to secondary school availability. Only 64 percent o f households reported that a secondary school is available in their residential area. As opposed to primary school availability, secondary school availability i s correlated with household wealth: 59 percent o f the poorest quarter of households reported a secondary school available in their area, while 69 percent o f the wealthiest did the same. Only 50 percent o f rural households reported availability, while 67 percent o f urban households did so. This situation i s further complicated by the presence of different types o f secondary schools (many vocational secondary schools with very low student-teacher ratios) that may not be attractive to large portions o f the population. Finally, Table IV.8 shows that $a school is available in the residential area, then the distance to the school does not vary much by urbadrural residence or by household wealth. 55 Table IV.8. Conditionalon SchoolAvailability inthe ResidentialArea Residence Wealth Quintiles(basedon a householdpossessions Access Urban Rural index) All Group 1(poorest) Group 2 Group 3 Group 4 SecondarySchool 500 400 500 400 500 500 500 [733] [526] [703] 16821 [7331 [6221 [7861 Universitiesand the Poor Poor children are unlikely to receive even a secondary education. Public spending pattems would, o f course, have direct implications for the poor, if they overwhelmingly emphasized higher education (from which the wealthier are more likely to benefit), but, as discussed previously, the distribution o f resources among different levels o f schooling is not particularly unbalanced (although overall public spending on education is low). InTurkey, entrance to university isprimarily basedona student's performance ina centrally administered examination; grades insecondary school are the other factor that determines the overall score. While this examination-based selection i s occasionally criticized, the system itself has important positive features. Perhaps most important i s the basic characteristic that the same questions are asked o f all students, and the evaluation i s undertaken in a consistent and centralized manner. Having said that, selective educational systems have been shown to cause significant inequality in other developing countries (see Mete, 2004). In the Turkish case, the solution to the inequality problem does not have much to do with the design o f the examination; rather, it has to do with increasing the enrollment o f poor children in quality public basic education and general (academic) secondary schools. Some insights into the status o f the few poor students who make it to the (two- or four-year) universities come from an analysis o f the 1997 University Student Survey, published by the Council o f Higher Education in 1998, under the title, "Parental Income, Educational Expenditures, Financial Aid and Job Expectations o f University Students." The survey, implemented by the State Institute o f Statistics (DIE), collected information on about 80,000 students enrolled in 51 public and 7 private universities during 1996-1997. The response rate was 99 percent. The study found that students coming from high-income families are much more likely to be enrolled in private universities, and they are more likely to be enrolled in "well-established" and "new and developing" institutions. Thus, university enrollment of students coming from poor households should be interpreted with the understanding that these students do not enroll in universities o f the same quality as wealthier students. Having said that, the observation that private universities tend to serve the wealthy is neither a surprise nor a negative factor in and o f itself-as long as quality public universities exist with a respectable non-wealthy student enrollment. One way to relax the capacity constraint inhigher education i s to allow and encourage the establishment o f private higher education institutions, as is happening in Turkey. Private tutoring plays a key role in determining who attends what type o f university, as acknowledged by 89 percent of the undergraduate students who view private tutoring as necessary for success in the university entrance examination. About 78 percent o f all undergraduate students report receiving private 56 tutoring (almost 90 percent of those through taking preparatory courses in private tutoring centers). As the main reason for not participating in private tutoring, 57 percent o f the surveyed undergraduate students mentioned lack o f economic resources, 23 percent mentioned availability o f high-quality education insecondary school, and 5 percent mentionedabsence o f private tutoring intheir neighborhood. Thus, the poverty linkage is revealed explicitly by the emergence o f the lack o f economic resources as a prevalent response, but also by the fact that the other common response is "availability o f high-quality education at secondary school," which this Poverty Assessment chapter shows as unlikely for poor children. Furthermore, the statistics from the 1997 University Student Survey reflect the private tutoring patterns o f those students who gained acceptance to university. The 1998 Council o f Higher Education report also contains statistics on receipt o f financial aid during undergraduate education. Household income status i s captured by the creation o f 15 income groups, which show that in public universities, 86 percent o f students coming from the least-wealthy households apply for financial aid, and among those who applied, 69 percent receive aid. For the students coming from wealthiest households, the percentages are 31 and 17 percent, respectively. Inprivate universities, the percentages for children coming from the least-wealthy households are 20 and 20 percent, respectively. For those coming from the wealthiest households, they are 20 and 17 percent, respectively. As a result, while it is true that public universities fare better in terms o f having a pro-poor financial aid distribution, improvements are needed, because a significant portion o f wealthy students receive financial aid while some o f the least-wealthy students are denied aid. Quality of Education (International Comparisons) Turkey has participated in the 1999 Trends in International Mathematics and Science Study (TIMSS), along with 37 other nations. Turkey ranked 31st and 33rd inmathematics and science achievement tests, respectively, administered to eighth-grade students, out o f a total o f 38 countries. While these results are not encouraging, it would be more productive to view them as baselines that need to be monitored and improved over time. For one thing, the countries that participate in this examination scheme are not randomly selected. Many o f them are wealthier than Turkey. Figures IV.6 and IV.7 present test scores by GDPper capita. Focusing on the countries that are inthe same GDP per capita neighborhood, Turkey still ranks at below-average performance, however. FigureIV.6. 1999TIMSS MathematicsAchievementof Eighth-GradeStudents a mu IWW 15ow mwo 2 5 ~ ~ 1 xm 30m 4oaw 4 m u m o a GDP p r sapits (in 1999) I Note: CounQ codes: Singapore (SGP), Republic of Korea (KOR), Taiwan (TWN). Hong Kong (HKG), Japan (JPN), Belgium (BEL), Netherlands (NLD). Slovak Republic (SVK). Hungary (HUN), Canada (CAN), Slovenia (SW).Russian Federation (RUS), AUShRh (AUS), Finland (FIN), Czech Republic (CZE), Mdaysia (MYS), Bulgaria (BGR), Lamia fLVA). United States (USA), United Kingdom (GBR), New Zealand (NZL), Lithuania (LTU), Italy (ITA). Cyprus (CYP). R0"ZiR (ROM). Moldova (MDA), Thailand (THA). Israel (ISR). Tunisia (TUN), Macedonia (MKD). Turkey (TUR), Jordon (JOR). Iran (IRNj. Indonesia (IDN), Chile (CHL), Philippines (PHLj. Morocco (MAR), SouthAfrica (ZAF). 57 Figure IV.7. 1999TIMSS Science Achievement of Eighth-Grade Students 250 + 200 ' 0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000 GDP per capita (In 1999) Note: Country codes: Singapore (SGP), Republicof Korea(KOR), Taiwan (TWN), Hong Kong (HKG), Japan (JPN), Belgium (BEL), Netherlands (NLD), Slovak Republic (SVK), Hungary (HUN), Canada (CAN), Slovenia (SVN), Russian Federation (RUS), Australia (AUS), Finland (FIX), Czech Republic (CZE), Malaysia (MYS), Bulgaria (BGR), Latvia (LVA), United States (USA), United Kingdom (GBR), New Zealand (NZL), Lithuania (LTU), Italy (ITA), Cyprus (CUP), Romania(ROM), Moldova (MDA), Thailand (THAI, Israel(ISR), Tunisia (TUN), Macedonia (MKD),Turkey (TUR), Jordan (JOR), Iran (IRN), Indonesia(IDN), Chile (CHL), Philippines(PHL), Morocco (MAR), SouthAfrica (ZAF). Household Survey Evidence on Variations in Quality of Education There are a number o f reasons poor parents do not send their children to school. Some may choose to restrict the lives o f females in a number o f ways, including not sending girls to school. In other cases, schools may not be available, which this report has shown to be the case especially in rural areas at the secondary school level. If schools are available, then the value o f children's time becomes a factor. As discussed, this i s likely to be the case especially in rural areas. It could also be that, despite highprivate returns to higher education, lack o f access to credit (that is, inability to borrow against future income) may force some children to quit school. A related issue i s access to "quality schooling" at the primary and secondary level: rates o f return to early levels o f schooling depend in part on quality o f schooling (Behnnan and Birdsall, 1983). Furthermore, the breadth o f training in early levels o f schooling determines which, ifany, universities poor children can attend. Dopoor children have access to quality schooling?21 Are wealthier parents more satisfied or less satisfied with the education their children receive? How do private schools fare when it comes to parental satisfaction with schooling? 21 Quality o f schooling can be defined in a number of ways. Input-based objective measures focus on actual observations o n school characteristics such as amenities; output-based measures often focus on standardized test results. These data are rarely available in a manner that can be linked to household characteristics o f children. The approach followed here is dictated by data availability, which is to rely on parental perceptions about the quality o f schools that their children are enrolled in. As with other alternatives, this approach has pros and cons. On the positive side, parents might be considering a large number o f factors that may not be fully captured or appropriately weighted by analytical approaches. On the negative side, a key issue might be what parents from different socioeconomic backgrounds consider quality education. For example, it could be that for illiterate parents, quality education means that teachers regularly show up and their children can read. Other parents might have more stringent quality criteria. There are two mediating factors here. First, the direction o f the possible bias is rather obvious. Ifpoorluneducated parents are more likely to say that school quality is bad, then the differences between schools that serve poorluneducated and wealthyleducated are probably even more pronounced than the survey data 58 Of those household members who were attending school at the time o f the 2001 HCIS, 2.3 percent attended private schools (2.6 percent in urban areas and 1.3 percent in rural areas). Among the least- wealthy quarter o f households, private school attendance was 1.3 percent, and among the wealthiest quarter o f households, it was 5.8 percent. Figure N.8 shows that household members are more likely to report problems with public schools compared to private schools. The leadingproblems are lack o fbooks and supplies, reportedas a problem for 15 percent of children enrolled in public schools and 10 percent of children enrolled in private schools. The next major problem, in both public and private schools, is poor teaching, which was reportedinabout 10percent o f cases. FigureIV.8. Problemswith Schools,PublicandPrivate i3.34 Total Public Private DNo problem .Lack of b00kSiSUpplieS DPWrteachiw ELack of teachers DFacilities in badandittons .Other Dmbiems Source: 2001 HCIS. Basedon a sample of2,638 childrenenrolled inpublic schools and 62 children enrolled in private schools at the time ofthe survey. The urbadrural differences are unequivocal (Figure IV.9). Only 45 percent o f responses indicated "no problem with school" inrural areas, compared to 68 percent inurban areas. Inrural schools, both lack o f bookshpplies (22 percent) and poor teaching (15 percent) are widespread. In fact, in more than 7 percent o f cases, "lack o fteachers" was reportedas a problem. reveal. Second, one can interpret the results, from a rather narrow perspective, as satisfaction from received education services (rather than as a proxy for actual quality o f schooling received). The differences between the replies o f poor and non-poor households would still be o f interest. 59 FigureIV.9. Problemswith Schools,Urban, andRural 70- 44 51 20- 10- 0 Total Urban Rural UNOproblem .Lack of taokdsuDDlies UPwrteaching ELack of teachen UFacilitiesinbad conditions .Other Dmblems Source: 2001 HCIS. Based on a sample of 2,200 children enrolled in urban schools and 537 children enrolled inrural schools at the time of the survey. Wealthier households are much less likely to report problems with schools (Table IV.9).22 Complaints about lack o f books and supplies decline drastically as household wealth increases. This i s not surprising, since wealthier households can afford to ensure their children have the necessary books and supplies. Complaints about facilities also decline with wealth, which may suggest preferential treatment o f schools located in wealthier neighborhoods (although, it may also be a result of higher informal contributions from wealthy families to the school). While complaints about poor teaching remains constant at around 10 percent for all wealth groups, complaints about "lack o f teachers" declines with the wealth o f households. Table IV.9. Problemswith School (Percent) Wealth Quintiles (basedon a household possessions index) Group 1(poorest) Group 2 Group 3 Group4 I I I 22This trend is not driven by the relatively small percentage of households that send their children to private schools. 60 Consistent with the reporting of problems with schools, satisfaction with schooling i s higher for private schools (Figure IV.10) and for urban schools (Figure V.11). Interestingly, though, satisfaction with schooling i s not that sensitive to changes inhouseholdwealth (Table IV.10). FigureIV.10. Satisfactionwith Quality of Education 60- 50- 0 All households Public Private 1DCompletely satisfied OSatisfied DNoidea DDissatisfied Completely dissatisfied (based on a sample of 2,670 currently enrolled students [2,610 public and 60 private]) Figure IV.ll. Satisfactionwith Quality of Education 45- 40- 35- 30 al 0 25 c (I) 2 20 a (I) 15 10 5 0 All households Urban Rural - ~ _ _ _ ~ _ _ _ OCompletely satisfied OSatisfied No idea BDissatisfied DCompletely dissatisfi I ised on a sample of 2,733 currently enrolled students [2,200 urban and 533 rural]) 61 Table IV.10. Satisfactionwith Quality of Education (percentages are reported) I Sample size 674 758 639 660 I Household heads that reported members enrolled in school also indicated perceptions about trends in the quality o f schooling. About 25 percent o f household heads believed primary school service improved in the last two years, 69 percent believed it remained the same, and 6 percent believed service worsened. This time, the wealth effects are visible. About 8 percent o f the poorest quarter o f the households believe the service has worsened in the last two years (in contrast, only 4 percent o f the wealthiest quarter o f households believe primary school service has worsened). Similarly, there are differences by urbadrural residence: in rural areas over 9 percent o f households felt primary schooling quality worsened, whereas only 5 percent o f urbanhouseholds reported decliningprimary school quality. A smaller percentage o fhouseholdheadsbelieves secondary school quality has improved inthe last two years (2 1percent). The differences inperceivedsecondary school quality are negligible among wealthy groups. Urbadruralresidencealso does not matter much. E. ECONOMIC GROWTH LABOR AND MARKET IMPLICATIONS OF A PUBLIC POLICY THAT UNDERVALUES EDUCATION Physical capital and human capital are complements. An optimized combination o f physical and human capital leads to highest economic growth: ifone i s too low, the other does not produce much.23 Both public and private rates o f return to schooling are high, although an empirical regularity i s that there i s a substantial gap between public and private rates o f return to higher education.24 Consistent with this evidence, emphasis on early levels o f schooling i s considered to be a key ingredient in the remarkable economic growth performance o f the Asian Tigers that spans three decades (with a brief halt at the end of the 1990s)-along with generally prudent macroeconomic policies (Mingat 1998). Investment in education also has a self-reinforcing dimension inthat the positive impact on economic growth translates into more resources that can be devoted to education (even if the share o f public spending on education remains the same), which inturn aids economic growth and so on. Indeed, with the exception o f the early decades following the foundation o f the republic, Turkey's failure to invest more aggressively on early 23 The pathways o f the relationship between population's educational attainment and economic development are complex and discussed thoroughly inthe literature. On the relationship between physical capital and human capital see, for example, Heckman(2003). 24 Various studies since then have confirmed these important findings, but the original reference is Psacharopoulos (1994). 62 levels o f schooling has meant missed opportunities both for unschooled individuals and for the economy ingeneral, aspartiallyrevealedbythe labormarkettrends that are reportednext. Returns to Schoolingin Turkey As mentioned, private returns to schooling are very highinTurkey. Table Iv.11presents the results from the estimation o f a Mincerian-type semi-log wage equation (Mincer 1974) using data from the 2002 HBS. The dependent variable i s the natural log o f hourly earnings. Explanatory variables are indicators for gender, potential experience (and its square divided by loo), schooling attainment, urban residence, and region o f residence. Individuals aged 25 to 64, and who report non-zero earnings in the reference month, form the sample. This section outlines key findings from this exercise. These estimates are provided here as the basis for highlightingkey trends inearnings that are of particular interest to this study. Readers interested in this topic should also refer to Dayioglu and Tunali (2004), who carry out a more elaborate investigation o f the determinants o f earnings (the main trends discussed here are consistent with their findings) by using data from 1988 and 1994 in addition to 2002, by employing Oaxaca-Blinder wage gap decomposition to better illustrate gender inequalities, by including provincial variables (such as share o f manufacturing, trade), and so forth. On average, males earn 45 percent more per hour than females of similar characteristics. Schoolinghas a robust, positive, and large impact on earnings. Vocational junior secondary school graduates earn more than general junior secondary school graduates; similarly vocational secondary school graduates earn slightly more than general secondary school graduates. There are a number o f selectivity issues that hinder a clear-cut interpretation o f these estimates. For example, one could argue that because "more able" general secondary graduates continue their education at the university level, the observed returns to general secondary education are underestimated. On the other hand, one could also claim vocational school students are disadvantaged in terms o f family contacts or other unobserved characteristics that depress the distribution o ftheir earnings. When separate models are estimated for males and females, we see that the coefficients for secondary and higher education are much larger for females. As discussed, on one hand, females earn less than males o f similar characteristics in the labor market. But on the other hand, the impact o f schooling on earnings i s even more visible for females. A final issue that deserves emphasis is the gender distribution o f those who reported non-zero earnings inthe reference month: 85 percent o f this group is male. These findings are supportive of the literature that argues that payoffs to educating girls are at least as high as to educating boys, especially in countries like Turkey, where the gender gap in schooling i s significant- because social rates o f return decline with schooling. Other reasons for emphasis on girls' schooling include the close linkage between a mother's schooling and children's health and education, and the empirical regularity that more educated women are more likely to participate in the labor force, thus broadening the tax base (Schultz 2002). 63 Table IV.11. Ordinary Least SquaresEstimates of Log-Hourly-Earnings (13.9) (11.4) (6.23) [.239] Higher Education 1.59 1.40 1.86 .125 UrbanResidence Marmara Note: It/-statisticsare in parentheses; individualsaged 25 to 64 are includedin the sample. Source: DIE 2002 HBS. 64 UnempIoymentby SchooIingAttainment The evolution of unemployment rates over time is shown in Figure IV.12, separately for males and females, and for younger individuals. Figures IV.13 to IV.20 display unemployment rates by schooling attainment. The data for these figures come from the State Statistics Institute(originaldata source is the 1990-2003 LaborForceSurveys). Figure IV.12 shows that (a) since 2000, unemployment rates have been on the rise; (b) youth unemployment rates are significantly higher than overall unemploymentrates; and (c) the probabilityof unemploymentis higher for males inrecentyears, especially amongyouths. While each of the remaining figures could be discussed separately, for brevity, three issues will be mentionedhere. First, the female unemployment rate is higheramong(vocationaland general) secondary school graduates. Second, the unemployment rate amongvocationalsecondary school graduatestends to be at about the same levelas the unemployment rate amonggeneral secondary school graduates-in fact, in some cases vocational school graduates are much more likely to be unemployed. Third, those with higher education diplomas are unique in that there is a clear distinction between their overall unemploymentrates (low comparedwith others inthe sample) andyouth unemploymentrates (very high at around 35 percent). Since this trend has persisted since 1990, one explanation is the possibilitythat reservation wages are higher for recent higher-educationgraduates, such that they are more selective when it comesto acceptingjobs, comparedto less-educatedindividuals. But overallunemployment rates are lower, as one would expect-skilled individualshave more opportunitiesinthe labor market, andthis effect seems to dominateother factors (such as differences in reservationwage) eventuallywhen it comes to determiningobserved unemployment rates. Finally, among individuals who have completed higher education, the female unemployment rate is much higher, despite the lack of a gender difference among the younger cohorts. This is a finding for whichwe haveno explanation at this stage. FigureIV.12. UnemploymentRates Over Time 250- 20 0- c al m 1501 -E, 8 0 n E 100- 3 C 50 65 FigureIV.13. Illiterate UnemploymentRates Over Time 25.0- 200 P zm 150 C 0 -0Q k l E 100- 3 C I 50- 001 1990 1931 1992 1993 1994 1995 1996 1937 1998 1999 2ooo 2M1 'XJlZ 2003 -+Me Lk"@o,nmRae-%Fenale LhE@qmertRae+Fn;leYu& UlnplOymrtFate uFenaleYadh LhE@u,m-d Rae ~ _ _ FigureIV.14. LiterateWithout DiplomaUnemploymentRatesOver Time I 1 0.0- 1990 1931 1992 1933 1994 19% 1996 1937 1998 1999 2Mx) 2M1 2002 xo3 66 FigureIV.15. Primary School Graduates' UnemploymentRatesOverTime I I J 200. - a 5m150. S a - 6 a 5 E 100- I 50 FigureIV.16. Junior Secondary Graduates' UnemploymentRatesOver Time 250- d 2 l d C 200. - E a E 150 E 3 100 67 FigureIV.17. VocationalJunior SecondaryGraduates' UnemploymentRatesOver Time 00. 500 e? :400 0 3i :330 0 3 200 1004I FigureIV.18. Secondary School Graduates'Unemployment Rates Over Time 400 350 > i;330 ? ;EO i i200 ) 50 68 FigureIV.19. Vocational Secondary School Graduates' Unemployment Rates Over Time FigureIV.20. Higher EducationGraduates'UnemploymentRates Over Time ?Lao- ? ' 150 69 CHAPTERV: HEALTH This chapter looks at health status, access to, and use of health care services, and private and public expenditure on health care, with the objective o f determining to what extent the lower-income groups have adequate health care protection. It i s largely based on data o f the health module o f the 2001 Household Consumption and Income Survey (HCIS), and on data on health care expenditure from the 2002 Household Budget Survey (HBS). Data o f the 2003 Demographic and Health Survey (DHS) were not available to support this chapter, nor were the data of a household health survey carried out as part o f the National Health Accounts exercise, which was completed in 2003.25 The chapter draws on work carried out inthe framework o f the World Bank's Turkey Health Sector Report (World Bank 2003~). A. THEHEALTH SYSTEM CARE Turkey's health care delivery and financing system has been characterized by fragmentation. Health services are supplied by a multitude o f public and private providers, with the main providers being the Ministryo f Healthy (MOH), the Social Insurance Organization (SSK), and the university hospitals. The MOH is the major provider o f primary and secondary health care services, and essentially the only provider o f preventive health services. MOH operates an extensive network o f health facilities, including rural health posts, health centers, dispensaries, and hospitals that provide outpatient specialist and inpatient care. SSK operates an extensive network o f hospitals that provide outpatient and inpatient care, and a smaller network o f primary care facilities. University hospitals provide tertiary inpatient and outpatient care. Until recently, access to SSK facilities was limited to those covered by SSK's health insurance scheme (for example, formal sector employees and their dependents), and to those covered under Bag-Kur Insurance (self-employed and their dependents). Under the government's recently adopted reformprogram, the distinction in access to SSK and MOHhospitals and health centers has been removed, and patients are now free to visit the facility o f their choice. Although still small compared to the public health care sector, provision o f health care services by the private sector i s gaining importance, particularly in urban areas and in the western parts o f the country. Private outpatient care i s provided in various settings, including private doctors' offices, polyclinics, and medical centers, private services provided in public hospitals, health services provided by occupational physicians incompanies with more than 50 employees, private hospitals, and foundation hospitals. 25Given the age o f the available D H S data and the fact that the 2003 DHS data will become available during the second half o f 2004, an analysis of D H S data by household welfare level, along the lines o f the one carried out by the World Bank using 1993 DHS data, was not undertaken. 70 Table V.l. Distributionof Primary Care Health Staff, by Region, 2002 Population/ Population/ Population/ Doctor Midwife Nurse Marmara 7,65 1 5,569 9,382 Aegean 3,565 2,333 4,597 Mediterranean 3,595 2,341 5,440 CentralAnatolia 3,985 4,339 5,409 Black Sea 3,747 2,952 4,770 EastemAnatolia 5,223 4,511 5,923 SoutheasternAnatolia 7,304 6,960 10,477 Turkey 4,708 3,672 6,196 Source: MOH, General Directorate of Primary Health Care web page. Table V.2. MOHHealthCenters and HealthPosts Lacking Key Staff, by Region, 2002 Region Number of Health % of All Number of Village YOof 'YOof Births Centers without Health CentersHealth Postswithout Health Posts Unattended Doctors Midwives by Health Staff Marmara 97 11 891 62 2% Aegean 129 13 814 55 6% Mediterranean 78 9 808 70 3% CentralAnatolia 151 14 1,353 80 4% Black Sea 130 13 2,326 77 4% EasternAnatolia 116 20 1,660 90 19% SoutheasternAnatolia 84 20 984 90 20% Turkey 785 13 8,836 75 6% Source: MOH, General Directorate of Primary Health Care web page. Although significant efforts have been made over the past two decades to expand the health care network and assure adequate physical access to health facilities across the country, the service delivery network remains highlyuneven. Health facilities and health care professionals remainconcentrated inurban areas, particularly in the three largest cities, Istanbul, Ankara, and Izmir, while many rural areas, especially in Eastem and Southeastem Anatolia, suffer fiom severe shortages of medical staff (Table V.l). The skewed distribution o f facilities, and particularly o f staff, has resulted insignificant regional differences in access to and use o f health care and, concomitantly, health outcomes (Tables V.2 and V.3). The chronic shortage o f health care staff in primary care facilities in many rural areas in general and in the East and Southeast in particular, has been compounded by a lack of operating budgets in many primary care facilities. This, in tum, has led to the perceived low quality o f care provided at the primary care level, as demonstrated by patients' frequently circumventing primary care and self-referring directly to higher- level specialist or private care. Unequal service provision has resulted in a marked regional difference in health sector performance and, consequently, also widely varying health outcomes. 71 Table V.3. RegionalDistributionof HospitalBeds,2002 Total Beds/10,000 Admissions/ Number of Beds Population 1,000 Population Marmara 49,655 27.5 76.6 Aegean 21,541 23.5 84.4 Mediterranean 17,306 19.2 71.4 Central Anatolia 31,679 26.6 87.6 Black Sea 22,303 26.4 83.1 EasternAnatolia 13,823 18.0 64.6 Southeastem Anatolia 5,928 10.9 47.0 Turkey 162,235 23.3 76.0 Note: Admissionfiguresare for 2000, beds for 2002. Source: MOH, In-patientTreatment Institutions, Statisical Yearbook (2003). B. HEALTH INSURANCE Several public health insurance schemes currently provide financial protection to various target groups. Ofthose, the healthinsurance o fthe SSK i s the most important, andprovides coverage to those employed in the formal sector and their dependents. Bag-Kur is the health insurance for the self-employed, including in principle the rural population and informal sector workers. Emekli Sandigi provides health insurance to retired civil servants, while those in active civil servant status are covered directly through their employers. In 1992, the Government introduced the green card system, designed to afford protection to low-income groups who are not covered otherwise. Private voluntary and supplementary insurance i s also available. To increase operational efficiency and improve health insurance coverage, the Government plans to shift from the currently multiple public insurance schemes to universal health insurance, which would operate on the principles o f solidarity and risk pooling, and provide coverage to the entire population. The Government would make contributions on behalf o f those who cannot afford to do so themselves, while others would contribute through the social insurance system. Implementation o f this system i s expected to be gradual, with completion around 2011. Both the 2001 HCIS and the 2002 HBS provide information about the population's health insurance coverage (Figure V.1). Both surveys suggest that over one-third (36 to 37 percent) o f the populationdoes not have access to health insurance, including the green card program.26 The data suggest that in rural areas, almost half the population has no health insurance, and in urban areas, about one-third has none. Because a larger share o f the lower-income groups is employed in the informal sector, the share o f those inlower-income groups that is covered byhealthinsurance (including green card) is significantly smaller 26 While the 2001 HCIS and the 2002 HBS show highly similar figures with respect to health insurance coverage, other data sources suggest a somewhat higher share of coverage. For example, the Draft National Health Accounts Report suggests that about one-third of the population remains without health insurance coverage, compared to 36 percent impliedby the 2002 HBS and 37 percent impliedby the 2001 HCIS. Furthermore, official records indicate that in 2001, about 75 percent of the population was covered by insurance, with this share rising to almost 83 percent in 2003 (data provided by Emekli Sandigi and Ministry o f Labor). It is, however, thought that official records include a significant amount o f double counting (see for example, World Bank 2003c) and are thus not an appropriate mechanism to determine the number o fpeople without coverage. 72 than among upper-income groups (42 percent inthe bottom quintile are covered, compared to 79 percent inthe top quintile). FigureV.1. Share ofPopulationwith HealthInsurance, by Quintile 90% 8 0 % 7 0 % 60% 50% 40% =Rural 30% 20% 10% 0% 1 2 3 4 5 Total Quintile Source: 2001 H C I S . The green card program, which was introduced in 1992 to provide insurance protect.-.l to low-income households, does provide protection to those who benefit from it, but fails to provide broad coverage to all those living inpaver$' (Table V.4). HBS data suggest that more than half(58 percent) of those who live below the poverty line, and over two-thirds o f the extremely poor, remain without any insurance coverage, while somewhat under one-fifth o f the extremely poor, and less than 10 percent o f the poor, benefit from the green card program. On the other hand, the data suggest that two-thirds of green card holders are not poor.28 Thus, an important share of the lowest-income households remains without any access to health insurance coverage, which in turn requires that they make significant out-of-pocket payments when seeking health care. Table V.4. HealthInsuranceCoverageandPoverty Extremely Insurance Status Poor Poor TOTAL % ofpoor with insurance Compulsory 28% 13% 54% Voluntary 3% 0% 4% Compulsory and Voluntary 3% 0% 2% Green Card 9% 18% 4% None 58% 68% 36% Total 100% 100% 100% Source: 2002 HBS. 27This analysis is based on the food poverty andpoverty lines usedthroughout thisreport. 28 Green card program eligibility criteria stipulate that a Turkish citizen, not covered by any social insurance and with a monthly income o f less than one-third o f the net minimumwage, is eligible for green card coverage. In2002, the net minimumwage was TL174 million, making the cutoff point for qualification under the green card program TL58 million, which i s below the adult equivalent per capita poverty line used inthis report. 73 C. HEALTH OUTCOMES Health outcomes in Turkey are poorer than would be expected in a country with Turkey's income level. Despite considerable progress achieved in the recent past, Turkey continues to rank far behind most middle-income and European Union (EU) accession countries on key health indicators. Life expectancy is 9 years lower than the average for women in Organization for Economic Co-operation and Development (OECD) countries, and 7.5 years lower for men (Table V.5). Infant and maternal mortality rates are among the highest in the Europe and Central Asia (ECA) region and among middle-income countries. Turkey also continues to have a relatively high death toll from preventable infectious diseases, and scores relatively poorly interms of vaccination coverage. TableV.5. HealthStatusComparisons, 2002 Southeast and Central Turkey Europe Europe OECD Life Expectancy Male 66.2 69.7 69.4 73.8 Female 70.9 77.8 77.2 79.9 Infant Mortality per 1,000 life births 36.0 9.7 10.5 7.3 Maternal Mortality 100.0 18.1 13.0 Sources: WHO-HFA, OECD Health Indicators, Turkey MOH. Note: Data are for 2002 or most recent available prior to 2002. Keyhealth outcome indicators vary markedly across urban and rural Turkey and across regions, reflecting the uneven supply of and access to health care across regions. Infant and child mortality rates are substantially below the national average in urban areas and in Western and Southern Turkey, but almost 40 percent above the national average in rural areas and in Eastern Turkey, and this gap has widened over the past decade (Figure V.2).29 Similarly, vaccination coverage o f infants and pregnant women is significantly lower in the poorer Eastern and Southeastern Provinces than in the rest of the country, although the gap has begun to narrow over the past five years (Table V.6). FigureV.2. RegionalVariationinInfant and Child Mortality Rate; and RegionalVariation in PrenatalCare and HomeBirths RegionalVariation in Infant and Child Mortality RegionalVariation in Prenatal Care and Home Births 80 80 0%without pre-natal 60 60 care 40 40 20 0%homebirths 20 0 0 I North South Center Wen East Rural Urban . Source: DHS(1998). 29Health Outcome Indicators on a regional basis are not readily available. The only data available for this report were the 1998 DHS data as data from the 2003 DHS were not yet available for inclusion. 74 Table V.6. Vaccination Coverageof InfantsandPregnantWomen (TT-2T), byRegion, 1998-2002 BCG DBT-3/OPV-3 Measles HBV-3 TT-2T 1998 2002 1998 2002 1998 2002 1998 2002 1998 2002 Marmara 95 88 90 79 88 86 69 73 27 25 Aegean 86 88 85 80 85 85 87 82 57 58 Mediterranean 86 90 86 85 83 89 64 83 55 53 Central Anatolia 87 87 88 83 86 87 84 84 47 48 Black Sea 85 88 88 82 89 85 83 78 45 47 Eastem Anatolia 42 57 63 68 67 71 42 55 19 24 SoutheastemAnatolia 46 61 54 63 52 67 23 49 12 18 Turkey 77 82 79 78 70 82 65 72 36 37 Source: MOH webpage. The Turhsh health system faces a dual challenge. Significant parts o f the country and the population continue to be afflicted by a high burden o f disease from preventable infectious diseases, and high maternal and infant mortality rates typical o f developing countries. At the same time, a growing share of the population is affected by noncommunicablediseases prevalent indeveloped countries. Morbidityand mortality associated with heart and cerebrovascular diseases have increased sharply over the last two decades as, for example, reflectedby a fourfold increase inhospital discharges for ischemic heart disease and cerebrovascular problems between 1988 and 2002, and an increase o f over 150 percent inregistered cancer cases (WHOHFA 2004). D. SELF-REPORTED MORBIDITY Information on self-reported morbidity i s available from the 2001 HCIS. The data indicate that 10.6 percent o f the population reported having been sick or illduringthe month prior to the survey, and 3.6 percent reported having suffered from an illness that required hospitalization during the six months prior to the survey. Both the share of people reporting an illness or injury duringthe past month, and the share o f people reporting an illness requiring hospitalization, are higher among the lower-income groups, particularly inrural areas (Figure V.3 and Table V.7). FigureV.3. IncidenceofIllnessIncomeQuintilesandLocation Ulllness/in]u!y past 30 20 0 days urban 15 0 MIllness/injury past 30 days rural 10 0 Oillness requiring 5 0 hospitalization past 0 0 6 months urban Total Oillness requiring hosoltalization Dast Per Capita Income Quintile 6 months rural ' ~~ ~~~~~ ~~~ Source 2001 HCIS 75 Table V.7. Self-ReportedMorbidity by IncomeQuintile Quintiles of Per Capita ircome 1 2 3 4 5 Total % reporting being sick/injured in past 30 days 13.0% 9.7% 10.0% 9.1% 11.0%10.6% % reporting illness requiring hospitalization in past 6 months 5.6% 3.6% 2.9% 3.3% 2.8% 3.6% Severity of ///ness (% of those reportingIllnesshjury last Month) Life-threatening 10% 5% 8% 5% 4% 6% Very Serious 15% 21% 20% 16% 16% 17% Serious 54% 45% 41% 53% 41% 47% Not Serious 21% 29% 31% 26% 39% 29% Source.2021 HCIS Similarly, a significantly higher share o f those in low-income households report that the illness or injury from which they suffered duringthe month prior to the survey was life threatening. The share of those indicating that the illness or injury was not serious i s almost twice as high among people in the highest income group than among those inthe lowest quintile. The overall higher reportedincidence and severity of illness among the lower income groups, particularly inrural areas, may suggest poorer quality of care and more limited access to and use o f health care among these groups, resulting indelayed treatment and thus more severe disease incidences. A look at reported incidence o f disease by health insurance category indicates that those without any insurance report a slightly lower disease incidence than those with insurance (Figure V.4). This i s likely explained by the fact that the absence o f insurance creates access barriers to health care, which then lead those without insurance to less readily admit to being sick. International evidence has shown that self- reported morbiditytends to be associated with access to health care. FigureV.4. Incidenceof Morbidity by HealthInsuranceStatus I N o Insurance OOther % reporting I G r e e n t Card being 0Private sick/injured OBagkhur I S S K 114% I DES 0% 5% 10% 15% Source: 2001 HCIS. E. ACCESS AND USEOFHEALTH TO CARE Ofthose reporting an illness over the past month, 72 percent report seeking some form of health care for the reported illness, with the share being lower inrural (66 percent) than inurban areas (74 percent) (2001 HCIS). The likelihood o f seeking care when illis significantly lower inthe bottom two quintiles than in the upper quintiles, with the difference being particularly marked in rural areas (Figure V.5). On a 76 regionalbasis, the share of those seekingcare in Southeasternand EasternAnatolia is significantly lower than inthe rest ofthe country(Figure V.6). FigureV.5. Share of ThoseReportingIllnessWho SoughtCare by Quintile and Location 100% 80% 1 60% U%sought care rural , 40% ~ 20% 0% 1 2 3 4 5 PerCapita IncomeQuintile I Source: 2001 HCIS FigureV.6. Propensityto Seek Treatment (by regionand location) 130% 80% 30% 0Rural 0Urban Region . ~~ Source: 2001 HCIS. Even among those who report that their illness was life threatening or very serious, the share of those seeking care remains a low 55 percent in the lowest-incomegroup, while it is over four-fifths in the top income group. This suggests significant access problems among the low-income groups. Those with insurance, including a green card, are significantly more likely to seek health care when illthan those without insurance, underliningthe importanceof insuranceto improvingaccess to healthcare. Even among those who report that their illness was life threatening or very serious, the share of those seekingcare remains a low 55 percent in the lowest-incomegroup, while it is over four-fifths in the top income group (Table V.8). This suggests significant access problems among the low-income groups. Those with insurance, including a green card, are significantly more likely to seek health care when ill than those without insurance, underliningthe importance of insuranceto improvingaccess to health care (FigureV.7). 77 Figure V.7. Propensity to Seek CareWhen I11by Insurance Status FigureV.7. Propensityto Seek Care'VVhen111byInsurance status Private EmekliSandigi Eagkur SSK Green Card None 0O h 20% 40% 60Yo anyo 100% Source; 2001 HCIS Table V.8. Health Care Utilization by Severity of Illness Life- Threatening Very Serious Serious Not Serious Total Quintile YOSeekingCare 1 55% 66% 70% 54% 63% 2 87% 58% 72% 54% 63% 3 80% 84% 79% 72% 79% 4 83% 86% 84% 77% 79% 5 83% 89% 84% 74% 78% Total 72% 76% 77% 67% 72% Source: 2001 HCIS Hospital Care. Of the 552 individuals(3.6 percent of population)who reportedhavingsufferedfrom an illness that required hospitalization over the past six months, only two-thirds reported that they had actually been hospitalized, with the incidence of hospitalization being inversely related to household income andto insurancestatus (FigureV.8). The data seem to suggestthat possessionof a green cardhas a significant impact on access to hospital care, because there is no marked difference in the hospitalizationrate between those with insurance and those with a green card, but the share among those without any insurance protection is 26 percentage points lower than among those with insurance or a green card. 78 FigureV.8. HospitalizationAmongThoseRequiringHospitalization,by Quintile Hospitalization among Those Requiring Hospitalization among Those Hospitalization, by Quintile RequiringHospitalization,by Insurance Status 5 4 no insurance ~~~~~ 3 2 2 G B greencard 1 g Quintile -c insuranw 7 2 t 0 0.2 0.4 0.6 0.8 '~ 0% 20% 40% 60% 80% Source: 2001 HCIS. Determinantsof Care Seeking. The 2001 HCIS showed that the most importantreason for not seeking outpatient care when sick, and for not seeking hospital admission when required, is the lack of affordability (Table V.9). Close to three-quarters of those who did not seek outpatient care when ill,and a similar share ofthose who did not seek hospitalcare when needed, reportedthat they couldnot affordto do so, with the share of those not being able to afford it expectedly falling rapidly with rising income levels. Table V.9. Reasonsfor Not SeekingOutpatient CareWhen I11 ~~ ~~ Reason for Not Poor Seeking Care Unaffordable Too Far Quality Bad Time No Doctor No Drugs Other Quintile 1 95% 0% 2% 0% 0% 2% 2% 2 79% 0% 0% 6% 2% 0% 13% 3 80% 0% 5% 5% 0% 0% 10% 4 32% 0% 16% 5yo 5% 5% 37% 5 25% 4% 14% 14% 7% 0% 36% Total 72% 0% 5% 5yo 2Yo 1Yo 14% Insurance Status EmekliSandigi 17% 0% 50% 17% 0% 0% 17% SSK 5 1% 2% 6% 10% 4% 2% 24% Bag-kur 40% 0% 10% 0% 10% 0% 40% Private 0% 0% 0% 0% 0% 0% 100% GreenCard 90% 0% 0% 0% 0% 10% 0% No insurance 85% 0% 3yo 3yo 0% 0% 8% Source: 2001 HCIS. Although two-fifths of the population indicated that outpatient facilities where they seek care are far away, physical access barriers do not appear to be a key determinant of not seeking care when ill, 79 regardless of household welfare level. The fact that a significant share o f those who sought care indicated that the place were they got care was far away is likely driven by people self-refening to outpatient hospital care rather than using primary care, which would be closer to their residence, but i s often perceivedto be o f low quality (Table V.10). Table V.10. Reasonsfor Not BeingHospitalizedWhenNeeded ~~ ________~ Quintile 1 2 3 4 5 Total Cannot Afford It 88% 83% 75% 40% 30% 78% Too far 4% 8% 17% 0% 0% 5% Poor Quality 5% 4% 0% 10% 17% 5% Other 3% 5% 8% 50% 53% 12% Source: 2001 HCIS. Information pertaining to problems that those who sought care encountered, confirm that affordability o f health care is a serious issue for lower-income households. One out o f 5 people from the lowest income quintile who sought outpatient care reported that the main problem with the care was that it was too expensive, and another 1 out o f 10 reported that lack o f access to drugs was a problem. The share of those reporting the same problems among the top quintile, on the other hand, i s only one-fifth of that among the bottom quintile (Table V.11). Table V.11. ProblemsEncounteredWhen SeekingOutpatient Care, by Quintile Quintile (per capita income) NatureofProblem 1 2 3 4 5 Total No problem 46% 50% 54% 61YO 68% 55% Not clean 2% 3% 3% 3% 1Yo 2% Longwait 12% 20% 18% 18% 13% 16% No specialists 0% 2% 2% 1% 2% 1% Too expensive 20% 10% 14% 10% 4% 12% N o drugs 9yo 4% 3yo 1Yo 2% 4% Treatment did not work 5% 3% 7% 4% 6% 5% Staff was rude 4% 6% 1% 2% 3% 3% Other 1% 1% 0% 0% 2% 1% Source: 2001 HCIS. The share o f the population that had to pay for outpatient treatment, drugs, and hospitalization is consistently higher among the lowest-income quintile than among the upper-income groups-a reflection of the lower insurance coverage among low-income households. Among those who paid for outpatient care, total payments (covering consultation, drugs, and gifts to staff) were highest among the lowest- income group, and dropped with rising income and associated increased insurance coverage (Table V.12). The situation i s less marked inthe case of hospital treatments; while the share o fthose who had to pay for hospital treatment was higher in the lowest-income groups, the average amount paid for inpatient care was lower among the lowest-income groups (Table V.13). 80 Table V.12. Average Amount PaidDuringLast OutpatientTreatmentby Those Who Paid Quintile Consultation Drugs Gift Total % Who Paid TL Million 1 34 24 4 62 60% 2 28 30 1 59 40% 3 27 19 3 48 35% 4 35 20 3 57 46% 5 30 17 1 48 37% Total 31 21 2 54 43% Source: 2001 HCIS. Table V.13. Share and Average Amount of Paymentsfor LastHospitalAdmission Quintile TL Million % Who Paidfor Treatment 93 170 68% 139 58% 67% 133 50% 161 66% Total 136 62% Note: Average paid is average for those who paidonly. Source: 2001 HCIS. Multivariate analysis of the determinants of health-care-seeking behavior confirms that income, insurance coverage (including green card), household size, gender of the household head, and severity of illness are the most important determinants of an individual's seeking health care (see Table V.14). Together with the above information, these findings thus confirmthat financial constraints constitute a significant access barrier to care for low-income groups, particularly those who have no financial protection. They underscore the importance of access to health insurance inthe decision to seek health care. Table V.14. Determinantsof HealthCare-Seeking Behavior,ProbitEstimates (basedon 2001 HCIS data) Employed head of household Female head o f household -0.363 -2.890 Insurance: EmekliSandigi 1.076 7.960 81 Variable Coefficient t-ratio Insurance: SSK 0.671 7.570 Insurance: Bag-Kur 0.683 5.590 Insurance: Private 1.261 2.110 Greencard 0.618 3.670 Severely ill 0.598 6.520 Household size -0.062 -3.160 Constant 0.024 0.120 Bank's Health Sector Report(2003c), particularly Chapter 2. This report also carriedout a more sophisticated analysis of the determinants of health-care-seeking behaviors using a nesting logit model to predict the probability of seeking health care conditional upon reporting morbidity. The results of that analysis similarly confirmed the importance of income, insurance, householdsize, gender of household head, and severity of illness as the most importantdeterminantso f an individual's decision to seek healthcare. Locationof Care. Lower-income groups are most likely to seek care at MOH facilities because they can gain access to them even ifthey have no insurance (or with green card coverage), and because the average cost of treatment there is lower than elsewhere, with the exception of SSK and military facilities, botho f which were, until recently, reserved for specific target groups (SSK beneficiaries and military personnel and their dependents). The upper-income groups, which are more likely to benefit from formal insurance protection (most notably SSK insurance), tend to seek care at SSK, university, and private facilities. However, a relatively important share of those in the lowest-income groups also seek care from private providers, suggesting that once a decision to seek care i s taken, perceived quality o f care is an important determinant o f the location o f care seeking (Tables V.15 and V.16).30 The data also confirmthat primary care facilities, particularly MOH ambulatories and health posts, which have been set up countrywide to provide the population with access to essential care and to serve as a first contact point, largely fail to do so. Almost three-quarters o f those who sought outpatient care did so at a hospital, rather than at a primary care facility, with the share o f those in lower-income groups not varying markedly from those in the upper-income groups. The Government's recently introducedpolicy changes, which provide patients free choice between MOHand SSK facilities regardless o f insurance affiliation, are likely to change this care- seelung pattern somewhat. It is unlikely, however, that the distribution between private and public providers will substantially change as long as waiting times and perceived quality of care at public facilities are not improved. 30 The World Bank Turkey Health Sector Report (2003~)found that the demand for health care is very price inelastic, even among low-income groups. Thus, once a decision to seek care is taken, particularly by those who have to pay for the entire cost of care themselves, quality considerations are likely to drive care seekers to seek care from private providers that are perceived as providing higher quality of care. 82 TableV.15. LocationofOutpatientCareby Quintile Quintile (per capita income) Location of Care 1 2 3 4 5 Total MOH Hospital 43% 38% 29% 18% 26% 31% MOHClinic 5% 3% 4% 5% 3% 4% SSK Clinic 6% 4% 5% 9% 3yo 5yo SSK Hospital 13% 30% 38% 29% 29% 27% University Hospital 3% 4% 5% 11% 8% 6% Military Hospital 2% 0% 1% 1% 1% 1% Private Clinic 6% 6% 6% 8% 13% 1% Private Hospital 11% 6% 5% 7% 8% 8% Consulting Room 8% 7% 2% 8% 7% 6% Pharmacy 0% 0% 0% 0% 1% 0% Other 4% 3% 5% 2% 2% 3% Source: 2001 HCIS. TableV.16. Average Amount Paidfor OutpatientTreatmentby FacilityType TL million MOHHospital 30.5 MOH Clinic 30.3 SSK Clinic 24.6 SSK Hospital 26.2 University Hospital 51.9 Military Hospital 3.9 Private Hospital 91.5 Private Clinic 55.2 Doctor's Residence 50.4 Pharmacists 22.5 Other 61.9 Note: Paymentsinclude consultation, drugs, and gifts. Source: 2001 HCIS. PreventiveHealthCare. Given accessbarriers to healthcare, preventivehealthcare is, not surprisingly, a luxury item inTurkey. The 2001 HCIS data show that only about 1percent o f the three lowest-income quintiles had a checkup in the past six months, while this share rises to somewhat over 5 percent and under 9 percent for the top two quintiles, respectively, with little difference between urban and rural households. Not surprisingly, the vast majority o f those in the bottom two quintiles indicated that they could not afford to have a checkup, while the main reason for not having a checkup in the top two quintiles was that there was no need (Table V.17). 83 Table V.17. Utilizationof PreventiveCare 1 2 3 4 5 Total % o fPopulation Who Hada Check-Up inPast 6 Months 1Yo 1% 1Yo 4% 9% 3% Reasonsfor Not Having a Check-Up Unaffordable 85% 80% 68% 53% 31% 64% Too Far 1% 1% 1Yo 1% 1Yo 1Yo Not Time 0% 0% 1% 2% 4% 1% Not Aware o f Need 2% 3% 2% 2% 2% 2% Not Necessary 10% 15% 27% 42% 59% 30% Other 1% 0% 1% 1% 3% 1% Source: 2001 HCIS. F. HEALTH EXPENDITURE CARE Out-of-Pocket Payments. Household expenditure on health care occurs in the form o f insurance contributions and direct out-of-pocket payments. The 2002 HBS provides information on the relative importance and composition o f out-of-pocket expenditures. The data suggest that households in Turkey allocate a relatively modest share o f their total expenditure to health care in the form o f out-of-pocket expenditures, but this share increases with income, suggesting that health care is considered a luxury good. The top quintile spends about twice as high a share o f total household expenditure on health care than the lowest quintile. More important, the top quintile spends over 12times more per capita on health care than the bottom quintile (Table V.18). Table V.18. Compositionof HealthCareExpenditure,byExpenditureQuintiles 1 2 3 4 5 Total 2002 Per Capita Health Expenditure/month (milTL) 1.2 2.9 3.4 6.1 14.8 3.3 Share o f Total Expenditure 1.7% 2.5% 2.2% 2.7% 3.5% 2.2% Compositionof Health CareExpenditure (YOof total OOP) Outpatient Consultation 9% 16% 16% 20% 19% 14% Inpatient care 1% 2Yo 1% 2% 5% 2% Dental Care 2% 4% 4% 4% 9% 4% Diagnostics 1% 2Yo 3% 6% 5% 3% Drugs 84% 73% 71% 65% 58% 74% Other Medical supplies 4% 4% 6% 5% 6% 5% 1. Total maynot addup to 100due to rounding. 2. Healthcare expenditures adjustedfor spatialpricevariations. Average unadjustedout-of-pocket(OOP) amountsto about TL2.7 millionper capitaper month. Source: DIE 2002 HBS. The largest share o f out-of-pocket expenditures on health by far i s allocated to the purchase o f drugs (74 percent), with payments for outpatient consultations ranking second across all income levels. The share o f expenditure allocated to drugs i s negatively correlated with household income. Poorer households spend a significantly higher share on drugs and lower shares on outpatient consultations, dental care, and diagnostics than the upper-income groups, an indication that poor households may resort to self-treatment 84 when ill,rather than seek professional care. There are larger differences inthe average absolute amounts spent on all types o f care across income quintiles: the top quintile spends almost 12 times more on outpatient care per capita than the bottom quintile, and almost four times more on drugs. Catastrophic Effect of HealthCareExpenditures. The catastrophic effect o f health care expenditures can be determined by looking at the ratio o f people who spend a higher share o f total expenditures on health care than a predetermined threshold considered to be catastrophic, and by looking at the extent to which health care expenditure shares surpass the catastrophic threshold (Wagstaff and Doorslaer 2002; Xu and others 2003). Thresholds o f 10 percent, 20 percent, and 40 percent o f a household's non-food expenditure were considered for this analysis. Table V.19 shows that there is on average 10 percent o f the population that spends more than 10 percent o f non-food expenditure on health care, with the share dropping to 0.7 percent for a threshold o f 40 percent. This compares favorably to many countries in the E C A region, and even in the EU (Xu and others 2003). The data suggest that the two-thirds o f households with insurance are indeed protected against catastrophic expenditures, while those without insurance often tend to forego health care altogether, and thus largely avoid catastrophic health care expenditures. There i s relatively little difference across income groups in the share o f those with catastrophic health care expenditures, though it tends to be somewhat higher among the second-lowest quintile than the rest. Comparing the poor to the non-poor, the share o f those with catastrophic expenditures i s somewhat higher among the non-poor, likely a reflection o f the fact that many o f the poor who do not have insurance simply forego health care. This is also confirmed by the fact that the share of those with catastrophic expenditure shares is not significantly higher among those without health insurance. Table V.19. ProportionofPeoplewith CatastrophicHealthCareExpenditure 10% or More 20% or More 40% or More of Non-Food of Non-Food of Non-Food Expenditure Expenditure Expenditure Quintile 1 9.3% 4.2% 0.5% 2 12.2% 5.6% 1.O% 3 11.0% 3.5% 0.1% 4 10.0% 3.6% 1.4% 5 11.0% 5.0% 1.9% Total 10.0% 4.0% 0.7% Poor 9.5% 4.3% 0.6% Non-Poor 10.8% 4.4% 0.8% Insurance Status Compulsory 8.7% 3.2% 0.7% Voluntary 15.6% 6.4% 0.8% GreenCard 12.9% 7.4% 0.7% None 11.6% 5.1% 0.9% Source: DIE 2002HBS. 85 Impoverishing Effects of Health Care Expenditures. To see to what extent health care expenditures can throw people into poverty by preventing them from using the money spent on health care on other essential items, poverty measures before and after health care expenditures were calculated (Table V.20).31 The results suggest that health care expenditures do not leadto a substantial increase inpoverty. These results are fairly consistent across the various regions, with the impact being lowest in Southeastern Anatolia and the Black Sea, and somewhat higher in the Marmara Region. While still relatively small, the poverty impact o f health care expenditure i s bigger among those with green cards and no insurance, than among those with insurance. The relatively modest overall impact o f health care expenditures on poverty can be explained by the fact that a significant share o f those who do not have insurance but live near the poverty line, simply forego health care (thus negatively affecting their health outcomes), while others do enjoy financial protection through insurance. An additional factor may be the existence o f informal support networks, particularly for drug expenditures whereby those covered by insurance obtain informally-paid prescriptions for uncovered members of their extended families. Table V.20. ImpoverishingEffectsof HealthCareExpenditures Pre-Payment Post-Payment Difference Poor 26.9% 27.9% 1.O% Extremely Poor 1.3% 1.4% 0.1% By Region Marmara 18.8% 17.6% 1.2% Aegean 13.4% 12.9% 0.5% Mediterranean 41.6% 40.4% 1.2% CentralAnatolia 26.7% 25.7% 1.1% Black Sea 26.7% 26.3% 0.4% EasternAnatolia 48.3% 47.4% 0.9% SoutheasternAnatolia 37.9% 37.5% 0.4% By Insurance Status Compulsory 13.7% 14.4% 0.7% Voluntary 19.9% 20.0% 0.1% GreenCard 65.9% 67.4% 1.5% None 43.6% 44.7% 1.2% Note: The post-paymentpoverty line was adjustedby the amount of healthcare expenditure. of those livingjust at the povertyline to account for lower overall spending Source: DIE 2002 HBS. Public Expenditure on HealthCare. Public expenditure on health care grew at an average annual rate o f 7.3 percent between 1999 and 2003 (Figure V.9). The Government's efforts to protect social sector expenditures during the crisis years in 2001-2002 and thereafter have been reflected in an increasing share o f health care spending as a share o f GDP, and o f consolidated government spending. The public financing system o f health care i s highly fragmented and complex, mirroring fragmented service provision. The main public financiers are the Central Government (48 percent o f public health care funding in 2003), and the social security institutions (50 percent in 2003) (Figure V.10). Central Government spending is distributed across four major areas, including health care programs and service delivery through MOH, the green card scheme, civil servants health benefits, and government financing o f social health insurance schemes, when the latter run deficits. Central Government funding i s 3'This follows the methodologydevelopedbyWagstaff andvan Doorslaer(2002). 86 supplemented by limited funding from the provincial administrations and municipal governments, with the latter operating their own facilities insome o f the larger cities. FigureV.9. EvolutionofPublic Sector SpendingonHealth 6.00 15,000,000 b 4.00 10,000,000 c c .--B 2.00 5,000,000 0.00 - '+Health 1999 2000 2001 2002 2003 I ExpenditureslGNP +Real Health Expenditure Source: World Bank Staff Calculations, based on data provided by MOF and SPO. FigureV.10. Compositionof Public Sector Spendingon Health, 2003 State Local Economic Governmen CentraI Governmen Social Securi 48% Funds 50% Source: World Bank Staff Calculations based on MOF, SPO figures. The recently completed Turkey National Health Accounts Study estimates that Central Government funding accounts for somewhat over one-third o f Turkey's health care expenditure, employer contributions account for somewhat less than one-fifth, and households pay for over two-fifths through out-of-pocket payments and contributions to social insurance and private health insurance (Figure ~.ii).~~ 32Turkey Ministryo fHealth School o fPublic Health, National Health Accounts 2000, Draft Report, 2003. 87 FigureV.ll. Compositionof TotalHealthSector Spending(2000) Composition of of Total Health Sector Spending [2000] I CentralGOV I LocalGOY 0SocialSecurityFunds 0Private SocialInsurnace PrivateHealthInsurance HouseholdOOP I Corporations Other I Benefits incidence analysis carried out within the framework o f the National Health Accounts found that public sector spending on health care i s skewed in favor o f the upper-income groups, particularly spending on outpatient care (Table V.21). The top quintile consumes about 23 percent o f total public spending on health care, while the lowest quintile consumes about 15 percent, with the average per capita consumption o f the top quintile being about 50 percent higher than that o f the bottom quintile. These findings are to be expected because the lower-income groups consume significantly less health care than the upper-income groups. This i s particularly the case for those among the lower-income groups without insurance protection. TableV.21. Distributionof PublicSector Spendingon HealthAcross IncomeQuintiles ~ ~~ Quintile Inpatient Outpatient Total % of spendingaccruing to quintiles Q1 16% 15% 15% Q2 19% 16% 17% 43 20% 19% 19% 44 25% 26% 26% Q5 20% 24% 23% Source: Turkey MOH School of Public Health: National Health Accounts, DraftReport, 2003. Budgetary funds are not particularly well targeted toward assuring equitable access o f the entire population. Less than one-tenth o f Central Government funding goes toward the green card system, which i s aimed at facilitating lower-income group access to health care, while over one-fifth goes toward providing civil servants with health care benefits, a population which is traditionally not among the lowest-income groups (Figure V.12). Another fifth goes in subsidies to Bag-Kur and Emekli Sandigi, neither o f which i s specifically targeted toward lower-income groups. Thus, overall, the relatively important public subsidies to health care are benefiting middle- and upper-income households more than the poor, who continue to face significant access barriers to health care. 88 FigureV.12. Destinationof CentralGovernmentFundingfor HealthCare Destinationof Central GovernmentFundingFor HealthCare wealthServices Delivery reen Card Rogram FIVII ServantHealh Benefib m o v Subsidiesto SSI Source:Ministy of Health,Turkey Nabnal Health Pccounts, Drat.Report.2003 89 CHAPTERVI: LABOR InTurkey, as in most countries, poverty is closely correlated with employment status and type ofjob, whether formal or informal. Informally employed or casual workers have a noticeably higher rate o f poverty. In Turkey, unemployment of the household head i s particularly associated with poverty. Education i s a key factor inexplaining employment, and therefore poverty outcomes. Certain sectors of the economy employ more poorly educated people, and poverty rates for those employed in these sectors are higher than average. This chapter uses two sources of information: the Labor Force Surveys (LFS) conducted by the State Institute of Statistics (DIE), and the Household Budget Survey (HBS). The Turkish LFS do not include any consumption information that can be used to determine poverty status, so any findings herein on poverty and the labor market are based on the HBS. Ingeneral, the HBS data confirm the overall trends o f the LFS, but the LFS should be viewed as definitive for measuring unemployment inTurkey, because this is what they are designed for, as opposed to the HBS, which i s not. A. OFFICIAL UNEMPLOYMENT AND LABOR FORCE PARTICIPATION RATES FROM THE LFS Unemployment in Turkey was not especially high, hovering around 8 percent o f the labor force since 1990, but increasedafter the 2001 economic crisis (Table VI.l) and remained at about 10percent in2003. However, unemployment i s affected to a large extent by low levels o f labor force participation, whereby those who do not have work, typically drop out of the labor force and, thus, are not captured in the unemployment rate figures, which are calculated as those reporting they are unemployed and looking for work, divided by the labor force. From one-fifth to one-fourth of males aged 15 and above are not inthe labor force, representing many discouraged workers.33 Table VI.1. Turkey: UnemploymentRate andLabor Force ParticipationRate 11990 11991 11992 11993 11994 11995 11996 11997 11998 11999 PO00 PO01 PO02 nemployment I Labor Force Participation Rate Total 56.6 57.0 56.0 52.2 54.6 54.1 53.7 52.6 52.8 52.7 49.9 49.8 49.6 Male 79.7 80.3 79.7 78.1 78.5 77.8 77.3 76.8 76.7 75.8 73.7 72.9 71.6 Female 34.2 34.1 32.7 26.8 31.3 30.9 30.6 28.8 29.3 30.0 26.6 27.1 27.9 Source: www:lldie.gov.trlENGLISH. 33 For DIElabor force information, note that the age limit o f the labor force was 12 and above for years before 2000, and was 15 and over starting with the 2000 LFS. 90 There i s no sharp difference in unemployment rates between males and females (Figure VI.l), with the interesting result that the female unemployment rate has typically been below that o f males since 1990. However, this result is primarily driven by the labor force participation rates. Unemployment among women i s lower since so few women are inthe labor force. Under 30 percent o f women are labor force participants, and the female rate o f labor force participation in2002 was significantly lower than in 1990. Figure VI.2 shows the low rate o f female labor force participation compared to the male participationrate. Figure VI.1. Turkey: UnemploymentRate 12 0 100 , ~ * 8 0 0Total e 8Male f 6 0 17Female 4 0 I 2 0 0 0 ~ Figure VI.2. Turkey: Labor Force Participation Rate __ * +Total e U +Male a' Female _. 91 B. UNEMPLOYMENT AND INACTIVITY Inthe 2002 HBS, 35 percent o f those aged 12 and above reportedthat they hadworked in a paidjob in the survey month. Of those reporting a paidjob, the poverty rate was 25 percent. Another 43 percent o f those aged 12 and above reported that they did not work, but their poverty rate was essentially the same as the employed (it was 24 percent poor). The difference here is that 22 percent o f the sample are children, and that poverty is concentrated among families with children, as demonstrated in the poverty profile chapter. Turkey's low rate of labor force participation, particularly o f women, i s detailed in the World Bank's, Labor Market Study (forthcoming), and was summarized above. The 2002 HBS results confirm these findings (Figures VI.3 and VI.4). Inthe 2002 HBS sample, 22 percent were children and 10percent were adults aged 60 and above, leaving an adult population of 68 percent. Of the adult population, about half (51.2 percent) reported that they had a paid job in the survey month, but women accounted for only 32 percent of all reporting paidjobs (adults and a few elderly). Figure VI.3. Turkey: Adults Employed Figure VI.4. Turkey: Adults Not Employed -~ 0.4 Male Female 69 92 Unemployment is measured inthe HBS as those who report that they did not work for income or inkind in the survey month, and that they were seelung a job. Only 7.2 percent of those aged 12 and above reported that they were looking for ajob, which again underscoresthe low labor force participationrate in Turkey, becausethose not looking for ajob are not inthe definition of the labor force. Households where the head was unemployed had a poverty rate o f 35 percent compared to 26 percent poor of households whose heads were employed. While households with unemployed heads are poorer, this relatively low difference relates mainly to the fact that so few people are loolung for work, and are thus able to be definedas unemployed, and also to the fact that only 8 percent o f households reported that the head was unemployed (by meetingthe definition ofbeinginthe labor force and looking for work). For example, 39 percent o f those aged 12 and above reported that they were not seelung a job (Table VI.2) for a variety of reasons ranging from factors related to age or family structure (students, housewives, elderly, family/personal reasons) to disability (disability or illness) or to seasonal employment. Table VI.2. Turkey: Poverty Rate of Reasonfor Not Seeking Job Percent o i Poor Non-Poor Percent of Valid* Population Subtotals Total Found ajob but waiting 16.8 83.2 0.1 37,568 0.1 Student 24.3 75.7 25.7 6,879,196 10.1 Housewife 22.7 77.3 42.2 11.373.103 16.6 * The percentof those who answeredthe question or for whom we have data. Notes: Poverty in percentages. Population Subtotals: Number of observations. Source: DIE2002 HBS. Some o f these categories were associated with an elevated risk of poverty (Figure VIS), although the categories were numerically small, such as family/personal reasons for not working, which was given by 6 percent o f the sample. Of those, however, the poverty rate was 37 percent compared to the average of 27 percent. Disability and illness accounted for 1and 2 percent of the responses, but with a poverty rate of 36 and 33 percent, respectively. However, for those very few households headedby a disabled person, almost all were poor (86 percent).34 _ _ _ _ ~ 34 However, these households were only 0.3 percent of those sampled, so this number is too low for robust estimation. It is presentedas indicative. 93 Figure VI.5. Turkey: Poverty and Inactivity c. QUALITY OFEMPLOYMENT In Turkey, there is a strong association between the type of employment and the poverty status of the individual or household. For example, the poverty rate of those aged 12 and above who reported that their job was permanent employment was only 12 percent, but for those who reported their job as casual, the poverty rate was 44 percent (and it was 30 percent for the very few people who described their employment as temporary) (Figure VI.6). The relative risk of poverty for casual work was thus 3.7 times greater than for permanent employment. Figure VI.6. Turkey: Poverty and Employment Situation 50.0 45.0 40.0 35.0 c 30.0 Poor 2 g225.0 20.0 _ _ _ _ _Poor~ +Total _ 15.0 10.0 5.0 0.0 Permanent Temporary Casual type of employment employmentI work under contract for a fixed period ~~ ~ 94 In a different question on the same theme, virtually the same percentages of poor and non-poor were observed for casual employees and regular employees (corresponding to permanent employment) (Figure VI.7). FigureVI.7. Turkey: PovertyandEmploymentStatus I I 1 ' 50.0 45.0 40.0 35.0 ' 30.0 22 25.0 20.0 15.0 10.0 5.0 0.0 Inaddition, povertyrateswere higher for the self-employed and unpaid family workers. Reflecting the finding on the low rate o f poverty for apprentices, it is important to note that the poverty rate o f trade union members was only 7 percent (but only 2 percent o f those aged 12 and above reportedmembership), while the poverty rate o f nonmembers was 20 percent. However, only 16percent o f the sample answered the question on trade unions, so the true rate o f poverty among non-trade-union members would be higher than the total poverty rate o f 27 percent. Poverty is sharply associated with a lack o fregistration at a social security institution. Conversely, formal employment as measured by enrollment in social security i s a strong bulwark against poverty in Turkey. It i s important to emphasize that reliable figures on who i s enrolled in social security are difficult to obtain. As discussed in the health chapter, the rate o f coverage o f social security i s approximately two- thirds o f the population. Inthe 2002 HBS, o f those 35 percent reporting employment, about 32 percent reported enrollment in social security (about 15 percent o f those aged 12 and above). All lunds o f social security enrollees had poverty rates well below the total rate o f those who responded to this question (Figure VI.8). 95 Figure VI.8. Turkey: Poverty and Social Security 40.0 ;%:835.0 'I $ 20.0 15.0 10.0 iI 5.0 0.0 Of these formal jobs, the most significant in terms o f employees facing a lower risk o f poverty are government employment and employment in state-owned enterprises. The poverty rate o f government employees was 10 percent, and for state-owned enterprises (SOEs), it was only 5 percent (Figure VI.9). However, it should be emphasized that few people surveyed have such empIoyment--onIy 4 percent are government employees-and less than 1 percent o f those aged 12 and over reported working in an SOE (Table VI.3). Figure VI.9. Turkey: Poverty and Type of Workplace 30.0 25.0 20.0 US 22 015.0 Poor +Total Poor- 10.0 5.0 0.0 Private PublicI State-Owned Government Enterprise 96 Table M.3. Turkey: Poverty Rate of Status of Workplace Notes: Povertyin percentages. dation Subtotals: Number of observations. Poverty i s associated with the size o f the enterprise. People employed inenterprises o f 1 to 9 people had a poverty rate o f 30 percent, but only 7 percent o f those employed in firms employing 50 people and above were poor (Figure VI.10). Seventy percent o f respondents aged 12 and over reported that they worked in a firm o f 1 to 9 people, so the higher poverty rate for this category means that it i s a major factor influencing the absolute number o f poor employees. Figure VI.10. Turkey: Poverty and Size of Firm 35.0 30.0 _-- -_____ I 25.0 Poor 11 g2 C 20.0 +Total Poor 1 n Q 15.0 10.0 5.0 0.0 119 10124 25/49 50 and persons persons persons over Another way to gauge informality i s to look at the legal status o f the respondent's workplace, assuming that more individually owned workplaces would be likely to be informal, while few incorporated companies would be. Usingthese criteria, it i s clear that the highest poverty rate o f those aged 12 and above (3 1percent) was found inthose who reportedthat their workplace status was individual ownership, and the lowest poverty rate o f 6 percent pertained to incorporated companies (Table VI.4). 97 Table VI.4. Turkey: Poverty Rate of Legal Status of Workplace InTurkey, poverty is associated more with lackofwork than with work. For example, the meannumber o f hours worked by the poor was 43.4 per week, but by the non-poor it was 46.3 hours. The poor also seem to persist inthe same low-payingjobs longer than the non-poor-mean duration o f employment was 12.9 years for the poor, but only 11.7 years for the non-poor. For the 9 percent o f the sample who reported employment o f more than 19 years, the poverty rate was essentially the same (28 percent) as the average o f 27 percent. However, the poverty rate for those injobs less than nine years was slightly below (24 percent) the average of 27 percent, suggesting that mobility i s slightly associated with lower risk o f poverty. D. SECTOR OFEMPLOYMENT The largest sector interms o f employment inTurkey i s agriculture (including forestry and hunting). Of the 35 percent aged 12 and above, fully 40 percent are engaged inagriculture (Table VIS). Agriculture is also the sector with the highest poverty rate o f those employed in it, at 36.5 percent. The next-highest poverty rate i s that o f construction, at 36 percent, but accounting for only 5 percent of employment, as reported inthe HBS.35 This latter finding is curious because construction accounts for 4 percent o f GDP, and i s generally measured as a larger share o f employment according to labor force survey data (see World Bank, Labor Market Study, forthcoming). The poverty rate is lowest for mining and quarrying, where it is under 3 percent poor, but less than 1percent o f employees are employed inthat sector. After agriculture, the other significant sectors in terms o f employment are manufacturing (15 percent), and wholesale and retail trade (14 percent), both o f which have poverty rates below the average o f 17 and 18 percent, respectively. Interms o f poverty, "other social, public, and personal services" have a higher povertyrate (30 percent) than average, but account for only 2 percent o f employment. 35 Note that according to the LFS, construction accounted for 4.5 percent of employment in2002 and 6.1percent in 2003. 98 Table VIS. Turkey: PovertyRate ofBasic Code ofMainActivity of Workplace I Percent I Population 1Percent Poor Non Poor of Valid* Subtotals of Total Agriculture, Hunting, Forestry, Fishing 36.5 63.5 40.5 9,786,383 14.3 Mining and Quarrying 2.6 97.4 0.7 168,655 0.2 Manufacturing 16.8 83.2 15.3 3,708,013 5.4 Electricitv. Gas and Water 16.6 83.4 0.5 I 121,490 0.2 r l Construction 35.9 64.1 5.3 1,280,230 1.9 Wholesale and retail trade and hotels and 19.2 80.8 17.0 4,101,936 6.0 restaurants Transportation communication and 14.1 I 85.9 I 5.3 I 1,285,5531 1.9 storage services I I Financial services and real estate I I lComunitv andDersonalservices 15.3 I 84.7 I 13.0 1 3,176:20914.6 Total Not reDortin2 emdovment I I 44,233,8571 64.7 * Thepercentofthose who answered the questionor for whom we havedata. Notes: Number of work hours per week 4 0 . Povertyinpercentages. PopulationSubtotals: Number of observations. Source: DIE 2002 HBS. 99 CHAPTERVII: SOCIAL PROTECTION Social protection inTurkey consists primarily o f limited formal systems of pensions and social assistance, supplemented greatly by informal mechanisms. The role o f informal coping mechanisms, particularly interhousehold transfers o f food and other assistance, is documented inthe World Bank's report, "Turkey: Poverty and Coping After Crises" (2003). For social insurance, the primary informal mechanism i s the extended family, with elderly members receiving significant support from children and other relatives. This mechanism works well to keep most elderly from poverty in general, as documented in the poverty profile chapter, but i s under increasing pressure, particularly in urban areas (World Bank 2003b, UNDP 2003). For social assistance, informal mechanisms are also important. Inrural areas, strong social solidarity usually results in families o f the "deserving" poor (usually widows with young children) receiving informal transfers that keep them from extreme poverty. These rural ties are strong enough to extend to the urban gecekondu (slum) areas, through networks o f people from villages o f origin, called hemseri. Additionally, religious charity plays an important role. Inqualitative interviews conducted for the Social Risk Mitigation Project, and in general poverty monitoring, some urban very poor reported charity, includingrent-free housing ingecekondu areas. Formal elements o f social protection in Turkey are the pension (social security) system, and the Social Assistance and Solidarity Encouragement Fund (SYDTF) and its affiliated 931 Social Assistance and Solidarity Foundations (SYDVs). A. TURKISHPENSION SYSTEM Turkey's current social security system is highly fragmented, with benefits and contributions dependent on a person's occupation. The bulk o f the covered labor force falls under the Social Insurance Organization (Sosyal Sigortalar Kurumu, SSK), the system that covers private sector workers and those public workers who do not qualify as civil servants. Civil servants are covered separately under Emekli Sandigi (ES), and the self-employed and farmers are covered by a third scheme, Bag-Kur (BK) (Box VII.1). There is a small noncontributorypensionscheme availableto those who reach age 65 andhaveno means o f support. This noncontributory scheme i s administered by ES, but i s financed by general revenue transfers to ES. Finally, various groups o f workers are covered by separate occupational schemes. These usually are voluntary, and additional to the existing public schemes. However, in the case o f some banks and the Central Bank, for example, their previously existing schemes were grandfathered so that workers contribute to the separate schemes in lieu o f contributing to the larger public schemes. Overall, 42 percent of the labor force is contributing to one or the other o f the schemes, with the bulk of the coverage in SSK. Of the 42 percent o f the labor force covered, 48 percent are covered in SSK, 22 percent each inES and inthe self-employed scheme under BK, and an additional 8 percent inthe farmers' scheme under BK. The total number o f contributors to all schemes i s around 11million. 100 Box VII.1. Parameters of the CurrentPensionSystem SSK Retirement age is 58160 for full and 58160 for partialretirement 7,000 days o f contribution (19.4 years) are required for full retirement Replacementrate i s 54 percent for full retirement Average earnings inwhole working life are considered incalculating pensions Present value o f earnings is calculated through inflating each year's earnings by associated Consumer Price Index (CPI) and gross domestic product (GDP) for the period Pensions are indexed to inflation Contribution rate is 20 percent Insurable earnings are indexedto CPI and GDP. Bag-Kur(self-employed) Retirement age is 58160 for full and 60/62 for partial retirement 25 years o f contribution are required for 111retirement Replacement rate is 65 percent for full retirement Average earnings inwhole working life are considered incalculating pensions Present value o f earnings is calculated through inflating each year's earnings by associated CPI an( GDP for the period Pensions are indexed to inflation Contributionrate is 20 percent Insurable earnings are indexedto CPI and GDP. Bag-Kur (farmers) Retirement age is 58160 for full and 60162 for partialretirement 25 years of contribution is required for full retirement Replacement rate is 70 percent for full retirement Last income step where contributions are paid i s considered incalculating pensions Pensions are indexed to civil servant salary increases Contributionrate is 20 percent Insurable earnings are indexed to increase incivil servant salaries. ES Retirement age is 58160 for full and 60162 for partial retirement 25 years o f contributionis requiredfor full retirement Replacement rate is 75 percent for full retirement Last salary i s considered incalculating pensions Pensions are indexedto civil servant salary increases Contribution rate is 20 percent Insurable earnings are indexedto increase incivil servant salaries. 101 On the beneficiary side, only 29 percent o f the population over age 65 is receiving an old-age pension from any o f the public schemes. Of that 29 percent, 47 percent receive pensions from SSK, 18 percent from ES, 30 percent from the self-employed scheme in Bag-Kur, and only 5 percent from the farmers' scheme inBag-Kur. A total o f 1.2million people over age 65 are receiving old-age pensions, with about 25 percent more receiving survivor and disability pensions. The differences in distribution between beneficiaries and contributors among the schemes largely arise from evasion in Bag-Kur, where individuals frequently do not pay contributions untiljust before retirement, and then make a large lump- sumpayment that is supposed to representthe cumulative value o fallpast-due contributions. On the basis o f these statistics, one would not suspect major problems with the Turlush pension system, aside from the evasion in Bag-Kur, given that the total number of contributors is far larger than the pensions going to the elderly. However, in 1991, the minimum retirement age was abolished, which, coupled with short minimumcontributory periods in SSK o f less than 15 years, and only slightly longer periods inES and Bag-Kur o f 25 years for men and 20 years for women, meant that individuals were able to retire as early as age 34 inSSK, and as early as age 40 inES and Bag-Kur. While there are 1.2 million people over age 65 receiving pensions, there are 3.1 million people below age 65 receiving old-age pensions, with an additional 882,000 people, at a minimum, below age 65 receiving disability and survivor pensions. Age data for ES and Bag-Kur survivor pensioners do not exist, so there are probably even more young people collectingpensions. The 1999 reform sought to remedy this problem by reestablishing a minimumretirement age. However, the reform i s to be phased inextremely slowly, with the retirement age as low as 38 for women and 43 for men for the first cohorts retiring after the reform. The retirement age will eventually be 60 for men and 58 for women, but these ages will not apply to cohorts retiring before 2034. Both the 1991 law and the retirement age aspect o fthe 1999 reformappliedto all three pension systems, although the impact on SSK i s far greater than on the other systems, since its shorter minimum contributory period had resulted in such low retirement ages initially. SSK. By far the largest system, SSK covers mostly private sector workers. Employers are required to contribute 11 percent o f wage to pensions, and employees contribute 9 percent o f wage. The total contribution for all social programs i s 15 percent for the employee, with 5 percent for health and 1percent for unemployment insurance, in addition to what i s paid for pensions and 21.5 percent from the employer who pays 6 percent for health, 1percent for maternity, 1.5 to 7 percent for work injury, and 2 percent for unemployment insurance on top o f what is paid for pensions (Table VII.l). Relative to even older Organisation for Economic Co-operation and Development (OECD) countries, these contribution rates are quite high. OECD average contribution rates are 19.4 percent o f wage for pension and 31.9 percent for all social programs, but include countries such as Canada, with rates as low as 6 percent for pensions. The average for Latin American countries, which are closer demographically to Turkey than the older OECD countries, i s 12 percent contributions for pensions and 21.6 percent overall. Table VII.1. ContributionRatesto SSK (includingunemploymentinsurancepremiums) Employee Employer Health Insurance 5% 6% Matemity 0 1% Work Injuryand Occupational Diseases 0 1.5%-7%' Pension 9% 11% Unemployment Insurance2 1% 2% Total 15% 21.5% 1. The rate is subject to risks of injury and disease in aparticular sector. 2. There is also a 1 percentstate contributionfor unemployment insurance. 102 The 1999 law changed most o f the SSK benefit parameters, but since the benefits received even by new pensioners consist o f a combination o f old system benefits and post-reform benefits proratedby years of service under each system, it i s necessary to describe the benefit structure both pre-reform and post- reform. Pre-reform, benefits used to be 60 percent o f the average o f salaries from the last five years plus 1 percent additional for each additional 240 days, with individuals able to retire having contributed for 5,000 days, or 13.9 years. Inaddition, workers were to have been members o f SSK for 25 years ifmale, and 20 years if female. A second retirement option existed whereby workers who had reached age 55 for males and age 50 for females could retire with 54 percent o f last salary, with only 3,600 days of contributions, or 10 years. Contribution ceilings and floors and maximum and minimumpensions were adjusted by Parliament in an ad hoc fashion, with some periods where the contribution ceiling actually fell below the minimumwage. Pensions post-retirement were adjusted by growth in civil servant wages. Duringperiods o f crisis, the adjustments could lag inflation considerably, while during boom periods, pensioners did quite well. In addition, in 1984, Parliament instituted a social assistance payment to all pensioners. The nominal amount o f this flat payment was decided annually by Parliament. By 1996, this flat payment had become so large that it exceeded the average pension level, completely unraveling any relationship between pensions and contributions. Since 1996, this flat payment was frozen in nominal terms, and is now only about 1percent o f pensions, but it i s still gwen to all pensioners, including those just retiring. Post-reform, the requirements are a little more stringent, with workers required to contribute 7,000 days, or 19.4 years, for full retirement, although this change is also being phased inslowly for current workers. Full retirement benefits are 54 percent o f lifetime career average, although the average is computed by revaluing the nominal wages by growth innominal GDP. Workers are still required to belong to SSK for 25 years if male, and 20 years if female, and the new minimum retirement ages are being phased in slowly. A partial retirement option i s still available, but at the normal retirement age o f 60 for men and 58 for women, effective immediately. Workers need only 4,500 days o f contributions, or 12.5 years, under this option, and will receive 40 percent o f their career average salary with 2 percent additional upto a total o f 25 years, and 1.5 percent subsequently. A major innovation o f the new law was the removal o f ad hoc adjustments, replaced by automatic indexation rules. The minimum salary on which contributions were to be paid was indexed to nominal GDP growth. The maximum salary on which contributions were to be paid was set at five times the minimum insurable earnings (three times in the original law, but subsequently raised to five). Pensions were to be automatically indexedto inflation on a monthlybasis. One issue that has emerged from the automatic indexation i s that a discrepancy has arisen between minimumwage and the minimuminsurable earnings. Particularly since the 2001 fiscal crisis, minimum wage, which is still adjusted on an ad hoc basis, has not risen as fast as nominal gross national product (GNP) growth. Thus, minimuminsurable earnings are almost 50 percent higher than minimum wage. Employers are currently required by law to pay the full contribution, both employer and employee portion, for the difference between the two. This has been remedied by legislation passed in June 2004 that makes the minimum insurable earnings equal to minimum wage, and the ceiling equal to 6.5 times minimumwage, to leave it at itsprevious level. Bag-Kur. Bag-Kur primarily covers the self-employed and some farmers. Contribution rates are 20 percent for pensions and 20 percent for health coverage. As with all systems that cover the self- employed, there i s the perennial problem o f evaluating the income earned. Turkey initiated a system of minimumearnings steps that are attributed to individuals regardless o f what they actually earn. Most people declare earning level 1 in their first year o f contribution, and are automatically raised to the next level the following year. This occurs for the first 12 years. Subsequently, automatic increases occur only 103 every two years. Workers o f course have the freedom to declare a higher earning level anytime they choose. The income levels associated with each step are different for the self-employed and for farmers, with farmers' income substantially below that of the self-employed. Prior to the reform, there were only 12 steps, with automatic increases in the first six years, and then only at the worker's discretion beyond that. The nominal earnings level associated with each step i s now also automatically indexed to nominal GDP growth. However, it should be noted that Bag-Kur has very l o w collection rates for its contribution revenue. Typically, workers pay very little if anything during their working years. Just prior to retirement, they pay Bag-Kur a lump sum equivalent to the past-due contributions, with interest, and then receive their retirement. Prior to reform, Bag-Kur maintained the same benefit structure for the self-employed and for farmers, with each getting 70 percent o fthe last earning step after 25 years o f service at any age, and 60 percent o f the last earning step after 15 years o f service at age 55 for males and age 50 for females. Pensions were indexed by growth in civil servant wages. A flat social assistance payment was provided in Bag-Kur as well, but its nominal amount had been frozen early on and only rose briefly in 1995 and 1996 before being frozen permanently. The reform affected the benefit structure for only the self-employed. The farmers were allowed to retain their previous structure, but with retirement ages imposed. The self-employed benefit structure became identical to that o f SSK, with individuals receiving 65 percent for 25 years o f service and 45 percent for 15 years o f service. Pensions were based on the full earnings history (as defined by the steps). Pensions were also now automatically indexedto inflation on a monthly basis. As with SSK, the earnings steps are rising with nominal GNP growth, which exceeds the rate o f wage growth, potentially causing difficulties for workers who try to comply with the system. EmekliSandigi. Emekli Sandigi is the program that covers civil servants, including military personnel. The financing o f Emekli Sandigi i s somewhat different from the other plans in that health insurance during working years is not covered by the pension fund. Rather, it is covered directly by the line ministries with which the civil servants are employed. The pension fund covers only retirement age benefits, pensions, and health services during retirement. For this, a 36 percent contribution is collected, of which 20 percent comes from the employer and 16percent from the employee. No distinction i s made between revenue for health or for pensions, but in 2003, about 30 percent o f expenditure came from provision o f health services, and 70 percent was derived directly from pensions, suggesting that about 26 points o f the 36 percent contribution finance pensions, with the rest financing health. Another distinction between Emekli Sandigi and the other schemes i s that the basis for contributions and the basis for benefits are different. Contributions are paid on the basis o f basic salary. Depending on the grade o f the worker, this basic salary may represent as little as 60 percent o f the worker's total cash remuneration. When pension benefits are paid, they are paid on the full remuneration, including all the bonuses, and so forth. Thus, from the initial design, there is botha financing gap inEmeMi Sandigi and an equity issue, whereby lower-grade workers pay contributions on a larger share o f their salary than higher-grade workers. While the retirement age change includes civil servants, the 1999 law otherwise left civil servants untouched. Their benefit as always i s 75 percent o f last salary, based on all remuneration, with increases after retirement indexedto growth incivil servant wages. Infact, it i s slightly more generous than that. If a worker retires from a particular grade, his or her pension i s tied to the growth in salary for that job. The retirement age change has had almost no impact yet because civil servants had to work a minimumo f 25 years for a man and 20 years for a woman before collecting a pension. Since the legal retirement age i s currently about 43 for a man and 40 for a woman, with the very slow phase-in period, virtually no one is affected by the change yet. 104 Noncontributory pensionbenefits. Turkey also provides a small noncontributory benefit to those over age 65 who earn below the level o f the benefit. Currently, Emekli Sandigi pays 1,050,852 noncontributory pensions, of which 207,578 are elderly men (over 65), 424,623 are elderly women, and 139,046 are elderly couples. In addition, 48,554 men over age 18 with 70 percent or more disability receive benefits, as do 24,082 women and 908 couples inthe same group. Finally, 136,949 men over age 18 with disability of 40 to 70 percent receive it, as do 61,313 women and 7,799 couples. The amount as of March 2004 was TL 54.63 million per month for a single person, and TL 81.945 million (50 percent more) for a couple. The amount i s paid quarterly through banks if in urban areas, but through the Postal Telephone & Telegraph (PTT) system if in rural areas. The amount was initially two-thirds o f minimum wage, but i s now about one-sixth. In addition, the person receives an identification card that provides outpatient care at state hospitals, but no medicines. Inpatient care i s provided by the hospital itself. The pensionis exempt from all taxes. Eligibility is on the basis o f need as defined by Law 2022 passed in 1976 and enacted in 1977. People have to apply through the district administrative council where their birth record is registered. They can do so by mail. A six-person board decides whether the applicant is needy. For the disabled, a health board of three physicians determines whether the person is truly disabled. Bothproperty and income are counted, but primarily property that generates income. For elderly living with younger family members, the income o f the son or daughter is taken into consideration. Ifthe son or daughter can afford to give the parent the equivalent o f the benefit, the parent will not receive it. If people are receiving the pension amount in other income, they will not be considered needy. If they receive even 1 lira less, they will receive the full amount from Emekli Sandigi. There are poor people who choose not to apply. One reason i s that they prefer to qualify under SSK or Bag-Kur ifthey have a son or daughter working, because they can get health benefits with pharmaceutical coverage under SSK. But they have to give them up if they qualify under Emekli Sandigi. When a person receives the pension, his or her birth office i s notified and automatically informs Emekli Sandigi o f death, since that also i s registered at the birthoffice. There i s some crosschecking with SSK and Bag- Kur databasesto verify that the individual is not receiving a pension from one o f those systems. Budget issues do not impact how many people are paid; ifthey qualify, they are paid. Issuesin TurkishPensions. The obvious main issue inthe Turkishpension system is its lack o f fiscal sustainability. Figure VII.l shows the fiscal situation o f the pension system, with overall deficits projectedto be 4.5 percent o f GNP in2004, after havinghit a low o f 2.57 percent of GNP in2000. FigureVII.1. Budgetary Transfersto Social Security Institutions % of GNP 1 1994 1996 1998 2000 2002 &SSK +Bag-Kur -&-- EmekliSandigt --rc-Total 105 While SSK is still slightly below its pre-reform deficit, the large growth in deficits comes from Emekli Sandigi, which as noted, had made few changes in its policy since the 1999 reform. While the picture improves somewhat as the slow retirement age is phased in, the increases in life expectancy in Turkey overcome the pace o f reform and, inthe longterm, the deficits increaseto 7 percent o f GDP. But along with the unsustainability of the pension system is the view that pensions are quite meager in Turkey. While seeming to be a contradiction with unsustainability, the system provides incentives that encourage individuals to contribute for too few years, and to retire early, so that a small pension is provided to many people, rather than focusing on those unable to work due to age or disability. As a result, most younger retirees continue to work while collecting a pension, and then feel the loss of the work income when they truly retire. Life expectancy at retirement i s about 28 years for men and 34 years for women. Thus, the men and women who are currently retiring from SSK with 15 to 17 years of contributions, can be spending almost twice as much time collecting pensions as they do contributing to them. Obviously, a system cannot work this way. Usually pension systems count on contributions from a large group o f contributors to sustain a small group o f beneficiaries. In SSK, there are not even two contributors per beneficiary, and the other systems are even worse. Since most o f the younger retirees work, but do not pay contributions, the practice o f work without contributions has become ingrained in Turkishsociety. While the retirement age issue is being addressed in the 1999 reform, although at a very slow pace, the number of years required to collect a pension is risingfrom 15 years to only 19.4 years. Even at that rate, people will spend as much time or more inretirement as inworking and contributing. Inmost countries, people contribute twice or three times as long as they collect benefits, so the Turkish situation is quite abnormal. On top o f that, the benefit i s front-loaded. Workers in SSK receive 35 percent for the first 10 years o f service-3.5 percent per year. This drops to 2 percent for the next 15 years o f service, resulting in workers who retire with 19 years o f service collecting a benefit equal to 53 percent of their average salary. Ifthey work an additional year, they get only 55 percent. It is not worth the effort for the worker. And if the worker works beyond 25 years, the additional benefit accrual drops to 1.5 percent, removing any incentive for workers to continue contributing beyond 25 years. This needs to be compared with workers in other countries who regularly accrue 35 to 45 years o f service. As a result, even as the retirement age increases, workers are going to choose to evade a large part o f their working career to avoid the fairly high employee contributions (21 percent), and the culture o f workers all contributing never gets established. Benefit systems usually pay 1to 2 percent, closer to 1percent per year o f service. Even ifpeople are allowed to retire with 20 years o f service, iftheir benefit is only 20 percent o f average salary, they might think twice about retiring as early as possible. Along the same lines, the target replacement rate for 25 years o f work in Turkey i s 65 percent o f salary. The International Labor Organization (ILO) targets 40 percent replacement after 30 years o f contributions. Thus, by intemational standards the pension system in Turkey appears to be overly generous. Comparing Turkish workers with Turlush pensioners, out of a salary of 100 percent o f gross wage, workers pay 9 percent for pensions, 5 percent for health insurance, and 1 percent for unemployment insurance. Minimum-wage workers pay inaddition 15 percent o f salary for income taxes, with higher rates on higher-income workers. Thus, the take-home salary o f a worker is only 70 percent o f gross wage. If the pension, which i s not subject to income tax, i s 65 percent o f gross salary, workers in retirement receive 93 percent o f their take-home salary. Duringworking years, workers usually support children, and perhaps parents. During retirement years, they usually support only themselves, so they require less income. Other expenses, like commuting and work clothes, also fall in retirement. Thus, pensions seem overly generous, by bothTurlush and intemational standards. 106 The one issue that i s consistent with both the perception o f low pensions and the generosity o f the system i s under-reporting o f earnings. More than 50 percent o f workers in SSK report the minimum insurable earnings. InBag-Kur, through the step system, workers cannot declare the same level o f earnings year after year, but no one voluntarily declares higher earnings than required. If workers are in fact earning substantially more, but together with their employers are under-declaring their earnings, they will o f course end up with low pensions. The current system appears to reward this behavior by awarding large pensions for minimumyears o f service. Inaddition, periodically, the Government grants additional social support payments to pensioners, which are flat amounts per pensioner. Despite the 1999 law, the Government did this in 2003, which raises the level o f pensions to acceptable levels without raising contributions, and sends the message to workers that it i s acceptable to under-declare earnings. Pensions will be provided at a reasonable level through the political process inany case. Relatedto this is the level of contributions. The overall labor charges are 36.5 percent of the wage bill. It is no wonder that workers and employers choose to evade. But as a result o f this evasion, the pension system i s not able to fulfill its primary role, which is to replace a worker's income during retirement. Workers then look to the political process to deliver on income replacement, which further undermines the system. Finally, it has to be noted that given the large segments o f the population not receiving pensions, fiscal resources spent to bail out the pension system, which covers primarily upper- and middle-income individuals, end up with a regressive impact on the overall distribution o f income, with resources drawn from a broader tax base being used to support pensions for a narrower tax base. Ideally, a pension system with less than full coverage should be fiscally sustainable through employee and employer contributions alone. While the Turkish pension system may fulfill the role o f reducing poverty in old age to a limited extent, the above discussion should make clear that the system has significant problems that prevent it from playing an even more constructive role. The Government is in the process o f unifying all the pension systems, and in doing so reviewing the parameters that would apply to the new unified system. This process provides an opportunity for improving the system by removing some of its more severe flaws. B. SOCIAL SOLIDARITY FUND The Social Assistance and Solidarity Encouragement Fund (SYDTF) was established in 1986 as an umbrella organization and financing entity for 931 regional affiliate Foundations (under Law No. 3294, which entered into effect on May 14, 1986). The Foundations are called vakzjZar in Turkish and in acronym, SYDVs. The purpose o f the SYDTF was defined as, "TO aid poor and destitute citizens in circumstances o f need and, as necessary, those who have been accepted in Turkey or have traveled here by whatever means, to ensure the distribution o f wealth in an equitable fashion by taking measures to improve social justice and to encourage social assistance and solidarity." The Committee o f the SYDTF i s made up o f a Prime Ministry Undersecretary, an Interior Ministry Undersecretary, a Health Ministry Undersecretary, and the General Director o f Foundations under the chairmanship of a State Minister appointed by the Prime Minister. Decisions o f the Committee enter into effect upon their ratification by the Prime Minister. The Fund Administration carries out its services through the 931 local Social Assistance and Solidarity Foundations established throughout Turkey under the chairmanship o f Provincial and Sub-Provincial Governors. The SYDTF i s an extrabudgetary fund financed by earmarked taxes and administered by a Cabinet Minister. The regional affiliate foundations provide a variety o f social assistance programs (usually in 107 kind) to the needy. The needy are determined at the discretion o f the regional affiliates. The SYDTF, together with its local affiliates, i s the largest social assistance program in Turkey in terms o f number o f beneficiaries (4.2 million from July 1, 1997 through March 26, 1999).36 The type o f assistance provided by the SYDTF and its regional affiliates is shown in Table VII.2. Project assistance refers to employment-generating projects, shown inTable VII.3. General revenues and expenditures are presented inTables VII.4 andVIIS. Table V11.2. Types of SYDTFAssistance, July 1,1997-March 26,1999 (in trillionTL) Other 1.3 1 Total 128.4 100 4,232,022 --- = available. Not Source: Cumhuriyet hukumetleri doneminde soysal hizmetler, ozurluler ve sosyal yardimlar. Ankara: Nisan, 1999. 36Turkey also has a program of Old Age and Disability assistance formulated under Law 2022, which provides benefits to the elderly aged 65 and above, and to those who are more than "40 percent" disabled, discussed in the pension section. In September 1998, there were over 900,000 beneficiaries, 80 percent of whom were elderly and 20 percent of whom were disabled(WorldBank 1999 Living Standards Assessment). 108 Table VII.3. SYDTF Employment-GeneratingProject Assistance, July 1, 1997-March26, 1999(millionTL) Type of Assistance Amount Share No. of People (percent) Greenhouse 2,467,5 17 16.34 9,857 Poultry 315,03 1 2.09 27,656 Carpedtapestry 611,328 4.05 16,598 Confection 469,350 3.11 2,500 Small handicrafts 55,865 0.37 449 Milk 1,171,004 7.75 11,345 Fishery 206,175 1.36 587 Beekeeping 1,436,048 9.5 1 4,957 Growing h i t s 898,051 5.95 36,943 Cowbreeding 3,634,746 24.06 13,434 Sheep breeding 2371,015 17.02 11,873 Miscellaneous 512,043 3.39 12,688 Disabledpeople 121,480 0.80 979 Startingownbusiness 69,740 0.46 185 Growing vineyard 175,039 1.16 1,434 Growingplants 239,190 1.58 9,602 Culturemushrooms 151,865 1.01 672 TOTAL 15,105,487 100 161,759 Central Social Solidarity Fundrevenues are comprised o f 0 A 10percent sum to be transferred from funds that exist or are to be established by statute or decreewith the force o f law under the Decision of the Council of Ministers 0 Payments inserted into the budget 0 Halfthe revenue o f fines for motoring offenses A 2.8 percent share o f the total payments madeby income and corporationtaxpayers 0 A 15 percent share to be transferred from TurlushRadio andTelevision's advertisingincome 0 All types of donations and assistancefrom outside 0 Other income. 109 TableVII.4. SYDTFRevenues Fuel consumption tax )229,000[ 0) 0 Fromnational budget 01 01 01 0t230,OOO Incomes ofprevious years I505,000[ 7,7951 2,0941 1,8381 1,303 I I I I I Loan Account of World Bank 0 0 0 3,313 7,803 Other Income 17,114 102,050 2,090 4,380 9,155 Total income 454,7241519,627 841,698 890,217 826,583 Source: SYDTF TableVIIS. SYDTFExpenditures 1999 2000 2001 2002 2003 Source: SYDTF. 110 Local Social Solidarity Foundations. The Social Assistance and Solidarity Foundations (SYDVs) established in the provinces and sub-provinces are legal entities governed by specific law. The chairpersons o f the local SYDVs are senior local government officials. The Board o f Trustees, which i s chaired inprovinces by the Provincial Governor, and insub-provinces by the Sub-Provincial Govemor, i s composed in provinces of the Municipal Mayor, the Security Director, the Provincial Head of Finance, the Provincial Director of National Education, the Provincial Health Director, the Provincial Director of Social Services and Child Care, and the Provincial Director of Religious Affairs. In the sub-provinces, it is composed of the Municipal Mayor, the senior security officer of the sub province, the Finance Director, the Sub-Provincial Director o f National Education, the senior Health Ministry official o f the sub province, and the Sub-provincial Director o f Religious Affairs. For each period o f activity, three members o f the Administrative Committee are appointed by the Provincial Governor from among philanthropic citizens. The implementation o f assistance programs in the provinces and sub-provinces enters into effect through the decision of the Board o f Trustees. The staff working within the Foundations are appointed by the Board o f Trustees in accordance with Labor LawNo. 1475. Each Foundation has a separate legal entity and no hierarchical connection with the central Fund Administration. Foundations are independent intheir decision-making. Taking into consideration the population structure of the province and sub-province, the socioeconomic development index, and other social factors, each monthresources are transferred to the Social Assistance and Solidarity Foundations (SYDVs) in the provinces and sub-provinces to meet the daily needs (for foodstuffs, clothing, housing, health, fuel, and so forth) o f economically and socially deprivedpeople and families. Support is provided for business establishment, occupational training, and employment projects (in fields such as beekeeping, fruit cultivation, hothouse cultivation, carpet and rug malung, and handicrafts) directed toward bringing people and families who are unable to participate in production by reason of economic deprivation to a state inwhich they can do so. The medical expenses (for medicine) of outpatients with green cards who are unable to meet health expenditures from their own means, and the costs of medicine and treatment of patients not covered by social insurance, are met by the SYDTF. In addition, all types of equipment needs of physically handicapped people (for hearing aids, prostheses, wheelchairs, and so forth) fall within the scope of health assistance. Social assistance to destitute citizens who suffer losses due to natural disaster, fire, and terrorism, and social support programs encompassing the provision o f fuel, clothing, foodstuffs, and housing aid, are implemented. Funds are transferred to all the SYDVs at the start of the winter season for fuel, at the start of the school year for education, and before religious festivals for foodstuffs and clothing. Soup kitchens are opened inareas of intensive migration, and where there is obvious unemployment and poverty, As a new element under eight-year basic education, in order to make a contribution to the fundamental problems such as those o f accommodation of and food for needy students who travel to the centers where schools are located in the implementation of education requiringtransportation, funds are transferred for the purpose of providing these students with a mid-day meal. The assistance is passed on through the SYDVs. 111 For the purpose o f solving the accommodation problems of students in middle-school education, an educational service i s provided in various provinces and sub-provinces by constructing middle-school student hostels. With contributions provided from the SYDTF, the SYDVs have provided funds for both the construction and furnishing and decorating of hostels, thus assuming an important responsibility within the education system. The scholarship program for young people in universities has continued from 1989to the present. Contributions are also madeby the SYDTF to orphanages, old people's homes, and rehabilitation centers aimed at the accommodation and rehabilitation of the weak and destitute. C. CONDITIONALCASHTRANSFERS Turkey's social assistance system, provided by the SYDVs with financing from the SYDTF, recently underwent an important innovation. Under a loan financed by the World Bank (the Social Risk MitigationProject, SRMP), Turkey began a nationalprogramo f conditional cashtransfers (CCTs). CCTs are payments made to the mothers of poor children, provided that they attend school or health clinics (Table VII.6). CCTs have proved to be highly effective in raising the enrollment rates of girls in secondary school in Mexico, and in increasing attendance in general. There are large CCT programs in Latin America-the largest i s in Brazil, followed by World Bank-supported programs in Colombia and several other Latin American countries. Turkey is the first country in Europe to have adopted CCTs as a formal program supported by the Bank, although the social protection systems in most European countries provide for child allowances, often conditioned, as in France and Hungary, on school attendance. CCTs are an important tool inthe arsenal of poverty-fighting interventions inTurkey. They are targeted to the poorest of the poor, many o f whom stated in qualitative interviews that they would not be able to afford the out-of-pocket expenses o f sending their childrento school. The State Minister for the SYDTF and SYDVs, and the Ministries of Education and Health, view CCTs as an important intervention that will help Turkey achieve multiple objectives: amelioration o f extreme poverty; secondary school enrollment of girls; improved health care access for poor children, including immunization inurban areas (rates are high in rural areas); and improved prenatal care (pregnant women are to be included in the program shortly, as i s the case in Jamaica and other Latin American countries). Secondary benefits include improvement o f the registry for national identification numbers and the registration o f birthsand marriages. As of May 2005, there were 601,400 eligible families, with 1,627,000 child beneficiaries. The CCT program was fully operational across Turkey in 2004. A national evaluation o f the effectiveness of the CCT programwill be undertaken. 112 Table VII.6. Quarterly Figures for CCT NUMBEROFBENEFICIARIES PAYMENTS EDUCATION HEALTH TOTAL EDUCATION HEALTH TOTAL 4 2 11,233 6,625 17,858 135,053 79,65 1 214,704 2003 4 3 25,697 14,647 40,344 233,934 180,228 414,162 Q4 59,206 28,027 87,233 731,472 294,2 18 1,025,690 Q1 225,324 87,919 313,243 3,194,2 14 847,457 4,041,671 2004 4 2 305,666 113,103 418,769 3,979,778 1,389,752 5,369,530 Figures for number ofbeneficiaries are cumulative. Payment amounts are not cumulative. Q1 includesJanuaryand Marchfigures. 42 includesMay figures. 43 includesJulyandSeptemberfigures. 44 includesNovemberfigures. Source: SRMP ProjectCoordinatingUnit. D. LOCAL INITIATIVES Supported under the SRMP, the Government o f Turkey has undertaken a significant expansion o f the microprojects traditionally done by the SYDVs with approval from the SYDTF, along with a sharpening of procedures. In 2004, 250,000 people will benefit from income generation, employment, and social service opportunities under the SRMP Local Initiatives component, which seeks to provide these people with sustainable livelihoods, thereby liftingthem permanentlyout o fpoverty. Incidence of Government Programs Unfortunately, the 2002 HBS questionnaire was not well designed to capture information about government social protection program receipt. For example, there i s a question asking, "Do you receive any transfers from persons (neighbor, relative, etc.) or foundations (SYDVs)?" but this question is not further subdivided, so government programs cannot be analyzed separately from private charity or transfers. While there are detailed questions about transfers inkind in which government programs can be separated from private programs, this is not the case for cash, which is the predominant form o f government social assistance. And, unfortunately, in the income section, again public and private sources of cash transfers are either confounded, or questions about monetary aid from the SYDVs were simply not asked, Given the absence o f detailed questions on cash assistance from the SYDVs, the incidence of spending from other government programs is presented in Table VII.7. All government transfers except old-age income and transfers in kind are regressive. This analysis is based on nominal transfers reported on an annual basis. It i s not clear how best to deflate these data, since respondents were asked to report their annual income by type, so it i s not clear what reference period would pertain. 113 Table VII.7. Turkey: Incidenceof NominalTransfers(annual, undeflated) institutions inthe last 12 months 14,024,605 1 11,869,626 I 27.99 114 ANNEX I:METHODOLOGY This appendix summarizes the methodological approaches used to analyze poverty and inequality in the Joint Poverty Assessment Report for Turkey. This methodology was approved at the World Bank Concept Paper Review meeting, and is presented here in greater detail. The source o f data i s the 2002 HouseholdBudget Survey (HBS) data of the State Institute of Statistics (Devlet Istatistik Enstitusu, DIE). The methodologyis well established inthe poverty literature and inWorld Bank researchand operational policies (Coudouel and Hentschel 2000; Deaton 1997; Deaton and Zaidi 2002; Foster, Greer, and Thorbecke 1984; Lanjouw and Ravallion 1995; Ravallion 1992, 1994, 1998, 2000; World Bank 1990, 2001). First, a welfare indicator must be devised from the data, then analytic choices must be made about how to distribute that welfare level across household members, and finally, poverty lines must be established and applied to the data. Indicators of Well-Being Two measures o f economic welfare are constructed in this report4onsumption expenditure and income-although our preferred measure of welfare i s consumption expenditure. It should also be stated that income and expenditure aggregates are very well correlated, and in principle the income aggregate could also be used for the poverty and inequality analysis. However, there are several arguments for usingconsumption as the welfare indicator, as is standardWorld Bank practice. For Turkey, it shouldbe noted that an extensive informal sector and flows of private remittances, coupled with a low level o f tax declaration compliance and enforcement, make money income an inferior measurement o f household welfare. Inaddition, a very important source o f food for not only rural households, but also many urban households inthe gecekondu areas, i s food produced ingardenplots, includingboth crops and livestock. Consumption Aggregate The consumption aggregate constructed from the 2002 HBS data set follows standard practices as well established in the literature (Deaton 1980; Deaton and Zaidi 2002). This consumption aggregate will be referred to as the "new" methodology, and draws most from Deaton and Zaidi (2002). A version o f the consumption aggregate for 2002 was also constructed with exactly the same algorithm used in the previous World Bank poverty assessment. This consumption aggregate will be referred to as the "previous" methodology. Drawing this distinction i s very important, because the poverty and inequality findings will differ depending on which consumption aggregate is used. In particular, the previous methodology included expenditures on consumer durables which, following Deaton and Zaidi (2002), have been excluded from the new methodology. The previous methodology did not includethe imputed value o f owner-occupied housinginthe consumption aggregate. New Methodology The 2002 HBS data were used to construct the consumption aggregate. Food consumption includes expenditures on food, consumption o f home-produced food, and food received as a gift. Consumption of non-food items includes expenditures on personal care and hygiene items, clothing, utilities, transportation, and other frequently or not-so-frequently purchased non-food items, and the estimated value of services renderedby durables. The expenditures on durables are excluded from the consumption aggregate. The definition o f durable goods i s made according to the Classification o f Individual Consumption According to Purpose (COICOP). The purchase o f semidurable goods and services i s included in the consumption aggregate. The information on the reported current value o f each durable 115 item owned by a householdwas not available, nor was the age or purchase price o f the durable item. That is why the imputed or reported value o fdurable goods i s not included inthe consumption aggregate. The housingrental market i s well developed inTurkey (according to the 2002 HBS data set, 20.2 percent of households were renting their housing), malung it possible to apply hedonic housing regressions in a meaningful way to derive the estimates o f the rental value o f housing. The consumption aggregate does include a monetary assessment o f the value o f consumptionreceived by the household from occupying its own dwelling. The significant share o f households paying rent makes possible the imputation o f the value o f owner-occupied housing. Both methodologies, previous and new, include the self-reported monthly rent, but the new one includes the imputed value o f owner-occupied housing, where this imputation is done by a semi-log regression o f monthly rent on housing attributes, and then the regression coefficients were used to impute rent based on the characteristics o f owner- occupied housing. Income Aggregate The household income aggregate was constructed using the income section o f the HBS. The information on income is collected on a monthly basis for the survey month, and on a yearly basis (the income o f the household for the last 12 months). Total household income includes market incomes, public and private transfers, and imputed value o f home-produced goods. Market incomes include salaries in cash and in kind, net income from self-employment (including in agriculture), and capital income. Public transfers include pensions, stipends, unemployment benefits, social assistance, and other social insurance benefits. Private transfers include financial assistance from relatives living incountry and remittances from abroad. Methodology of Making HouseholdWelfare Indicators Comparable Ranking households according to their welfare, and ranking them below or above a single national level poverty line, requires some adjustments for prices with the nominal welfare measures o f households, such as income or consumption. This adjustment i s necessary because all households were not interviewed in a single month. Rather, interviews were done across the 12 months o f 2002, with approximately one- twelfth o f the total sample o f 9,555 interviewed in each month. Inflation was significant in Turkey, averaging 29.7 percent during2002. Turkey i s a country with significant differences in consumer prices across the regions, and considerable urbadrural differences. Therefore, to make household welfare comparisons, accounting for price differences usingthree-dimensional price indexes by regions, type o f settlement, and months is one of the possible solutions to deflate the nominal welfare measures for the different times and areas, and to make households comparable to each other and to any single national poverty line. Thus, real consumption can be compared to a single poverty line.37 37 The alternative approach to the solution o f the problem o f nominal values needing to be deflated is to set up different poverty lines for households based on region, settlement type, and month, so that nominal consumption would be compared to nominal poverty lines. Comparison o f the nominal per capita or per adult equivalent measures with deflated poverty lines would give a correct picture o f poverty, but inthis case, the nominal welfare measures o f households surveyed in different regions and urbadrural localities during different survey months would not be comparable to each other, and therefore could not serve as a base, for instance, for the analysis o f inequality. As a result, this second approach is not usedinthis report. 116 Household consumption and income, as well as other monetary measures, were deflated using three- dimensional price indexes. There are two choices for which price index to use: one calculated from inside the survey database, and one from another source o f information, namely the DIE price statistics department. The DIE official Consumer Price Index (CPI), based on both food and non-food items, 2002 price differences across regions, urbadrural location, and months were provided by DIE. The average CPI for Turkey for 2002 i s taken as a base, and i s set equal to 1.00. The official CPI varies across all three dimensions: regions, urbadrural location, and months. The total number o fCPIis equal to 168 (7 regions times 2 settlement types times 12 months). However, it is more preferable to use a price index from inside the survey database, based on survey prices, because all other information used for the poverty analysis i s based on survey estimates, and introducing another data source could lead to bias (Glewwe and Hall 1992). The survey food price index was estimated based on price differences o f food across regions, urbadrural location, and months. An appropriate price index for non-food items appears to be impossible to develop on the basis o f survey information, because o f the insufficient number o f purchases o f such items in terms of providing statistical significance. A sensitivity analysis o f poverty depending on the price index used for deflation (that is, survey price index or official CPI) was undertaken. The results are different, but they are very close. There is no significant difference between the poverty figures based on deflation by these two different price indexes. Appendix 3 contains the findings o fthe sensitivity analysis. Using Equivalent Measures To make welfare comparisons across households with different demographic compositions, some way is needed o f adjusting the welfare measures to account for different sizes and compositions o f households, Simply dividing household consumption by the number o f members (per capita consumption) leads to overstatement o f poverty among large households (Ravallion and Lanjouw 1995). The basic idea behind equivalence i s that household members are not alike, so using per capita as average consumption per member obscures the fact that children need less food than adults, and adult males typically have expenditures relating to employment (like clothing and commuting) that elderly females do not, for example. To take into account these differences, equivalence scales have been proposed. An example scale would be one in which children are counted as 50 percent o f adult male consumption, females and elderly are 70 percent o f adult male consumption, and the adult male i s the reference adult. To illustrate, a 5-member family with 2 children and 1 grandparent and 2 parents would have an equivalent household size not o f 5 (which i s per capita), but o f 3.4 (1 for the male plus .5 for each child and .7 for the female and elderly). For many years, the so-called Engels approach (Deaton and Muellbauer 1980, 1986) was used to estimate equivalence from survey data. However, Deaton has since demonstrated that this approach i s not correct (Deaton and Paxon 1998). Since then, the literature has not revealed an empirical way to estimate equivalence. However, several normative equivalence scales (which are not empirically verifiable) have come into common use, including one used by the Organization o f Economic Co-operation and Development (OECD) for its income-based poverty comparisons, and one from the United Nations Food and Agricultural Office (FAO) based on dietary information. I17 Various equivalent measures for income and consumption are considered in this appendix. Different scales can have major impacts on poverty measurementand the profile of the poor (Lanjouw, Milanovic, and Patemostro 1998). Adult Equivalent Size General Formula To measurethe effects of economies o f scale and the different consumption needs by different household members, household size i s converted into aduZt equivalent (AE) using the following formula for the household i: parameters. Children are individuals aged 14 and below. Inthis appendix a value o f e= 0.6 was adopted where Ai is the number of adults in the household, Ci is the number o f children, and a and 6' are to ensure comparability with other World Bankregionalpoverty findings. Two different values of a were considered in this analysis (a= 0.9 and a = 1). The batch mode programs developed for the analysis allowed examinationof the impact o f other possible values of the parameters on the poverty statistics. Adjusted Adult Equivalent Size Based on Modal Household Composition However, as pointed out by Deaton and Zaidi (2002), this adjustment would overestimate the total consumption unless all households were single-adult households. They suggest using an adjusted adult equivalent size o fthe household usingthe formula shown below. Adjusted adult equivalent size of the household i (AE-ADJ;) i s definedas AE-ADJ, = A0 +co (A0 +acole AEi where Auand Cuare the numberof adults and children inthe "pivotal" household, respectively, andAiand Ciarethe number of adults and children inthe i" household. The modal or pivotal household inTurkey is a 4-member household with 2 adults and 2 children (Au= 2 and C,= 2). AEi here i s from the general formula above. OECDEquivalence Scale AE-OECD = 1+ [number of children under age 141x 0.5 + ([number of adults] - 1) x 0.75. FA0Equivalence Scale + nmZ0- x 0.88 +nf40- x 0.76) x 0.75 AE FA0=(n-ch5 x 0.64 +nch5-11 +nm12-17 +nfl2-17 x 0.84 +nm18-39 + nfl8-39 x 0.84 where n-ch5 is the number of childrenunder age 5, nch5-11 i s the number o f children aged 5 to 11, nm12-17, nf12-17 arenumber of male and female children, respectively, aged 12to 17, and nm40- and nmf40- are the number o f males and females aged 40 and above, respectively. 118 Table A.I.1describes the relationbetween the adult equivalent sizes and the real size o f the household, for a =0.9 and 8 =0.6: Table A.I.1. Household Size by the Set-Type of Settlement rypeof ettlement rurkey Rural Urban Householdsize 4.30 4.07 4.16 Adult equiv. size generalformula 2.29 2.23 2.26 AE adjustedby modalHH(2 ad+2 ch) 4.12 4.01 4.05 Adult equiv.size OECD 3.21 3.02 3.09 Adult equiv. size FA0 2.63 2.55 2.58 Source. DIE, Turkey 2002 HBS. The variation o f adult equivalent size o f households depending on equivalence scale type and parameters aand 8 arepresentedinAppendix 6. Since there is no guidance from the literature on valid empirical determination, once the sensitivity of findings to the scale has been demonstrated (Appendix 6), a normative choice must be made as to which to use. This analysis has chosen the adjusted adult equivalence scale based on the modal household (Deaton and Zaidi) approach for its basic approach, and compared it to the main poverty line (explained below) usinga = 0.9 and 8 = 0.6. UsingDay-BasedMeasures To make welfare comparisons across households surveyed duringmonths with different numbers o f days, some way o f adjusting the welfare measures i s needed. The daily measures were calculated by dividing the monthly measures into the number of days in each survey month. The daily welfare measures are used for the poverty and inequality analysis. Average monthly measures are constructed by multiplying the daily welfare o f the household by 30.41'7 (the average number o f days inthe month). Structureof IncomeandConsumption Income The average household disposable income according to the survey data was around TL672 million per month for 2002, o f which TL639 million i s the average cash income and the other TL33 million is the average household income inkind. Structureof Cash Income The main source of income for Turkish urban households i s wages and salaries (43.4 percent), and for rural households it i s the income from agricultural activities (32.5 percent). 119 Table A.1.2 describes the structure of cash income. TableA.I.2. The Structureof CashIncome ttlement Urban Share 1 0.4340 0.2206 0.0144 0.0198 Total net income incash from all previous jobs inthe last 12 months 133,371,432 76,957,193 168,881,759 0.2086 0.1416 0.2414 Rent fromreal estateincome in cash inthe survey month 26,197,746 10,929,818 35,808,248 0.0410 0.0201 0.0512 Interest income incash from movable propertyinthe survey month 31,794,266 15,809,179 41,856,189 0.0497 0.0291 0.0598 Transfer income incash inthe survey month 116,345,155 87,212,289 134,683,036 0.1820 0.1605 0.1925 The Structureof Consumption The main item o f consumption o f Turkish households is the category of housing and water consumption (28.7 percent). This estimate includes the value of real housing rents for households renting the apartment or house, and the value of imputed rents for house and apartment owners. The second important item is food consumption (25.45 percent). For urban and rural households there is a slight difference in the importance of consumption items. The main item for rural households i s food consumption (26.96 percent), and the second item i s housing (22.73 percent), while for the urban households the main item of consumption i s housing (3 1.53 percent), and the second item i s food consumption(24.73 percent). 120 Table A.I.3 describes the structure o f consumption. Foodreceived as income inkindfrom activities and real estate 2,954,25 1 2,272,086 3,383,645 0.0053 0.0049 0.0055 Non-foodgoods and servicesreceivedas income inkindfrom activities and real estate 23,219,048 13,191,239 29,531,122 0.0418 0.0284 0.0481 Foodreceived as a present 5,384,374 5,513,449 5,303,127 0.0097 0.0119 0.0086 Non-food goods and services received as a present 6,797,208 5,596,940 7,552,724 0.0122 0.0121 0.0123 Inequality Measures L o w income and consumption, and inequality in their distribution, are the key determinants o f the well- being of the population. One o f the common indicators o f inequality is the decile 90/10. Inthis analysis, the decile 90/10 i s the adjusted (by the modal household) per adult equivalent deflated consumption (PEC) o f the 90thpercentile divided by the PEC of the 10th percentile, that is, the PEC o f the poorest person inthe richest decile over the PEC o f the richest person inthe first or poorest decile. This indicator i s easy to interpret, but it does not reflect the situation in the middle o f the distribution. The decile ratio 90/10 i s estimated as 4.38 for PEC. 121 The PEC and adjusted per equivalent income (PEI) by deciles are presented inTable A.I.4. Table A.I.4. AverageMonthlyConsumptionPer CapitaandPer Adult Equivalent,byDeciles TL Per Month Adult Equivalent Adjusted with ModalHH 55,348,090 100,4363 17 96,184,245 6 115.135.441 112.996.782 Turkey 132,137,395 I . . 133,701,603 Source: DIE, Turkey 2002 HBS. Another common inequality measure i s the Gini coefficient, which is sensitive to all the parts o f distribution. The Gini coefficient i s bounded between 0 and 1; it i s 0 in the case o f absolute equality, when the per equivalent consumption o f each person is the same, and it i s 1 in the case o f absolute inequality, when one person consumes everything and others consume nothing. The Gini coefficient is gven by: where there are n individuals indexed by i, their equivalent consumption i s given by ci, mean equivalent consumption i s denoted by p, and where ri i s household's i rank in the equivalent consumption ranlung (that is, for the household with lowest equivalent consumption, r; equals 1, while for the household with the highest equivalent consumption, ri equals n). 122 Figure A.I.1describes the inequalitybased on PEC. FigureA.I.l. CumulativeDistributionof PEC ~' 0.9 0.8 0.7 0.6 0.5 0.4 1 0.3 i,0.2 0.1 0 The Gini concentration coefficient i s calculated on the basis o f individual information o f the HBS database, which i s more accurate than ifwe use consumption or income by deciles. The Gini coefficient is not an additive or decomposable measure, meaning that it may not be broken down by population groups or income sources or in any other dimension. The Theil index o f inequality i s given by: The Theil index i s most sensitive to inequality inthe top o f the distribution, while the mean log deviation measure, also called the Theil-T index, i s most sensitive to inequality in the bottom range o f the distribution. The Theil-T formula is: Neither the Theil index nor the mean log deviation measure i s easy to interpret, except in reference to other countries or the same country at different points in time. Both measures are zero for perfect equality. For complete inequality (one person consumes everything), Theil-T goes to infinity, and Theil-L reaches nln(n). The other inequality decomposable measures presented along with the Gini coefficient inTable A.I.5 are the Theil T index and Theil L index. 123 Table A.1.5. InequalityMeasures Both indexes indicate that inequality is higher in urban areas than in rural areas. Inequality based on income i s greater than inequalitybased on consumption. Poverty Poverty Lines and Poverty Rates Where the poverty line i s drawn depends at least on how the following questions are answered: what indicator i s used to measure the well-being o f a household, which poverty line i s used, and which minimal consumer basket i s used in case the "food poverty line" is chosen. It should be emphasized that there is no perfect indicator identifying the living standard. World Bank practice is to set at least two poverty lines, one for extreme or food poverty, the other for overall poverty. However, the World Bank also uses several absolute poverty lines for international comparisons, and relative poverty is used in the OECD countries for comparison. This appendix lays out several possible lines, and decides on a main line (the complete poverty line) that can be subdivided into overall poverty and food or extreme poverty. Absolute Poverty Lines The following absolute poverty lines were used for the analysis: US$l.OO, US$2.15, US$4.30 by purchasing power parity (PPP) for international comparisons. PPP exchange rates reflect the purchasing power o f national currencies, and differ (sometimes substantially) from current market exchange rates. According to DIE, US$l.OO PPP in Turkey is estimated as TL663,575 for 2002. This is DIE'Scurrent estimate o f PPP. InWorld Bank publications, such as the World Development Indicators, PPPs are often estimated by taking a base year and extending it by the CPI. In the comparisons chapter, two different kindso fPPPs are used, one the extended-by-CPI version, and the other the current DIEPPPs. Usingthe current DIEPPP, the absolute poverty lines used for the analysis here are TL663,575 (US$l.OO PPP), TL1,426,686 ($2.15 PPP), and TL2,853,373 ($4.30 PPP) per day per adult. Relative Poverty Line The relative poverty line used for the analysis is drawn as 60 percent o f median consumption per capita. It is estimated as TL2,377,139 per day per adult. The relative approach is a common practice in OECD countries, where the notion o f ability to share in increased general prosperity, rather than absolute survival, i s probably more relevant. 124 Food Poverty Line The food poverty line was developed using the actual quantities for the most popular 80 products consumed in the third and forth deciles of the population, pricedout by using the country average survey prices for 2002. The calorie intake information from the 2002 HBS survey was calculated using the UnitedStates Department of Agriculture (USDA)nutritional database. The composition of the minimum food basket was calculated on this base to reach 2,100 Kilocalories per day (Kcal) per averageperson per day (a nutritional minimum accepted internationally according to FA0 and World Health Organization [WHO] recommendations). By using the price information from the 2002 HBS survey, it is estimated that the amount necessary for attaining the minimum food consumption i s TL1,082,359 per person per day inan average Turkishfamily. Table A.I.6 describes the structure o f a minimumfood basket based on the consumption patterns of the reference population, which arepopulationdeciles 3 and4 by food consumption. TableA.I.6. The Structureof aMinimumFoodBasketEstimatedBasedon Survey Data Detailedinformation about the structure and cost o f a minimumfood basket is found inAppendix 5. CompletePoverty Line Individuals have non-food needs in addition to food needs. Taking into account the need for non-food consumption requires adding an allowance for non-food goods and services to the food poverty line. The lower-bound methodo fRavallion (1994) i s usedhere to determine the value of the general poverty line. To determine the allowance for non-food consumption, usingthe survey data itself, first those individuals whose total consumption i s just above the value of the food poverty line are selected. This part of the sample will now constitute the reference group for the derivation o f the general poverty line. The share of total consumption that goes to non-food consumption i s calculatedfor this reference group. This share i s the "allowance" for non-food consumptionthat i s added to the value o f the food poverty line to get the complete poverty line. 125 The share o f non-food consumption among those whose total consumption isjust above the value of the food poverty line is 57 percent, and food consumption represents43 percent. The value of the complete poverty line i s thus: Complete Poverty Line (CPL) =Value of FoodConsumption +Value of Non-FoodConsumption where: FoodConsumption =Value o f FoodPovertyLine=TL1.08Mlm =43% of CPL Non-food Consumption = 57% of CPL CPL =TL1.08Mln divided by 0.43 =TL2.5Mln =TL1.08 Mln+TL1.42 Mln. The CPL is estimated as TL2,510,930 per day per person, which includes a 43.1 percent food component (TL1,082,359) and 56.9 percent non-food component (TL1,428,571). This structure is based on the consumption pattems o f population whose adjusted PEC is just above the food poverty line. These households are inthe first decile of PEC. This method of deriving the complete poverty line i s the simplest way to assess the value o f the minimum consistent with the actual consumption pattern of the population. We have chosen the simplest method described above, because it is the most transparent, most easily replicable, and most intuitive. It could be arguedthat other lines would be more accurate. However, if a certain way to set the line i s not commonly understood, its use will not help the national poverty diagnostics. Given the fact that any poverty line is a matter o f convention and includes in itself a technical judgment, the team considered the CPL thus derived as the most accurate for the use with the 2002 HBS data set, and for the analysis of poverty inTurkey. The value o f the CPL is presented inTable A.1.7. Household Size Average PovertyLine for Household Deflated 1 137.055.990 2 207,561,617 4 309,55 1,477 I 7 1 431,521,387 I 8 466,737,285 9 500.4 14.569 I 10 I 532,742,054 I 126 Poverty Statistics Three different poverty measures are used in this analysis, all o f which are members o f the class o f additive and decomposable measures proposed by Foster, Greer, and Thorbecke (1984). The first measure i s the Headcount Index of Poverty, given by the proportion o f the population for which total per capita household consumption (income) y i s less than the poverty line z. It is the most frequently used poverty measure. The main advantage o f this statistic i s its simplicity. Ifq i s the number o f poor people inthe populationo fsize n, thenthe headcount is givenby: PO=-4 n However, the headcount measure i s totally insensitive to differences in the depth o f poverty. A way to look at the poverty deficit o f the poor relative to the poverty line i s to use the Poverty GapIndex. Let Q be the subgroup o fpoor; the poverty gap i s then given by: The poverty gap also allows an interpretationin terms o f the potential fiscal cost for eliminating poverty by targeting transfers to the poor. Summing all the poverty gaps inthe sample population and taking the average provides an estimate o f what would be the minimum cost o f eliminating poverty in the society, assuming perfect targeting. One shortfall o f the poverty gap measure i s that it may not adequately capture differences in the severity o f poverty. A way to tackle this problem i s to include the Severity of Poverty Index in the poverty analysis. This measure gives more weight to the consumption (income) gap of those households located further below the poverty line, and i s defined as: The severity index has the main advantage o f comparing policies that aim to reach the poorest, but it is more difficult to interpret and i s less intuitive than the two previous poverty measures. Table A.I.8 represents the poverty headcount, gap, and the severity index based on adjusted and deflated PEC and complete various poverty lines. Note that in this table, poverty measures are not in percent as presented inthe main text, but are inlevels. To obtain percentages, multiplyby 100. Table A.I.8. PovertyHeadcount, Gap, and SeverityIndex, Basedon Adjusted and Deflated PEC and CompleteVarious PovertyLines Food Complete Poverty Poverty Poverty Line Poverty Line Re'ative Poverty PovertyLines1 Line=$2.15 Line=%4.3 1.082.359 2.510.930 Line 1.980.949 ppp=663575 TLlday PPP=1426686 PPP=2853373 TLlday TL/day TLlday TLlday TLlday PO 0.0135 0.2696 0.1474 0.0021 0.041 1 0.3414 P I 0.0026 0.0688 0.03 13 0.0004 0.0083 0.0973 P2 0.0008 0.0253 0.0104 0.0001 0.0027 0.0384 127 Other poverty figures based on other welfare indicators, CPI, parameter o f economies o f scale, and adult equivalence are presentedinAppendix 3. Sensitivity of Poverty Statistics The poverty indexes are sensitive to the methodology o f constructing the welfare aggregate, choosing the reference population for defining the food basket, choosing adult equivalence, and economies o f size parameters. Table A.1.9 represents the poverty headcount index, poverty gap, severity, and shortfall index depending on several key parameters o fhow they were calculated: CPI DIE CPI DIE CPI Survey CPI Survey Imputed Reported Imputed Reported Rent by Rent by Rent by HH Regression I Rent by HH Regression Food Poverty PO 0.0438 I 0.0462 0.0379 0.0388 ~~ PI 0.0083 0.0090 0.0067 0.0073 P2 0.0026 0.0028 0.0021 0.0022 Shorkfall 0.1894 0.1943 0.1775 0.1877 CompletePoverty PO 0.2896 0.2810 0.2696 0.2618 PI 0.0764 0.0749 0.0688 0.0691 P2 0.0288 0.0287 0.0253 0.0259 More detailed information about the results o f the sensitivity analysis can be found inAppendix 3. Comparisonwith PreviousWorld BankEstimates There are significant differences between the main poverty line used in this report and the poverty line and consumption aggregate o f the World Bank's previous Poverty Assessment (PA), which used the 1994 data (World Bank 2000), and the two figures are not directly comparable. In the previous PA, the food basket was developed by Turkish academic institutions, and it i s much richer in terms o f quantity and quality o f food. The caloric value o f the food basket used for the previous PA i s more than 3,000 Kcal/day. This poverty line i s muchhigher than the main one usedherein. The data collected inboth 1987 and 1994 seem to adequately address most o f the problems inmeasuring well-being. Unfortunately, it was impossible to follow the new methodology to recalculate basic results for 1994, because the 1994 data are not ina format that could be used without going back to the line item codes, which would be prohibitively costly in terms o f staff time. Thus the decision was to apply to the 2002 data the old methodology to obtain roughly comparable results. The current consumption indicator for measuring living standards and poverty based on Turkishhousehold data includes: All monetary non-business and non-investment expenditures the (2002 version would exclude all durables; the 1994 version would exclude only expenditures on selected items) Gifts, earnings, and transfers inkind Consumption from stocks Consumption from own production 128 0 Imputedrents from owner-occupied housing (imputed by regression in 2002 methodology and self-reported in 1994). Inaddition, the methodologydeveloped for the analysis of 1994datareliedon spatial and time indexes of prices from the official CPI statistics, while the new methodology i s based on survey unit values. To obtain comparable figures, the 1994 methodology was applied to the 2002 HBS, including the use of the CPI instead o f the survey price index. Data are broadly comparable, but some important differences are noticeable inthe detailed structure of bothaggregates. Table A.I.10 shows the consumption of food per household. While overall the change o f total per- household consumption is within the plausible range, some important changes in the structure reflect the comparability problem. Inthe 1994 consumption aggregate, both in-kindconsumption of own food and consumption from stock were included. The DIEwas concerned that this might lead to double counting. Inthe new version of the questionnaire, much more detailed information on consumption from stocks is included, and only own production i s included inthe food consumption. Table A.I.lO. FoodConsumptionandExpenditures, 1994 and2002 TotalFoodper Household, ConsumedIn TL/Month Earningsor Outside Production Gifts (aid) Rural 203.759.737 73.1% I 1I I 4.9% I II 6.1% ~ I , Urban 200,831,846 87.6% 0.7% 4.6% 7.0% 201.962.890 82.0% 6.5% I 4.8% 6.7% 3.344.540 , , 62.8% I 30.4% I 3.1% I 3.8% 3,988,774 79.1% 7.8% 3.9% 9.1% 3,676,828 72.6% , , 17.5% 3.6% 6.3% 4-2002. Times 60.9 71.0 31.8 I 96.5 I 99.6 Urban 50.3 55.7 4.8 59.9 38.6 54.9 62.0 20.5 73.1 58.4 This factor, with some economic changes, explains the very small change inthe consumptionof own food for urban households. On the other hand, the consumption of meals outside of the home has clearly not kept pace with the overall consumption of food, especially in urban areas, suggesting that the current questionnaire forms for collecting these types of expenditures may needrevisions. Creating the "previous" methodology consumption aggregateincludedthese adjustments: 0 Householdconsumption includes the purchase o f all durable goods, excluding cars. 0 The FA0adult equivalence scale was used. 0 The official CPIwas usedfor deflation. 0 The value ofthe minimumfood basket of the PA was usedas a food poverty line. 0 The quantities from the Hacateppebasket were pricedout by usingthe 2002 survey prices. 0 The same non-food share is addedto the poverty line for different regions and settlement types.38 38Note that the PA generated these constants by regression. W e simply use them "as is." 129 The 2002 poverty line comparable with the PA i s TL2,361,383 =US$3.60 PPP at 2002 prices. Poverty rates according to the poverty line basedon the minimal food basket used for P A 1994 and using comparable methodologyare giveninTable A.I.11. TableA.I.ll. ComparablePovertyMeasures,1994and 2002 Complete p l P2 Poverty Complete Complete 2002 HBS, using 1994 foodbasket andprevious 2002 0.352 0.108 0.045 algorithmfor povertyline andwelfare measure PA 1994 0.363 0.109 0.046 Thus, there was a very slight decreaseinpovertybetween 1994and 2002. Comparison of the Food Basket Used for the 1994 Poverty Assessment and the Food Basket Developed Based on Survey Data There are significant differences among the poverty line, the consumption aggregate, and the methodology usedinthis analysis of the 2002 HBS compared to what was done for the previous PA using the 1994 data. Thus, the poverty figures presented above are not comparable with poverty figures o f the 1994 PA. Inthe previous PA, the food basket was developed by a Turkishacademic institution, and it i s muchricher interms of quantity and quality of food. The caloric value of the food basket used for the PA (1994 data) i s around 3,000 Kcal/day. It i s important to note that the new survey-based basket gives a muchmore realistic structure of consumption of the poor population than the one used for the 1994 data analysis. Table A.I.12 lists the approximate comparison. TableA.I.12. FoodBasketfor EquivalentAdult: 2002 Survey-BasedandHacettepeUniversity (kilogramsper day per adult) ...= Not available. n.e.c. = Not elsewhere classified. The food poverty line based on the food basket developed by Hacettepe University is TL2,361,383 per adult per day =US$3.60 PPP at 2002 prices. 130 Talung into account the parameter o f the approximate difference between the food consumptiono f adults and children (0.9), the food line per child in the modal family i s TL2,125,244 per day, and the food line per average person in the modal family i s TL2,243,314 per day, which i s significantly more than the estimated poverty line based on the consumption of the third and fourth deciles of population (TL1,082,359 per person per day). Ifwe price out the minimumfood basket developed by Hacettepe University, add the same 57 percent non-food share for modal households, and use the same adjusted equivalence scale, then the complete poverty line becomes TL2,243,314 divided by 0.43, or equal to TL5,217,009 per person per day in the modal family (or US$7.86 PPP). The poverty headcount index based on this poverty line is 75 percent, which i s far too highto be usable for either public policy applications or research. 131 Appendix 1: IncomeandConsumptionbyPoverty Status andUrbadRural Dimension 1 Household's Average Monthly Consumption (TL) I Type of Settlement I Poverty Status Turkey Rural Urban Non-Poor Poor 1 Households 1 I Households ~ I Consumption 02,387,828 510,357,187 660,317,142 700,347,775 264,094,536 1 i I Food and nonalcoholic beverages 41,54 1,355 125,083,823 151,900,660 157,345,897 86,962,205 I Alcohol and tobacco 24.158.598 123.370.600 I 24.654.608 I 27,146,480I 13,840,292: r Clothing and footwear 36,289,947 30,474,680 39,950,407 43,977,902 9,740,489 I 1 I I Housing, water supply 59,589,070 105,441,687 193,672,510 I II 185,916,570 68,670,110 I , Furniture HHappliances and home care ~ services 43,2 16,912 40,37 1,957 45,007,688 52,530,573 11,053,266' I I I i I I I I I Health 14.127.7121 13.955.395 14.236.1781 16.921.274 4.480.470 ! 7 Transportation 46,917,368 33,023,943 55,662,679 58,045,979 8,486,002 1 I Communication 27,362,109 20,820,121 31,480,008 32,599,264 9,276,204 Entertainment and culture 14,862,268 12,101,713 16,599,919 18,169,140 3,442,372 I Education 8,387,213 4,497,613 10,835,548 10,675,340 485,429 I Restaurants and hotels 15,037,810 13,358,012 16,095,170 18,077,677 4,539,983, I Various goods and services 25,330,134 22,389,654 27,18 1,039 30,953,323 5,911,102 ~ I Ownproduced food for self consumption 14,465,225 35,028,585 1,521,478 13,837,979 16,631,345 ~ 1 Ownproducednon-food for self- 1 consumption 3,646,997 9,041,038 251,681 3,466,240 4,271,221 I Foodreceived as income inkindfrom activities and real estate 2,954,25 1 2,272,086 3,383,645 3,328,750 1,660,964 r I Non-food goods and services receivedas i income inkindfrom activities and 23,219,048 13,191,239 29331,122 28,219,806 5,949,515 I real estate ~ I Foodreceived as a present 5,384,374 5,513,449 5,303,127 5,418,279 5,267,284 I Non-food goods and services received as a present 6,797,208 5,596,940 7,552,724 7,798,547 3,339,198 Source: DIE. Turkev 2002 HBS. 132 Household'sAverageMonthlyFood Consumption(TL) Turkey Poverty Status Yon-Poor HH Poor HH Foodand nonalcoholic beverages 1413 41,355 157,345,897 <6,962,205 Bread and cereals 34,248,017 36,866,463 !5,205,5 18 Meat and meat production 19,760,842 23,512,042 6,806,509 Fish 1,893,908 2,198,002 843,758 Milk,cheese, andeggs 17,324,07 1 19,687,699 9,161,555 Fats and oils 10,313,034 11,238,373 7,117,483 Fruits 13,248,611 14,937,249 7,417,097 Vegetables 21,112,786 22,659,346 15,771,920 Sugar, jam, honey, chocolate, and confectionery 12,376,600 13,5 10,851 8,459,596 Nonalcoholic beverages 8,388,184 9,5 11,660 4,508,388 Source: DIE, Turkey 2002 HBS. Household'sAverageMonthlyAlcohol and Tobacco Consumption(TL) Turkey Type of Settlement ' Poverty Status Rural Urban Non-Poor HH Poor HH Alcohol and tobacco 24,158,598 23,370,600 24,654,608 27,146,480 13,840,292 Alcohol 1,962,920 2,029,541 1,920,985 2,362,809 581,949 Tobacco 22,195,678 21,341,060 22,733,623 24,783,671 13,258,343 Source: DIE, Turkey 2002 HBS. Household'sAverageMonthly Clothing and footwear Consumption(TL) Turkey Type of Settlement Poverty Status Rural Urban Non-Poor HH Poor HH Clothing and footwear 36,289,947 30,474,680 39,950,407 43,977,902 9,740,489 Fabrics 1,066,185 1,164,060 1,004,577 1,278,976 331,337 Clothing (man, woman, children) 24,619,709 20,230,937 27,382,252 30,215,310 5,295,952 Accessories and other clothing 1,426,808 1,253,045 1,536,184 1,64 1,346 685,926 Drycleaning, mending, andhiring 416,676 286,582 498,565 526,456 37,565 Footwear 8,636,495 7,460,496 9,376,736 10,17 1,666 3,334,960 Repair and hiring o f footwear 124,073 79,560 152,092 144,148 54,747 Source. DIE, Turkey 2002 HBS. 133 Household'sAverageMonthlyHousingConsumption(TL) Type of Settlement PovertyStatus Turkey Rural Urban Non-poor HH PoorHH Housing, water supply 159,589,070 105,44 1,687 193,672,5 10 185,9 16,570 68,670,110 Housingrents (real for owners and imputed for non-owners 77,976,501 44,322,664 99,160,139 87,538,065 44,956,754 Maintenance andrepair 8,355,128 8,859,876 8,037,411 10,603,146 591,859 Various services relatedto water supply and dwelling 12,432,638 5,827,295 16,590,4 17 14,8 84,063 3,966,929 Electricity, gas, and other fuel 49,925,03 1 41,256,505 55,323 1,495 58,810,050 19,241,652 Source: DIE, Turkey2002 HBS. Household'sAverage MonthlyExpenditureson FurnitureHHAppliances and Home Care Services (TL) Type of Settlement PovertyStatus Turkey Rural Urban Non-Poor HH PoorHH FurnitureHHappliances and home care services $3,216,912 40,371,957 45,007,688 52,530,573 11,053,266 Furniture, householdtextile, carpetand other floor coverings 15,451,131 16,994,088 14,479,905 19,270,358 2,261,871 Householdtextile 3,568,869 3,792,127 3,42 8,3 38 4,536,687 226,622 Householdappliances 10,376,276 9,091,674 11,184,877 12,358,762 3,529,990 Glassware 2,692,338 1,853,838 3,220,138 3,240,661 798,769 House andgarden appliances 972,847 1,219,101 817,840 1,108,330 504,97 3 Nondurablehouseholdgoods and services 10,155,452 7,42 1,129 11,876,590 12,015,775 3,731,041 Source. DIE, Turkey 2002 HBS. Household'sAverage MonthlyExpendituresonHealth(TL Turkey Type of Settlement Poverty Status Rural Urban Non-Poor HH Poor HH Health .4,127,712 13,955,395 14,236,178 16,92 1,274 4,480,470 Medicalproducts, appliances, andmaterials 6,155,298 6,276,510 6,079,001 7,036,015 3,113,846 Treatment without keepinginhospital 6,366,685 6,560,875 6,244,45 1 7,832,358 1,305,156 Treatment inhospital 1,605,728 1,118,010 1,912,726 2,052,902 61,468 Source DIE, Turkey 2002 HBS. 134 Household's Average Monthly Expenditures on Transportation(TL) Type of Settlement Poverty Status Turkey Rural Urban Non-Poor Poor Households Households Transportation 46,917,368 33,023,943 55,662,679 58,045,979 8,486,002 Vehicle purchase 6,286,030 1,436,997 9,338,287 8,060,823 156,990 Operating o f the personal vehicles 22,862,932 19,292,156 25,110.58 1 28,851,25 1 2,182,969 Transportation services 17,768,406 12,294,789 21,213,811 21,133,906 6,146,043 Source: DIE, Turkey 2002 HBS. Household's Average Monthly Expenditures on Communication (TL) Type of Settlement Poverty Status Turkey Rural Urban Non-Poor Poor Households Households Communication 27,362,109 20,820,121 31,480,008 32,599,264 9,276,204 Postal services 40,906 19,079 54,644 51,168 5,468 Telephone and telefax equipments 3,380,155 3,773,542 3,132,535 4,192,931 573,328 Telephone and telefax services 23,941,048 17,027,500 28,292,828 28,355,165 8,697,408 Source: DIE,Turkey 2002 HBS. Household's Average Monthly Expenditu es on Entertainment and Culture(TL) Type of Settlement Poverty Status Turkey Rural Non-Poor Poor Urban Households Households Entertainment and culture 14,862,268 12,101,713 16,599,919 18,169,140 3,442,372 Visual, auditory, and photography equipment 5,858,250 7,012,750 5,131,541 6,986,606 1,961,602 Other main durable recreational and cultural goods 374,641 719,760 157,403 482,242 3,05 1 Other recreational equipment, goods, and services for gardening and pets 898,017 510,499 1,14 1,943 1,118,226 137,549 Recreational and sport services 2,787,295 1,23 1,942 3,766,323 3,492,486 351,999 Newspaper, book, and stationery 3,878,372 2,617,93 1 4,67 1,764 4,715,980 985,787 Tours 1,065,695 8,831 1,730,945 1,373,599 2,384 Source: DIE,Turkey 2002 HBS. 135 Household'sAverageMonthlyExpenditureson Education(TL) Turkey Type of Settlement Poverty Status Rural Urban Yon-Poor Households Poor Households Education 8,387,2 13 4,497,613 10,835,548 10,675,340 485,429 Pre-primary and primary education 1,478,850 527,594 2,077,624 1,867,793 135,678 Secondary education 1,057,236 1,260,763 929,125 1,323,594 137,399 Pre-university education 3,609,717 1,748,586 4,78 1,218 4,607,833 162,840 University education 1,822,2 12 883,182 2,4 13,291 2,349,872 0 Education not definable by level 419,198 77,488 634,289 526,248 49,512 Source: DIE, Turkey 2002 HBS. Household'sAverage MonthlyExpenditureson Education(TL) Turkev Type of Settlement Poverty Status Rural Urban Yon-Poor Households Poor Households Restaurants and hotels 15,037,810 13,358,012 16,095,170 18,077,677 4,539,983 Meal services 13,459,082 12,491,193 14,068,326 16,059,038 4,480,438 Accomodation services 1,578,728 866,819 2,026,844 2,018,640 59,545 Source: DIE, Turkey 2002 HBS Household'sAverage MonthlyExpendituresonEducation(TL) Turkey Type of Settlement Poverty Status Rural Urban Yon-Poor Households 'oor Households Various goods and services 25,330,134 !2,389,654 !7,18 1,039 30,953,323 5,911,102 Personal care 10,020,723 6,681,370 12,122,701 11,979,547 3,256,151 Other personal care Materials 5,373,555 7,150,584 4,254,992 6,608,523 1,108,737 Social services 30,364 0 49,477 39,156 0 Insurance 1,467,226 586,990 2,02 1,297 1,874,328 61,349 Financial services 207,427 235,172 189,962 266,150 4,635 Other services 8,230,839 7,735,537 8,542,610 10,185,620 1,480,230 Source: DIE, Turkey2002 HBS. 136 Household'sAverage MonthlyIncome(TL) Turkey Type of Settlement Rural Urban Total monthly disposable income 672,103,224 564,545,784 739,805,987 Total monthly income incash 639,211,405 543,466,796 699,478,501 Total monthly income inkind 32,891,819 21,078,988 40,327,486 Source: DIE, Turkey 2002 HBS. Household'sAverageMonthlyCASH Income(TL) Type of Settlement PovertyStatus Turkey Rural Urban Non-Poor Poor Households Households Total monthly income incash 639,211,405 j43,466,796 j99,478,50 1 738,542,973 296,181,381 Total netcash income as wage or salary inthe survey month 241,134,556 141,984,036 303,545,527 281,879,866 100,425,364 Net disposable income incash inthe survey month 125,712,674 80,282,906 154,308,750 148,329,645 47,607,591 Total net agricultural income incash inthe survey month 74,346,597 176,424,673 10,092,856 74,172,091 74,949,230 Total net income incash from additionaljob inthe survey 19,295,668 27,982,891 13,827,436 21,735,790 10,868,989 month Total net income incash from all previous jobs inthe last 12 133,371,432 76,957,193 168,881,759 127,474,292 153,736,516 months Rentfromreal estate income incash inthe survey month 26,197,746 10,929,818 35,808,248 32,685,823 3,79 1,927 Interest income incash from movable property inthe survey month 31,794,266 15,809,179 41,856,189 39,973,071 3,549,714 Transfer income incash inthe survey month 116,345,155 87,212,289 134,683,036 136,385,222 47,139,116 Source. DIE,Turkey 2002 HBS 137 Household'sAverage MonthlyIncome InKind L) Type of Settlement PovertyStatus Turkey Rural Urban Non-Poor Poor Households Households Total monthly income inkind 32,891,819 :1,078,988 ,0,327,486 38,174,736 14,647,880 Total income inkindas wage or salary in 27,322,675 5,766,196 the survey month 22,482,203 2,43 1,164 18,808,898 Total income inkindas entrepreneurship 4,013,223 1,73 1,892 income inthe survey month 3,500,954 2,78 1,567 3,953,778 Total net agricultural income inkindin 9,782 0 the survey month 7,585 14,474 3,249 Total income inkindfrom additional job 202,875 112,392 inthe survey month 182,557 236,119 148,842 Total income inkindfrom all previous 4,710,377 jobs inthe last 12 months 4,823,847 2,042,993 6,574,275 4,856,705 Rent from real estate income inkindin 3,345 2,647 0 the survey month 2,053 0 Imputedrent for the survey month 0 0 0 0 0 Transfer income inkind inthe survey 6,413,460 6,723,191 month 6,483,009 5,298,640 7,2283 19 Source: DIE, Turkey 2002 HBS. Household'sAverageAnnualIncome (TL) Type of Settlement Poverty Status Turkey Rural Urban Non-Poor Poor Households Households Total annual disposable 7,65,260,790 6,056,834,352 3,025,913,OS 1 8,406,3 16317 3,324,756,431 Total annual income in 8,06 1,387,046 3,176,547,7 17 cash 6,964,504,409 5,879,567,064 7,647,425,6 17 Total annual income in kind 300,756,381 177,267,287 378,487,433 344,929,77 1 148,208,714 Source. DIE, Turkey 2002 HBS 138 Household'sAvera Annual CashIncome(TL) Type of Settlement Poverty Status Turkey Rural Urban Non-Poor HH Poor HH Total annual income incash 5,964,504,409 j,879,567,064 7,647,425,6 17 3,061,387,046 3,176,547,7 17 Total net cash income as wage or salary inthe last 12 months !,3 17,576,913 1,3 12,237,637 2,950,394,574 !,719,991,977 927,883,264 Total bonus income inthe last 12 months l13,693,73C 62,025,938 146,2 16,374 143,762,894 9,853,367 Total premium, allowance income for bayram, etc., inthe last 12 35,172,034 12,356,576 49,533,379 44,111,001 4.302.351 months Total expertise, counseling income inthe last 12months 2,337,690 213,510 3,674,769 3,013,178 4,969 Total per diem(travel allowance) income inthe last 12 months 0 0 0 0 0 Total net disposable entrepreneurship income incash .,391,840,185 842,269,3 85 1,737,771,271 1,656,452,040 478,033,898 inthe last 12months Total net agricultural income incash inthe last 12months 892,143,203 !,117,055,050 121,114,091 890,05 1,084 899,368,096 Total net income incash from additional job inthe last 12 182,262,486 267,491,791 128,614,3 19 205,369,279 102,465,863 months Total net income incash from all previousjobs inthe last 12 133,371,432 76,957,193 168,881,759 127,474,292 153,736,516 months Rentfromreal estate income incash inthe last 12months 296,235,392 121,396,133 406,289,156 370,04 1,275 41,355,353 Profit income frombank account in the last 12 months 232,4 16,447 72,899,704 332,825,348 295,033,799 16,174,699 Interest income from foreign currency bank accounts inthe 15,921,198 22,954,920 11,493,774 19,978,704 1,909,075 last 12 months Interest income from bond, debenture inthe last 12 months 62,046,915 31,452,401 81,304,841 76,204,O 16 13,157,012 Interest income from capital association inthe last 12 months 44,757,623 1,504,454 71,983,626 57,702,776 53,043 Interest income fromprivate finance council inthe last 12 months 10,975,697 582,618 17,517,691 14,151,698 7,744 Pension from government inthe last 12months 826,171,920 559,878,543 993,792,102 982,267,466 287,114,095 Tax (returned) from government in the last 12 months 57,553,657 32,206,320 73,508,7 10 69,285,744 17,038,255 139 Household'sAverar Annual CashIncome(TL) Type of Settlement PovertyStatus Turkey Rural Urban Non-Poor HH PoorHH Scholarship from governmentinthe last 12 months 3,206,961 2,708,054 3,521,002 3,721,242 1,430,950 Old-age income from governmentin the last 12 months 43,412,524 96,626,467 9,9 16,645 40,071,717 54,949,613 Unemploymentpension from governmentinthe last 12 1,425,121 2,805,205 556,417 1,811,657 90,264 months Orphanand widower pensionfrom governmentinthe last 12 93,558,512 65,086,869 111,480,182 110,877,439 33,749,612 months Disability (war veteran) pension from government inthe last 12 10,035,59 1 4,762,177 13,354,978 12,057,892 3,051,809 months Pensionfrom abroadinthe last 12 months 35,115,146 52,714,058 24,037,391 42,914,302 8,181,668 Foreigncurrency from abroadinthe last 12 months 8,254,063 10,332,200 6,945,965 10,400,923 840,131 Scholarship, alimony, etc., from abroad inthe last 12months 21,380,827 16,066,068 24,726,239 22,7 15,410 16,772,002 Alimony income fromother private persons or institutions inthe last 4,970,810 0 8,099,720 5,992,960 1,440,936 12months Scholarship, alimony, etc., from other private persons or 128,668,333 94,983,796 149,871,294 135,932,282 103,583,13 1 institutions inthe last 12 months Source: DIE,Turkey 2002 HBS. 140 Household'sAverage Annual IncomeInKind (T Type of Settlement Poverty Status Turkey Rural Urban Non-Poor Poor Households Households Total annual income inkind 300,756,381 177,267,287 378,487,433 344,929,771 148,208,714 Total income inkindas wage or salary inthe last 12months !04,588,895 96,291,576 !72,757,380 246,762,2 17 58,948,229 Total net entrepreneurship income in kindinthe last 12months 39,046,736 29,247,897 45,2 14,68 1 44,308,530 20,875,740 Total net agricultural income inkindin the last 12 months 414,796 589,529 304,809 458,947 262,326 Total income inkindfrom additional job inthe last 12months 3,205,989 5,337,639 1,864,208 2,909,242 4,230,773 Total income inkindfrom all previous jobs inthe last 12 months 4,823,847 2,042,993 6,574,275 4,856,705 4,710,377 Rent from real estate income inkindin the last 12 months 854,378 1,7538,184 285,470 820,816 970,279 Imputedrent for the last 12 months 0 0 0 0 0 Income inkindfrom government inthe last 12 months 6,021,880 9,452,001 3,862,767 5,565,361 7,598,415 Transfer income inkindfrom abroad in the last 12 months 491,855 102,777 736,763 569,959 222,132 Transfer income inkindfrom other private persons or institutions inthe40,002,97 1 29,925,532 46,346,284 37,145,3 10 49,871,573 last 12 months Source: DIE,Turkey 2002 HBS. 141 Appendix 2: Poverty Figures for Various Poverty Lines The Poverty Headcount According to Various Poverty Lines, Based on Consumption Per Capita Relative Absolute Food Zomplete Poverty Absolute Poverty 4bsolute Poverty Poverty i n e 50% Poverty Line Poverty Line TL Line TL of Line $1 Line $4.3 Median PPP Day i2.15 PPP PPPDay TL Day ?raction Fraction Fraction Fraction Fraction Fraction of Poor of Poor of Poor of Poor of Poor of Poor rurkey .0379 .3167 .2030 .0038 .0918 .3893 Lower thanprimary .lo35 S902 .4701 .0158 .2380 .6639 Education ofHH 'primary .0324 .3240 .1887 .0021 .Of303 .4105 lead ' Secondarylvocational ,0054 .lo83 .0541 .oooo ,0209 .1466 University .0012 ,0228 .0111 .oooo .0012 .0284 Rural .0556 .3989 .2700 .0056 .1289 .4822 rype of settlement Urban .0261 .2619 ,1583 .0026 .0671 ,3275 No .0365 .2832 .1781 .0044 ,0820 .3500 Access to land Yes ,0407 .3842 .2532 .0027 .1117 .4687 .0487 .3850 .2537 .0050 .1179 .4665 Own automobile ` No ,Yes .0013 .0871 .0323 .oooo .0042 .1298 Zero .0019 .1283 .0594 .0008 .0126 ,1792 One .0107 .2304 ,1167 .oooo .0321 .3044 Children under age 14 Two .O 179 .3291 .1909 .0067 .0554 .4254 Three .0906 S789 .4151 .0040 .2440 .6705 4 andmore .1958 .7670 .6277 .0151 .3827 3276 Zero .0342 .2989 .1891 .0038 .0874 .3708 Number ofHH members age >60 One .0298 .3718 .2266 .oo18 .0700 .4550 Two and more .0820 .3569 .2697 .0078 .1677 .4160 .3185 .2030 .0036 .0925 .3913 Gender of HHhead Male .0380 Female .0352 .2905 .2028 .0071 .0823 .3613 Dependency ratio n 0-0.25 ,0543 .3802 .2545 .0057 .1204 .4616 o f employed (EXCLunpaid 0.25-0.5 .0117 ,2273 ,1332 ,0007 .05 10 ,2887 family workers)/HHsize 0.5 and more ,0016 .1144 .0243 .oooo .0035 .1532 Source: DIE, Turkey 2002 HBS. 142 The Poverty Gap Index According to Various Poverty Lines, Based on Consumption Per Capita Relative Absolute Food Complete Poverty Absolute Poverty 4bsolute Poverty Poverty Line Poverty Line Poverty ,ine TL line TL 50% of Line $1 ,ine $4.3 Median PPPDay $2.15 PPPDay TL PPP Day Poverty Poverty Poverty Poverty Poverty Poverty Gap Gap Gap Gap Gap Gap Index Index Index Index Index Index Turkey .0067 .0982 .OS48 .0007 .0200 .1288 Lower than primary .0188 .2223 .1395 ,0028 .OS58 .2717 Education o f HH Primary .0057 .0925 .0483 .0004 .O 169 .1254 head Secondarylvocationa .0006 .0264 .0123 .oooo ,0030 .0384 University .oooo .0051 ,0024 .oooo .0003 .0076 Type of settlement Rural .0096 .1304 .0745 .OO 13 .0284 .1676 Urban ,0048 .0767 .0416 .0003 .0144 .1029 Access to land No ,0075 ,0882 ,0502 .0007 .0195 .1156 Yes .0052 .1184 .0640 .0006 ,0211 ,1554 No Ownautomobile .0086 .1225 .0696 .0009 ,0257 .1589 Yes .0003 .0165 .0049 .oooo .0009 .0275 Zero ,0006 .0286 .0110 .oooo .OO 17 .0433 One ,0015 .OS65 .0248 .oooo .0062 ,0817 Children under age 14 Two .0051 .0883 .0425 .0017 .0116 .1228 Three .0127 .2044 .1244 .0010 .0485 .2559 4 and more .0343 .3092 .2066 .0016 .0945 .3684 Zero .0068 ,0917 .os10 .0006 .O 187 .1210 Number o f HH members age >60 One .0048 .lo36 .os01 .0002 .0162 .1406 Two and more .0094 .1401 .0931 .0022 .0375 .1693 Gender o f hhhead Male .0065 .0983 .OS47 .0007 .0200 .1290 Female .0094 .0972 ,0558 .0005 .0202 .1253 Dependency ratio G0.25 .0096 .1236 .0715 .0011 ,0280 .1593 nofemployed (EXCL unpaid 0.25-0.5 .0021 .0623 .0308 .OOOO .0076 .0857 family workers)/HHsize 0.5 and more ,0002 .0186 .0037 .OOOO .0006 ,0322 Source: DIE, Turkey 2002 HBS. 143 The Poverty Severity Index According to Various Poverty Lines, Basedon Consumption Per Capita, Based on HHConsumption per Capita Relative Absolute Food Complete Poverty Absolute Poverty Absolute Poverty Poverty Ane 50% Poverty Line Poverty Line T L Line T L of Line $1 Line$4.3 Median PPPDay i2.15 PPP PPPDay TL Day Poverty Poverty Poverty Poverty Poverty Poverty Severity Severity Severity Severity Severity Severity Index Index Index Index Index Index rurkey ,0021 .0423 .0214 .0002 .0068 .0584 Lower than primary ,0065 .lo48 .OS72 .0008 .o195 .1371 Education o fHH Primary .OO 16 .0381 .0185 .0001 .0056 .0541 head Secondarylvocational .ooo1 ,0098 .0041 .oooo .0007 .0149 University .oooo .0018 .0007 .oooo .0001 ,0028 Rural Type o f settlement .0032 .0575 .0299 .0004 .0098 .0783 Urban .0013 .0322 .0158 .oooo .0048 .0452 Access to land No .0023 .0389 .0203 .0002 .0070 .0531 Yes .OO 16 .0493 ,0237 .0002 .0063 .0692 Own automobile No .0026 .0535 .0274 .0002 .0087 .0733 Yes .0001 .0047 .OO 13 .oooo .0003 .0086 Zero .0002 .0092 .0031 .oooo ,0006 .0153 One .0003 .0202 .0081 .oooo .0017 .0312 Children under age 14 Two .0023 .0339 .0148 .0005 .0047 .0501 Three ,0036 .0937 .0499 .0004 .0146 .1246 4 and more .0095 .1555 .0913 .0003 ,0331 .1971 Zero .0021 ,0396 .0201 .0002 .0065 .0547 Number o f HH members age >60 One .oo12 ,0402 .0179 .oooo .0051 .0588 Two and more .0033 .0683 .0385 .0006 .0119 .0879 Gender ofHHhead Male .0020 .0423 ,0214 .0002 .0067 .0584 Female .0030 .0430 .0222 .0001 .0080 .0585 Dependency ratio n 0-0.25 ,0030 .0551 .0289 .0003 .0096 ,0746 of Employed (EXCL unpaid 0.25-0.5 .0005 .0239 .0102 .oooo ,0022 .0352 family workers)/HHsize 0.5 and more .oooo ,0044 .0011 .oooo .0002 .0092 Source: DIE, Turkey 2002 HBS. 144 The Relative Poverty Risks According to Various Poverty Lines, Based on Consurr tioi Per Capita Relative Relative Relative Re1ative Relative Relative Risk Risk Risk Risk Risk Risk rurkey .oo .oo .oo .oo .oo .oo Lower than primary 1.73 -86 1.32 3.13 1.59 .71 Education of HHhead Primary -.14 .02 -.07 -.45 -.13 .05 Secondarylvocational -.86 -.66 -.73 -1.00 -.77 -.62 University -.97 -.93 -.95 -1.00 -.99 -.93 Rural .47 .26 .33 .47 .40 .24 rype o f settlement Urban -.3 1 -.17 -.22 -.3 1 -.27 -.16 -.lo Access to land No -.04 -.11 -.12 .14 -.11 Yes .07 .21 .25 -.28 .22 .20 Own automobile No .29 .22 .25 .30 .28 .20 Yes -.97 -.72 -.84 -1.00 -.95 -.67 Zero -.95 -.60 -.7 1 -30 -.86 -.54 One -.72 -.27 -,42 -1.oo -.65 -.22 Children under age 14 Two -.53 .04 -.06 .75 -.40 .09 Three 1.39 3 3 1.os .03 1.66 .72 4 and more 4.17 1.42 2.09 2.95 3.17 1.13 Zero -.lo -.06 -.07 -.01 -.05 -.os ' Number ofHH members age >60 One -.2 1 .17 .12 -.52 -.24 .17 Two and more 1.17 .13 .33 1.03 .83 .O? .oo .o1 .oo .oo Gender ofHHhead Male -.06 .01 Female -.07 -.08 .oo .86 -.lo -.07 Dependency ratio nof 0-0.25 .43 .20 .25 S O .31 .19 -.69 -.28 -.34 -.82 -.44 -.26 workers)/ HHsize 0.5 and more -.96 -.64 -38 -1.oo -.96 -.61 Source: DIE, Turkey 2002 HBS. 145 The Relative Poverty Risks According to Various Poverty Lines, Based on Consumption Per Capita Relative Relative Relative Relative Relative Relative Risk Risk Risk Risk Risk Risk Index Index Index Index Index Index Turkey 1.oo 1.oo 1.oo 1.oo 1.oo 1.oo Lower thanprimary 2.73 1.86 2.32 4.13 2.59 1.71 Education o f HH Primary .86 1.02 .93 .55 .87 1.05 head Secondarylvocational .14 .34 .27 .oo .23 .38 University .03 .07 .05 .oo .o1 .07 Type ofsettlement Rural 1.47 1.26 1.33 1.47 1.40 1.24 Urban .69 .83 .78 .69 .73 .84 Access to land N o .96 .89 .88 1.14 .89 .90 Yes 1.07 1.21 1.25 .72 1.22 1.20 N o Ownautomobile 1.29 1.22 1.25 1.30 1.28 1.20 Yes .03 .28 .16 .oo .05 .33 Zero .05 .40 .29 .20 .14 .46 One .28 .73 .58 .oo .35 .78 Children under age 14 Two .47 1.04 .94 1.75 .60 1.09 Three 2.39 1.83 2.05 1.03 2.66 1.72 4 andmore 5.17 2.42 3.09 3.95 4.17 2.13 Zero .90 .94 .93 .99 .95 .95 NumberofHH members age >60 One .79 1.17 1.12 .48 .76 1.17 Two and more 2.17 1.13 1.33 2.03 1.83 1.07 Gender ofHHhead Male 1.oo 1.01 1.oo .94 1.01 1.oo Female .93 .92 1.oo 1.86 .90 .93 Dependency ratio n 0-0.25 1.43 1.20 1.25 1.50 1.31 1.19 o femployed(EXCL unpaid family 0.25-0.5 .31 .72 .66 .18 .56 .74 workers)/HH size 0.5 andmore .04 .36 .12 .oo .04 .39 Source: DIE, Turkey 2002 HBS. 146 The PovertyHeadcountAccordingto VariousPovertyLines, Basedon ConsumptionPerAdult EquivalentAdjusted PO, Poverty PO, Poverty PO, Poverty PO, Food PO, Complete PO, Relative Line=$l Line=$2.15 Line=$4.3 PovertyLine, PovertyLine, PovertyLine, PPP=663515 PPF1426686 PPP=2853313 Adjusted Ad. Adjusted Adjusted TLlday, TL/day, TLIday, Eq. Measures Ad. Eq. Ad. Eq. Adjusted Adjusted Adjusted Measures Measures Ad. Eq. Ad. Eq. Ad. Eq. Measures Measures Measures Fractionof Fractionof Fractionof Fractionof Fractionof Fractionof Poor Poor Poor Poor Poor Poor Turkey .0135 ,2696 ,1474 ,0021 ,0411 ,3414 Lower than primary .0304 .5245 ,3238 .0084 .0807 .5985 Educationof Primary .0133 ,2709 ,1406 ,0013 ,0429 ,3568 HHhead Secondary /vocational ,0019 .0915 .0423 .oooo .0060 .I243 University .oooo .O 160 .0095 .oooo ,0012 ,0267 Type of Rural ,0201 ,3448 ,1986 ,0046 ,0533 .4313 settlement Urban .0092 ,2195 ,1133 ,0005 ,0330 ,2815 Access to No ,0145 ,2466 .I391 ,0018 .0456 ,3102 land Yes .01I 6 .3I62 .I641 .0027 ,0320 .4044 OWn No ,0174 ,3313 .I852 .0028 .0525 .4129 automobi1e Yes .0007 ,0623 .0202 .oooo ,0028 .IO1 1 Zero ,0043 .I594 ,0735 ,0008 ,0131 ,2201 One ,0077 ,2010 ,0925 .oooo .0178 .2741 Children under age 14 Two ,0125 .2702 ,1277 ,0066 .0347 ,3526 Three ,0237 ,4678 ,2946 .0027 .0869 S458 4 and more ,0460 ,5395 .3781 .0003 ,1426 .6112 Number of Zero ,0150 ,2588 ,1404 ,0018 .0436 ,3244 HH One .0075 ,2800 ,1291 .0005 ,0315 .3867 membersage >60 Two and more ,0129 ,3366 .2367 .0078 ,0386 ,3939 Gender of hh Male ,0128 ,2662 ,1458 ,0022 ,0389 .3388 head Female ,0235 .3197 ,1699 .0008 ,0729 ,3792 Dependency C0.25 ,0186 ,3163 ,1746 ,003I .0535 ,3880 ratio nof employed 0.25-0.5 ,0059 ,2034 ,1122 ,0006 ,0212 ,2823 (EXCL unpaid family 0.5 and more .oooo ,1241 ,0450 .oooo ,0135 ,1604 workers)/HH size Source: DIE, Turkey 2002 HBS. 147 The Poverty Gap Index Accordingto Various Poverty Lines, Basedon ConsumptionPer Adult EquivalentAdjusted P1, Food P1, Poverty P1, Poverty P1, Poverty Poverty P1, Complete P1, Relative Line=$l Line=$2.15 Line=$4.3 Line, Poverty Line, PovertyLine, PPP=663575 PPF1426686 PPP=2853373 Adjusted Adjusted Adjusted TLlday, TLlday , TLlday, Ad. Eq. Ad. Eq. Ad. Eq. Adjusted Adjusted Adjusted Measures Measures Measures Ad. Eq. Ad. Eq. Ad. Eq. Measures Measures Measures Poverty PovertyGap Poverty Gap Poverty Gap Poverty Gap PovertyGap Gap Index Index Index Index Index Index I'urkey ,0026 .0688 .0313 .0004 .0083 .0973 Lower than primary .0081 .1475 ,0696 ,0012 .0200 .1975 Educationof Primary ,0021 .0667 ,0301 .0003 ,0078 ,0964 HH head Secondary1 vocational ,0001 ,0191 ,0073 .oooo .0009 ,0298 University .oooo .0036 .0010 .oooo .ooo1 ,0059 Type of Rural ,0043 ,0923 ,0431 .0009 .0117 .1282 settlement Urban ,0015 .0531 .0234 .oooo .0060 ,0767 Access to No ,0029 ,0649 .0310 ,0004 ,0092 .0907 land Yes ,0021 .0765 .0318 ,0003 ,0064 ,1107 OWn No ,0034 .0862 .0398 ,0005 ,0106 .1207 automobile Yes ,0001 .0102 ,0029 .oooo ,0004 .0185 Zero ,0009 ,0345 ,0119 ,0001 ,0024 ,0530 One .0009 ,0444 .0184 .oooo .0037 .0679 Children under age 14 Two .0041 ,0640 .0273 ,0011 .0086 ,0939 Three ,0046 ,1315 ,0665 ,0007 .0162 ,1763 4 andmore ,0062 .1695 ,0897 .0001 ,0268 .2191 Number of Zero ,0028 ,0667 ,0312 ,0004 ,0087 ,0935 HH members age One ,0011 ,0635 ,0249 .oooo ,0054 .0970 >60 Two and more .0041 .0953 ,0442 ,0009 ,0095 ,1280 Gender of Male .0025 .0673 ,0304 ,0004 ,0078 ,0957 HHhead Female ,0043 .0897 ,0446 .0001 ,0148 ,1212 Source: DIE, Turkey 2002 HBS. 148 The Poverty Severity Index Accordingto VariousPoverty Lines, Basedon ConsumptionPer Adult EquivalentAdjusted P2, Poverty P2, Poverty P2, Poverty P2, Food P2, Complete P2, Relative Line=$l Line=$2.15 Line=$4.3 PovertyLine, Poverty Line, ?overtyLine, PPP=663575 PPP=1426686 'PP=2853373 Adjusted Adjusted Adjusted TLlday, TLlday, TLIday, Ad. Eq. Ad. Eq. Ad. Eq. Adjusted Adjusted Adjusted Measures Measures Measures Ad. Eq. Ad. Eq. Ad. Eq. Measures Measures Measures Poverty Poverty Poverty Poverty Poverty Poverty ieverityIndex Severity Index ieverityIndex everity Index Severity Index everity Index Turkey .0008 .0253 .0104 .0001 ,0027 ,0384 Lower than primary ,0028 ,0562 .0242 ,0002 .0075 .0826 Educationof Primary .0006 .0243 .0099 .0001 .0023 ,0373 HHhead Secondary /vocational .oooo ,0061 ,0018 .oooo .0002 .0102 University .oooo .0010 ,0002 .oooo .oooo .oo19 Type of rural .0016 .0347 .0147 .0002 ,0042 .0519 settlement urban ,0003 ,0190 .0076 .oooo .0017 ,0294 Access to No ,0009 ,0249 ,0109 .0001 ,0030 .0368 land Yes ,0007 .0262 ,0095 .oooo ,0021 ,0415 Own No ,0011 ,0320 ,0133 .0001 .0035 .0482 automobile Yes .oooo ,0028 ,0007 .oooo .0001 ,0054 Zero ,0003 .0109 ,0035 .oooo ,0008 ,0183 One .0001 ,0151 .0056 .oooo ,0010 ,0243 Children under age 14 Two .0018 ,0232 .0100 ,0002 .0037 ,0358 Three .OO 14 ,0508 ,0215 .0002 ,0050 .0742 4 andmore ,0012 ,0690 .03 13 .oooo ,0074 ,0975 Numberof Zero ,0008 ,0250 .O 107 .0001 ,0029 ,0375 HH One ,0003 ,0212 ,0076 .oooo ,0014 ,0345 members age >60 Two and more .0017 ,0352 .0136 .0001 ,0038 ,0525 Gender of male ,0008 ,0246 .o100 .0001 ,0026 ,0375 HHhead female ,0011 ,0356 ,0165 .oooo .0045 ,0514 Dependency 0 - 0.25 ,0012 .0313 ,0136 ,0001 .0038 .0466 ratio n of employed 0.25 - 0.5 ,0002 ,0169 .0057 .oooo ,0010 ,0270 (EXCL unpaid family 0.5 andmore .oooo ,0063 ,0019 .oooo .0002 ,0116 workers)/HH size Source: DIE, Turkey 2002 HBS. 149 The Relative Poverty Risks According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted Relative telative Relative Relative telative belative Risk Risk Risk Risk Risk Risk Turkey .oo .oo .oo .oo .oo .oo Lower thanprimary 1.24 .95 1.20 2.93 .96 .75 EducationofHHhead Primary -.02 .oo -.05 -.40 .04 .04 Secondary/vocational -.86 -.66 -.71 -1.00 -.85 -.64 University -1.00 -.94 -.94 -1.00 -.97 -.92 Rural .35 1.16 .30 Type ofsettlement .49 .28 .26 Urban -.32 -.I9 -.23 -.78 -.20 -.18 -.14 Accessto land No .07 -.09 -.06 .I1 -.09 Yes -.I4 .I7 .I1 .29 -.22 .18 Own automobile No .28 .23 .26 .30 .28 .21 Yes -.95 -.77 -.86 -1.00 -.93 -.70 Zero -.68 -.41 -.50 -.64 -.68 -.36 One -.43 -.25 -.37 -1.oo -.57 -.20 Childrenunderage 14 Two -.08 .oo -.I3 2.10 -.I5 .03 Three .75 .73 1.oo .25 1.11 .60 4 andmore 2.40 1.oo 1.57 -.87 2.47 .79 Zero .ll -.04 -.05 -.16 .06 -.05 Number of HHmembers age >60 One -.45 .04 -.I2 -.76 -.23 .13 Two and more -.05 .25 .61 2.65 -.06 .15 Gender ofHHhead Male -.05 -.01 -.01 .04 -.05 -.01 Female .74 .19 .15 -.61 .77 .11 Dependencyratio nof 0-0.25 .37 .I7 .18 .45 .30 .14 employed (EXCL unpaid 0.25-0.5 -.57 -.25 -.24 -.71 -.48 -.17 family workers)/HHsize 0.5 andmore -1.00 -.54 -.69 -1.oo -.67 -.53 Source: DIE, Turkey 2002 HBS. 150 The Relative Poverty Risks According to Various Poverty Lines, Based on ConsumptionPer Aduk Equivalent Adjusted Xelative ielative Zelative Relative ielative telativc Risk Risk Risk Risk Risk Risk Index Index Index index Index Index Turkey 1.oo 1.oo 1.oo 1.oo 1.oo 1.oo Lower than primary 2.24 1.95 2.20 3.93 1.96 1.75 Primary .98 1.oo .95 .60 1.04 1.04 Education o f HHhead Secondary1 vocational .14 .34 .29 .oo .15 .36 University .oo .06 .06 .oo .03 .08 Type o f settlement Rural 1.49 1.28 1.35 2.16 1.30 1.26 Urban .68 .81 .77 .22 .80 .82 Access to land No 1.07 .91 .94 .86 1.11 .91 Yes .86 1.17 1.11 1.29 .78 1.18 1.28 1.23 1.26 1.28 Own automobile No 1.30 1.21 Yes .05 .23 .I4 .oo .07 .30 Zero .32 .59 s o .36 .32 .64 One .57 .75 .63 .oo .43 .80 Children under age 14 Two .92 1.oo .87 3.10 .85 1.03 Three 1.75 1.73 2.00 1.25 2.11 1.60 4 andmore 3.40 2.00 2.57 .13 3.47 1.79 Zero 1.11 .96 .95 .84 1.06 .95 ' Number of HHmembers age >60 One .55 1.04 .88 .24 .77 1.13 Two and more .95 1.25 1.61 3.65 .94 1.15 Gender ofHHhead Male .95 .99 .99 1.04 .95 .99 Female 1.74 1.19 1.15 .39 1.77 1.11 Dependency ratio no f 0-0.25 1.37 1.17 1.18 1.45 1.30 1.14 employed (EXCL unpaid 0.25-0.5 .43 .75 .76 .29 .52 .83 family workers)/HH size 0.5 and more .oo .46 .3 1 .oo .33 .47 Source: DIE,Turkey 2002 HBS. 151 The PovertyHeadcountAccording to VariousPovertyLines, Basedon ConsumptionPerAdult EquivalentAdjusted PO, Food PO, '0, Relative PO, Poverty PO, Poverty PO, Poverty Poverty Complete Poverty Line=$l Line=$2.15 Line=$4.3 Line, Poverty Line, PPP=66357 PPP=14266 PPP=2853373 OECD Line, OECD 5 TLIday, 86 TLIday, TLIday, Ad. Eq. OECD Ad. Eq. OECD OECD OECD Measures Ad. Eq. Ad. Eq. Ad. Eq. Ad. Eq. Measures Measures Measures Measures Measures Fractionof Fractionof Fractionof Fractionof Fractionof Fractionof Poor Poor Poor Poor Poor Poor Turkey ,0044 .I597 .0814 ,0002 .0204 .2151 Lower than primary ,0189 ,4082 .2186 ,0015 .0518 .4910 Educationof Primary .0023 ,1403 .0696 .oooo .0188 ,2026 HHhead Secondary /vocational .oooo ,0385 ,0170 .oooo .0013 .0534 University .oooo .0059 ,0012 .oooo .oooo ,0116 Type of Rural .0065 ,2201 .I214 .0006 ,0264 ,2926 settlement Urban ,0030 .I195 .0548 .oooo .O163 ,1634 Access to No ,0047 ,1382 ,0700 ,0004 ,0235 .I850 land Yes ,0038 ,2032 ,1045 .oooo ,0141 ,2758 Own No .0057 ,2010 ,1044 .0003 ,0261 ,2687 automobile Yes .oooo .0208 .0039 .oooo ,0012 ,0349 Zero ,0019 ,0658 .0253 .oooo ,0031 .1103 One .0005 ,0956 .0430 .oooo ,0089 ,1473 Children under age 14 Two ,0071 ,1400 ,0506 .oooo .0136 ,1940 Three .0043 ,3114 ,1992 ,0021 .0398 .3898 4 andmore ,0151 ,4679 ,2786 .oooo ,0926 ,5423 Number of Zero .0046 .1445 ,0719 ,0003 ,0205 ,1983 HH membersage One .0019 .I708 ,0755 .oooo ,0173 .2394 >60 Two andmore ,0078 .2612 ,1685 .oooo ,0248 .3044 Gender of Male ,0042 ,1587 ,0824 .0003 ,0202 .2147 HHhead Female ,0078 .1754 ,0673 .oooo ,0233 ,2206 Dependency 0-0.25 .0062 ,1965 .I062 ,0004 .0285 ,2572 ratio n of employed 0.25-0.5 ,0017 .I129 ,0460 .oooo .0074 ,1610 (EXCL unpaid family 0.5 andmore .oooo .O 176 ,0049 .oooo ,0016 ,0545 workers)/HH size Source DIE, Turkey2002 HBS. 152 The PovertyGapIndexAccordingto VariousPovertyLines, Basedon ConsumptionPer Adult EquivalentAdjusted P1, Food PI, PI, P1, Poverty P1, Poverty P1, Poverty Poverty Zomplete Relative Line=$l Line=$2.15 Line=$4.3 Line, Poverty Poverty PP=663575 'PP=1426686 PP=2853373 OECD Line, Line, TLIday, TLIday, TLIday, Ad. Eq. OECD OECD 3ECD Ad. OECD Ad. OECDAd. Measures Ad. Eq. Ad. Eq. Eq- Measures vleasures Measures :q. Measures :q.Measures Poverty Poverty Poverty 'overty Gap PovertyGap 'overty Gap Gap Gap Gap Index Index Index Index Index Index Turkey .0010 ,0384 ,0166 .oooo .0039 .0563 Lower than primary ,0045 .lo24 .0467 .0002 .0119 .1440 Education .0005 .0331 .0139 .oooo .0031 .0497 ofHH Primary head Secondaryivocational .oooo .0078 ,0022 .oooo .0001 .0124 University .oooo ,0010 .0003 .oooo .oooo .0021 Type of Rural ,0019 .0553 .0246 .0001 .0061 ,0793 settlementUrban .0005 .0272 .0112 .oooo .0024 .0410 Access to No .0010 .0345 .0157 .oooo .0041 .0496 land Yes .0011 .0465 .0183 .oooo .0034 .0698 OWn No .OO 13 ,0489 .0213 .oooo .0050 .0713 automobile Yes .oooo ,0031 .0007 .oooo .0002 .0059 Zero .0003 .0124 ,0036 .oooo .0007 .0212 Children One .0001 .0206 .0075 .oooo .0014 .0330 under age Two .0024 ,0279 .0107 .oooo .0042 .0446 14 Three ,0013 .OS57 .0389 .0002 .0073 ,1174 4 and more .0021 .1297 .0653 .oooo .0143 .1747 Numberof Zero .0010 .0349 .0152 .oooo .0038 .0511 HH members One ,0003 .0368 .0144 .oooo .0028 .0577 age >60 Two and more .0030 ,0700 .0316 .oooo .0061 .0953 Gender of Male .0010 .0386 .0166 .oooo .0038 .OS64 HHhead Female ,0010 .0363 .0165 .oooo .0053 ,0554 Source: DIE, Turkey 2002 HBS. 153 The Poverty Severity Index According to Various Poverty Lines, Basedon Consumption Per Adult Equivalent Adjusted P2, Food P2, Poverty P2, Poverty P2, Poverty Poverty P2, Complete P2 Relative Line=%l Line=$2.15 Line=%4.3 Line, Poverty Line, Poverty PPP=663575 ?PP=1426686 PPP=2853373 OECD OECD Ad. Line, OECD TLlday, TLIday, TL/day, OECD Ad. Eq. Eq. Measures Ad. Eq. OECD OECD Ad. Eq. Measures Measures Ad. Eq. Ad. Eq. Measures Measures Measures Poverty Poverty Poverty Poverty Severity Severity Severity Severity Poverty Poverty Severity Index Index Index Index ieverity Index Index Turkey ,0003 ,0135 .0052 .oooo ,0012 ,0211 Lower than primary ,0014 .0374 ,0153 .oooo ,0042 .0568 Educationof Primary ,0002 ,0114 .0043 .oooo ,0008 ,0182 HHhead Secondary /vocational .oooo ,0022 ,0005 .oooo .oooo .0040 University .oooo .0002 ,0001 .oooo .oooo .0005 Type of Rural .0006 ,0199 ,0079 .oooo .0020 ,0306 settlement Urban .0001 ,0093 ,0034 .oooo .OOM ,0149 Access to No ,0003 .0126 ,0052 .oooo ,0012 .O 193 land Yes ,0004 ,0153 .0052 .oooo ,0011 ,0249 O W No ,0004 .O 173 .0067 .oooo .0015 ,0270 automobile Yes .oooo ,0008 .0002 .oooo .oooo ,0016 Zero .ooo1 ,0035 ,0010 .oooo ,0003 ,0065 One .oooo ,0065 .0021 .oooo ,0003 .0110 Children under age 14 Two ,0008 .0096 ,0044 .oooo .0019 ,0156 Three ,0005 .0309 .0115 .oooo .0020 .0469 4 andmore ,0004 ,0494 ,0200 .oooo ,0033 ,0727 Numberof Zero ,0003 .0124 ,0049 .oooo .001I .0193 HH members age One ,0001 ,0119 ,004I .oooo .0006 .o199 >60 Two andmore ,0011 ,0253 ,0096 .oooo .0025 .0383 Gender of hh Male ,0003 ,0136 ,0052 .oooo .0011 .0212 head Female ,0002 ,0131 ,0058 .oooo .0014 ,0205 Dependency Cb0.25 .0005 .0180 ,0073 .oooo ,0016 ,0275 ratio n of employed 0.254.5 .0001 ,0068 ,0020 .oooo ,0004 .0120 (EXCL unpaid family 0.5 andmore .oooo ,0012 ,0003 .oooo .oooo ,0021 workers)/HH size Source: DIE, Turkey 2002 HBS. 154 The Relative Poverty Risks According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted Relative Relative Relative Relative Relative Relative Risk Risk Risk Risk Risk Risk Turkey .oo .oo .oo .oo .00 .oo Lower than primary 3.28 1.56 1.69 5.36 1.54 1.28 Education o fHHhead Primary -.49 -.12 -.14 -1.00 -.08 -.06 Secondary1 vocational -1.00 -.76 -.79 -1.oo -.93 -.75 University -1.00 -.96 -.99 -1.00 -1a 0 -.95 Type o f settlement Rural .47 .38 .49 1.50 .30 .36 Urban -.3 1 -.25 -.33 -1.00 -.20 -.24 -.14 Access to land No .07 -.13 -.14 S O .I5 Yes -.14 .27 .28 -1.00 -.3 1 .28 Own automobile N o .30 .26 .28 .30 .28 .25 Yes -1.oo -.87 -.95 -1.00 -.94 -.84 Zero -.58 -.59 -.69 -1.00 -.85 -.49 One -.89 -.40 -.47 -1.00 -.56 -.32 Children under age 14 Two .61 -.12 -.38 -1.00 -.33 -.10 Three -.02 .95 1.45 7.72 .95 .81 4 and more 2.41 1.93 2.42 -1.00 3.55 1.52 Zero .04 -.lo -.12 .36 .01 -.08 Number o fHHmembersage One -.56 .07 -.07 -1.oo -.15 .I1 >60 Two and more .76 .63 1.07 -1.00 .22 .42 Gender o f hhhead Male -.05 -.01 .01 .07 -.01 .oo Female .76 .10 -.17 -1.oo .14 .03 Dependency ratio no f 0-0.25 .40 .23 .30 .59 .40 .20 employed (EXCL unpaid 0.25-0.5 -.61 -.29 -.43 -1.00 -.64 -.25 family workers)iHH size 0.5 and more -1.00 -.89 -.94 -1.00 -.92 -.75 Source DIE, Turkey 2002 HBS. 155 The Relative Poverty Risks According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted Relative Relative Relative Relative Relative Relative Risk Risk Risk Risk Risk Risk Index Index Index Index Index Index Turkey 1.oo 1.oo 1.oo 1.oo 1.oo 1.oo Lower than 2.69 6.36 2.54 2.28 Primary 4.28 2.56 .oo EducationofHHhead Primary .5 1 .88 .86 .92 .94 Secondary1 vocational .oo .24 .21 .oo .07 .25 University .oo .04 .01 .oo .oo .05 2.50 1.30 1.36 Type of settlement Rural 1.47 1.38 1.49 Urban .69 .75 .67 .oo .80 .76 1.50 1.15 Access to land No 1.07 .87 .86 .86 Yes .86 1.27 1.28 .oo .69 1.28 1.30 1.28 1.25 Own automobile No 1.30 1.26 1.28 Yes .oo .13 .05 .oo .06 .16 Zero .42 .41 .31 .oo .15 .5 1 One .11 .60 .53 .oo .44 .68 Childrenunder age 14 Two 1.61 .88 .62 .oo .67 .90 Three .98 1.95 2.45 8.72 1.95 1.81 4 andmore 3.41 2.93 3.42 .oo 4.55 2.52 Zero 1.04 .90 .88 1.36 1.01 .92 Number of HHmembers One .44 1.07 .93 .oo .85 1.11 age >60 Two and more 1.76 1.63 2.07 .oo 1.22 1.42 Male .95 .99 1.01 1.07 .99 1.oo Gender ofhhhead Female 1.76 1,lO .83 .oo 1.14 1.03 Dependencyratio nof C0.25 1.40 1.23 1.30 1.59 1.40 1.20 employed(EXCL unpaidfamily workers)/ 0.25-0.5 .39 .71 .57 .oo .36 .75 HHsize 0.5 and more .oo .ll .06 .oo .08 .25 Source: DIE, Turkey 2002 HBS. 156 Appendix 3: Sensitivity of Poverty to the MinimumFood Basket Caloric Requirements, CPI Source, Size Economies' and EquivalenceScales Parameters C P I DIE kciden.Poverty P1 Poverty Gap Index Poverty Shortfall Index 157 Alpha = 0.9 Theta =0.6 Kcal =2,100 PO Poverty Incident t-- :. I Poverty Index Food Adult equivalent 0.0011 0.0013 0.0008 0.0010 P2 Poverty Adjusted adult Poverty Line Severity equivalent OECD 0.0005 0.0006 0.0003 0.0004 Index Per capita 0.0468 0.0461 0.0423 0.0426 Complete Adult equivalent 0.0288 0.0287 0.0253 0.0259 Poverty Line . Adjustedadult equivalent OECD 0.0155 0.0155 0.0135 0.0139 Per capita 0.1894 0.1943 0.1775 0.1877 Food Adult equivalent 0.1870 0.2125 0.1939 0.2037 Poverty Poverty Line Adjusted adult Shortfall eauivalent OECD 0.2171 0.2367 0.2337 0.25 10 Index Per capita 0.3148 0.3 180 0.3101 0.3123 Complete Adult equivalent 0.2638 0.2664 0.2551 0.2638 Poverty Line Adjusted adult equivalent OECD 0.2430 0.2409 0.2406 0.2438 . 158 I Alpha = 1 Theta =0.6 Kcal=2450 CPI DIE CPI Survey ImputedRent Reported ImputedRent Reported by Regression Rent by HH by Regression Rentby HH Per capita 0.0714 0.0752 0.0596 0.0653 Food Adultequivalent 0.0306 0.0342 0.0252 0.0274 ~Povertyte n n ~ i d Line Poverty Adjusted adult equivalent OECD 0.0153 0.0147 0.0131 0.0143 1 Per capita 0.4324 0.4251 0.4 184 0.4083 Complete Adult equivalent Povertv 0.3906 0.3788 0.3652 0.3570 P1 Poverty Gap Index P2 Poverty Severity Index 159 Alpha = 1 Theta =0.6 Kcal =2100 1 t- CPI DIE CPI Survey ImputedRent Reported Imputed Rent Reported I by Regression Rent by HH byRegression Rent by HH Per capita 0.0438 0.0462 0.0379 0.0388 PO Poverty Incident Poverty Gap Index I Poverty Severity ~ n ~ e ~ Line equivalent OECD 0.0155 0.0158 0.0136 0.0139 I Per capita 0.1894 0.1943 0.1775 0.1877 Food Adult equivalent 0.1870 0.2080 0.1881 0.1961 'Over@ Poverty Line Adjusted adult Shortfall eauivalent OECD 0.2171 0.2367 0.2337 0.2510 Index Per capita 0.3 146 0.3 153 0.3097 0.3119 Complete Adult equivalent 0.2624 0.2658 0.2570 0.2613 Poverty Line Adjusted adult I equivalent OECD 0.2430 0.24 15 0.2415 0.2435 160 Appendix4: RegressionAnalysis DescriptiveModel Variables Enteredmemoved (b,c) Model Variables Entered Variables Removed Method Do you have in your dwelling-Closedgarage, DHEAT-4, DDWEL-5, DDWEL-4, Do you have in your dwelling-Electricity, secondary education of HH head, DDWEL-7, DDWEL-1, DDWEL-9, Do you have in your dwelling-Heatingfrom ground, South eastern Anatolia, HH head is female, DDWAGE-3, hh head is unemployed, DDWEL-8, Do you have in your dwelling-Satellite antenna, Do you have in your dwelling-Waste disposal, Eastern Anatolia, DDWEL-10, higher education of HH head, Central Anatolia, DDWAGE-1, University, master, 1 doctorate education of HH head, How many rooms are there in your dwelling, Enter DHEAT-3, Aegean, DDWAGE-2, DDWEL-3, Mediterranean, DDWAGE-5, Do you have inyour dwelling-Pipedwater system, childrenunder age 14 ,Do you have inyour dwelling-Kitchen,RuralHousehold,DHEAT-2, Black sea, Do youhave in your dwelling-Hot water, Do you have in your dwelling-Toilet (indoors), Do you have in your dwelling-Bathroom, Do you have in your dwelling-Natural gas, AE adjusted by modal hh (2ad + 2 ch), How many square meters is the utilized area (d), DDWEL-2, Do youhave inyour dwelling-Centralheating(a) 4. All requestedvariables entered. 3. DependentVariable: LNF. Z. Weightedleast squaresregression-weighted by populationweight. Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate 1 .681(a) .464 .464 57.89849 A Predictors: (Constant), Do you have in your dwelling-Closedgarage, DHEAT-4, DDWEL-5, DDWEL-4, Do you have in your dwelling-Electricity, secondary education of HH head, DDWEL-7, DDWEL-1, DDWEL-9, Do you have inyour dwelling-Heatingfrom ground, SoutheastemAnatolia, HHheadis female, DDWAGE-3, hhhead i s unemployed, DDWEL-8, Do you have in your dwelling-Satelliteantenna, Do you have in your dwelling-Waste disposal, EasternAnatolia, DDWEL-10, higher education of HHhead, Central Anatolia, DDWAGE-1, University, master, doctorate education of HH head , How many rooms are there in your dwelling, DHEAT-3, Aegean, DDWAGE-2, DDWEL-3, Mediterranean, DDWAGE-5, Do you have in your dwelling-Piped water system, children under age 14 ,Do you have in your dwelling-Kitchen, Rural Household, DHEAT-2, Black sea, Do you have in your dwelling-Hot water, Do you have in your dwelling-Toilet (indoors), Do you have in your dwelling- Bathroom, Do you have in your dwelling-Naturalgas, AE adjusted by modal hh (2ad + 2 ch), How many square meters is the utilizedarea (d), DDWEL-2, Doyouhave inyour dwelling-Centralheating 161 ANOVA(b,c) Model Sumof Squares Df Mean Square F Regression 198,585,252,725,405 44 4,5 13,301,198,305 1,346,355,937 1 Residual 229,269,350,347,546 68,392,987 3,352,235 Total 427,854,603,072,951 68,393,03 1 a. Predictors:(Constant), Do youhave in your dwelling-Closedgarage, DHEAT-4, DDWEL-5, DDWEL-4, Do you have in your dwelling-Electricity, secondary education of HH head, DDWEL-7, DDWEL-1, DDWEL-9, Do you have in your dwelling-Heating from ground, South eastern Anatolia, HHhead is female, DDWAGE-3, hh head is unemployed, DDWEL-8, Do you have in your dwelling-Satellite antenna, Do you have in your dwelling-Waste disposal, Eastern Anatolia, DDWEL-10, higher education of HH head, Central Anatolia, DDWAGE-1, University, master, doctorate education of HH head , How many rooms are there in your dwelling, DHEAT-3, Aegean, DDWAGE-2, DDWEL-3, Mediterranean, DDWAGE-5, Do you haveinyour dwelling-Piped water system,children under age 14 ,Do you have inyour dwelling-Kitchen, Rural Household, DHEAT-2, Black sea, Do you have in your dwelling-Hot water, Do you have inyour dwelling-Toilet (indoors), Do you have inyour dwelling-Bathroom,Do you have in your dwelling-Natural gas, AE adjusted by modalhh (2ad + 2 ch), How many square meters i s the utilized area (m2),DDWEL-2, Doyou have inyour dwelling-Centralheating. b. Dependentvariable: LNF. c. Weightedleast squares regression-weighted by populationweight. Coefficients(a,b) Unstandardized Standardize 95% Confidence Coefficients d Coefficients t Sig. Intervalfor B Model B Std. Beta Lower Upper Error Bound Bound 1 (Constant) 14.444 .002 9,064.796 .ooo 14.440 14.448 RuralHousehold -5.424E-02 .ooo -.045 -362.625 .ooo -.055 -.054 Aegean .145 .ooo .084 747.964 .ooo .144 .145 Mediterranean -.323 .ooo -.200 -1,553.420 .ooo -.323 -.322 CentralAnatolia -2.221E-03 .ooo -.001 -12.147 .ooo -.003 -.002 Black sea -.135 .ooo -.099 -846.849 .ooo -.135 -.134 EasternAnatolia -5.622E-02 .ooo -.026 -227.808 .ooo -.057 -.056 Southeastern Anatolia 2.584E-02 .ooo .011 90.725 .ooo .025 ,027 Childrenunder age 14 -5.752E-02 .ooo -.181 -1,322.171 .ooo - . O S -.057 Secondaryeducationof HH head .161 .ooo ,082 886.785 .ooo .160 .161 Higher educationofHHhead ,264 ,000 .136 1,391.592 .ooo .263 .264 University, master, doctorate educationofHHhead .399 .ooo .I30 1,322.195 .ooo ,399 .400 HHheadis female -1.487E-02 .ooo -.006 -61.519 .ooo -.015 -.014 162 Coefficients (a,b) Unstandardized Standardize 95% Confidence Coefficients d Coefficients t Sig. Interval for B Lower lode1 B Std. Beta Upper Error Bound Bound IHheadisunemployed -.143 .ooo -.073 -757.275 .ooo -.144 -.143 9E adjustedby modalHH -178.748 .ooo -.011 -.010 2ad 2 ch) + 1.059E-02 .ooo -.027 3DWEL-1 .359 .002 .021 233.354 .ooo .355 .362 3DWEL-2 8.491E-02 .ooo -.073 -444.759 ,000 -.085 -.084 DDWEL-3 -.176 .ooo -.098 -753.416 .ooo -.176 -.175 3DWEL-4 8.633E-02 ,001 -.009 -104.229 .ooo -.088 -.084 DDWEL-5 5.205E-02 .ooo .014 152.366 ,000 .051 .053 DDWEL-7 .164 .001 .028 300.844 .ooo .163 ,165 DDWEL-8 .140 .oo1 .011 118.521 .ooo .137 .143 DDWEL-9 -.190 .ooo -.049 -499.557 .ooo -.191 -.189 DDWEL- 10 -.35 1 .oo1 -.063 -640.854 ,000 -.352 -.349 DDWAGE- 1 1.053E-02 .ooo .005 45.498 .ooo .010 ,011 DDWAGE-2 1.670E-02 .ooo -.010 -91.800 .ooo -.017 -.016 DDWAGE-3 7.986E-02 .ooo .068 643.960 .ooo .080 .080 DDWAGE-5 9.034E-02 .ooo .051 497.510 .ooo .090 .091 DHEAT-2 .254 .001 .lo8 244.697 .ooo ,251 ,256 DHEAT-3 .260 .001 .073 242.991 .ooo .258 .263 DHE AT-4 .459 .001 .039 399.724 .ooo .456 .462 How many rooms are there in .ooo 360.789 .ooo .027 .027 your dwelling? 2.697E-02 .054 How many square meters is the 717.584 ,000 .002 .002 utilized area (m2)? 1.737E-03 .ooo .lo7 Do you have inyour dwelling- .ooo -.010 -78.179 ,000 -.016 -.015 Bathroom? 1.557E-02 Do you have inyour dwelling- 1,003.096 .ooo .147 .147 Toilet (indoors)? ,145 ,000 .125 Do you have inyour dwelling- .ooo -.041 -348.0 14 .ooo -.073 -.072 Kitchen? .7.218E-02 Doyou have inyour dwelling- .oo1 109.885 .ooo .143 .149 Waste disposal? .14t .010 Do you have inyour dwelling- -79.875 .ooo -.OS4 -.079 Centralheating? 3.139E-02 .001 -.04 1 Do you have inyour dwelling- 298.03 1 .ooo .347 .353 Heatingfrom ground? .350 .001 .027 163 Coefficients (a,b) Unstandardized Standardize 95% Confidence Coefficients d Coefficients t Sig. Interval for B lode1 B Std. Beta Lower Upper Error Bound Bound Do you have inyour dwelling- Pipedwater system? j.122E-02 .ooo .035 312.081 .ooo .061 .062 Do you have inyour dwelling- Electricity? .326 .002 ,019 207.743 .ooo .322 ,330 Doyou have inyour dwelling- Natural gas? .I25 .ooo ,039 306.671 ,000 .I24 .I26 Do you have inyour dwelling- Hot water? ,126 .ooo .I13 936.297 ,000 .125 .I26 Do you have inyour dwelling- Satellite antenna? .126 .ooo .035 346.933 .ooo .125 .I27 Do you have inyour dwelling- Closedgarage? .287 ,000 .091 977.939 .ooo .287 .288 a. Dependent Variable: LNF. b. Weighted least squares regression-weighted by population weight. 164 Appendix 5: FoodBasket 165 Item Name 166 IItemCode I ItemName ICaloriesin1 Price I Q2450 IC2450IVAL2450 I 167 168 0115201010 Margarine (packaged) 719 2,209,605 0.0099 71.1736 21,873 0115302010 Sunflower oil 884 1,812,505 0.0393 347.1712 71,182 0115303010 Maize oil 884 2,072,678 0.0058 51.2739 12,022 0116101020 Orange 47 592,783 0.0496 23.2889 29,373 0116102010 Grape (having seeds) 67 952,293 0.0062 4.1285 5,868 169 IItemCode I ItemName /Caloriesin 100 grlPriceI 43100 IC3100 [VAL31001 170 Appendix 6: Estimationof Average HouseholdAdult EquivalentSize, Dependingonthe Formula and Parameters The Average HouseholdAdult Equivalent Size Alpha 0.8 0.8 0.8 0.8 0.8 Theta 0.6 0.7 0.8 0.9 1 Average adult equivalent size ofhousehold 2.22 2.55 2.94 3.39 3.92 Average adjusted adult equivalent size of household 4.11 4.16 4.22 4.28 4.36 Alpha 0.7 0.7 0.7 0.7 0.7 Theta 0.6 0.7 0.8 0.9 1 Average adult equivalent size of household 2.18 2.50 2.87 3.31 3.81 Average adjusted adult equivalent size o f household 4.19 4.25 4.32 4.40 4.49 Alpha 0.6 0.6 0.6 0.6 0.6 Theta 0.6 0.7 0.8 0.9 1 Average adult equivalent size o f household 2.14 2.45 2.81 3.22 3.70 Average adjusted adult equivalent size of household 4.27 4.34 4.43 4.52 4.63 171 RelationshipbetweenHHSize, Composition,andAdult EquivalentSize for P?ha=0.9 and Theta = 0.6 Adult Adult Equivalent Adult Adult Equivalent Adjusted by Equivalent Equivalent Size Modalhh Size OECD Size FA0 (2ad +2 ch) HHSize Adult Child 1 1 0 1.oo 1.80 1.oo .85 2 1 1 1.47 2.64 1s o 1.50 2 0 1.52 2.72 1.75 1.47 1 2 1.85 3.33 2.00 1.97 3 2 1 1.89 3.40 2.25 2.05 3 0 1.93 3.47 2.50 2.03 1 3 2.19 3.94 2.50 2.55 4 2 2 2,23 4.00 2.75 2.59 3 1 2.26 4.06 3.00 2.57 4 0 2.30 4.13 3.25 2.56 1 4 2.50 4.49 3.00 3.14 2 3 2.53 4.54 3.25 3.06 5 3 2 2.56 4.60 3.50 3.07 4 1 2.59 4.66 3.75 3.04 5 0 2.63 4.72 4.00 3.03 1 5 2.78 4.99 3.50 3.50 2 4 2.81 5.05 3.75 3.52 6 3 3 2.84 5.10 4.00 3.53 4 2 2.87 5.16 4.25 3.47 5 1 2.90 5.21 4.50 3.46 6 0 2.93 5.26 4.75 3.49 2 5 3.07 5.52 4.25 3.95 3 4 3.10 5.57 4.50 3.98 7 4 3 3.13 5.62 4.75 3.92 5 2 3.16 5.67 5.00 3.87 6 1 3.19 5.72 5.25 3.91 7 0 3.21 5.77 5.50 3.93 172 2 6 3.32 5.97 4.75 4.23 3 5 3.35 6.01 5.00 4.35 4 4 3.38 6.06 5.25 4.42 8 5 3 3.40 6.11 5.50 4.32 6 2 3.43 6.16 5.75 4.36 7 1 3.46 6.21 6.00 4.28 8 0 3.48 6.25 6.25 4.32 2 7 3.56 6.39 5.25 4.73 3 6 3.59 6.44 5.50 4.63 4 5 3.61 6.48 5.75 4.69 5 4 3.64 6.53 6.00 4.75 9 6 3 3.66 6.58 6.25 4.75 7 2 3.69 6.62 6.50 4.75 8 1 3.71 6.67 6.75 4.73 9 0 3.74 6.71 7.00 4.74 2 8 3.79 6.80 5.75 5.13 3 7 3.81 6.84 6.00 5.12 4 6 3.84 6.89 6.25 5.15 5 5 3.86 6.93 6.50 5.16 10 6 4 3.88 6.98 6.75 5.14 7 3 3.91 7.02 7.00 5.09 8 2 3.93 7.06 7.25 5.03 9 1 3.96 7.11 7.50 5.18 10 0 3.98 7.15 7.75 5.30 2 9 4.00 7.19 6.25 5.54 3 8 4.03 7.23 6.50 5.31 4 7 4.05 7.28 6.75 5.54 5 6 4.08 7.32 7.00 5.40 11 6 5 4.10 7.36 7.25 5.41 7 4 4.12 7.40 7.50 5.40 8 3 4.15 7.44 7.75 5.43 9 2 4.17 7.49 8.00 5.37 10 1 4.19 7.53 8.25 5.62 173 3 9 4.24 7.61 7.00 5.67 4 8 4.26 7.65 7.25 5.86 6 6 4.31 7.73 7.75 5.75 12 7 5 4.33 7.77 8.00 5.93 8 4 4.35 7.81 8.25 5.85 9 3 4.37 7.85 8.50 5.83 I O 2 4.40 7.89 8.75 5.81 6 7 4.5 1 8.09 8.25 6.43 7 6 4.53 8.13 8.50 6.23 13 8 5 4.55 8.17 8.75 6.05 9 4 4.57 8.21 9.00 6.17 10 3 4.59 8.25 9.25 6.56 7 7 4.72 8.48 9.00 6.48 8 6 4.75 8.52 9.25 6.56 9.50 14 9 5 4.77 8.56 6.74 I O 4 4.79 8.60 9.75 6.56 11 3 4.81 8.63 10.00 6.57 12 2 4.83 8.67 10.25 6.50 4 11 4.85 8.71 8.75 7.30 7 8 4.91 8.82 9.50 6.94 15 8 7 4.93 8.86 9.75 6.74 9 6 4.95 8.90 10.00 7.02 11 4 5.00 8.97 10.50 6.85 6 I O 5.08 9.12 9.75 7.33 16 7 9 5.10 9.15 10.00 7.13 9 7 5.I 4 9.23 10.50 7.16 7 10 5.28 9.48 10.50 7.58 11.00 17 9 8 5.32 9.55 7.83 I O 7 5.34 9.58 11.25 7.62 14 3 5.42 9.72 12.25 7.83 8 1 1 5.65 10.14 11.75 8.28 19 9 10 5.66 10.17 12.00 8.12 10 9 5.68 10.20 12.25 8.07 9 11 5.83 10.47 12.50 8.64 12 5.89 10.57 13.25 8.59 20 8 13 7 5.91 10.61 13.50 8.99 14 6 5.92 10.64 13.75 8.87 174 Appendix 7: GeneralDescriptiveStatistics PopulationStructureby Sex Turkey Type of Settlement Rural Urban Gender Male 48.7% 48.5% 48.7% Female 51.3% 51.5% 51.3% Turkey 100.0% 100.0% 100.0% Source: DIE, Turkey2002 HBS. PopulationStructure by Age Type of Settlement Rural Urban rurkey 0-5 8.9% 10.8% 10.0% 6-14 17.9% 18.8% 18.4% 15-19 10.9% 10.5% 10.7% 20-24 7.2% 8.8% 8.2% 25-29 6.6% 8.1% 7.5% 30-34 6.3% 7.4% 7.0% Age Group 35-39 7.0% 7.7% 7.4% 4 w 4 6.6% 6.7% 6.7% 45-49 5.8% 5.9% 5.9% 50-54 4.9% 4.9% 4.9% 55-59 3.8% 3.1% 3.4% 60-64 3.7% 2.5% 2.9% 65+ 10.5% 4.8% 7.0% Turkey 100.0% 100.0% 100.0% Source: DIE, Turkey 2002 HBS. 175 Population Structure by Age and Sex Gender Turkey Male Female YO Y O Y O YO 0-5 52.2% 47.8% 10.0% 100.0% 6-14 50.1% 49.9% 18.4% 100.0% 15-19 48.6% 51.4% 10.7% 100.0% 20-24 42.0% 58.0% 8.2% 100.0% 25-29 46.4% 53.6% 7.5% 100.0% 30-34 47.4% 52.6% 7.0% 100.0% Age Group 35-39 49.7% 50.3% 7.4% 100.0% 4 w 4 49.6% 50.4% 6.7% 100.0% 45-49 50.6% 49.4% 5.9% 100.0% 50-54 50.7% 49.3% 4.9% 100.0% 55-59 46,8% 53.2% 3.4% 100.0% 60-64 49.0% 51.0% 2.9% 100.0% 65+ 46.9% 53.1% 7.0% 100.0% Turkey 48.7% 51.3% 100.0% 100.0% Source DIE, Turkey 2002 HBS Population (age 6 and over) Structure by Education by the Type of Settlement Type of Settlement Turk 1 Rural Urban Count Count Y O Y O Count Y O Educational Illiterate 3,798,846 15.3% 3,910,208 10.7% 7,709,053 12.5% Institutions Graduated Literate without diploma 5,491,283 22.0% 7,721,572 21.1% 3,212,854 21.5% From Primary school (5 years) 10,666,730 42.8% 12,291,816 33.6% '2,958,546 37.3% Primary education(8 years) 1,295,683 5.2% 1,920,310 5.2% 3,215,992 5.2% Junior high school (8=5+3) 1,2 14,386 4.9% 2,791,097 7.6% 4,005,484 6.5% Vocational school at Jr. high (8-9 = 5+ 3-4) 59,108 .2% 115,567 .3% 174,674 .3% High school 11-12= 8-9 + 3-4 1,526,118 6.1% 4,193,597 11.5% 5,7 19,715 9.3% Vocational school at high 464,725 1.9% 1,470,172 4.0% 1,934,897 3.1% 176 Population(age 6 and over) StructurebyEducationby the Type of Settlement Typeof Settlement Turkey Rural Urban Count YO Count Count O/O Y O 2 year higher education (univ.) 183,938 .7% 676,758 1.8% 860,696 1.4% 4 year higher education (univ.) 200,058 .8% 1,404,934 3.8% 1,604,993 2.6% Master and doctorate 8,012 .O% 121,175 .3% 129,186 .2% rurkey 14,908,885 100.0% 36,617,206 100.0% 61,526,091 100.0% Yource: DIE, Turkey 2002 HBS. ~ Population(age 14-65) EmploymentStatusby Type of Settlement Type of Settlement Turkey Rural Urban Col Yo Col Yo Col Yo Employed insome way 57.7% 37.2% 45.5% Unemployed 1.6% 3.4% 2.7% Economically inactive 40.8% 59.4% 51.8% Turkey 100.0% 100.0% 100.0% Source: Turkey 2002 HBS. Population(age 12+) EmploymentStatus by Sex Gender Turkey Male Female ColO/O Col Yo Col Yo Employed insome way 64.2% 28.4% 45.5% Unemployed 4.0% 1.4% 2.7% Economically inactive 31.8% 70.2% 51.8% rurkey 100.0% 100.0% 100.0% iource: DIE, Turkey 2002 HBS. 177 Population(age 12 +) Employment Status by Education Education Turkey Lowerthan Secondary Primary Primary Vocational University Col Yo Col Yo Col Yo C O l % Col% Employedin someway 27.5% 49.9% 49.3% 70.2% 45.5% Unemployed .4% 2.3% 6.3% 6.4% 2.7% Economicallyinactive 72.1% 47.8% 44.4% 23.4% 51.8% Turkey 100.0% 100.0% 100.0% 100.0% 100.0% Source: DIE, Turkey 2002 HBS. Population(age 14-65) Employment by Type of Settlement Type of Settlement Turkey Rural Urban Col Yo Col Yo Col Yo Regular employee 17.0% 60.5% 38.0% Casual employee 5.8% 9.2% 7.4% Employment Status in Apprentice .2% .3% .2% Workplace Employer 2.0% 5.8% 3.8% Self-employed 36.5% 16.6% 26.9% Unpaidfamily worker 38.6% 7.6% 23.6% Turkey 100.0% 100.0% 100.0% Source. Turkey 2002 HBS. Population(age 12+) Emp iment by Sex Gender Turkey Male Female Col Yo Col Yo Col Yo Regularemployee 45.8% 21.9% 38.0% Casual employee 7.8% 6.7% 7.4% EmploymentStatus in Apprentice .3% .1% .2% Workplace Employer 5.5% .4% 3.8% Self-employed 31.4% 17.4% 26.9% Unpaidfamily worker 9.2% 53.5% 23.6% Turkey 100.0% 100.0% 100.0% Source: DIE, Turkey 2002 HBS. 178 Population (age 12 +) Employment by Education Education rurkey Lower than Secondary Primary Primary Vocational Jniversity Col Yo Col Yo Col Yo Col Yo Col Yo Regularemployee 6.8% 32.2% 67.7% 81.8% 38.0% Casualemployee 8.9% 9.1% 2.6% .6% 7.4% status in Apprentice .3% .3% .l% .2% workplace Employer .?yo 3.8% 5.o% 7.3% 3.8% Self-employed 40.0% 30.3% 11.5% 6.5% 26.9% Unpaidfamily worker 43.3% 24.2% 13.0% 3.8% 23.6% Turkey 100.0% 100.0% 100.0% 100.0% 100.0% Source: DIE, Turkey 2002 HBS. HHHead'sGender by the Type ofSettlement Type of Settlement Turkey Rural Urban Col Yo ROW Yo ColYo ROW% Col Yo &ow Yo Gender Male 92.0% 39.5% 88.8% 60.5% 90.0% 100.0% Female 8.0% 31.1% 11.2% 68.9% 10.0% 100.0% Turkey 100.0% 38.6% 100.0% 61.4% 100.0% 100.0% Source: DIE, Turkey 2002 HBS. 179 HHHead'sEducationby the Type ofSettlement Type of Settlement Turkey Rural Urban Col Yo ROWYo Col % ROWYo Col Yo ROWYo Illiterate 10.5% 49.7% 6.7% 50.3% 8.2% 100.0% Literate without diploma 9.3% 53.2% 5.2% 46.8% 6.8% 100.0% Primary school (5 years) 61.O% 44.3% 48.2% 55.7% 53.1% 100.0% Primary education(8 years) .l% 48.0% .l% 52.0% .l% 100.0% Educational Junior highschool (8=5+3) 7.4% 30.9% 10.3% 69.1% 9.2% 100.0% Institutions Vocational school at Jr. high Graduated (8-9 =5+ 3-4) .l% 13.7% .5% 86.3% .4% 100.0% From High school 11-12= 8-9 +3-4 5.6% 22.9% 11.8% 77.1% 9.4% 100.0% Vocational school at high 3.3% 26.6% 5.8% 73.4% 4.8% 100.0% 2-year higher education (univ.) 1.O% 18.3% 2.8% 81.7% 2.1% 100.0% 4-year higher education(univ.) 1.6% 11.7% 7.7% 88.3% 5.4% 100.0% Master and doctorate .l% 6.0% .9% 94.0% .6% 100.0% Turkey 100.0% 38.6% 100.0% 61.4% 100.0% 100.0% Source: DIE, Turkey 2002 HBS. Dwelling Type by the Type of Settlement Type of Settlement Turkey Rural Urban YO Y O Y O Luxurious house .2% .l% .2% Separated 74.1% 22.9% 42.7% Semi detached house 12.2% 12.2% 12.2% Basement .O% 1.4% 3 % What is the type of your Ground floor 1.9% 6.1% 4.5% dwelling? Regular floor 9.0% 51.3% 35.0% Attic .6% 1.5% 1.1% Double floor .l% .6% .4% Shanty 1.O% 3.7% 2.7% Other (Indicate) 3 % .l% .4% Turkey 100.0% 100.0% 100.0% Source. Turkey2002 HBS. 180 Housing Rent Type o f Settlement Turkey Rural Urban Mean 60,287,195 100,553,2 13 94,554,897 Actual Rents Median 50,000,000 80,000,000 75,000,000 Source: Turkey 2002 HBS. Number of Rooms (frequency distribution ) Type of Settlement Turkey Rural Urban eo1Yo Col Yo eo1Yo 1 .92 .71 .79 I 2 11.80 9.86 10.61 38.90 43.48 41.71 35.84 40.79 38.88 H o w many rooms are there inyour 8.26 4.19 5.76 dwelling? 2.63 .58 1.37 7 1.07 .16 .5 1 8 .so .18 .30 9 .04 .03 10 .07 .01 .03 Turkey 100.00 100.00 100.00 Source Turkey2002 HBS. Number of Rooms Type of Settlement Rural Urban Turkey H o w many rooms are there inyour Mean 3.5 3.4 3.5 dwelling? Median 3.0 3.0 3.0 Source Turkey 2002 HBS. 181 PersonlRoomRatio Statistics Type of Settlement Rural Urban Turkey Ratio numberof members/ number of rooms 1.301 1.277 1.286 Source: Turkey 2002 HBS. Amenities Type of Settlement hrkey Rural Urban Do you have inyour dwelling: Bathroom? .836 .966 .916 Do you have inyour dwelling: Toilet (indoors)? .621 .943 .819 Do you have inyour dwelling:Kitchen? .883 .980 .942 Doyouhave inyour dwelling: Waste disposal? .001 .004 .003 Doyouhave inyour dwelling: Central heating? .OS4 .235 .165 Do youhave inyour dwelling: Heatingfrom ground? .ooo .006 .004 Do you have inyour dwelling: Pipedwater system? 3 8 8 .994 ,953 Do youhave inyour dwelling: Electricity? .999 .999 .999 Do youhave inyour dwelling:Naturalgas? ,002 .149 .092 Do you have inyour dwelling: Hot water? SO9 .781 .676 Do you have inyour dwelling: Satellite antenna? .002 .093 .058 Do you have inyour dwelling: Closedgarage? .033 ,036 ,035 Source: Turkey 2002 HBS. 182 DurableGoods: AverageNumber PerOneHousehold rype of Settlement Rural Urban rurkey How many telephones do you have? ,869 ,900 .888 How many cellular phones do you have? ,446 ,781 .652 How many computers do you have? .027 .lo9 .077 How many emets do you have? .008 .043 .030 How many televisions do you have? 1.009 1.207 1.130 How many videos do you have? ,044 .096 .076 How many DVDs or VCRs do youhave? .066 .151 .118 How many video cameras do you have? .009 ,025 .019 How many satellite antennas do you have? .296 .160 .212 How many HI-FIsystems do you have? ,232 ,434 .356 How many CD players do you have? .033 .092 .070 How many refrigerators do you have? .954 .981 .971 How many deep freezers do you have? .062 .061 .061 How many dishwashers do youhave? .073 .281 .201 How many gas stoves with oven do you have? .480 .629 ,572 How many electric ovens do you have? .379 .437 .415 How many microwave ovens do you have? .017 .053 .039 How many automatic washing machines do you have? .595 ,842 .746 How many dryers do you have? .006 ,011 .009 How many presses(irons) do you have? .005 ,026 ,018 How many vacuum cleaners do you have? .592 ,815 .729 How many carpet washing machines do you have? .051 .161 .119 How many air-conditioners do you have? .008 ,044 .030 How many water heaters do you have? .391 .625 .535 How many Jacuzzis do you have? .003 .005 .004 How many aspirators do you have? .143 .318 .250 Source: Turkey 2002 HBS. 183 Vehicles :Average Number Per One Household rype of Settlement Rural Urban Turkey Numberofautomobiles-owned .182 .275 .239 Numberofautomobile-provided by employer .004 .015 ,011 NumberofJeeps .ooo .001 .001 Number of minibus-caravans .016 .013 ,014 Number ofmotorcycles .051 .038 .043 Number of motorized-motorized sea vehicle .001 .ooo .oo1 NumberofYachts .ooo .ooo .ooo Source: Turkey 2002 HBS. 184 ANNEX 11: POVERTYINTURKEY: A LITERATUREREVIEW This literature review summarizes the methodology andresults o f studies that focus on poverty inTurkey. Most o f these studies use 1987 and 1994 Household Budget Survey (HBS) data. One o f them is a sociological study that examines poverty in Turkey on the basis o f qualitative data that come from 160 interviews conducted with the poor. There are various methods that can be applied to assess poverty, and the resulting poverty measures are extremely sensitive to the type o f method used. Therefore, a briefreview o fthe methodology used should be an essential part o f any poverty assessment report. In this review, it i s found that one problem with some Turkishpoverty studies i s that they fail to give sufficient information about their methodology, and that there i s a lack o f a unified framework. This makes comparison o f results across studies difficult. Therefore, in this review, there i s some emphasis on the comparison o f the methodology and poverty measures used in the studies. In general, if there is information on the method used in the study, this i s reviewed indetail herein. StudiesBasedon 1994HBS Alici (2000) assesses poverty and examines its determinants using the 1994 Household Budget Survey (HBS) conducted by DIE. An absolute poverty line based on a minimumrequired level o f calories, and a relative poverty line, are used in this study to measure poverty. In the construction o f the absolute poverty line, actual consumption habits are taken into account. The approach to setting the poverty lines in this study distinguishes it from other related studies. The minimum required level o f calories i s set by first taking the actual consumption o f households in the bottom quintile of total consumption expenditures and subsequently computing the total calorie content o f the food items consumed for each household in this group. The total calories for each household are then transformed into per equivalent adult units. The average calorie intake per equivalent adult inthis reference group i s taken as the minimumrequired level o f calories. The resulting minimumrequired level o f calories i s not reported in the study. In total, 164 different minimum food baskets-that differ in their composition but not in their total calorie content- are obtained for the rural areas and 200 for the urban areas. Households with consumption expenditures per equivalent adult lower than the imputed cost o f the minimum food basket are considered extremely poor (Food Poverty). The Complete Poverty Line (CPL)--which takes into account consumption o f non- food goods and services-is constructed using the share o f non-food consumption among individuals whose total consumption i sjust above the value o f the food poverty line. It i s found that the cost o f the basket that also includes the non-food components i s 1.78 times the cost o f the minimumfood basket in urban areas, and 1.5 times inrural areas. Households with consumption expenditures per equivalent adult lower than the CPL are considered poor (Cost o f Basic Needs Approach). The relative line i s set at one- half o f the monthly median income per equivalent adult. According to this approach, households with monthly income less than the corresponding relative line are considered poor. Inthis study, household sizes are adjusted usingan appropriate economies-of-scale parameter for all the calculations. The results for different values o f this parameter are reported separately. Table A.II.l presents results usingdifferent poverty measures and different economies-of-scale parameters. 185 Table A.II.l. DifferentPoverty MeasuresandResults Consumption-Based I I Yo Measures Poverty) e = 0.75 Basic Needs, CPL (Food +Non-Food) e = 0.75 OECDscale I33 economiesof scale 8 = 0.90 ~ ~ Income-Based 1Measures BasicNeeds* IMediumeconomiesofscale 1 5.09 1 e =0.75 OECDmeasure** Higheconomiesof scale 8 = 15.69 (Relative Poverty) 0.5 ** Comparisonof incomefor equivalentadultwith one-halfofthe monthlymedianincome. Note: Basedon 1994 HBS. Source: Alici (2000). The main focus o f the study i s the poverty profile in Turkey. Table A.II.2 presents the poverty rates by region inTurkey. East Anatolia is the region facing highest poverty risk (11.25 percent). Table A.II.3 presents the poverty rates in urban and rural areas. Urban and rural areas do not differ significantly in terms o f poverty risk. This result is different from the general findings in other studies that conclude that rural poverty rates are higher than urban poverty rates. Regions Share of Basic Extreme Total Needs poverty Population (CPL)(YO) (Food ("A) Poverty) ( Y O ) Marmara 24.7 3.86 0.36 Aegean 13.6 1.88 0.14 Mediterranean 12.8 3.78 0.30 Central 17.2 6.38 1.14 Rlack Sea 13 5 2 74 O 16 Southeast 9.6 11.05 1 1.36 TOTAL 100.00 I1 5.20 I 0.69 186 Table A.II.3. Urban andRuralPovertyRates Shareof Basic Extreme Total Needs Poverty Regions Population (CPL) (Food ("/) ("/.I Poverty) ( Y O ) Rural 46.46 5.21 0.80 Urban 53.54 5.20 0.59 While Alici (2000) i s a study that mainly focuses on poverty across various dimensions in Turkey in general, Erdogan (2000) is a study that puts more emphasis on the structure of poverty in Turkey. The Erdogan (2000) study consists o f an introduction to poverty and alternative choices o f setting poverty lines, reviews o f relevant poverty literature inTurkey, and poverty results based on the 1994 HBS. The main feature of the study i s its emphasis on the detailed examinationo f the structure o fpoverty inTurkey. Erdogan's (2000) choice o f poverty measures is similar to Alici (2000). A FoodPoverty Line based on a minimumrequired level of calories, and the Complete Poverty Linebased on consumption of both food and non-food goods and services, are constructed. The results are reported for botho f the measuresused. The important distinction between the poverty measures used by Alia (2000) and Erdogan (2000) is the difference inthe methods they use insetting the minimum required level o f calories. Alici (2000) sets the minimumrequirement of calories usingthe actual survey data,39whereas Erdogan calculates the calorie requirements for a four-person family accordingto Table A.II.4. Table A.II.4. Calorie Requirementsby Age and Gender Source: Erdogan(2000). Inthe Erdogan (2000) study, the FoodPoverty Line is set as the estimated local cost ofthe minimumfood basket that meets the food energy requirements, and households with monthly monetary incomes below the cost of the minimum food basket are considered poor. As the Complete Poverty Line (Cost of Basic Needs Approach), cost of a consumption bundle deemed to be adequate for basic needs (which include food and non-food components) is estimated, and households with monthly monetary incomes below the cost of this bundle are considered poor. The second approach takes into account housing, clothing, transportation, and furniture expenditures, as well as food expenditures. Erdogan takes into account the differences inneeds inbothapproaches, according to location of residence and household size. 39For more details onAlici's (2000) method, see previousexplanation. 187 According to the first method, 8.37 percent of the individuals and 5.66 percent o f the households are poor. Ingeneral, the higherpoverty rates of individualsrelative to households can be explained bythe fact that households with more members face more poverty compared to households with fewer members. In the rural areas, 11.82 percent of individuals are poor, and inthe urban areas 4.60 percent are poor. There i s a significant difference between urban and rural areas in terms of poverty rates, and this difference holds after disaggregating according to certain demographic and socioeconomic characteristics. The highest percentage o f poor is in Southeast Anatolia rural areas. Also, household size is a significant determinant factor inpoverty. As the household size gets bigger, poverty risk increases. InTurkey, 32.19 percent of the households with 13 members are poor, while 27.32 percent of households with 14 members, and 24.89 percent for households with 15 or more members, are poor. According to the secondmethod, 24.30 percent o f individuals and 19.31percent of households are poor. Although the cost of basic needs inrural areas i s lower than in urban areas, the poverty rates in the rural areas tend to be higher. The highest percentageofpoor is again inthe SoutheastAnatolia rural areas. The Erdogan (2000) study places strong emphasis on the structure of poverty. Dimensions and characteristics o f poverty are examined in detail. Distribution of age, gender, education, marital status, employment status, and economic activities among the poor inTurkey are important to understand who is poor inthe country, and this study thoroughly analyzes this issue. According to the first method, among the poor inTurkey, 51.49 percent are women and 48.5 1percent are men. Some 72.67 percent o f the poor reside in rural places. Individuals aged 15 to 64 and 0 to 14 are characterized by high poverty rates compared to individuals age 65+. According to the second method, 52.78 percent o f those aged 15 to 64 are poor, 42.26 percent of those aged 0 to 14 are poor, and 4.96 percent of those 65+ are poor. Tables A.II.5 and A.II.6 presentErdogan's findings regarding the structure of poverty by age and gender. Table A.II.5. Structure of Poverty by Age and Gender (Food Poverty) 65+ 2.96I 2.94 I 2.98 188 Table A.II.6. Structureof Povertyby Age and Gender (Basic Needs) Note: Basedon 1994 HBS. Source: Erdogan(2000). InTurkey, thepopulationaged 12andaboveconstitutes76percent ofthe totalpopulation. Tables A.II.7 and A.II.8 present the poverty of this group by employment status. Analysis o f the distribution o f the poor according to employment status reveals that according to the first method, among the poor in Turkey, 53.67 percent are employed and 46.33 are unemployed (or out o f the labor force). According to the second method, among the poor, 46.71 percent are employed and 53.29 percent are unemployed. There is greater disparity in the distribution o f employment status among the poor when rural and urban areas are examined separately. In the urban areas, 34.02 percent o f the poor are employed, and in the rural areas 64.66 percent of the poor are employed. Within the group of individuals who are not working, housewives, students, and the elderly face higher poverty risk relative to other groups. Table A.II.7. Structureof Povertyby EmploymentStatus(Food Poverty) Sick 2.05 2.70 1.10 Retired 1.08 0.55 1.86 Elderly 9.54 12.09 5.83 Other 0.43 0.57 0.22 Working 53.67 62.49 29.69 Casual worker 16.56 11.50 45.52 24.67 25.57 19.50 Family worker w l 49.86 57.50 6.15 Source; Erdogan(2000). 189 Table A.II.8. Structure of Povertyby Employment Status (Basic Needs) Employment YOAmong YOAmong YOAmong Status Poor Rural Poor Urban Poor Total 100.00 62.89 37.11 Out ofwork 46.71 35.34 65.98 Unemployed 14.95 17.11 13.00 Student 22.56 19.49 25.33 Hoiisewife 46 99 42 69 50.89 Handicapped 1.36 1.48 1.07 Sick 2.18 3.24 1.23 Retired 2.93 2.62 3.21 I nopay Source: Erdogan (2000). Within the group of poor individuals who are economically active, a majority (73.2 percent) i s in the agnculture and forestry sector, according to the first method. According to the second method, 65.60 percent of the poor are in the agriculture and forestry sectors, 8.67 percent are in the manufacturing sector, and 8.32 percent are in construction. In rural areas, the poor are mainly in the agriculture and forestry sectors. Tables A.II.9 and AX.10 summarize the distribution according to economic activities among the poor. Table A.II.9. Structure of Povertyby EconomicActivity (Food Poverty) --- =Negligible. Source: Erdogan(2000). 190 Table A.II.lO. Structure of Poverty by Economic Activity (Basic Needs) trade Transportation 3.03 1.62 7.57 Financial institutions 0.41 0.18 1.16 Source: Erdogan(2000). Despite the recent shift of poverty from rural to urban areas, rural poverty remains an important dimension o f poverty inTurkey. However, there are not many studies that focus on the determinants of rural poverty. Pamuk (2000) attempts to address this issue in her study. She uses rural area data o f the 1994 Household Income Distribution Survey. There are three poverty measures used in this study: the Headcount Index, the Poverty Gap Index (PGI), and the Foster-Greer-Thorbecke P2 Index. The second measure, the PGI, is the arithmetic mean o f the difference between the income o f individuals in poor households and the poverty line; the third measure, P2, is the arithmetic mean o f the square o f this difference, and it captures the severity o fpoverty. Pamuk shows that rural poverty differs significantly across regions. Table AX.11 summarizes the measurements o f poverty and the structure o f poverty across regions. One important point to keep in mind for all the tables in Pamuk's study is that the rural population is the group being examined. Therefore, all the numbers are indexes that compare certain subgroups within the rural population only, not the total populationinTurkey. Table A.II.ll. Regional Poverty Regions Head Count Poverty Severity of Index Gap Index Poverty Marmara 6.50 21.76 0.53 Aegean 7.67 23.36 0.68 Mediterranean 17.42 27.77 1.94 Central Anatolia 15.79 26.73 1.84 Black Sea 16.44 29.44 2.05 East 12.97 26.17 1.27 Southeast 30.46 26.93 3.26 26.92 1.62 191 Pamuk also looks at how the poverty risk faced by the members o f a household changes according to the socioeconomic status of the household head. Table A.II.12 summarizes Pamuk's findings for poverty by gender of the householdhead. Ingeneral, households with female heads face higher poverty risk, and the severity of poverty i s higher for suchhouseholds. Table A.II.12. Poverty and Gender of HouseholdHead 1 Gender of Head Poverty Severity of Household Head Count Gap Poverty Index Index 14.46 26.91 1.58 IMale Female 22.11 27.15 2.41 Total 14.80 26.92 1.62 Source: Pamuk (2000). Inthe rural areas, just likeinurbanareas, household size andyears ofeducation ofthe householdheadare determinant factors of poverty. As household size increases, the poverty risk faced by the household members increases, and as years of education o f the household head increases, the poverty risk decreases. Some other socioeconomic variables she uses to disaggregate poverty are by employment status of household head, number of working individuals in the household, source o f income, and types o f economic activities the household head is engaged in (the study also looks at some of these statistics at the individual level). The findings are similar to the statistics obtained from the total population. An important aspect o f the study is that poverty is examined according to certain factors that play an important role in determining the relative standing of households in rural places. Its analysis o f poverty profiles that are geared solely toward the analysis of the rural areas i s what gives this study its unique place in the poverty literature in Turkey. For example, Pamuk looks at poverty according to the agricultural activity the household is engaged in, and ownership of agricultural land and equipment. Table A.II.13 summarizes Pamuk's findings on poverty rates according to the type o f agricultural activity rural households engage in. Inrural areas, households not engaged in agricultural activities face above- averagepoverty risk. Table A.II.13. Poverty and Agricultural Activity of the Household Type of Agricultural Activity of Head Count Index Total Poor the Household Population (Yo) (YO) ( Y O ) Do not engage in agricultural 17.07 29.42 33.94 activity Farmer 15.88 15.03 16.12 Stockbreeder 15.79 4.99 5.32 Both farmer and stockbreeder I 13.07 50.56 44.62 Total 14.80 100.00 100.00 192 Table A.II.14 summarizes poverty rates according to the size and ownership o f agricultural land in rural areas. It i s observed that the highest level o f poverty riski s faced by households with agricultural land o f 6 to 10 decares (1 decare = 0.247 acres). As the size of the agncultural land that is owned increases, poverty risk decreases. One general conclusion i s that in rural areas, households that do not engage in agricultural activities, and those that engage in agricultural activities with land o f less than 20 decares, face higher poverty risk. Table A.II.14. Povertyand Agricultural Land Size Head Count Total Poor Agricultural Land Size Index Population Population (decares) ("/.I ( Y O ) ( Y O ) Donot have land 16.72 36.26 40.95 Ck5 17.88 7.77 9.38 6-10 21.36 8.97 12.95 11-20 16.80 13.24 15.03 21-50 13.80 18.27 17.04 51-100 5.64 9.71 3.70 101-500 2.54 5.60 0.96 501+ 0.00 0.18 0.00 Total 14.80 100.00 100.00 Note: Based on 1994HBS. *1 decare = 0.247 acres. Source: Pamuk(2000). Table A.II.15 shows the poverty rates according to the size of agricultural land owned. One striking observation i s that the poverty rate in Southeast Anatolia i s 30.46 percent, while the poverty rate in the group o f individuals living inhouseholds with agncultural land o f 0 to 5 decares is 48.23 percent, in this region. The regional breakdown of the poverty rates reveals that those households engaging in agricultural activity but that own relatively smaller amounts o f land face higher poverty risk than households not engaging in agricultural activity inmost of the regions, except for Central Anatolia. This observation contradicts the results from the previous table, that is, that households engaging in agricultural activity face lower poverty riskrelative to those that do not (Table A.II.14). This is a result o f the fact that lower poverty risk of those households engaging in agricultural activity i s mainly driven by those who own more land. 193 Table A.II.15. Poverty and Agricultural Land Size Across Regions Agricultural Marmara Aegean Mediterr- Central Black Sea East South- Total Land Size anean Anatolia east O* 8.23 9.47 20.39 17.04 10.71 18.90 39.45 17.07 0** 5.36 13.84 18.20 15.06 22.68 15.36 20.27 15.79 0-5 8.20 4.61 19.88 27.18 21.84 19.52 48.23 18.60 6-10 15.37 21.84 23.22 11.47 22.24 22.50 42.65 22.40 11-20 4.67 8.97 18.05 19.56 19.67 20.37 34.86 17.36 21-50 4.33 1.75 12.93 23.74 17.57 5.40 22.35 12.05 51-100 0.52 0.00 3.56 6.22 0.00 3.16 15.79 4.17 101-500 0.00 0.00 0.00 3.44 0.00 0.00 5.79 2.09 501 + 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Total 6.50 7.67 17.42 15.79 16.44 12.97 30.46 14.80 ***Householdsthat ** engage only in stockbreedkg as an agricultural activity. 1 decare = 0.247 acres. Another important factor in rural places i s the ownership of tractors. Table A.II.16 depicts the high poverty risk faced by households that do not own a tractor. An analysis o f the structure of poverty by tractor ownership (Table A.II.17) reveals that while the proportion of ruralhouseholds without a tractor is 84.42 percent, their proportionwithinthe poor population i s 92.32 percent. Table A.II.16. Poverty andNumber of Tractors Owned Number of Marmara Aegean Mediterr- Central Black Sea East South- Total Tractors anean Anatolia east Do not 8.73 10.94 16.97 17.48 16.86 13.12 30.74 15.47 own one 1 0.88 1.26 9.04 12.68 8.35 8.71 15.76 7.03 2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Total 7.05 9.42 16.02 16.45 15.73 12.59 29.19 14.15 194 Table A.II.17. Structureof PovertyAccordingto Number of Tractors Owned Number of Marmara Aegean Mediter Central Black Sea East South- Total Tractors r-anean Anatolia east Donot own 97.36 97.94 93.25 84.59 92.98 91.67 94.39 92.32 one 1 2.64 2.06 6.75 15.41 7.02 8.33 5.61 7.68 2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Total I 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 ource: Pamuk (2000). StudiesBasedon 1987HBS Dumanli (1996) looks at dimensions of poverty in Turkey using the 1987 Household Income and Expenditures Survey (HBS). The poverty lines are computed based on a minimum required level of calories taken as 2,450 calories per day. The cost o f a minimumfood basket that meets the daily calorie- intake requirement is taken as the food poverty line. Table A.II.18 presents the poverty lines used inthis study. Table A.II.18. PovertyLinePer Capita (Annual) Years TL US$ 1987 284,700 332.2 1988 473,405 332.2 1989 777,085 365.0 1990 1,388,433 532.4 1991 2,135,740 513.4 1992 , 3.830.809 , , 556.2 1993 I 14,696,360 1 541.4 TL =TurkishLira. Source: Dumanlr (1996). Table A.II.19 summarizes Dumanli's findings on poverty. Here, he takes an approach of comparing average household income to the poverty line across households grouped according to annual income. In his study, Dumanli puts a stronger emphasis on income inequality than poverty. TableA.II.19. Comparisonof Incomeand PovertyLinesby 5 Percentof Population Income 5% 1.156 915 L 5% 1,176 1,127 5yo 1,264 1,298 5yo 1,293 1,483 5yo 1,361 1,591 5yo 1,361 1,677 195 I Household Poverty Line I Household 1 Percentage II Income 5% 1.330 1.863 5% 1,347 2,249 5% 1,361 2,454 5% 1.341 2.692 I 5?'o I 1.463 I I - 3 - 2.960 - 7 -- - - 5% 1,426 3,259 5% 1,415 3,623 5% 1,449 4,063 5% 1,469 4,627 5% 1,572 5,332 5% 1,495 6,377 5yo 1,515 8,102 5yo 1,461 16,947 *Valuesin 1,000 TL. Source: Dumanli(1996). ComparativeStudies Due to the difficulty o f comparing poverty across time, there are not many comparative studies that examine how poverty has changed over the years in Turkey. Dagdemir (1999) is unique inthis respect, and analyzes data from the 1987 and 1994 Household Income and ConsumptionExpenditures Surveys (HBS) to assess changes inpoverty during 1987-1994. For the 1994 poverty lines, Dagdemir uses the poverty lines that Erdogan (2000) calculates using the minimumfood cost (MFC) and cost-of-basic needs (CBN) methods. The 1987poverty lines are obtained after deflating the 1994 poverty lines with an appropriate index that takes into account increase in per capita income between 1987 and 1994. The issue o f comparability o f poverty between these two years is not addressed in detail in this study. It is expected that results are very sensitive to the price indexes that are used, and a more detailed analysis o f the comparability o f poverty lines between the two periods i s essential for a comparative analysis o f poverty. The changes inpoverty inTurkey duringthis period are summarized inTables A.II.20 and A.II.21. Table A.II.20. PovertyinTurkey, 1987-1 994 (Minimum-Food-CostApproach Poverty (YO) Marmara I Mediterranean 127 13.3 13.6 Central I 85 I 108 I 10.1 I 10.1 I Black Sea 109 143 14.2 18.8 East 141 144 --- 14.7 Southeast 129 131 --_ 19.6 Urban 95 136 6.9 8.7 Rural 126 135 21.2 20.2 Turkey 105 138 11.5 11.5 196 Table A.II.21. PovertyinTurkey, 1987-1994, (Cost-of-Basic-Needs Approach) Regions Poverty Line Poverty (%) (US$) 1987 1994 1987 1994 Marmara 165 220 --- 23.4 Aegean 168 208 --- 20.8 Mediterranean 177 254 41 5 46 7 Central I 134 I 170 I 23.2 I 27.2 Black Sea I 148 I 194 I 23.0 32.2 Dagdemir finds that the cost of a minimum food basket increased from US$l05 to US$138 (monthly) between 1987 and 1994. For the CBN approach, the minimumcost o f meeting basic food and non-food requirements increased from US152 to US$198. It i s observed that the household income of the poor has followed an increasing trend parallel to the increase in the minimum food cost, and therefore the percentage of households that cannot meet the minimumfood cost has not changed significantly during this period. On the other hand, the percentage of households whose income is below the CBN poverty line has increased from 27 to 29 percent. According to the MFC method, the poverty rate inurban areas increased from 6.9 to 8.7 percent between 1987 and 1994. In contrast, in rural areas the poverty rate decreased from 21.2 to 20.7 percent. According to the CBN method, the poverty rate in urban areas increased from 14.3 to 20 percent, and in rural areas it increased from 41.5 to 42.5 percent. One conclusion from these observations i s that there i s a sharp contrast inthe change inpoverty rates inrural areas compared to change inpoverty rates inurban areas according to the CBN method. On the other hand, accordingto the MFC method, inbothurban and rural areas, the change in poverty rates during this period is not as significant. According to the CBN method, it is observed that the poverty rates inurban areas have increased significantly duringthis period. A closer look into the structure of poverty reveals the share of the urban poor in the total population of poor has increased from 27.5 to 37.6 percent, while the share of the rural poor has decreased from 72.5 to 63.4 percent. Based on these observations, it can be concluded that poverty slowly shiftedfrom urban to rural areas between 1987 and 1994. However, rural poverty still remains significantly higher than urban poverty inTurkey. Dagdemir also analyzes the changes in regional poverty during 1987-1994. In Southeast Anatolia, the poverty line increased from US$135 monthly to US$137, and the poverty rate decreased from 21.9 to 16.5 percent. In 1994, the region with the highest poverty rate was the Black Sea, followed by East and Southeast, Mediterranean, Central Anatolia, Aegean, and Marmara regions. Ingeneral, it is observed that poverty rates have increased most significantly inthe Aegean, Marmara, and Black Searegions. Perceptionsof Urban Poverty Perceptions of Urban Poverty in Turkey (Erdogan 2002) i s a sociological study on poverty. Its focus i s the urban poor, and it is based on 160 interviews conducted with extremely poor households. The study consists of articles by various authors, and a selection o f the 160 interviews. Its main purpose i s to 197 identify how the social hierarchy reflects in the self-image o f the poor, and the "deep scars," other than hunger and physical hardships, that poverty inflicts upon them. Such identification is achieved through using the interviews with the poor as the main source. The following are some o f the issues on which the authors focus: H o w the poor feel about their place in society; what their daily experiences are and how they choose to express and define themselves; the conditions that the poor live in; how the poor perceive class and cultural hierarchy in the society; how this affects the way the poor perceive themselves and the "others" (the "rich"); the process o f the marginalization and isolation o f the poor; and differences in the way men and women experience poverty. The study seeks to go beyond listing the effects o f "material poverty." The main concern i s the other dimension o f poverty that i s more subjective, personal, and not quantifiable. Using the interviews, it i s shown that the experience o f poverty takes various forms. One form is the "visual experience" o f the poor. The social hierarchy are reflected in the way the poor feel they are being "looked" at by the "others" (the rich). The poor feel that the rich look at them from above and that they humiliate the poor even with their looks. Therefore, "looks" constitute the first feature o f the individual's experience o f his or her poverty. Another feature o f the experience of poverty is the conflict between "wanting to speak up" while "never speaking up." Inother words, while the poor have the desire to speak up about their painful experiences, they are also reluctant to speak at all, especially because they do not see themselves ina position to speak. Erdoganemphasizes the fact that the poor are ashamed of their way of speaking. Therefore, listening to the poor should consist not only o f listening to their words-it should also be listening to their silence and interpreting it correctly. Another feature of the self-image of the poor is the reflection of the poverty experience inthe "physical body." One example i s the general belief o f the poor that the rich see them as "animals." Inthe interviews, itisobservedthat thisself-image isalsoreflected in their body posture. For example, the respondents usually refuse to sit at the same level as the interviewer; they want to sit on the floor or on a lower chair, instead. It is observed in the interviews that the extent o f informal assistance that is in general believed to exist among relatives, fiiends, neighbors, and hemseri is rarely mentioned by the poor as a source of help. Most o f the interviewees say that they never get support fiom friends, relatives, neighbors, or hemseri. For the poor households interviewed, the major source o f help comes from various government and private organizations (the Social Services and Child Protection Organization [Sosyal Hizmetler ve Cocuk Esirgeme Kurumu, SHCEK], Deniz Feneri, municipalities, and so forth). Although some mention a couple o f instances where their relatives or neighbors helped them find a job or gave them food and money, the general consensus i s that such sources o f aid are rare and unreliable. It is concluded that the social assistance providedby such sources in general excludes the "extremely poor." For example, when asked whether they help each other among relatives and neighbors, most o f the respondents say, "Nobody i s in the [financial] position to help anybody." It is concluded that support from govemment or private organizations i s a major source o f help for the extremely poor. The woman's role in seeking such support i s very important for the coping strategies of the poor households. Inmost interviews, the women say that their husband i s too proud to ask for help. Therefore, it i s usually the woman who tries these altematives. The interviews provide a different perspective on the lives o f the extremely poor households. The picture o f poverty that i s revealed here i s very different from the one usually observed in the popular media. Poverty i s not romanticized. Moreover, the common tendencies o f the popular media, such as attaching a different identity to the poor or looking for the characteristics o f the "culture o f poverty," are severely criticized in this study. Cultural schemas in relation to poverty are studied in a framework that i s not restrictive in its implications. This study i s unique in the way it sheds light on these cultural schemas by measuring or quantifying the perceptions of the poor o ftheir daily life. 198 Conclusion The only way to combat poverty is to eliminate the factors that create it. This can be done only if policymakersknow what those factors are. The studies reviewedhere aim to provide such informationby examining various dimensions o f poverty. It i s important to bringthis information together and compare the studies in terms of the different aspects of poverty they look at, and this review is an attempt to achieve that goal. It shouldbe notedthat there are some poverty studies that have not been reviewedhere mainly due to lack of access. 199 REFERENCES Alici, Sema. 2000. "Turkiye de Yoksullugun Sosyo Ekonomik Analizi" ("A Socioeconomic Analysis of Poverty inTurkey"). In YoksulluklaMucadele Stratejileri. Ankara: Hak-19 Konfederasyonu Yayini 2002. Behrman, Jere R., and Nancy Birdsall. 1983. "The Quality of Schooling. Quantity Alone I s Misleading." American Economic Review, 73(5):928-46. Bourguignon, Francois. 2002. "The Growth Elasticity o f Poverty Reduction: Explaining Heterogeneity Across Countries andTime Periods." UnpublishedManuscript. February. Brahmbhatt, Milan. 2001. "Lessons for Turkey from International Experience with Financial Crisis." Manuscript. Washington, D.C.: World Bank. Coudouel, Aline, and Jesko Hentschel. 2000. "Poverty Data and Measurement." A Sourcebookfor Poverty Reduction Strategies. Washington, D.C.: World Bank. Council o f HigherEducation. 1998. "Parental Income, EducationalExpenditures, Financial Aid and Job Expectations o f UniversityStudents." YOK (Turkish)Council o fHigherEducation. Dagdemir, Ozcan. 1999. "Turkiye EkonomisindeYoksulluk Sorunu ve Yoksulluk Analizi 1987-1 994" (Poverty Problem and the Analysis of Poverty inthe TurkishEconomy 1987-1994). H.U. Iktisadi ve Idari Bilimler Fakiiltesi Dergisi, Ciltl7, Say1 1. Datt, Guarav, and MartinRavallion. 1992. "Growth and RedistributionComponents of Changes in Poverty Measures: A Decompositionwith Applications to Brazil and India inthe 1980s." Journal of Development Economics, 38:275-95. Dayioglu, Meltem, and InsanTunali. 2004. "Falling Behind While CatchingUp: Changes inthe Female-Male Wage Differential inUrbanTurkey." Unpublishedmanuscript, Koc University. Deaton, Angus. 1997. TheAnalysis of Household Surveys: Microeconomic Analysisfor Development Policy. Baltimore: Johns Hopkins University Press for the World Bank. Deaton, Angus, and Chstina Paxon. 1998. "Economies of Scale, Household Size, and the Demand for Food." Journal of Political Economy, 106:897-930. Deaton, Angus, and John Muellbauer. 1980. Economics and Consumer Behavior. Cambridge: Cambridge UniversityPress. . 1986. "On MeasuringChild Costs: WithApplications to Poor Countries." Journal of Political Economy, 94:72&44. Deaton, Angus, and Salman Zaidi. 2002. "Guidelines for Constructing ConsumptionAggregates for Welfare Analysis." LSMS WorkingPaper 135. Washington D.C.: World Bank. htt~://pove1-ty.worldbank.orrr/librarylview/43 36/. Dulger, Ilhan. 2004. "Turkey: Rapid Coverage for Compulsory Education Program." Unpublished manuscript. 200 Dumanli, Recep. 1996. "Yoksulluk ve Tiirkiye'deki Boyutlan" ("Poverty and its Dimensions in Turkey"). Erdogan, Guzin. 2000. "Tiirkiye'de ve Diinya'da Yoksulluk Olqumleri Uzerine Degerlendirmeler" ("Poverty Assessment inTurkey and inthe World"). InDIE, Iygiicii Piyasalan Analizleri. Ankara: DevletIstatistik EnstitiisiiMatbaasi. Erdogan, Necmi (ed). 2002. Yoksulluk Halleri (Perceptions of Urban Poverty). Istanbul: Demokrasi Kitapligi Yayinevi. Foster, James, J. Greer, and Eric Thorbecke. 1984. "A Class of Decomposable Poverty Measures." Econometrica, 52:761-66. Glewwe, Paul, and Gilette Hall. 1992. "Poverty and Inequality duringUnorthodox Adjustment: The Case of Peru, 1985-90." World Bank Living Standardsand Measurement Study Working Paper No. 86. Washington, D.C. Heckman, James J. 2003. "China's Investment inHuman Capital." Economic Development and Cultural Change, 51(4):795-804. Lanjouw, Peter, BrankoMilanovic, and Stefan0 Paternostro. 1998. "Poverty and the Economic Transition: How Do Changes inEconomies of Scale Affect Poverty Rates for Different Households?" Policy Research WorkingPapers 2009. Washington, D.C.: World Bank. Lanjouw, Peter, and Martin Ravallion. 1995. "Poverty and HouseholdSize." Economic Journal, 105:1415-34. Liviatan, Nissan. 2002. "Lessons for Turkey from International Experience with Financial Crisis and Disinflation." TurkeyCountry Economic Memorandum (2003) Background Paper. Washington, D.C.: World Bank. Mete, Cem. 2004. "The Inequality Implications of Highly Selective Promotion Practices." Economics of Education Review (forthcoming). Milanovic, Branko, and Sholomo Yitzhaki. 2001. "Decomposing World Income Distribution: Does the World Have aMiddleClass?" Policy ResearchWorking PaperNo. 2562. Washington, D.C.: World Bank. Mincer, J. 1974. "Schooling, Experience and Earnings." National Bureauo f Economic Research. New York: Columbia UniversityPress. Mingat, Alain. 1998. "The Strategy Usedby High-performingAsian Economies inEducation: Some Lessons for DevelopingCountries." WorldDevelopment, 26(4):695-7 11. Ministry of Health. 2003. National HealthAccounts 2000. Draft Report. Turkey. Ministry of National Education. 2003. "Milli EgitimSayisal Veriler 2002-2003 ." Ankara. Ogut, Menevis, and R.Funda Barbaros. "Regional DevelopmentInequalitiesinTurkey: An Assessment on the Distribution of Investment Incentives." Draft. Undated. 201 Pamuk, Muzeyyen. 2000. "Kirsal Yerlerde Yoksulluk" ("Poverty inRural Places"). InDIE, Iggucu PiyasalariAnalizleri. Ankara: Devlet Istatistik EnstitiistiMatbaasi. Psacharopoulos, George. 1994. "Returns to Investment inEducation: A Global Update." World Development, 22(9): 132543. Ravallion, Martin. 1992. "Poverty Comparisons: A Guide to Concepts and Methods." LSMS Working Paper 88. Washington, D.C.: World Bank. . 1994. Poverty Comparisons: Fundamentals of Pure and Applied Economics. Chur: Harwood Academic Press. . 1998. "Poverty LinesinTheory andPractice." LSMS WorkingPaper 133. Washington, D.C.: World Bank. . 1999. "Poverty Lines inTheory andPractice." Living Standards Measurement Study Worlung PaperNo. 133. Washington, D.C. . 2000. "Poverty inTheory andPractice." LivingStandardsMeasurement StudyWorkingPaper No. 133. Washington, DC: World Bank. Schultz, T. Paul. 2002. "Why Governments ShouldInvest More to Educate Girls." WorldDevelopment, 30(2):207-25. Shorrocks, A. F. 1999. "Decomposition Procedures for Distributional Analysis: A UnifiedFramework Based onthe Shapley Value." Mimeo, University o f Essex. TC MilliEgitimBakanligi (Republic o fTurkey Ministryof Education). 2001. "2002 Yili Basinda Milli Egitim" ("National Educationat the Beginningo f 2002"). Ankara, Turkey, December. TurlushGovernment. 2003. "Preliminary NationalDevelopment Plan." Ankara, December. TurkiyeIs Bankasi. 1999. "75 Yilda Egitim" ("Education duringthe Last 75 Years"). Istanbul, Turkey, 1999. UNDP (UnitedNations Development Programme). 2001. "Human Development Report." 2003. "New Poverty and the ChangingWelfare Regimeo f Turkey." Ankara. Wagstaff, A., and E.Doorslaer. 2002. "Catastrophe and Impoverishment inPaying for Health Care: With Applications to Vietnam, 1993-1 998." Washington, D.C.: World Bank. World Bank. 1990. WorldDevelopment Report 1990:Poverty. Washington, D.C. . 2000. "Making TransitionWork for Everyone: PovertyandInequality inEuropeandCentral Asia." Washington, D.C. . 2000. "Turkey: CountryEconomicMemorandum-Structural Reformsfor Sustainable Growth." Report No. 20657-TU, September 15. 202 . 2000. "Turkey: Economic Reforms, Living Standards, and SocialWelfare Study." ReportNo. 20029-TU. Washington, D.C., January 27. . 2001. WorldDevelopment Report 2000/2001: Attacking Poverty. Washington, D.C.: World Bank and OxfordUniversityPress. . 2003a. "Turkey: Country Economic Memorandum-Towards Macroeconomic Stability and Sustained Growth." ReportNo. 26301-TU, July 28. . 2003b. "Turkey: Poverty andCopingAfter Crises." ReportNo. 24185-TR. Human DevelopmentUnit, Europe and Central Asia Region. Washington, D.C., July 28. . 2003c. "Turkey: Reformingthe Health Sector for ImprovedAccess andEfficiency." Report No. 24358-TU. Washington, D.C., March. ------_-. 2004. "Turkey: A Review ofthe Impact ofThe Reformo fAgricultural Sector Subsidization." Washington, D.C. ,March. . Forthcoming. "Labor Market Study." Washington, D.C. WHO (World HealthOrganization). HFA. 2004. Ke Xu, David, B.Evans, KeiKawabata, Riadh Zerramdini, Jan Klaus, J. Chnstopher, and L.Murray. 2003. "Household Catastrophic Health Expenditure: Multicounty Analysis." n e Lancet, 362(12):111-17. 203 TURKEY JOINT POVERTY ASSESSMENT REPORT (JPAR) PROVINCE CAPITALS* PROVINCE BOUNDARIES NATIONAL CAPITAL INTERNATIONAL BOUNDARIES TURKEY MAIN ROADS This map was produced by the Map Design Unit of The World Bank. The boundaries, colors, denominations and any other information RAILROADS shown on this map do not imply, on the part of The World Bank *Province names are the same as their capitals. Group, any judgment on the legal status of any territory, or any endorsement or acceptance of such boundaries. 28° 32° 36° 44° B U L G A R I A Black Sea To Malko GEORGIA Turnovo To Kapitan Andreevo and Kharmanli Edirne Kirklareli Sinop To To Batumi Akhaltsikhe and T'bilisi Bosporus Bartin GREECE To Zonguldak Ergene Kizil Samsun Artvin Kipoi Ardahan Tekirdag Istanbul Karabük Kastamonou Irmak Kocaeli TrabzonRize Sea of Marmara Sakarya Ordu To Leninakan (Marmara Denizi) (Izmit) (Adapazari) Irmak Kars Yalova Giresun Çoruh ARMENIA Dardanelles Bolu Çankiri Irmak Amasya Gümüshane L. Manyas Nilüfer Delice Bursa Kelkit Çayl To Yerevan 40° Çanakkale Bilecik Çorum Tokat Bayburt 40° L. Apolyont Igdir Simav Sakar ya Nehri ANKARA Ankara Kirikkale Balikesir Eskisehir Yozgat Sivas Erzincan Erzurum Agri Sakar To Maku Kütahya and Reza'iyeh ya Euphrates Nehri ISLAMIC Kirsehir Tunceli Sea Keban Bingöl REP. OF Reservoir Manisa IRAN Gediz Afyon Mus Lake Van Nehri Tuz Elazig To Sharafkhaneh Gölü Van Izmir Usak Nevsehir Kayseri Bitlis Aksaray Malatya Hoyran Aegean Aci Gölü Egridir Gölü Diyarbakir To Aydin Gölü Siirt Reza'iyeh Denizli Konya Nigde Batman Isparta Tigris BurdurBurdur CeyhanNehri Adiyaman Euphrates Sirnak Hakkari Gölü Beysehir Kahraman Gölü Maras Mardin To Zakhu Mugla Karaman Osmaniye Gaziantep Sanliurfa To Tall To Tall Antalya Kushik Birak Icel Adana Kilis SEPTEMBER Gulf of Göksu (Mersin) I R A Q IBRD Antalya To Aleppo 36° 36° Hatay (Antakya) 0 100 200 To Al Atarib SYRIAN ARAB REPUBLIC 33581 2004 and Aleppo KILOMETERS 28° Mediterranean Sea 32° 40° 44° To Latakia TURKEY JOINT POVERTY ASSESSMENT REPORT (JPAR) GROSS DOMESTIC PRODUCT BY STATISTICAL REGION STATISTICAL REGION BOUNDARIES PROVINCE CAPITALS* PER CAPITAL GROSS DOMESTIC PRODUCT, TURKEY BY STATISTICAL REGION, 2001 AT CURRENT PRICES NATIONAL CAPITAL PROVINCE BOUNDARIES 2,500 .00 US$ This map was produced by the Map Design Unit of The World Bank. 2,000 .00 US$ INTERNATIONAL BOUNDARIES The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of The World Bank 1,500 .00 US$ Group, any judgment on the legal status of any territory, or any *Province names are the same as their capitals. 1,000 .00 US$ endorsement or acceptance of such boundaries. 28° 32° 36° 44° B U L G A R I A Black Sea GEORGIA Edirne Kirklareli Sinop Bartin Bosporus GREECE Zonguldak Samsun Karabük Artvin Tekirdag Istanbul Kastamonou TrabzonRize Ardahan Sea of Marmara Ordu Kocaeli Sakarya (Marmara Denizi) Yalova (Izmit) (Adapazari) Giresun Kars ARMENIA Bolu Çankiri Amasya Bursa Gümüshane 40° Bilecik Tokat 40° Çanakkale Çorum Bayburt Igdir ANKARA Kirikkale Balikesir Erzurum Eskisehir Yozgat Sivas Erzincan Agri Kütahya Euphrates Kirsehir ISLAMIC Tunceli Sea Keban Bingöl REP. OF Reservoir Manisa Mus IRAN Afyon Tuz Lake Van Van Usak Gölü Nevsehir Elazig Kayseri Izmir Eup hrates Bitlis Malatya Aksaray Aegean Aydin Konya Nigde Diyarbakir Batman Siirt Denizli Isparta Kahraman Adiyaman Tigris Burdur Hakkari Maras Beysehir Sirnak Gölü Mardin Mugla Karaman Osmaniye Sanliurfa Antalya Gaziantep Icel Adana Kilis SEPTEMBER Gulf of (Mersin) I R A Q IBRD Antalya 36° 36° Hatay (Antakya) SYRIAN ARAB REPUBLIC 0 100 200 33582 2004 KILOMETERS 28° Mediterranean Sea 32° 40° 44°