Turkey Joint Poverty Assessment Report (Volume I) by World Bank Turkey Office

Public Disclosure Authorized

Report No. 29619-TU

Turkey Joint Poverty Assessment Report (In Two Volumes) Volume I:

Main Report

Human Development Sector Unit

State Institute of Statistics

Europe and Central Asia Region

Turkey

Public Disclosure Authorized

Joint Poverty Assessment Report Volume I

Public Disclosure Authorized

Turkey

Public Disclosure Authorized

August 8, 2005

Document of the World Bank and the State Institute of Statistics, Turkey

HBS HCIS HDR HH ILO

IMF

JPAR Kcallday Kg LFS MFC MOH NA n.e.c. NUTS OECD PA PEC PEI PGI PPP PSBR PTT REER SHCEK SME SOEs SRMP SSK SYDTF SYDVs TIMSS TL UNDP USD USDA UNESCO WB WDI WHO

Household Budget Survey Household Consumption and Income Survey Human Development Report Household International Labor Organization International Monetary Fund Joint Poverty Assessment Report Kilocalories per day Kilograms Labor Force Surveys MinimumFood Cost Ministry o f Health National Accounts N o t Elsewhere Classified Nomenclature o f Statistical Territorial Units Organisation for Economic Cooperation and Development Poverty Assessment Per Equivalent Consumption Per Equivalent Income Poverty Gap Index Purchasing Power Parity Public Sector Borrowing Requirement Postal Telephone & Telegraph Real Effective Exchange Rate Social Services and Children Protection Organization, Sosyal Hizmetler ve Cocuk Esirgeme Kurumu Small and Medium Enterprises State-owned Enterprises Social Risk Mitigation Project Social Insurance Organization, Sosyal Sigortalar Kurumu Social Assistance and Solidarity Encouragement Fund, Sosyal Yardimlasma ve Dayanisma Tesvik Fonu Social Assistance and Solidarity Foundations, Sosyal Yardimlasma ve Dayanisma Vakiflar Trends in International Mathematics and Science Study Turkish Lira United Nations Development Programme US. Dollars United States Department o f Agriculture United Nations Educational, Scientific and Cultural Organization World Bank World Development Indicators World Health Organization

TURKEY JOINT POVERTY ASSESSMENT REPORT Table o f Contents PREFACE ACKNOWLEDGEMENTS EXECUTIVE SUMMARY .............................................................................................................

CHAPTER I:D A T A COMPARABILITY, 1994-2002................................................................. A . Real Income and Consumption o f Households Between 1994 and 2002 .............................

1 1

A . Data Comparability ............................................................................................................................. i B. Recent Economic Developments ........................................................................................................ ii ... C . Poverty Profile ................................................................................................................................... 111 D. Education........................................................................................................................................... iv E. Health................................................................................................................................................. vi ... F. Labor Market in Turkey ................................................................................................................... viii G. Social Protection in Turkey ............................................................................................................... ix B. Inequality in Consumption o f Households Between 1994 and 2002: HBS........................................ C . Poverty Lines: 1994 and 2002 Methodology ..................................................................................... D. Poverty in 1994 and 2002: HBS......................................................................................................... E. Using the 1994 HBS and 2002 HBS to Decompose Changes in Poverty ....................................... F. Conclusions .......................................................................................................................................

5 5 7 10 13

CHAPTER 11: MACROECONOMIC CONTEXT ......................................................................

CHAPTER 111: POVERTY PROFILE.........................................................................................

CHAPTER IV: EDUCATION.....................................................................................................

CHAPTER V: HEALTH .............................................................................................................

A . Macroeconomic Instability Has Helped Keep Growth Below Potential .......................................... 15 B. Fiscal Imbalance Has Been the Root o f Chronic Macroeconomic Instability in Turkey ................. 15 C. Turkey’s Exchange Rate-Based Disinflation Program..................................................................... 16 D. Exchange Rate-Based Disinflation Programs A r e Vulnerable in a Globalized World .................... 17 E. The Government Responded Quickly to the 2001 Crisis ................................................................. 18 F. Turkey’s Crisis Response Program Incorporated the Experience o f Other Emerging Markets ....... 18 G. Strong Recovery Has Been Under Way Since Early 2002 ............................................................... 20 H. Macroeconomic Outcomes in 2003 Were Favorable ....................................................................... 21 I . Turkey Has Achieved Sizable Fiscal Adjustment ............................................................................. 21 J. Deterioration in External Balance Has Created Volatility in Domestic Financial Indicators ........... 22 K. Ongoing Structural Reforms Are Expected T o Stimulate Growth................................................... 23 A . Household Size and Composition .................................................................................................... B. Household Head Characteristics ....................................................................................................... C . Spatial Characteristics ...................................................................................................................... D. Non-Income Aspects o f Poverty ...................................................................................................... E. Inequality and Regional Differences ................................................................................................ F. Multivariate Analysis ........................................................................................................................

24 29 31 34 37 40

A . The Context ...................................................................................................................................... 43 B. Public Spending on Education.......................................................................................................... 45 C . Traces o f Inefficient Resource Allocation in the Education Sector ................................................. 47 D. Distributional and Education-Quality Implications o f Public Policies on Education ...................... 50 E. Economic Growth and Labor Market Implications o f a Public Policy that Undervalues Education 62

A . The Health Care System................................................................................................................... B. Health Insurance............................................................................................................................... C . Health Outcomes .............................................................................................................................. D. Self-Reported Morbidity .................................................................................................................. E. Access T o and U s e o f Health Care ................................................................................................... F. Health Care Expenditure...................................................................................................................

70 72 74 75 76 84

CHAPTER VI: LABOR ..............................................................................................................

CHAPTER VII: SOCIAL PROTECTION ................................................................................

100

REFERENCES ...........................................................................................................................

200

B o x VII.l. Parameters ofthe Current Pension System ...........................................................................

101

A. B. C. D.

Official Unemployment and Labor Force Participation Rates from the LFS................................... Unemployment and Inactivity .......................................................................................................... Quality o f Employment .................................................................................................................... Sector o f Employment ......................................................................................................................

A. B. C. D.

Turkish Pension System ................................................................................................................. Social Solidarity Fund .................................................................................................................... Conditional Cash Transfers ............................................................................................................ Local Initiatives ..............................................................................................................................

90 92 94 98

100 107 112 113

List of Figures Figure I .1. Consumption Per Capita with Various Deflators. 1990 .100 ................................................... 4 Figure 1.2. Price Indexes in Turkey. 1990.2002. Annual Percentage Changes ........................................... 4 Figure 1.3. Turkey: Poverty and GDP Growth ........................................................................................... 8 Figure 1.4. Poverty Changes ...................................................................................................................... 10 Figure 11.1. Per Capita GDP Compared to EU Average ............................................................................ 14 Figure 11.2. GDP and GDP Per Capita Growth Rates in Emerging Economies (1965-2001) ................... 14 Figure 11.3. Turkey: Volatility o f REER and GDP Growth, 1990-2000 .................................................. 15 16 Figure 11.4. Adjusted Public Sector Borrowing Requirement (% o f GNP) ............................................... Figure 11.5. Total Net Public Debt (% o f GNP) ......................................................................................... 19 Figure 11.6. GNP Growth........................................................................................................................... 20 Figure 111.1. Turkey: Poverty and Household Size ................................................................................... 25 26 Figure 111.2. Turkey: Poverty and Number o f Children............................................................................ Figure 111.3. Turkey: Poverty and Number o f Elderly .............................................................................. 27 Figure 111.4. Turkey: Share o f Poor in Percent by Household Consumption ........................................... 28 Figure 111.5. Turkey: Poverty Rate and Age ............................................................................................. 29 Figure 111.6. Turkey: Poverty and Gender o f Head .................................................................................. 30 Figure 111.7. Turkey: Poverty and Education o f Household Head............................................................ 30 Figure 111.8. Turkey: Poverty and Location.............................................................................................. 31 Figure 111.9. Turkey: Household Size and Location ................................................................................. 32 Figure 111.10. Rural Employment by Sector .............................................................................................. 33 Figure 111.11. Urban Employment by Sector ............................................................................................. 33 Figure 111.12. Rural Employment by Sector .............................................................................................. 34 Figure 111.13. Turkey: Poverty and Dwelling ........................................................................................... 35 35 Figure 111.14. Turkey: Poverty and U s e o f Dung ..................................................................................... 37 Figure 111.15. Turkey: Poverty and Discretionary Activities .................................................................... 37 Figure 111.16. Turkey: Poverty and Shopping........................................................................................... 41 Figure 111.17. Per Capita GDP Index ......................................................................................................... Figure 111.18. Socioeconomic Development Levels (2003) ...................................................................... 41

Figure 111.19. Female Literacy Rates (2000) ............................................................................................. 42 Figure IV.1. Public Spending on Education as Percentage o f Gross Domestic Product ........................... 45 Figure IV.2. Number o f Students by School Type .................................................................................... 48 Figure IV.3. Number o f Classrooms Over Time ....................................................................................... 48 Figure IV.4. Number o f Students Per Classroom by School Type ............................................................ 49 50 Figure IV.5. Student-Teacher Ratios in the 1997-1998 and 2002-2003 School Years ........................... 57 Figure IV.6. 1999 TIMSS Mathematics Achievement o f Eighth-Grade Students .................................... 58 Figure IV.7. 1999 TIMSS Science Achievement o f Eighth-Grade Students ............................................ 59 Figure IV.8. Problems with Schools, Public and Private .......................................................................... Figure IV.9. Problems with Schools, Urban, and Rural ............................................................................ 60 61 Figure IV.10. Satisfaction with Quality o f Education ............................................................................... Figure IV.11. Satisfaction with Quality o f Education ............................................................................... 61 Figure IV.12. Unemployment Rates Over Time ....................................................................................... 65 Figure IV.13. Illiterate Unemployment Rates Over Time ......................................................................... 66 Figure IV.14. Literate Without Diploma Unemployment Rates Over Time ............................................. 66 67 Figure IV.15. Primary School Graduates’ Unemployment Rates Over Time ........................................... Figure IV.16. Junior Secondary Graduates’ Unemployment Rates Over Time ........................................ 67 Figure IV.17. Vocational Junior Secondary Graduates’ Unemployment Rates Over Time ...................... 68 Figure IV.18. Secondary School Graduates’ Unemployment Rates Over Time ....................................... 68 Figure IV.19. Vocational Secondary School Graduates’ Unemployment Rates Over Time..................... 69 Figure IV.20. Higher Education Graduates’ Unemployment Rates Over Time ........................................ 69 Figure V.l. Share o f Population with Health Insurance, by Quintile ........................................................ 73 Figure V.2. Regional Variation in Infant and Child Mortality Rate and Prenatal Care and Home Births 74 Figure V.3. Incidence o f Illness Income Quintiles and Location .............................................................. 75 Figure V.4. Incidence o f Morbidity by Health Insurance Status ............................................................... 76 Figure V.5. Share o f Those Reporting Illness Who Sought Care by Quintile and Location..................... 77 Figure V.6. Propensity to Seek Treatment ................................................................................................ 77 Figure V.7. Propensity to Seek Care When Illby Insurance Status .......................................................... 78 Figure V.8. Hospitalization Among Those RequiringHospitalization, by Quintile ................................. 79 87 Figure V.9. Evolution o f Public Sector Spending on Health..................................................................... Figure V.10. Composition o f Public Sector Spending on Health, 2003 .................................................... 87 Figure V.11. Composition o f Total Health Sector Spending (2000) ......................................................... 88 Figure V.12. Destination o f Central Government Funding for Health Care ............................................. 89 Figure VI.1. Turkey: Unemployment Rate............................................................................................... 91 Figure VI.2. Turkey: Labor Force Participation Rate............................................................................... 91 Figure VI.3. Turkey: Adults Employed.................................................................................................... 92 92 Figure VI.4. Turkey: Adults Not Employed............................................................................................. 94 Figure VI.5. Turkey: Poverty and Inactivity ............................................................................................ Figure VI.6. Turkey: Poverty and Employment Situation........................................................................ 94 Figure VI.7. Turkey: Poverty and Employment Status ............................................................................ 95 Figure VI.8. Turkey: Poverty and Social Security ................................................................................... 96 Figure VI.9. Turkey: Poverty and Type o f Workplace ............................................................................. 96 Figure VI.10. Turkey: Poverty and Size o f Firm...................................................................................... 97 105 Figure VII.l. Budgetary Transfers to Social Security Institutions .......................................................... List of Tables

Table 1.1. Table 1.2. Table 1.3. Table 1.4.

Macroeconomic and Household Survey-Based Data on Living Standards. 1994-2002 ............. 2 3 Indexes for Deflating Nominal Values: 1994 and 2002. 1994 = 1.00......................................... Change in M a i n Indicators for Real Incomes and Consumption. 1994.2002 ............................. 3 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 in Turkey Between 1994 and 2002 ....................................................... 7 Table 1.8. Comparable Poverty Incidence ................................................................................................... 8 Table 1.9. Absolute Poverty Rates o f Europe and Central Asia .................................................................. 9 Table I.10. Decomposition o f Poverty Change Between 1994 and 2002.................................................. 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 in Household ...................................................... 25 Table 111.3. Turkey: Age Structure o f H B S 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 o f Age ................................................................................................. 28 Table 111.7. Turkey: Poverty Rate o f Consumer Durables ........................................................................ 36 Table 111.8. Comparable Inequality ........................................................................................................... 38 Table 111.9. N e w 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.1. Number o f Schools, Students, and Teachers in the 2001-2002 School Year ........................ 44 Table IV.2. Incidence o f Public Spending o n Education in 1994 and in 2001.......................................... 46 Table IV.3, Incidence o f Household Expenditures on Education in 1994 and in 200 1............................. 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 ............................................................... 52 Table IV.6. Estimated Distance Conditional o n School Availability in the Residential Area .................. 53 Table IV.7. Predictors o f Continued Education after Completion o f 8-Year Primary School .................. 54 Table IV.8. Conditional on School Availability in the Residential Area .................................................. 56 Table IV.9. Problems with School ............................................................................................................ 60 Table IV.10. Satisfaction with Quality o f Education ................................................................................ 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 K e y Staff, by Region, 2002 ........................ 71 Table V.3. Regional Distribution o f Hospital 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 N o t Seeking Outpatient Care When I11................................................................ 79 Table V.10. Reasons for N o t Being Hospitalized When Needed.............................................................. 80 Table V.11. Problems Encountered When Seeking Outpatient Care, by Quintile .................................... 80 Table V.12. Average Amount Paid During Last 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 ....................................... 83 Table V .17. Utilization o f Preventive Care ............................................................................................... 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

Table V I.1. Turkey: Unemployment Rate and Labor Force ParticipationRate....................................... 90 Table VI.2. Turkey: Poverty Rate o f Reason for Not Seeking Job .......................................................... 93 Table VI.3. Turkey: Poverty Rate o f Status o f Workplace....................................................................... 97 Table VI.4. Turkey: Poverty Rate o f Legal Status o f Workplace ............................................................ 98 Table VI.5. Turkey: Poverty Rate o f Basic Code o f Main Activity o f Workplace .................................. 99 Table VII.1. ContributionRates to SSK .................................................................................................. 102 Table VII.2. Types o f SYDTF Assistance, July 1, 1997-March 26. 1999.............................................. 108 Table VII.3. SYDTF Employment-GeneratingProject Assistance, July 1, 1997-March26, 1999......... 109 Table VII.4. SYDTF Revenues ............................................................................................................... 110 Table VIIS . SYDTF Expenditures.......................................................................................................... 110 113 Table VII.6. Quarterly Figures for CCT .................................................................................................. Table VII.7. Turkey: Incidence ofNomina1Transfers ........................................................................... 114 List of Annexes

ANNEX I:METHODOLOGY................................................................................................................ ANNEX 11: POVERTY IN TURKEY: A LITERATUREREVIEW.....................................................

115 185

List of Annex Figures

Figure A.I.l. CumulativeDistributiono f PEC ........................................................................................

123

List of Annex Tables

Table A.I. 1. Household Size by the Type o f Settlement ......................................................................... Table A.I.2. The Structure o f Cash Income............................................................................................. Table A.I.3. The Structure o f Consumption ............................................................................................ Table A.I.4. Average Monthly Consumption by Deciles ........................................................................ Table A .1.5. Inequality Measures............................................................................................................ Table A.I.6. The Structure o f a Minimum Food Basket EstimatedBased on Survey Data .................... Table A.I.7. Average Value o f Monthly, Complete Poverty Line, by Household Size .......................... Table A.I.8. Poverty Headcount, Gap, and Severity Index ..................................................................... Table A.I.9. Sensitivity o f Poverty Statistics to CPI and Housing ImputedRent ................................... Table A.I.lO. Food Consumption and Expenditures, 1994 and 2002 ...................................................... Table A.I.ll. Comparable Poverty Measures, 1994 and 2002 ................................................................ Table A.I.12. Food Basket for Equivalent Adult ..................................................................................... Table A.II.1. Different Poverty Measures and Results............................................................................ Table A.II.2. Poverty by Region ............................................................................................................. Table A.II.3. Urban and Rural Poverty Rates.......................................................................................... Table A.II.4. Calorie Requirementsby Age and Gender ........................................................................ Table A.II.5. Structure o f Poverty by Age and Gender (Food Poverty).................................................. Table A.II.6. Structure of Poverty by Age and Gender (Basic Needs) ................................................... Table A.II.7. Structure of Poverty by Employment Status (Food Poverty) ............................................ Table A.II.8. Structure of Poverty by Employment Status (Basic Needs) .............................................. Table A.II.9. Structure of Poverty by Economic Activity (Food Poverty) .............................................. Table A.II.10. Structure o f Poverty by Economic Activity (Basic Needs) ............................................. Table A.II.l 1. Regional Poverty.............................................................................................................. Table A.II.12. Poverty and Gender o f Household Head ......................................................................... Table A.11. 13. Poverty and Agricultural Activity o f the Household ...................................................... Table A.II.14. Poverty and Agricultural Land Size ................................................................................. Table A.II.15. Poverty and Agricultural Land Size Across Regions.......................................................

119 120 121 122 124 125 126 127 128 129 130 130 186 186 187 187 188 189 189 190 190 191 191 192 192 193 194

Table A.II.16. Table A.II.17. Table A.II.18. Table A.II.19. Table A.II.20. Table A.II.21.

Poverty and Number o f Tractors Owned ......................................................................... Structure o f Poverty According to Number o f Tractors Owned ...................................... Poverty Line Per Capita (Annual) .................................................................................... Comparison o f Income and Poverty Lines by 5 Percent o f Population ........................... Poverty in Turkey, 1987-1994 (Minimum-Food-Cost Approach ................................... Poverty in Turkey, 1987-1994, (Cost-of-Basic-Needs Approach) ..................................

194 195 195 195 196 197

List of Appendices

Appendix 1: Appendix 2: Appendix 3: Appendix 4: Appendix 5: Appendix 6: Appendix 7:

Income and Consumption by Poverty Status and UrbadRural Dimention ........................ Poverty Figures for Various Poverty Lines ........................................................................ Sensitivity o f Poverty ......................................................................................................... Regression Analysis ........................................................................................................... Food Basket ........................................................................................................................ Estimation o f Average Household Adult Equivalent Size.................................................. General Descriptive Statistics .............................................................................................

132 142 157 161 165 171 175

This volume i s a product o f the staff o f the State Institute o f Statistics. Turkey and the International Bank for Reconstruction and Development/The World Bank. The findings. interpretations. and conclusions expressed in this paper do not necessarily reflect the views o f the Executive Directors o f the World Bank, or the govemments they represent.

T h e State Institute of Statistics and the World Bank do not guarantee the accuracy o f the data included in this work . The boundaries. colors. denominations. and other information shown on any map in this paper do not imply any judgment on the part of the State Institute o f Statistics and the World Bank concerning the legal status o f any territoty or the endorsement or acceptance o f such boundaries.

PREFACE This report represents the first time that the State Institute o f Statistics (DIE) o f Turkey and the World Bank have issued a joint report on poverty in Turkey. The findings o f the Volume One Report have been subject to thorough technical review by both institutions, and are considered to be joint findings. While this report presents a comprehensive analysis of poverty in Turkey, considerable research into the subject has already been done by the World Bank (World Bank 2000,2003), the Turkish Government (2003), the United Nations Development Programme (UNDP 2001), and academicians and scholars (see Annex I1for a review o f the literature). Under the institutional development component o f the Social Risk Mitigation Project (SRMP), the World Bank has delivered technical assistance to DIE o n poverty measurement and analysis, leading to this report. The basic data used in this report are from the official 2002 Household Budget Survey (HBS), conducted by DIE. Comparisons over time are made to the 1987 and 1994 official DIE H B S 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 (which was not conducted by DIE) are used as secondary sources in some o f the chapters. Administrative data from other Turkish Government 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. I t was not designed to be regionally representative. Thus, only qualitative data and secondary-source data are used herein to discuss the regional dimensions o f poverty in Turkey. 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 line’ methodology for Turkey as the basic measure o f poverty in the country. However, several poverty 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 published in a press release by DIE (April 13, 2004), and this report provides the underlying analysis and methodology for these figures. In 2002, 27 percent o f 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 in this 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.

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 o f the poor in Chapter Three, and considers education (Chapter Four), health (Chapter Five), labor (Chapter Six) and Social Protection (Chapter Seven).

’

The poverty line i s the minimum amount o f consumption needed for an individualor household to cover i t s basic needs for food and non-food goods.

ACKNOWLEDGEMENTS This task was managed by Jeanine Braithwaite (ECSHD) for the W o r l d Bank and by Ozlem Sarica for State Institute o f Statistics (DIE). The first volume i s a joint report o f the World Bank and DIE. The second volume i s a World Bank report. The team i s grateful for the support of Andrew Vorkink, Country Director, W o r l d Bank, and Omer Demir, President, DIE. The team benefited from comments o f peer reviewers Jeni Klugman (AFTP2), Peter Lanjouw (DECRG), and Edmundo 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 in Turkey. B o t h volumes 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 in some o f the chapters. Administrative data from other Turkish Government 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. I t was not designed to be regionally representative. Thus, only qualitative data and secondary source data are used herein to discuss the regional dimensions o f poverty in Turkey. 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. 2

This report sets out a new poverty line methodology for Turkey as the basic measure o f poverty in the country. However, several poverty 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 published in a press release by DIE (April 13, 2004), and this report provides the underlying analysis and methodology for these figures. In 2002, 27 percent o f the Turlush 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 found in Volume One). The analysis in this report refers generally to the new poverty l i n e 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 o f 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 in Turkey 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 methodology3declined from 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 o f 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 n o measurement o f the effect o f The poverty line i s the minimum amount o f consumption needed for an individual or household to cover i t s basic needs for food and non-food goods. This i s the updated poverty line for 1994, but the consumption aggregate for 1994 i s 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 f Volume One.

sustained growth on poverty. DIE i s n o w conducting surveys t o measure poverty annually, so the measurement problem should abate in the future.

B. RECENT ECONOMIC DEVELOPMENTS Despite many advantages ranging fkom i t s 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 emerging economies. Turkey has suffered from an exceptional degree of macroeconomic instability characterized by high inflation and sharp swings in business cycles. Inflation was higher and growth lower in the 1990s than in the 1980s. In this period, unsustainable fiscal policy has repeatedly put pressure on the Turkish L i r a (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.

In 2000, a crawling peg exchange rate regime was launched to rid the economy o f inflation. K e y 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 Turkeyâ&#x20AC;&#x2122;s underlying fiscal and financial sector weaknesses built up over decades o f instability and delayed reform. By 200 1, persistent doubts about the peg, and underlying fiscal instability, 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 TL. The TL immediately lost 40 percent o f its value in a single day. In response to the crisis, following the collapse o f 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 t o restore financial stability and ensure public debt sustainability; (b) a rapid restructuring o f the banking sector; (c) a public sector reform program; (d) renewed privatization; and (e) enhanced social assistance.

The Turkish economy started t o 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 o f GNP in 2003, but was easily financed by short-term capital inflows, public sector borrowing abroad and reverse currency substitution. Inflation fell t o 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 t o 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 in the f i r s t quarter o f 2003, to about 25 percent early in 2004.

However, the deterioration in global 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 t o 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, i t i s also likely to affect inflationary expectations and domestic interest rates. Higher domestic interest rates, in tum, together with the impact o f 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, if persistent, could disrupt the virtuous cycle that the economy has experienced over the last year and a half.

C. POVERTY PROFILE 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 i s 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 o n the poverty status of his or her household, through the employmenb‘inactivity nexus and the amount o f income she or he can contribute t o 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 i s 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 i s nearly 35 percent for the rural population, but only 22 percent for the urban population. The factors for high rural poverty are the same as for poverty overall. Rural areas are characterized by limited employment opportunities, and rural households where the head i s 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 t o 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. In Turkey, there i s a marked difference in the kind o f dwelling. About half the sample population lives in a house, and another 27 percent lives in apartments. While individual houses are primarily in rural areas, apartments are almost exclusively an urban phenomenon. Only 6.5 percent o f the apartment dwellers are poor, while 36 percent o f those who live in houses are poor. In terms o f land ownership, 27 percent o f 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 o f the non-poor fields. More than one-fifth o f 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 o f 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 o f 27 percent poor. Shopping patterns also vary between poor 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 i t s inequality i s high. Both consumption and income indexes indicate that inequality i s higher in urban areas than in rural areas, but not much. Other data confirm overwhelmingly that there i s a sharp East-West divide in Turkey where the Southeastern and Eastem Anatolia regions are much poorer and have sharply lower human development indicators than Westem Turkey.

D. EDUCATION In 1923, the year in which the Republic o f Turkey was founded, the adult literacy rate was approximately 10 percent. Such a l o w starting point (not uncommon for that era) gave the Republic a major cause to introduce a series o f ambitious reforms that included a move toward secular education and the adaptation of the Latin alphabet. W h i l e the schooling environment in Turkey gradually improved over time, the 1997-1998 school year marked another major leap forward in that 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 Turkeyâ&#x20AC;&#x2122;s public spending on education increased significantly after 1998, both in real terms and as a percentage o f gross domestic product (GDP). Consequently, as o f 2000, Turkeyâ&#x20AC;&#x2122;s public spending on education as a percentage o f GDP has become comparable to the spending patterns observed in countries of similar levels o f economic development. The expansion o f compulsory schooling to 8 years had an extremely positive impact on the distribution o f public spending on education across r i c h and poor households. In 2001, 21.7 percent o f public spending on basic education reached the poorest 20 percent of households (as opposed to 15.8 percent in 1994). But there i s significant room for improvement when i t comes to secondary education, since only 13 percent o f public spending reached the poorest 20 percent of the population in 200 1. Household expenditures on education strongly reinforce the disadvantageous situation o f poor children: in 2002, the wealthiest 20 percent o f households spent 6.4 times more on education than did the poorest 20 percent o f households. The current situation represents a significant improvement compared to 1994, when the wealthiest households spent 28.8 times more o n education than did the poorest households.

Inefficient Resource Allocation 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 o f vocational school classrooms increased significantly in 2002-2003 (reflecting previous investments) even though enrollment in vocational schools did not increase in recent years; and (ii) the number o f general secondary classrooms remained stagnant, even though general secondary school enrollment increased visibly in recent 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 t o 14 are either enrolled in school or have completed the basic education cycle. Parental schooling i s a very good predictor o f childrenâ&#x20AC;&#x2122;s enrollment level: about 70 percent o f the children who are not attending school (but who are o f school age) have at least one parent who has not completed primary school.

An econometric model o f enrollment in secondary schools reveals that males are 7 t o 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 10 percentage p o i n t s - o n l y 64 percent o f households reported that a secondary school i s available in their residential area. Other important correlates o f secondary school enrollment are household wealth and presence o f a mother in the household. Universities and The Poor In Turkey, entrance to universities i s primarily based on a studentâ&#x20AC;&#x2122;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 t o be enrolled in private universities and well-established institutions. Thus, (the few) students from poor households who are enrolled in universities do not enroll in universities o 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 Views About 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 o f children enrolled in public schools and 10 percent o f children enrolled in private schools. The next major problem i s poor teaching (in both public and private schools), reported in roughly 10 percent o f cases. The

urbadrural differences are pronounced-only 45 percent o f rural responses indicated “no problem with school” as opposed to 68 percent in urban areas. The wealthier are much less likely to report problems with school. Complaints about lack o f books and supplies decline rapidly with increases in household 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 Selected Labor Market Outcomes Private returns to schooling are very high in Turkey. The gender gap in earnings i s 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. If one estimates separate earning regressions for males and females, the impact o f schooling on earnings i s found t o be more pronounced for females. In other words, while females earn less on average, the variation in earnings by schooling attainment i s more significant for females when compared with males. These findings, jointly with the finding that 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 l o w 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 unemployment rate among general secondary school graduates.

E. HEALTH The Health Care System 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 in western and urban parts o f the country. Despite significant efforts, the service delivery network remains highly uneven, with major concentrations in urban areas, particularly in the western part o f the country. This skewed distribution has l e d to significant regional differences in access 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 t o those employed in the formal sector. The green card system, introduced in 1992, i s intended to provide coverage t o l o w income groups who are not cove4red otherwise. The Government soon plans t o 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 H B S 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 h a l f the population in rural areas remains without any coverage. The green card program fails to provide broad coverage to a l l those living in poverty. Thus an important share o f the lowest-income households remains without access to health insurance coverage even if they may be eligible for it, and makes significant outHCIS i s an unofficial survey which was not conducted by DIE.

of-pocket payments when seeking health care. The lack o f insurance coverage leads many l o w income households to forego health care, which in turn negatively affects their health outcomes.

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 o f the population continue to be afflicted with a high burden o f disease fi-om preventable infectious diseases, and high maternal and infant mortality rates as i s typical o f developing countries. At the same time, a growing share o f the population i s affected by non-communicable diseases prevalent in developed countries.

Access To and Use of Health Care People with health insurance, including a green card, are more likely t o seek health care when illthan those without insurance. The likelihood o f seeking care when illi s lower among the bottom two quintiles than among the upper quintiles, with the difference being particularly marked in rural areas. Low-income groups suffer fi-om significant access problems.

Determinants of Care Seeking 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. L a c k o f a facility nearby did not appear t o be a prime determinant o f not seeking care when ill. The share o f the population that had to pay for outpatient treatment, drugs, and hospitalization i s 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 Care Expenditure 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 o n health care as the lowest quintile. The largest share o f out-of-pocket expenditure o n health i s allocated to the purchase o f drugs, with payments for outpatient consultation ranlung second across a l l 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 t o the recent National Health Accounts study estimates in Turkey, Central Government funding accounts for over one-third o f Turkeyâ&#x20AC;&#x2122;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 o f 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. LABOR MARKET IN TURKEY In Turkey, as in most countries, poverty i s closely related to employment status and the type o f job, whether formal or informal. Informally employed and casual workers have a higher rate o f poverty. Education plays a key role in explaining employment and poverty outcomes. Unemployment and Labor Force Participation Unemployment in Turkey was 10.3 percent in 2002. Labor market outcomes are mostly driven by l o w levels o f 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 in labor market participation rates between men and women, with extremely l o w rates o f female labor market participation in Turkey and even decreases in the rate for females since the 1999 level o f 30 percent, down to 27.9 percent in 2002. The male labor force participation rate was 72 percent in 2002, or more than t w i c e that o f womenâ&#x20AC;&#x2122;s. Female unemployment rates have typically been slightly lower than the male unemployment rate. T h i s i s primarily because so few women are in the labor force

Unemployment and Inactivity

In 2002, 35 percent o f those aged 12 and above reported that they had worked in a paid j o b in the month o f the survey. The poverty rate o f these 35 percent was 25 percent. Another 43 percent o f those aged 12 and above who reported that they did not work had almost the same poverty rate (24 percent poor) as the employed. In terms o f unemployment, only 7.2 percent o f those aged 12 and above reported that they were looking for a job. Households with employed heads had a poverty rate o f 26 percent compared to 35 percent where the household head was unemployed. Reasons for not seeking a j o b varied f i o m factors relating to age or family structure (student, housewife, elderly) to disability or seasonal employment.

Quality of Employment In Turkey, there exists a strong association between the type o f employment and the poverty status o f the 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 o f 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 in social security. The poverty ...

Vlll

risk for people with formal jobs was the lowest for those employed by the government and in state-owned enterprises. However, i t should be emphasized that very few people have such employment. Poverty i s associated with the size o f the enterprise. People employed in larger f i r m s were less likely to be poor compared to people fkom firms with 1 to 9 people. Almost 70 percent o f the respondents (aged 12 and over) reported that they worked in a firm o f 1 to 9 people, and thus had a higher poverty rate. The mean number o f hours worked by the poor was 43.4 per week, whereas by the non-poor, i t was 46.3 hours. Also, the mean duration o f employment o f the poor in the same low-paying j o b was longer than o f the non-poor.

Sector of Employment

The largest sector o f employment in Turkey i s agriculture. Agriculture i s also the sector with the highest poverty rate o f those employed in it. O f the 35 percent aged 12 and above, 40 percent are engaged in agriculture. T h e next-highest poverty rate i s that o f construction. The poverty rate i s lowest for mining and quarrying. After agriculture, the other significant sectors in terms of employment are manufacturing, and wholesale and retail trade.

G. SOCIAL PROTECTION IN TURKEY Social protection in Turkey consists primarily o f limited formal systems in pensions and social assistance, supplemented greatly by informal mechanisms. Formal elements o f social protection are the pension (social security) system, and the Social Assistance and Solidarity Encouragement Fund (SYDTF) and i t s 93 1 affiliated Social Solidarity Foundations (SYDVs).

Turkish Pension System Turkeyâ&#x20AC;&#x2122;s social security system i s highly fragmented. Benefits and contributions depend on oneâ&#x20AC;&#x2122;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 covered by Bag-Kur (BK). Also, there are separate occupational schemes that cover various other groups. Overall, 42 percent (1 1 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) o f the population over age 65 i s receiving an old-age pension. However, almost 3 million individuals below the age o f 65 are receiving pensions, primarily old age pensions, with many of the recipients considerably younger than 65. As a result, the pension system i s showing large fiscal losses each year and i s in need o f transfers from the government to cover those losses expected to be around 4 percent o f GDP in 2004.

SSK i s by far the largest system, and covers mostly private sector employees. Employers contribute 11 percent o f 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 in the 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. T o combat the perennial problem o f evaluating income eamed, Turkey uses the system o f minimum earnings steps, which are attributed to individuals regardless o f what they actually eam.

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 then receive their retirement.

Emekli Sandigi covers civil servants, including military personnel. Financing o f Emekli Sandigi i s somewhat different from the other plans in that health insurance during working years i s not covered by the pension fund. Instead, i t i s 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 o f 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 o f the benefit. Social Assistance and Solidarity Encouragement Fund The Social Assistance and Solidarity Encouragement Fund (SYDTF) i s 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 program o f social assistance in Turkey in terms o f number o f beneficiaries.

Conditional Cash Transfers Conditional Cash Transfers (CCTs) are a national program recently introduced in Turkey, supported by a loan from the World Bank, the Social Risk Mitigation 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 M a y 2005, 1.6 million children and 7,000 pregnant women benefited from the program. Local Initiatives

Supported under the SRMP, the Government o f Turkey has undertaken a significant expansion o f the microprojects traditionally handled by the SYDVs with approval from the SYDTF, along with a tightening o f 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 o f poverty.

CHAPTER I:DATA COMPARABILITY, 1994-2002 The State Institute o f StatisticsiWorld Bank ( D I E M ) team analyzed the data from the 1994 and the 2002 Household Budget Survey (HBS) to (a) assess comparability o f consumption measures o f welfare between the two main surveys datasets, (b) establish comparable poverty lines, (c) and discem trends in poverty and inequality in Turkey between 1994 and 2002. The 1994 and 2002 HBS are the largest and the only nationally representative household surveys in Turkey for 1987-2002. The in-depth analysis has shown that the results o f these surveys are broadly comparable, and offer a solid base for analyzing dynamics o f living standards and poverty in Turkey. 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 o n the indicator o f inflation. The cost o f living in Turkey increased more rapidly than the GDP-based deflator would suggest. The use o f appropriate deflators and consumer baskets reflecting the consumption pattems o f the population reveals that living standards between 1994 and 2002 remained almost unchanged. Inequality, on the other hand, increased only marginally, remaining high by 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-based measures 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 becoming cheaper, 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 with respect to measurement assumptions.

R E A L INCOME AND

CONSUMPTION OF HOUSEHOLDS BETWEEN1994 AND 2002

The timing o f the two major surveys in Turkey-1994 and 2 0 0 2 4 i c t a t e s the frame for the comparisons. Table 1.1 shows main macroeconomic indicators o f 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; i t i s a measure o f the total output in the economy, which includes many items such as investment goods that do not make a contribution to the household welfare. Moreover, i t has no direct relevance for poverty measurement. That i s why another measure-per capita personal consumption-will be used in the table. Table 1.1 presents two measures o f 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. I t i s also more appropriate to take into account economies o f scale in the household, which are influenced by the household size.

The table shows that there i s a gap (as in all countries) between consumption measured in NA and consumption measured in the surveys (this i s due in part to different definitions, and to underreporting). However, while this gap existed in both 1994 and 2002, i t has narrowed over time: in 1994, the total consumption estimate based on the survey was around 55 percent o f the NA estimate; in 2002, it

increased t o 66 percent. That i s why the comparison o f 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. Macroeconomic and Household Survey-Based Data on Living Standards, 1994-2002

1994 GDP, current prices, million TL* Total personal consumption (from NA), million TL* Per capita GDP, current TL Per capita personal consumption, current TL

4,027,176,295 2,706,262,470 67,729,991 453 14,554

annual consumation INA method). current TL 24.917.311 Total household consumation (PA 1994 method). million TL ,I 1.500.338.527 , , , Per capita consumption (PA 1994 method), current TL 25,233,044 Consumption per equiv. adult (PA 1994 method), current TL** 40,427,172

Per caoita .~

~~~~~

Memo: Population Number o f households Average household size Average equivalent size o f household**

I !

59,459,277 13,342,055 4.456 2.71

2002

277,163,385,986 184,036,488,295 4,052,509,121 2,690,866,056

1 I

1.762.864.991 I , I 116.349.713.174 , , , 1,701,192,501 2,667,781,670 68,393,032 16,446,644 4.15848 2.58

Change,

69 68 60 59

I I

71 78 67 66 1.15 1.23 0.93 0.95

The main factor affecting direct comparability o f poverty fkom various years in Turkey i s inflation. Table 1.1 shows that consumption increased 66 times as measured by the survey. O f 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 C P I to exceed indexes with variable weights (Paasche), such as GDP deflators, especially during high 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 t o such an estimate. I t i s our preferred measure o f changes in real cost o f living in Turkey. Also worth noting i s the difference between the increase in food prices and total CPI, suggesting changes in relative prices. Exchange ratebased indexes show much slower change in prices than proper price indexes. This calls for extreme caution in using various measures o f prices for making comparisons over time.

Table 1.2. Indexes for Deflating Nominal Values: 1994 and 2002,1994 = 1.00 1994

CPI all CPI food GDP deflator GDP private consumption deflator Current exchange rate, to USD Current PPP exchange rate, to USD

1 1 1 1 29,826 12,096

2002

66.5 58.2 53.0 59.6 1,504,119 663,575

Change, Times

66.5 58.2 53.0 59.6 50.0 55.0

deflator data are from World Bank 2003a.

Based o n 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 i s extremely sensitive to the price indexes used, and that the preferred measures based o n macroeconomic data and based on surveys consistently show practically zero growth o f real consumption between 1994 and 2002.

Table 1.3. Change in Main Indicators for Real Incomes and Consumption, 1994,2002 1994

2002

Change, Times

** 1994 equivalence scale (nutrition based).

Sources: SIMA unified survey 2002, the World Bank for real GDP and gross national income (GNI) per capita data, other data

provided by DIE.

The baseline survey-based measure-per capita consumption (highlightedhincreased by only 1 percent 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 in real 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 populationâ&#x20AC;&#x2122;s living standard-private consumption per capita deflated with the GDP personal consumption deflator-shows modest growth between 1994 and 2002. But once the C P I price index i s

used, the result i s the opposite: real consumption f e l l between 1994 and 2002. The current dollar consumption shows even greater volatility. Overall, in 2002 i t appears to have increased only when compared with the through figures o f 1994 and 2001 (Figure 1.1). Figure 1.1. Consumption Per Capita with Various Deflators, 1990 - 100

-Private consumption per

capita, constant prices, private constant deflator

Private consumption per capita, current USD

Private consumption per capita, constant prices, CPI

I t i s evident that such an outcome i s driven by different trends in price indexes. Figure 1.1 illustrates this

point. For the entire period, the CPI index was above both GDP deflators.

Figure 1.2. Price Indexes in Turkey, 1990-2002, Annual Percentage Changes

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

-Inflation,

CPI

GDP private consumption deflator

pr ev ye ar

=I nn

-GDP

Deflator

Exchange rate chain index

INEQUALITY I N CONSUMPTION OF HOUSEHOLDS BETWEEN 1994 AND 2002: HBS

Inequality as measured by consumption from H B S 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 W o r l d Bank with the 1994 data (see Table 1.4 and Annex I).

Table 1.4. Inequality

Inequality, Gini for 1994 consumption per equiv. adult* Inequality, Gini for 1994 consumption per capita

1994

2002

0.385

0.390

Change, Times 1.01

0.408

0.413

1.01

c. POVERTY LINES:1994 AND 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 minimum food 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 H B S provide u s with expenditures and quantities of each food item consumed by the households. The 2002 H B S 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 o n the World Wide Web), the team also obtained the total calorie content o f these baskets. The calorie requirements (needs) of individuals may be used as a starting point for constructing food poverty lines. These requirements depend o n 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 K c a l 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 â&#x20AC;&#x153;best practiceâ&#x20AC;? 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 minimum food basket, the non-food component was estimated based on the actual consumption structure o f the poor. This part o f the methodology i s very similar t o 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. T o obtain the real value o f a 2002 basket o f food and non-food goods for 1994, components o f 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 t o 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 1994 baseline comparisons more accurate.

Table 1.5. Food Basket for Equivalent Adult: 2002 Survey-Based and Hacettepe University (kilograms per day per adult)

n.e.c. = Not elsewhere classified,

T o 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$l-a-day poverty line estimated using the 1985 purchasing power parity (PPP) exchange rate. Since, in the late 1990s, 1993 PPP exchange rates were available for a large number o f developing countries, the World Bank re-estimated the international poverty line. Local poverty lines were used for 33 countries, converted to 1993 PPP dollars. The median poverty line o f 10 countries with lowest poverty lines was calculated to be equal to US$l.OS per person per day, and i t was adopted as the new international poverty line. I t should be noted that this new poverty line i s not strictly comparable to the U S $ l -a-day poverty line in 1985 PPP dollars.

The US$1-a-day poverty line i s 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 maintain the same metabolism. Their basic needs are obviously very different from other countries, so their poverty lines should also be different. That i s why E C A 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 o f 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 i t i s not a correct application o f the World Bank methodology), one could use each yearâ&#x20AC;&#x2122;s estimate o f PPP. I t 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 in the PPP baskets o f 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 using either (a) the 1996 PPP, as done by ECAPOV, converted to current Turkish Lira (TL) using the CPI to 1994 and 2002 prices; (b) or the 1993 PPP, converted to current TL using the CPI.

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 o n methodology proposed in joint DIE/WB work o n the 1994 H B S data, and the new methodology used in this report t o 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. Poverty Lines, 1994-2002

Food basket, 1994 (Hacateppe), TLlday* Poverty line, 1994 methodology, TL/day**

38,794 69,142

2,159,424 4,712,879

56 68

Food basket, 2002 methodology. TLlday**** Poverty line, 2002 methodology, TLlday***

18,571 38,928

1,082,2 11 2,510,930

58 65

USD current PPP US$1 a day, TLlday**** USD 1985 PPP US$1 a day, CPI to update, TLlday* * * * USD 1993 PPP US$1.08 a day, CPI based, TLlday*** * USD 1996 PPP US$1.08 a day, CPI based, TLlday****

12,096

663,575

14,803

984,375

13,221

879,195

15,266

1,015,205

60% o f the median 1994 methodology cons, TL/day**

50,246

3,3 19,177

D. POVERTY IN 1994 AND 2002: HBS I t i s impossible to apply all elements o f the methodology in the same way t o 1994 as t o 2002. The consumption index developed for the 2002 H B S i s more accurate, but i t i s impossible to replicate this methodology with 1994 H B S data. The decision was therefore t o recalculate the consumption indicator in the 2002 dataset to make i t 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 i s presented in Table 1.7 by â&#x20AC;&#x153;Xâ&#x20AC;?. Table 1.7. Comparison o f Poverty in Turkey Between 1994 and 2002, with National Absolute Poverty Line POVERTY LINEMETHODOLOGY

Consumption 1994 Consumption 2002

Poverty Line 1994 PA (2000): 1994 HBS (baseline) + Est. 2002 HBS X

Poverty L i n e 2002 Est. 1994 HBS + Est. 2002 HBS X New report, only 2002 HRS

Based on comparable consumption measures (see Annex I)and comparable poverty lines, Table 1.8 presents results.

Table 1.8. Poverty Incidence with Comparable Consumption Aggregate, Equivalence Scale and Poverty Lines

National poverty lines Economic vulnerability, 1994 Complete PL Poverty, 1994 Food PL

1 t

1994 36.3% 7.3%

2002 34.5% 4.9%

1 t

Times

0.95 0.67

Poverty, 2002 complete PL Extreme poverty, 2002 food PL

28.3% 2.9%

27.0% 1.4%

0.95 0.48

World Bank international lines WDR 1990 method: 1985 PPP US$l a day Updated WB: US$1.08 at 1993 PPP a day Updated WB: US$2.15 at 1993 PPP a day

2.5% 1.7% 15.9%

3.2% 2.0% 16.8%

1.28 1.18 1.06

ECAPOV US$2.15 at 1996 PPP a day U S 4 . 3 0 at 1996 PPP a day

22.1% 6 1.O%

22.6% 60.6%

1.02 0.99

Current PPP exchange rate Poor, US$l per day Poor, US$2.15 per day Poor, US$4.30 per day

1.1% 15.3% 5 1.7%

0.2% 9.2% 38.9%

0.18 0.60 0.75

Using the 1994 methodology with the 1987 data leads to the result that 38.5 percent o f the 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. Figure 1.3. Turkey: Poverty and GDP Growth

The relative poverty line can be used to make some comparisons o f consumption poverty. The relative line i s fixed as a constant share o f 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. Using 60 percent o f the current year per equivalence adult (using the 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 o f their survey data, the table ranks all countries by the ratio o f survey-based consumption mean to the estimates o f personal consumption from national accounts. Turkey seems to occupy a unique position with respect to this indicator. The closest comparator countries, in terms o f both 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 in Turkey as a driver o f poverty. High inequality in Turkey worsens poverty considerably.

Table 1.9. Absolute Poverty Rates of Europe and Central Asia (ECAPOV methodology)

Survey Date-

I Headcount I Headcount I

Ratioof Survey

US$2.15/Day US$4.30/Day T o N A Private Cons. 22.1 60.9 0.55

1998GNP

Atlas Method 2,640**

Per Capita

2,500**

5,285"

TURKEY

1994

TURKEY

2002

Belarus

1999

10.4

0.99

2,180

6,318

Tajikistan

1999

68.3

95.8

1.02

370

1,040

22.6

60.6

0.63

5,873"

23.5

IAzerbaijan

* For Turkey, GDP data for corresponding years are used.

64.2

1.39

480

2,168

** 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 consumption data are not available for Tajikistan, Turkmenistan, and Kazakhstan. GDP per capita in current prices i s 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: â&#x20AC;&#x153;Making Transition Work for Everyone.â&#x20AC;? Turkey-estimates based on 1994 and 2002 HBS. E.

USING THE

1994 HBS AND 2002 HBS

TO DECOMPOSE CHANGES IN POVERTY

Methodology Poverty can change over time depending o n two factors. The first factor i s the magnitude o f economic growth rate; the larger the growth rate, the greater the poverty reduction. The second factor i s the distribution o f benefits o f growth; if the 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 a l l members o f society, or the share o f the poor in total consumption increases due to shifts in the distribution o f total welfare. On the other hand, poverty may rise i f either consumption falls or the distribution shifts against the poor. Often the two processes-growth and redistribution-operate simultaneously, and they reinforce each other if they w o r k in the same direction, o r weaken each other.

Decomposition of poverty changes. Changes in poverty can be decomposed into changes due to growth and changes due to an increase in inequality. T h i s decomposition i 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 0.5

-initial -new ----only

distr. distr. m e a n shift

0.4 0.3 0.2 0.1 0

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 o f the distribution-only shift in mean). Area B i s 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 o f income growth and redistribution. In the situation portrayed in Figure 1.4, the two effects reinforce each other to produce a significant reduction in the headcount poverty rate, but the same analysis can also be applied in less-favorable circumstances. Formally, we calibrate the distribution o f well-being relative to the poverty line.5 Poverty (P) i s then a function o f mean consumption and the Lorenz curve, or P(z/p,L), where z i s poverty line, p i s mean consumption, and L i s Lorenz curve, showing the distribution o f consumption among individuals. This procedure allows the change in poverty for Turkey between, say, 1994 and 2002 to be expressed as: p2002-p1994= g -k d + where the contributions o f growth (g) and redistribution (d) are defined as:

and r i s a residual due to interaction between growth and distribution.

The problem with this specification i s that g indicates the marginal effect o f the change in mean income with the distribution held constant at the initial configuration (L1994) while d computes the marginal impact o f 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 i s no logical reason for preferring one configuration over the other, symmetry arguments suggest that the two effects should be averaged to 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 expressions turn out to be the contributions associated with the level and distribution o f income in a two-way Shapley decomposition o f the change in poverty.6 We apply this, as well as a traditional decomposition, in what follows. Every poverty measure can be decomposed in a simple way to quantify the relative importance o f growth and changes in the distribution, as shown in Datt and Ravallion (1992). The Shapley decomposition i s inspired by the classic cooperative game theory problem o f dividing a p i e fairly, the Shapley solution to which assigns t o each player his o r her marginal contribution averaged over a l l possible coalitions o f agents. The reinterpretation described in Shorrocks (1999) considers the various factors (n in total, say) that together determine an indicator such as the overall level o f poverty, and assigns t o each factor the average marginal contribution taken over a l l the n! possible ways in which the factors m a y be â&#x20AC;&#x153;removedâ&#x20AC;? in sequence, The

Results: Turkey between 1994 and 2002

The annual results o f such decomposition for Turkey are presented in Table 1.10. I t i s evident that economic growth was a main driving force in poverty reduction, while inequality was acting in the opposite direction t o reduce the effect o f growth. Table 1.10 shows that poverty incidence decreased over the period by 1.77 percentage points, from 36.2 percent o f the population to 34.5 p e r ~ e n t . ~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 changed between 1994 and 2004 and i t s distribution between 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 in poverty to 38.7 percent. Table 1.10. Decomposition of Poverty Change Between 1994 and 2002 Into Growth and RedistributionComponents Percent Poor

1994 36.2%

2002 34.5%

Growth Component Redistribution Component Residual

-0.046% +0.025% +0.003%

Full (Shapley Decomposition) Growth Component Redistribution Component

-0.044% +0.026%

200211994 -0.0 18

Source: Calculated from DIE 2002 HBS.

This simple simulation shows both the dominance o f growth as a key factor in poverty changes, and the importance o f distribution, especially in the periods o f l o w growth that Turkey experienced between 1994 and 2002, when i t becomes a significant explanatory factor o f poverty dynamics. Unfortunately, overall changes in mean consumption and distribution between 1994 and 2002 were quite marginal, and thus the difference i s only marginally significant. I t should be kept in mind that 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 i t occurred.

particular attractions o f this technique are that the decomposition i s always exact, and the factors are treated symmetrically. As shown in Chapter Ion the comparison o f the 1994 HBS, baseline comparability can be assured only when 1994 poverty lines and methodology are applied to the 2002 data.

F. CONCLUSIONS The timing o f two major surveys in Turkey-1994 and 2 0 0 2 4 i c t a t e s 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 t o produce any sizable reduction in poverty, and the impact o f the little growth there was, was dampened by an increase in inequality.

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 leading emerging market economies.’ N o r 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 During the same period, Poland and Hungary made significant progress in reducing the per capita income differences with the EU. On average, the Turkish economy grew slightly under 3 percent per year over the past decade-respectable, but well below the best-performing emerging economies (Figure 11.2).

Figure 11.1. Per Capita GDP at PPS: 1991-2002 Compared to EU Average

T-

Spain

a a

.-c

3 W

Portugal

An __

Hungary

40 30

Poland

Turkey

Source: World Bank

Figure 11.2. GDP and GDP Per Capita Growth Rates in Emerging Economies (1965-2001)

8 - 6

- 4 I

2 0

.. .

Source: World Bank

This chapter draws o n “Turkey: Country Economic Memorandum-Towards Macroeconomic Stability and Sustained Growth,” July 28, 2003, W o r l d Bank Report No. 26301-TU. These figures do not reflect the informal economy, w h i c h is likely to be substantially larger as a percentage of G N P in Turkey relative t o the EU average.

A. MACROECONOMIC INSTABILITY HAS HELPED K E E P GROWTH BELOWPOTENTIAL Analysis suggests that macroeconomic instability-among many factors-has played an important role in Turkey’s inability t o realize i t s full growth potential. Cross-country comparisons and analytical w o r k 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 in human capital, and did more to improve government effectiveness and the business climate. O f a l l 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. M a n y emerging market countries have experienced large fluctuations in either 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 high growth was never sustained for long. Inflation was higher and growth lower, o n 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 f r o m 2.7 percent t o 5.5 percent. The boom-bust cycle has continued into the n e w decade, with a record contraction o f real gross national product (GNP)o f over 9.5 percent in 2001, followed by a strong recovery, with estimated growth o f 7.9 percent in 2002.

Figure 11.3. Turkey: REER and GDP Growth, 1990-2000

Turkey

S 6

Thailand

Ch*

0.5

0.15

1.25

1.5

CV: GDP (Ch%)

S o u r c e s : W o r l d B a n k , JPM o r g a n ; v o l a t i l i t y m e a s u r e d b y c o e f f i c i e n t o f v a r i a t i o n a p p l i e d to g r o w t h r a t e s o f t h e v a r i a b l e s v a r i a t i o n a p p l i e d to

B. FISCAL IMBALANCE HAS BEENTHE ROOT OF CHRONIC MACROECONOMIC INSTABILITY IN TURKEY The 2000 Country Economic Memorandum (World B a n k 2000a)” demonstrated that fiscal imbalances are key to understanding Turkey’s macroeconomic instability. Unsustainable fiscal policy has repeatedly put pressure o n the TL, fueled inflation, and undermined financial stability. Fiscal policy has been unable to act as a smoothing influence o n 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 o n macroeconomic stability has been magnified by Turkey’s open capital account and, until recently, its poorly regulated lo “Turkey: Country Economic Memorandu-Structural Report No. 20657-TU, World Bank.

Reforms for Sustainable Growth,” September 15, 2000,

banking system. Short-term capital flows have fluctuated widely as investors responded to the boombust 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 k e y to macroeconomic stability lies in sustained fiscal adjustment underpinned by credible structural reforms (Figure 11.4, data from W o r l d Bank 2003a).

Figure 11.4. Adjusted Public Sector Borrowing Requirement (YOo f GNP)

25 20

15 10 5

0 1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

C. TURKEY'S EXCHANGE RATE-BASED DISINFLATIONPROGRAM In 2000, an exchange rate-based disinflation program was launched in Turkey in a b o l d attempt t o rid the 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 b a n h n g were initiated. In fact, 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 o f 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 built up over decades o f instability and delayed reform.

Internal factors combined with unfavorable external developments started to undermine the exchange rate peg by mid-2000. On the external side, rising o i l prices and a prolonged slide in the Euro contributed to 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 t o finance their expanding domestic operations, including growing purchases o f government securities and consumer

lending. The expansion in domestic credit contributed to the consumption boom as banks quickly built up 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 t o 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 L i r a (TL) deposits remained extremely short because confidence in the TL remained fragile.

A first bout o f 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. A s banks came under increasing pressure from shrinking profit margins o n 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 board rules added to the disinflation program just prior to launch o 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 U S $ l O billion in financial assistance. This additional financing was conditioned o n the Government’s commitment to strengthen the p r o g r a m in particular, to accelerate financial sector restructuring and privatization. The Government introduced an 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 f these events, interest rates declined and a precarious degree of financial stability returned, but this proved to be a temporary respite. In early 2001, persistent doubts about the peg and underlying fiscal sustainability led t o 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 L i r a on February 21, 2001. The Lira immediately lost 40 percent o f i t s value in a single day.

D. EXCHANGE RATE-BASED DISINFLATIONPROGRAMS ARE VULNERABLE I N A 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 o n a downward long-term trend in chronic-inflation countries-have often been associated with currency crises. M a n y 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 h a l f 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 o f globalization. Exchange rate-based stabilization also generally results in sizable 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.

E. THEGOVERNMENT RESPONDED QUICKLY TO THE 2001 CRISIS 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 k e y structural and social elements o f the program were: (a) a macroeconomic framework designed t o 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 o n large resource transfers from the budget; (c) a more ambitious program o f public sector reforms centered on deeper structural and institutional reforms t o improve fiscal management and public governance; (d) a renewed privatization drive-in combination with further liberalization measures focused o n 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 RESPONSE PROGRAM INCORPORATED THE EXPERIENCE OF OTHER 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 o f bank restructuring was launched, backed by extensive fiscal resources. Bank restructuring was complemented by additional structural reforms in the financial sector designed t o further strengthen prudential regulation, adopt internationally accepted financial reporting standards and practices, and enforce compliance. In parallel 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 placed within a mediumterm programmatic framework in an effort to bolster investor confidence by demonstrating the Government’s intent t o 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 i t s 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 restructuring were bome by the budget. The price spike following the initial devaluation in February 200 1 was contained, but inflationary pressures persisted, with inflation reaching 69 percent by the end o f the year (Table 11.1). Following the decision t o 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 w e l l above the projected program path throughout 200 1, mainly due to the need to r o l l over large amounts o f short-term public debt in the face o f 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 in the recession was the sharp turnaround in the 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 high real 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

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 Economic Indicators Actual 21 2000 2001

1999

iMAIN M A C R O INDICATORS G N P Growth C P I Inflation (Dec-Dec) N o m i n a l Interest Rate Real ex ante Interest Rate 4/ Unemployment Rate U n i t Wage Index (1997 = 100)

2002

Est. 21 2003

Prop. 3 1 2004

-6.1 68.8 106.2 32.0 7.7 113.7

6.3 39.0 38.0 -9.5 6.5 103.0

-9.5 68.5 99.1 35.5 8.4 73.9

7.9 29.7 63.5 30.3 10.3 73.1

5.9 18.4 44.1 28.6 10.5 86.3

5.1 12.1 29.1 11.:

P U B L I C SECTOR Primary Balance (% G N P ) Overall D e f i c i t (% G N P ) N e t Public Debt (% G N P ) I / o f which net external debt (% G N P ) Privatization ($ bn)

-0.2 22.3 61.0 20.1 0.1

2.7 18.9 58.3 19.0 3.3

5.5 21.1 93.9 37.7 2.8

4.1 12.1 78.8 32.1 0.5

6.1 9.9 70.1 22.2 0.3

6.: 6.1 65.' 20.' 3.1

EXTERNALBALANCE Current account balance (% G N P ) Exports (fob, S bn) 51 Tourism (%bn) External Debt (% G N P ) C B T Foreign Exchange Reserves ($ bn)

-0.7 28.8 5.2 55.0 24.3

-4.9 30.7 7.6 59.0 23.2

2.4 34.4 8.1 79.0 19.8

-0.8 40.1 8.5 71.3 28.1

-3.4 51.2 13.2 61.1 35.2

-3.4 55. 9.8 48. 32.

G N P ( T L quadrillion) 78.3 125.6 176.5 275.0 356.7 I/I M F , includes government securities issued to recapitalize SDIF and state banks. 2/ Government figures as adjusted b y I M F and W B estimates. 2002 and 2003 G S P figures are as announced b y S I S in A p r i l 2004. 31 Updated as o f t h e 7th I M F review. 4/ Average o f monthly nominal interest rate divided b y 12-month ahead C P I inflation. 5 / Includes shuttle trade. Sources: Government, IM F and W B estimates.

415.

Figure 11.5. Total Net Public Debt (YOo f GNP)

100

"^"-

" ^ - ^

Cost o f rertructurmg o f state banks

60 40 20

1994

1995

1996 ~-

1997

1998

1999

2000

2001

2002

2003

STRONG RECOVERY HAS BEENUNDER WAY SINCE EARLY 2002

Economic activity rebounded strongly in 2002 and the recovery continued into 2003. Real GNP growth reached 7.9 percent in 2002, exceeding program projections by a wide margin (Figure 11.6). The recovery was led by robust export performance and exceptionally large inventory rebuilding in the first h a l f o f the year. The recovery was further buoyed by a sharp rise in public 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 in the 2002 growth accounting was offset by even faster growth in imports as the recovery gained steam. Importantly, private consumption and investment led the 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 i s that the recovery has been export led, with exports in US. dollar terms increasing by some 30 percent in 2003. 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 o f about 1 percent o f GNP in 2002 was easily financed. The extent o f the recovery, and i t s basis in export growth, place Turkey squarely in the 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 o f 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 i t s crisis lows. The strengthening TL did not hurt export performance because i t was counterbalanced by a very sharp drop in real wages. M o r e problematic were slippages in the fiscal program during the run up to the November elections. A gap o f about 2.5 percent o f 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

c i v i l 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 o f net public debt fell t o an estimated 80 percent o f GNP by the end o f 2002, helped by the rebound in the 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 t o weaken investor confidence in mid-December. Ad hoc increases in pensions in early January and other populist measures raised concems about the Govemment's political will to implement tough economic reforms. The average yield o f the benchmark government paper moved up t o the 60 percent range, and the L i r a came under some pressure. Financial market volatility continued during the first quarter o f 2003. However, the winding down o f hostilities in Iraq and approval by the U.S. Congress in early April o f a scaled-down assistance package for Turkey (a grant equivalent o f US$l billion, potentially convertible to up to US$8.5 billion in loans) eased some o f the market tension once again.

H. MACROECONOMIC OUTCOMES IN 2003 WERE 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 i t s weight in the 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 continued to play an important role in the recovery. The current account deficit widened t o 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 in 2003, and the latest data suggest that inflation i s falling toward the important singledigit level for the first time since the 1970s. Employment declined in 2003 following public and private sector restructuring, which, together with three years o f 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 s h n k a g e in the labor force. In urban 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 needed to strengthen j o b creation.

I.TURKEYHAS ACHIEVED SIZABLE FISCAL ADJUSTMENT Strong fiscal performance has been the comerstone o f the economic program. Fiscal gains were significant in 2003, and the primary surplus rose from 4 percent o f GNP in 2002 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 in December 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. In addition, 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 t o 5 percentage point adjustment in the public debt-toGNP ratio. The real exchange rate path i s given in Table 11.1,

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 o n 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 o n ad hoc, short-term measures, and a focus o n sustainable fiscal adjustment. Monetary policy followed a policy o f implicit inflation targeting, with the Central Bank occasionally intervening in the foreign exchange market t o dampen what was deemed t o be excessive fluctuation in the exchange rate. The decline in inflation, which was aided by the strength o f the TL, l e d to a commensurate decline in interest rates from a nominal 60 percent in the f i r s t quarter o f 2003, to about 25 percent early in 2004.

On the external front, despite appreciation o f the TL, rising productivity 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 o f the increase in imports during 2002 and 2003. Consumption goods also rose sharply in the second h a l f o f 2003, and continued in early 2004. M u c h o f the increase in consumption goods imports was driven by automobile imports, which benefited from a temporary tax credit o n automobile purchases introduced in August 2003. Tourism receipts were maintained in 2003 despite the Istanbul bombings and uncertainty caused by the Iraq war. The current account deficit increased t o 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 o f goods and services imports. The combination o f strong real and financial market performance had a favorable impact o n the public debt burden. The net public debt-to-GNP ratio fell sharply during 2002-2003 from i t s 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, Turkeyâ&#x20AC;&#x2122;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 n o problem servicing the debt. Nevertheless, the high rollover rate, 88 percent in 2003, indicates the continued dependence on market sentiment, and thus vulnerability to external developments.

J. DETERIORATION IN EXTERNAL BALANCE HAS CREATED VOLATILITY I N DOMESTIC FINANCIAL INDICATORS The deterioration in global financial indicators in early M a y 2004, combined with the higher-thanexpected current account deficit figures, has l e d to a sharp weakening in domestic financial indicators. While the TL depreciated about 14 percent, taking i t to i t s 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 high o f 30

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 i s an adjusting factor to the deteriorating current account balance, it i s also likely t o 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 o f 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 ARE EXPECTED To STIMULATE

GROWTH

In response t o 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 o n 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 in the 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 o f meeting the inflation target o f 12 percent, which would outperform targets for the third year in succession. Medium-term projections demonstrate that sustained implementation o f economic reform i s necessary i f Turkey i s to attain i t s macroeconomic stability and growth objectives. Under the structural reform program, the economy i s 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 in productivity 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 in 2003 to 5 percent o f GNP in 2006.

CHAPTER 111: POVERTY PROFILE In 2002, 27 percent o f the Turkish population was poor, as determined by the complete poverty l i n e methodology detailed in Annex One. This poverty profile i s based on that definition o f poverty. Poverty in Turkey i s strongly associated with age and household composition, where children and families with children are poorer than average. This association i s 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 o f poverty are standard, but s t i l l 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 o f the well off, including the ownership o f many assets and consumer durables and many housing attributes.

A. HOUSEHOLD SIZE AND COMPOSITION Poverty increases somewhat monotonically with additional household members (Table 111.1,Figure 111.l), starting at three members. There i s a small â&#x20AC;&#x153;blipâ&#x20AC;? whereby households comprised o f one member or two members are slightly poorer than those with three, but there are few such households in Turkey-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

Total

47.5

52.5

3.1

2,100,168

3.1

1o+

55.9

44.1

5.8

4,022,830

5.9

27.0

73.0

100.0

68,393,031

100.0

Figure 111.1. Turkey: Poverty and Household Size

910+

Household Size

Larger households are poorer than smaller households, and this result i s 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 in Tables 111.2, III.3, 111.4, and 111.5 and Figures 111.2, 111.3, 111.4, and 111.5, respectively). Households with n o 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 i s one o f two parents and two children, and was used in the poverty line methodologyhouseholds with two children have a poverty rate slightly below average. Table III.2. Turkey: Poverty and Number of Children in Household

* The percent o f those who answered the question or for whom we have data. Notes: Poverty in percentages. Population Subtotals: Weighted number o f observations.

Source: DIE 2002 HBS.

Figure 111.2. Turkey: Poverty and Number o f Children

Poor +Total Poor

Household Size

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 risk o f poverty, but not as much as having three or more children.

Table 111.3. Turkey: Age Structure o f HBS and Census

45-49 50-54 55-59 60-64 65+ Subtotal

2,035,002 1,571,553 495,889 401,241 1,280,363 33,307,408

1,978,808 1,786,466 1,8 10,571 1,615,043 3,533,030 35,085,627

3.0 2.3 0.7 0.6 1.9 48.7

2.9 2.6 2.6 2.4 5.2 51.3

Number of Elderly in Household 0 1

Poor 25.9 28.0

Non-Poor 74.1 72.0

Percent o f Valid* 73.7 17.1

Population Subtotals 50,394,138 11,684,961

33.7 27.0

66.3 73.0

9.2 100.0

6,3 13,935 68,393,034

2+ Total

The percent o f those who answered the question or for whom we have data.

Percent o f Total 73.7 17.1

9.2 100.0

Notes: Poverty in percentages.

Population Subtotals: Weighted number o f observations.

Source: DIE 2002 HBS.

Figure 111.3. Turkey: Poverty and Number of Elderly 40.0 35.0

__llll.__.-lll._l

30.0 -

25.0

20.0 -

15.0

Poor --t Total

Poor I I

10.0 5.0 0.0

The finding that having children increases poverty risk more than having elderly members i s 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 o f the population lives in such households, while 66 percent o f households have children but n o elderly, and only 4 percent have elderly members but n o children. Households with n o dependents are rarely poor (only a 12 percent poverty rate), while households with b o t h 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 i s comprised o f families with children (Figure 111.4).

Households with N o Elderly and N o Children No children, no elderly Children. no elderlv Elderly, no children

Percent Poor Non-Poor o f Valid* 11.9 88.1 12.3 28.5 71.5 65.7

18.2 37.6 27.0

Both children and elderly Total

81.8 62.4 73.0

4.4 17.6 100.0

Population Subtotals 7,915,326 42.478.812

Percent of Total 11.6 62.1

2,865,057 11,345,848 68,393,033

4.2 16.6 100.0

The percent o f those who answered the question or for whom we have data.

Notes: Poverty in percentages.

Population Subtotals: Weighted number of observations.

Source: DIE 2002 HBS.

Figure 111.4. Turkey: Share of Poor in Percent by Household Consumption

5.3

No Children, No Bderly

2.9

Children, No Bderly 0 Elderly, No Children [3 Both Children &

Bderly

In Turkey, poverty i s 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 i s the prime age bracket for adults t o have children.

Table III.6. Turkey: Poverty Rate o f Age

Age 45-49 50-54 55-59 60-64 65+ Total

Poor Non-Poor Percent o f Valid* 20.5 79.5 5.9 18.1 81.9 4.9 21.5 78.5 3.4 19.9 80.1 2.9 26.6 73.4 7.0 27.0 73.0 100.0

Population Subtotals 4,013,810 3,358,019 2,306,460 2,016,284 4,813,393 68,393,035

Percent o f Total 5.9 4.9 3.4 2.9 7.0 100.0

Figure 111.5. Turkey: Poverty Rate and Age

40.0 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0

+Poor l-Total

Poor

B. HOUSEHOLD HEAD CHARACTERISTICS The household head has an enormous impact o n the poverty status o f her or his household, through the employmenthnactivity nexus and the amount o f income she or he can contribute to the household. Employment and labor market variables are covered in the chapter o n labor (but it bears mentioning that the poverty rate for unemployed heads was 35.4 percent compared with total (average) poverty o f 27 percent), while education i s treated below. Other household demographic characteristics that are associated with poverty include whether the household i 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 w o m e n - o n l y 6.5 percent o f the sample population lives in a female-headed household.

Figure 111.6. Turkey: Poverty and Gender o f Head

25 20 Z 15 n 10 5 0 c

=Poor i1 +Total

Poor __

Male

Female

The education o f household head i s an even more significant influence o n household poverty than gender or unemployment for the illiterate and those who did not receive a primary school diploma (Figure 111.7), with relative r i s k 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 o f 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 o r doctorates were poor-but less than 1 percent 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 with these sorts o f heads were only 9 vocational education was a bulwark against poverty-households percent and 8 percent poor, respectively.

Figure 111.7. Turkey: Poverty and Education of Household Head

Poor +Total

Poor

Turkey recently changed the length o f mandatory education f r o m 5 to 8 years, so very few household heads have completed 8 years, because this would make these heads quite young.

C. SPATIAL CHARACTERISTICS The 2002 H B S sample was not designed to be regionally representative o f Turkeyâ&#x20AC;&#x2122;s seven regions, and therefore n o cross-tabulations are presented on the regional level. The 2003 survey, which was recently completed, was designed t o be regionally representative. When the 2003 data are cleaned and available, DIE and the World Bank intend to 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 much poorer than other regions o f Turkey, and Marmara being the 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 in the 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 in urban areas (three-fifths o f the sample population). Figure 111.8. Turkey: Poverty and Location

40.0

, ~

35.0

30.0

E 25.0 1 ~

2 20.0

15.0 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 in rural 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 i s 4.30 members, whereas in urban areas i t 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).

Figure 111.9. Turkey: Household Size and Location

Total

ILkban

Non-Fbor

Wral

Fbor 0

Number of Members

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 h e a d s - o n l y 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 i s 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 reported their 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 in rural areas (only 6 percent o f the population) compared to 14.5 percent o f the urban population. 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 o n labor, poverty rates for those employed in agnculture are sharply above those for other, urban sectors. In 2001, the Government introduced a Direct Income Support (DIS) program for agriculture, which involved a payment o f about US$90 per hectare to farmers in compensation for the removal o f agricultural subsidies. DIS was estimated to cover about h a l f the income loss t o 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 H B S 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). In rural areas, 69 percent of households produce their own food, but the differences in poverty rates among those that do (36 percent poor) and those that do not (3 1 percent poor) are not as pronounced as for urban households.

Figure 111.10. Rural Employment by Sector

Agriculture, huntinq, and

Manufacturing I3Construction

I 9.1

Wholesale and retail trade W Remainder

Figure 111.11. Urban Employment by Sector

Agriculture, hunting, and forestry Manufacturing I

9.7

0 Construction

ElWholesale and retail trade I Remainder

22.3

Education appears to have identical effects in both urban and rural settings, whereby those who are illiterate or limited t o primary school have sharply higher poverty rates than average, and higher education graduates are much less likely to be poor (Figure 111.12). In both areas, poverty rates steadily decrease as years o f education increase, and always, at every education level, the poverty risk premium o f rural location persists until the very highest levels for which there are essentially n o observations in rural areas.

Figure 111.12. Rural Employment by Sector 50 0 45 0

40 0 35 0 u

2 Q n

300 250 ZOO

,+Rural

Pwr

p u r a 1 Total Urban Pwr

"mwU*an

Total

15 0

10 0 50 00

Nonmaterial aspects o f poverty also confirm that poverty i s 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 OF POVERTY Poverty restricts the poor f r o m accessing many goods and services. The 2002 HBS collected some information o n 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 o n assets that uniformly shows that assets are distributed to the non-poor, which, o f course, i s t o 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. In Turkey, following o n f r o m the spatial differences, there i s a pronounced difference in the kind o f dwelling. About h a l f the sample population lives in a house, and another 27 percent lives in apartments. However, while individual houses are primarily in rural areas, they are also s t i l l 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 o f those who live in houses 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 a l l gecekondu dwellings are houses.

Figure 111.13. Turkey: Poverty and Dwelling 40

- __

-___.

______l____l___

35 30 E

Poor

20 15 10

--t Total

0 â&#x20AC;&#x2122;

Poverty

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 located in gecekondu areas. For example, 7.7 percent o f 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 o f the population that reported their heating source was a stove had a poverty risk o f 3 1 percent, which i s more than the total poverty rate o f 27 percent. Electricity i s virtually universal in Turkey; only 0.1 percent o f the sample population reported that they lived without it-but all o f these were poor (too few observations for reliable inference). Only 4 percent o f those surveyed reported that they owned an additional home, apartment, or summer resort, or that their home was a â&#x20AC;&#x153;luxuryâ&#x20AC;? one, but o f these, very few were poor, with average poverty rates in the single digits. Conversely, the use o f dung for heating was strongly correlated with poverty (Figure 111.14).

Figure 111.14. Turkey: Poverty and Use o f Dung

--!

$ 2

45 -40 35 -

! I I

, I

20 15 10 5 -

-Poor -

Yes

Land i s an important asset in Turkey, particularly for the poor. Twenty-seven percent o f the sample population reported that they had a field, but these households were poorer than average, with a poverty rate o f 34 percent. The mean size o f the fields for poor households was 75 percent the size o f the nonpoor fields. More than one-fifth (23 percent) o f the sample population reported owning a car, but only 6 percent o f these were poor. Other consumer durables showed similar results.

Table 111.7. Turkey: Poverty Rate of Consumer Durables

Water heater Aspirator

11.9 5.5

88.1 94.5

49.1 22.7

33,594,478 15,535,335

49.1 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 medication regularly are all lower, and often substantially lower, than the overall poverty rate o f 27 percent (Figure 111.15). However, the participation in each o f these activities vanes greatly, from a high o f 67 percent o f the sample population reporting that someone smokes, to a l o w o f 1.6 percent for nursery schoolkindergarten participation. Fifteen percent o f households reported use o f public transportation, but more non-poor than poor accessed this-the poverty rate for usage was 17 percent compared to the total poverty rate o f 27 percent.

Figure 111.15. Turkey: Poverty and Discretionary Activities

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 c a r d - o n l y 2 1 percent o f households reported using a credit card, but o f those, the poverty rate was only 5 percent.

Figure 111.16. Turkey: Poverty and Shopping 40

"_-- .

2 20

I-----

7 A

n. 15 10

-Poor +Total

Poor

5 -

0 -

i E.

MarketYes

Market--no

BazaarYes

Bazaar-no

INEQUALITY AND 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 t o the 2002 data, as were used by the World Bank with the 1994 data (see Annex I).

Table 111.8. Comparable Inequality ~

Inequality, Gini for 1994 consumption per equiv. adult* Inequality, Gini for 1994 consumption per ~~

1994 0.385

2002 0.390

Change, Times 1.01

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 in consumption o f the o l d methodology in terms o f handling o f consumer durables (Table 111.9). The new methodology shows n o change in income inequality, but a lower level o f consumption inequality. This lower level o f consumption inequality i s not due to any improvement in the distribution, but i s simply the result o f the new methodology.

Table 111.9. New Methodology Inequality Measures

B o t h 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 i s a middle-income country, and i t s inequality i s high. The worst inequality i s in the poorest countries (Milanovic, 200 l), 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 high income inequality).

In previous W o r l d Bank research and the Turkish literature, the major driver for Turkey’s high level o f 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 during the past 30 years.

This finding i s corroborated by every major study o n 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 in its “Human Development Reports,” the Turkish academic literature (for example, Ogut and Barbaros undated), reports f r o m the State Planning Organization, and numerous qualitative assessments undertaken for W o r l d Bank projects in Turkey 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. In a l l these, much greater poverty in Eastern 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

indicators by region. Note that in Figure 111.18, the index o f socio-economic development used i s lowest for the best-off regions and highest for the worst-off.

Table 111.10. Per Capita Gross Domestic Product, by Statistical Regions, 2001* Statistical Regions

US%

Turkey

2,146.00

Source: DIE.

*At Current Prices Ranked in Descending Order

Table III.ll. Turkey: Social and Human Development Indicators 1997 ~

Under-weight Adult Children Literacy Under 5 (YO) Rate (YO)

Life Expectancy at Birth (years)

Southeastem Anatolia Eastem Anatolia

71.59 71.53 67.22 65.52 66.63 65.89 62.05

Turkey

68.90

Marmara Aegean Black Sea Central Anatolia

IMediterranean

6.20 7.25 9.29 10.54 11.29 17.00 19.00

10.00

89.29 85.16 81.45 84.60 84.28 62.34 70.99

83.20

F. MULTIVARIATE ANALYSIS The multivariate analysis confirms almost a l l o f the findings o f the bivariate analysis presented above in poverty cross-tabulations. A probit analysis demonstrates that the probability o f 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 o f the household being poor. The coefficients themselves represent the percent change in poverty likelihood. For example, rural location increases the chance that the household i s poor by 9 percent, controlling for the other factors.

Table 111.12. Probit Estimates Number o f obs =

Log l i k e l i h o o d = - 4 0 4 1 . 3 8 5 8

3,555

LR c h i 2 ( 1 5 ) Prob > c h i 2 Pseudo R2

=1821.66 = 0.0000 = 0.1839

....._........__________________________~-

dP/dx

hunemp*/ dhigher*/ dprimar*/ dsocsec*/ dsecjob*/ elder65 I adult child unpaid*l rural*/ femalhd*/ self*/ employer*/ casual*/ temporar*i

,068221 -.1346925 ,0938267 -.1123116 -.0054283 ,0077759 -.0250377 ,0667864 ,0049726 ,096574 -.0043401 .012003 -.lo4834 ,1033531 .0436523

POcPla

Std. Err.

P>lz/

x-bar

0.000 0.000 0.000 0.000

,083203 ,090215 ,640502 ,428362 ,045107 ,223653 2.45313 1.57415 ,106541 .153218 .lo0576 ,241444 ,052747 ,084458 ,078074

95% C . I .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 1

.0169959 ,0125338 .0089331 .0086383 .0131209 ,0078997 ,00369 ,0026853 .0158795 .0143081 .0133545 ,0111668 ,0134159 ,0168735 ,0153984

4.37 -6.53 3.82 -12.52 -0.28

0.98

0.779 0.325

0.000 0.000

-6.78 25.73 0.32 7.42 -0.32 1.03

0.747 0.277

-5.58 6.88

0.000 0.000

2.90

0.004

0.752

0.000

.03431 -.153258 ,076201 -.123361 -.042905 -.007707 -.03227 ,061523 -.026151 ,068531 -.030514 -.009878 -.131123 .070282 ,012297

,101532 -.110127 ,111453 -.095262 ,032047 ,023259 -.017805 .072049 ,036096 .124617 ,021834 .033896 -.078539 .136425 .075003

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

obs. P pred. P

1 ~

,2132915 ,1648576

(at x-bar)

..............................................................................

(*)

d F j d x i s f o r d i s c r e t e change o f dummy v a r i a b l e f r o m 0 t o 1 z and P > / z i a r e t h e t e s t o f t h e u n d e r l y i n g c o e f f i c i e n t b e i n g 0

The greatest impact o n 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 o f dummy variables relating t o 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 o f employer reduces the risk o f poverty by 10 percent. Factors that increase the probability o f poverty, other than rural location, include the number o f children in the family (each additional child increases the chance o f poverty by nearly 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 in number.

Figure 111.17. P e r Capita GDP Index

Nomenclature of S t a t l s t l c a l T e r r i t o r i a l U n l t s l l Regions

0 1987-1989AW3'.AGES

1990- 1 9 9 4 A m G E S

___-___

019952000A~GES

Figure 111.18. Socioeconomic Development Levels (2003)

Nomenclature of Statistical Territorial Units II Regions

Figure 111.19. Female Literacy Rates (2000)

Nomenclature of Statistical Territorial Units II Regions

CHAPTER IV: EDUCATION A. THECONTEXT In 1923, the year in which the Republic o f Turkey was founded, the adult literacy rate was approximately 10 percent. Such a l o w starting point in terms o f human capital (not uncommon for that era) i s certainly undesirable, but this allowed the founders o f the Republic to carry out major reforms that might otherwise have been more difficult to implement. Indeed, the key features o f the Turkish education system were established in the early years o f the Republic. The Tevhid-i Tedrisat law o f 1924 marked 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 (in rural area schools) to five years.13 The idea of 8-year primary education first emerged in 1970, with the establishment o f a working group to formulate the specifics o f the proposed changes. This led to a 1973 law (1973 sayili M i l l i Egitim Temel Kanunu), which declared the duration o f basic education to be 8 years. In practice, however, the current structure o f 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 o f the 8-year primary school cycle, students may enroll in general or vocational secondary schools, which i s three years in duration. Tertiary schooling i s provided by universities. A key implication o f 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 lowersecondary parts o f vocational schools that offered religious training and other lower-secondary schools had to be closed. The reform, however, also resulted in a renewed public and private commitment to the achievement o f universal enrollment levels in basic education. On the 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 o f supporting the schooling o f poor children. Enrollment rates increased soon after the reform. During the 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 in the 1997-1998 school year, enrollments in primary schools increased from 9.1 million students to over 10.3 million students in the 2001-2002 school year. A substantial school and classroom construction effort dominated this period, and the number o f 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 I C T and improvement o f the teachers’ competencies especially in the 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 o f schools, students, and teachers by education level and urbadrural residence are depicted in Table I V . ~ . ’ ~

l3 A

detailed historical overview can b e found in Turkiye I s Bankasi (1999). comprehensive review o f the events and issues that characterized the Turkish education system in the 1990s until 2004 can be found in Dulger (2004). l 5 Extracted f r o m TC Milli E g i t i m Bakanligi (2001). l4 A

T a b l e IV.1. N u m b e r o f Schools, Students, and Teachers in t h e 2001-2002 School Y e a r

Public Schools Number o f Schools

Schooling Level

Total

Urban

Rural

PrePrimary

10,554

7,361

Primary

34,993

9,906

Secondary Total General Vocational

6,065 2,637 3,428

5,463 2,276 3,187

Number o f Students

Number of Teachers

Total

Urban

3,193

256,392

216,625

39,767

25,087

10,310,844

7,500,373

2,810,471

602 361 241

2,312,271 1,490,376 821,895

2,212,658 1,427,845 784,813

Rural

99,613 62,531 37,082

Total

Urban

Rural

14,520

12,579

1,941

375,511

256,272

119,239

138,785 72,609 66,176

131,482 68,092 63,390

7,303 4,517 2,786

University

1,156,9 15

63,029

Schooling Level

Number o f Schools

Number of Students

Number o f Teachers

PrePrima

799

18,152

1,822

642

171,623

14,811

487 465 22

73,136 72,05 1 1,085

8,229 8,128 101

55,022

3,721

Secondary Total General Vocational University

The Turkish Government originally planned to increase compulsory schooling once again, from 8 to 12 years, fairly quickly. This policy proposal seems to be on hold, although i t 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 in recent years. I t also identifies key problem areas that emerge as challenges to the schooling o f poor children. Finally, the findings of this study w i l l serve as inputs into the ongoing Turkey Education Sector Study titled, â&#x20AC;&#x153;Sustainable 16

See, for example, TC Milli Egitim Bakanligi (2001).

Pathways to an Equitable, Effective, and Efficient Education System,” which aims t o build a technical foundation for policy analysis, and provide an open forum for a technically based dialogue in Turkey.

B. PUBLIC SPENDING ON EDUCATION How Does Turkey Compare with Other Countries? Turkey’s public spending o n education significantly increased after 1998, both in real terms and as a percentage o f gross domestic product (GDP) (Figure IV.1). However, even after such a dramatic increase, as o f year 2000, Turkey’s public spending o n 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).”

Figure IV.l. Public Spending on Education as Percentage o f Gross Domestic Product

2 1990

1991

1992

1993

1994

1995

1998

1999

Yea, I

- 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

Source: World Development Indicators (WDI). Years 1996 and 1997 are not included since the WDI data set does not provide spending figures for those two years. Distribution of Public Resources Across Different Levels of Schooling In terms o f distribution o f public expenditure on education across different levels, the year 2000 figures from the World Bank’s EdStats data base suggest an emphasis o n 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 3 1 percent goes to tertiary education. The low share o f spending o n the secondary level might signal a problem, and this i s a topic that will be examined in depth by the forthcoming Turkey Education Sector study which will focus o n spending trends in more recent years.

” The total spending o n education in Turkey i s estimated as 7.31 percent o f GDP, by the “Research on Turkey’s Spending on Education in 2002” report published by the State Institute o f Statistics in 2003. T h e same document reveals that public spending on education comprised o f 63.5 percent o f the total (i.e., 4.64 percent o f GDP in 2002).

Incidence ofpublic Spending on Education

The expansion o f compulsory schooling to 8 years had an extremely positive impact on the distribution o f public education spending across poor and r i c h households (Table IV.2). In 1994, only 15.8 percent o f public spending o n basic education reached the poorest 20 percent o f households. In 2001, 21.7 percent of public spending on basic education reached the poorest 20 percent o f households, because the enrollment rates o f children coming from poor households increased substantially. Table IV.2 also shows that while some pro-poor redistribution occurred in the distribution o f 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 o f population in 2001 (up from 8.7 percent in 1994). These estimates o f incidence o f 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 funding as the public schools attended by wealthy students. If public schools with higher poor-student presence receive less funding on average, then this means the figures reported by Table IV.2 overestimate the amount o f resources that 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 o f 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 2001 Household Consumption and Income Survey (HCIS). Table IV.2. Incidence of Public Spending on Education in 1994 (before the expansion o f compulsory schooling) and in 2001 (after the expansion o f compulsory schooling)

Basic education (8 years, primary + middle) Secondary education Total public expenditures

Primary Secondary Tertiary

Basic education (8 years) Secondary education Total public expenditures

1 (poorest)

5 (richest)

15.8% 8.7% 13.5%

21.1% 16.2% 19.5%

22.2% 22.3% 22.2%

20.6% 25.4% 22.2%

20.3% 27.5% 22.7%

25.4 14.0 4.5

22.4 17.4 10.0

20.0 21.3 14.4

18.4 23.3 25.5

13.7 24.0 45.5

2 1.7% 13.0% 19.2%

2 1.4% 14.6% 19.4%

21.0% 25.4% 22.3%

22.0% 22.8% 22.2%

13.9% 24.2% 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 compared with Turkey. As a group, however, they provide some insights into the general circumstances in lowermiddle-income countries.

In order to put the outcome o f the incidence analysis in context, the middle segment o f Table 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 o f public funding became more in line with the experience o f other countries at about the same level o f economic development. Incidence of Household Spending on Education

The distribution o f household expenditures on education i s 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 in basic 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 in the schooling environment experienced by poor children.

Household Income Quintiles

1994

znni

1 (poorest) 2.2% 6.2%

7.1% 14.1%

9.4% 16.6%

18.0% 23.5%

5 (richest) 63.3% 39.6%

c. TRACES OF INEFFICIENT RESOURCE ALLOCATIONI N THE EDUCATION SECTOR A full discussion o f inefficiencies in the education sector i s beyond the scope o f 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 l e f t 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 in schools, and under what conditions.

Classroom Size and Student-Per-Teacher Ratios in Primary, General, and Vocational Secondary Schools

In recent years, the number o f primary school students has increased significantly (Figure IV.2). The number o f students enrolled in general secondary schools increased slightly and enrollments in vocational secondary schools has remained about the same. In response to (and to enable) rising primary school enrollments, the number o f primary school classrooms increased sharply after 1998 (Figure IV.3).

Figure IV.2. Number of Students by School Type 12000000

4000000

2000000 I

-B b

1997-1998

1998-1999

I9942000

2000-2001

2001-2002

2002-2003

School Year I

+Primary

-General

high school

-Vocational

high school

Source. Ministry o f National Education Statistical Yearbooks.

Figure IV.3. Number of Classrooms Over Time -~

300000 ___I

01997-1998

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 o f National Education Statistical Yearbooks.

Starting with the positive development that the increase in the number o f classrooms at the primary cycle has been impressive in recent years, this has prevented increases in students per class at this level despite increased enrollments due to the expansion o f compulsory schooling. In fact, the number o f students per classroom at the primary level has declined from 43.2 in the 1997-1998 school year to 36.9 in the 20022003 school year.

The cohorts affected the most by the expansion o f 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 â&#x20AC;&#x153;openâ&#x20AC;? general secondary enrollments). As a result, average classroom size increased steadily after 1998, and peaked at 33 in the 2002-2003 school year. In contrast, 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 in the 2002-2003 school year, compared to more than 20 for general secondary schools. As a result, Figures IV.4 and I V . 5 reveal misallocation o f scarce resources at the secondary level, but they also show very encouraging results at the primary level.

Figure IV.4. Number of Students P e r Classroom b y School Type

1997-1998

iwa1999

19942030

_ _ -General ~ -Pnmaw

2m2001

~ hylh school

2001-2002

2002-2003

SchW Year

svocatoonal high school

Source: Ministry o f National Education Statistical Yearbooks.

Figure IV.5. Student-Teacher Ratios in the 1997-1998 and 2002-2003 School Years

; c

U 0

1997-1998

’

OPrimaw

School Year

mGT&fil high s c b l

2W2-2W3

mvocatlona hlgh smoOl

Source: Ministry o f National Education Statistical Yearbooks.

In 2002, about 32 percent o f secondary school students were enrolled in vocational schools. The intention i s to increase the share o f vocational school students even more (TC Milli Egitim Bakanligi ZOOl), even though there are unresolved questions about the efficiency o f and demand for vocational secondary education.

D. DISTRIBUTIONAL AND EDUCATION-QUALITY IMPLICATIONS OF PUBLIC POLICIES ON EDUCATION The question, “who suffers from inadequate public education funding?” i s 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 funding the most”? In the Turkish context, two leading contenders are (a) females, and (b) poor children. Despite a consistent decline in the percentage o f illiterate adults, as o f 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 11 percent in 1990 to 6 percent in 2002.’’ The situation o f the poor i s described in detail below. Characteristics of Children Not Enrolled in (Compulsory) Primary Education Although i t might have been natural to focus on pre-primary enrollments first, this section starts with an analysis o f determinants o f enrollment in primary education. The reason i s because o f extremely l o w preprimary enrollment (and lack o f relevant questions in household surveys). UNESCO estimated that only 6 percent o f children o f 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”) (Ministry o f National Education, 2003).

’* The source for these statistics i s the 2003 World Development Indicators database. 50

According to the 2002 HBS, 97 percent o f the relevant age group (ages 6 t o 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 w h o cannot afford t o send their children to school-under b o t h ad hoc social assistance and as supported by the W o r l d 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 might b e unwillingness to send girls to school in rural areas, the desire to supplement household income through child labor, and so forth). Table IV.4 displays the characteristics o f the â&#x20AC;&#x153;3 percentâ&#x20AC;? 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 in primary 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 a l l children who are primary-school age but who are outside the schooling system. About 50 percent o f the children who are n o t attending school reside in Eastern Anatolia or South Eastern Anatolia. However, the 2002 survey i s n o t representative at the regional level, and thus there m a y be some error in regional estimates. M o r e 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 in rural areas (67 percent) and they tend to be poor (53 percent).

Table IV.4. Characteristics of Children Aged 6 to 14, by Primary School Attendance Status (reported numbers are percentages) Attending or Already Completed (compulsory) Primary Schooling Male Female

GENDER

5 1.82 48.18

Not Attending and Never Completed (compulsory) Primary Schooling

27.79 72.2 1

MOTHERâ&#x20AC;&#x2122;S SCHOOLING

200,000) Rural HOUSEHOLD POVERTY STATUS Poor Non-poor

60.02

66.68

33.69

53.15

66.3 1

46.85

205

6,587

Sample size

I t 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 i s much higher. Parental schooling and gender o f the child alone may not b e good enough t o narrow down children who are not in school, but jointly, these two indicators are adequate. Table IV.5 shows that if one focuses o n fathers with n o schooling, in almost 1 out o f 10 cases, a c h i l d (aged 6 to 14) who does not attend school would b e identified. A systematic intervention that reaches a l l mothers and fathers without primary schooling would identify roughly 70 percent o f children w h o 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 b y Parental Schooling Percentage of Children Aged 6-14, Who D o Not Attend School (and who have not already completed primary) B o t h Genders

Male

Female

Father did not complete primary (illiterate or literate without diploma)

9.1%

12.5%

Mother did not complete primary (illiterate o r literate without diploma)

6.1%

2.7%

9.7%

School availability in the residential area might be one reason for non-enrollment. This issue seems to b e especially relevant in less-developed areas for female children, even if transport to the nearest school i s 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 i s 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 o r less resolved. In urban areas, even if a primary school i s not available in the immediate residential area, there would be one in an adjacent neighborhood. F o r small rural villages, o n the other hand, access to schooling i s 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 o n school availability in the residential area. The mean distance to primary school i s higher in rural areas and for the poor, but

the differences are not large. Having said that, long distances might be more significant in rural areas because o f physical barriers such as streams, hills, and poor road conditions. Table IV.6. Conditional on School Availability in the Residential Area, Estimated Distance Residence All Urban Rural

Access Primary School

300 [445]

500 [600]

300 [475]

Wealth Quintiles (based on a householdpossessions index)

Group 1 (poorest)

300 [5 141

Group 2

300 [4721

Group 3

300 ~4281

Group 4

300 ~4861

What Happens After Eight Years of Primary Education?

As shown, the demand for (noncompulsory) general secondary education has increased in recent years. This i s likely to be one o f the positive side effects o f 8-year compulsory schooling, since i t increased the number o f potential secondary school students, and also increased the amount o f 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 i s through the presentation o f marginal effects on enrollment, as estimated by a probit model (Table IV.7). This i s 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 insights from 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. Compared to the least-wealthy group o f households, the next wealth category increases enrollment probability by roughly 7 percentage points. This figure increases to over 11 percent for other wealth categories. 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 o f enrollment, which i s quite substantial in itself). 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 9 This might be because we focus on the schooling o f household head and his or her spouse’s children. A selection bias might be responsible for this peculiar finding: it could be that if the father i s 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. In other 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 i s not captured in this estimate (because the household survey questionnaire allows only household head’s children to be matched to their mother and father).

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 o f certain household characteristics o n 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 t o 20 percentage points in alternative specifications. Mother’s absence hurts girls’ schooling prospects more. The impact o f secondary school availability i s gender neutral. Urban residence does not have a statistically significant influence o n boys’ enrollment, but i t 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. Predictors o f Continued Education after Completion o f %Year Primary School

WEALTHGROUPS

MOTHER’S SCHOOLING

2o When high school availability i s excluded from the model, the “urban residence premium” increases to 10 percent, partly capturing the fact that high schools are more likely to be available in urban areas. While availability o f secondary schools has a clear impact on enrollments, i t 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 f boarding schools etc.)

Both genders Model1 Model2 More than primary

FATHERâ&#x20AC;&#x2122;S SCHOOLING No schooling Some primary Primary graduate More than primary Mother absent from household Father absent from household Secondary school availability in the residential area (1 if available, 0 otherwise)

Males Model3 Model4

Females Model5 Model6

Means & Std. Dev. .27 (.44)

.115 (3.11)

.099 (2.74)

.129 (2.16)

.121 (2.06)

.155 (2.22)

.124 (1.78)

-----

_----

.039 (1.12) .049 (1.55) .117 (3.13)

.030 (0.87) .036 (1.20) .lo2 (2.88)

.040 (0.75) .005 (0.10) .lo1 (1.81)

.04 1 (0.80) -.010 (0.20) .086 (1.59)

,070 (0.94) .127 (2.05) .196 (2.69)

.035 (0.47) .lo4 (1.77) .169 (2.44)

.14 (.35) .15 (.35) .29 (.45) .42 (.49)

-.159 (3.05) .048 (1.19)

-.158 (3.05) .032 (0.78)

-.152 (2.15) .011 (0.16)

-.156 (2.22) -.009 (0.12)

-.226 (2.00) .110 (1.51)

-.238 (2.02) .08 1 (1.07)

.07 (.26) .06 (.24)

-----

----.011 (0.29)

.133 (3.47) -.034 (0.94)

.259 (4.25)

.126 (2.74) .209 (3.46)

.66 (.47) .8 1 (.39)

-123.3 355

-116.4 346

-128.9 343

-117.7 332

792

Urban residence (1 if urban, 0 if rural)

(3.40)

.loo

.099 (4.15) .052 (1.91)

Log Likelihood Sample Size

-261.3 792

-244.2 768

-----

Since secondary school availability emerged as an important determinant o f enrollment that can b e influenced by policymakers in the short run (as opposed t o parental schooling, for example), the next step i s to see the extent to which certain households are disadvantaged when i t comes to access to secondary school education. Overall, there i s significant r o o m for improvement when i t comes to secondary school availability. Only 64 percent o f households reported that a secondary school i s 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 o f 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 o f different types o f secondary schools (many vocational secondary schools with very l o w 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 o r by household wealth.

Table IV.8. Conditional on School Availability in the Residential Area Access

Secondary School

Residence Urban Rural All

500 [733]

400 [526]

500 [703]

Wealth Quintiles (based on a household possessions index) Group 1 (poorest) Group 2 Group 3 Group 4

400 16821

500 [7331

500 [6221

500 [7861

Universities and 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 i s not particularly unbalanced (although overall public spending o n education i s low). In Turkey, entrance to university i s primarily based o n a student’s performance in a centrally administered examination; grades in secondary 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 a l l 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 m u c h to do with the design o f the examination; rather, i t 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 i t 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 o n 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 l i k e l y t o be enrolled in private universities, and they are more likely to be enrolled in “well-established” and “new and developing” institutions. Thus, university enrollment o f students coming f r o m 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 t o serve the wealthy i s neither a surprise nor a negative factor in and o f itself-as l o n g as quality public universities exist with a respectable non-wealthy student enrollment. One way to relax the capacity constraint in higher education i s to allow and encourage the establishment o f private higher education institutions, as i s happening in Turkey. Private tutoring plays a k e y role in determining who attends what type o f university, as acknowledged by 89 percent o f the undergraduate students who view private tutoring as necessary f o r success in the university entrance examination. About 78 percent o f a l l undergraduate students report receiving private

tutoring (almost 90 percent o f 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 in secondary school, and 5 percent mentioned absence o f private tutoring in their neighborhood. Thus, the poverty linkage i s 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 i s "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 f r o m 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 3 1 and 17 percent, respectively. In private 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 i t i s 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 3 1st and 33rd in mathematics 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, i t would be more productive t o view them as baselines that need t o be monitored and improved over time. For one thing, the countries that participate in this examination scheme are not randomly selected. M a n y o f them are wealthier than Turkey. Figures IV.6 and IV.7 present test scores by GDP per capita. Focusing on the countries that are in the same GDP per capita neighborhood, Turkey s t i l l ranks at below-average performance, however.

Figure IV.6. 1999 T I M S S Mathematics Achievement of Eighth-Grade Students

IWW

15ow

mwo

2 5 ~ ~ 1

30m

4oaw

4mu

moa

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 fL VA). 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), South Africa (ZAF).

Figure IV.7. 1999 T I M S S Science Achievement o f Eighth-Grade Students

250 200

+ 5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

GDP per capita (In 1999)

Note: Country codes: Singapore (SGP), Republic o f 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 (MY S), 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), South Africa (ZAF).

Household Survey Evidence on Variations in Quality of Education There are a number o f reasons poor parents do not send their children t o school. Some may choose t o 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 b e the case especially in rural areas. I t could also be that, despite high private 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 o n quality o f schooling (Behnnan and Birdsall, 1983). Furthermore, the breadth o f training in early levels o f schooling determines which, if any, universities poor children can attend.

Do poor children have access to quality schooling?21 A r e wealthier parents more satisfied o r less satisfied with the education their children receive? H o w do private schools fare when it comes to parental satisfaction with schooling? 21

Quality o f schooling can b e defined in a number o f ways. Input-based objective measures focus o n actual observations o n school characteristics such as amenities; output-based measures often focus o n standardized test results. These data are rarely available in a manner that can b e linked t o household characteristics o f children. The approach followed here is dictated by data availability, which i s t o rely o n 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 b e considering a large number o f factors that m a y not be fully captured o r appropriately weighted by analytical approaches. On the negative side, a k e y issue might be what parents f r o m 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 i s rather obvious. If poorluneducated parents are more likely t o 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

O f 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 leastwealthy quarter o f households, private school attendance was 1.3 percent, and among the wealthiest quarter o f households, i t 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 leading problems are lack o f books and supplies, reported as a problem for 15 percent o f children enrolled in public schools and 10 percent o f children enrolled in private schools. The next major problem, in both public and private schools, i s poor teaching, which was reported in about 10 percent o f cases.

Figure IV.8. Problems with Schools, Public and Private

i3.34

Total DNo problem .Lack

Public

Private

of b00kSiSUpplieS DPWrteachiw ELack of teachers DFacilities in bad andittons .Other

Dmbiems

Source: 2001 HCIS. Based on a sample o f 2,638 children enrolled in public schools and 62 children enrolled in private schools at the time o f the survey.

The urbadrural differences are unequivocal (Figure IV.9). Only 45 percent o f responses indicated “no problem with school” in rural areas, compared t o 68 percent in urban areas. In rural 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 f teachers” was reported as a problem.

reveal. Second, one can interpret the results, f r o m a rather narrow perspective, as satisfaction from received education services (rather than as a p r o x y for actual quality o f schooling received). The differences between the replies o f poor and non-poor households w o u l d still b e o f interest.

Figure IV.9. Problems with Schools, Urban, and Rural

44 51

20-

10-

Urban

Total UNOproblem .Lack

Rural

of taokdsuDDlies UPwr teaching ELack of teachen UFacilities in bad conditions .Other Dmblems

Source: 2001 HCIS. Based on a sample o f 2,200 children enrolled in urban schools

and 537 children enrolled in rural schools at the time o f the survey.

Wealthier households are much less likely t o 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 t o 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, i t may also be a result o f 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 â&#x20AC;&#x153;lack o f teachersâ&#x20AC;? declines with the wealth o f households.

Table IV.9. Problems with School (Percent)

Wealth Quintiles (based on a household possessions index) Group 1 (poorest) Group 2 Group 3 Group 4 I

T h i s trend i s n o t d r i v e n by the relatively s m a l l percentage o f households that send their c h i l d r e n to private schools.

Consistent with the reporting o f 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 in household wealth (Table IV.10). Figure IV.10. Satisfaction with Quality of Education 60 -

50 -

All households

DCompletely satisfied

Private

Public

OSatisfied

DNo idea

DDissatisfied

Completely dissatisfied

(based on a sample o f 2,670 currently enrolled students [2,610 public and 60 private])

Figure IV.ll. Satisfaction with Quality of Education

40 -

35 30 al

(I)

2 (I) a

25 20

15 10 5

All households

Rural

Urban

~ I

OCompletely satisfied

OSatisfied

No idea

BDissatisfied

DCompletely dissatisfi

ised on a sample o f 2,733 currently enrolled students [2,200 urban and 533 rural])

Table IV.10. Satisfaction with Quality of Education (percentages are reported)

I Sample size

674

758

639

660

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 urban households reported declining primary school quality.

A smaller percentage o f household heads believes secondary school quality has improved in the last two years (2 1 percent). The differences in perceived secondary school quality are negligible among wealthy groups. Urbadrural residence also does not matter much.

E. ECONOMIC GROWTH AND LABOR 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: if one i s too low, the other does not produce much.23 B o t h 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 in that the positive impact o n economic growth translates into more resources that can be devoted to education (even i f the share o f public spending o n education remains the same), which in turn aids economic growth and so on. Indeed, with the exception o f the early decades following the foundation o f the republic, Turkeyâ&#x20AC;&#x2122;s failure to invest more aggressively on early The pathways o f the relationship between populationâ&#x20AC;&#x2122;s educational attainment and economic development are complex and discussed thoroughly in the 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). 23

levels o f schooling has meant missed opportunities both for unschooled individuals and for the economy in general, as partially revealed by the labor market trends that are reported next.

Returns to Schooling in Turkey As mentioned, private returns to schooling are very high in Turkey. Table Iv.11 presents the results from the estimation o f a Mincerian-type semi-log wage equation (Mincer 1974) using data f r o m the 2002 HBS. The dependent variable i s the natural l o g 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 in earnings that are o f 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. Schooling has 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 f their earnings. When separate models are estimated for males and females, w e see that the coefficients for secondary and higher education are much larger for females. As discussed, o n one hand, females earn less than males o f similar characteristics in the labor market. But o n the other hand, the impact o f schooling o n earnings i s even more visible for females. A final issue that deserves emphasis i s the gender distribution o f those who reported non-zero earnings in the reference month: 85 percent o f this group i s male. These findings are supportive o f the literature that argues that payoffs t o educating girls are at least as high as to educating boys, especially in countries like Turkey, where the gender gap in schooling i s significantbecause social rates o f return decline with schooling. Other reasons for emphasis o n 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).

Table IV.11. Ordinary Least Squares Estimates of Log-Hourly-Earnings

Higher Education

(13.9) 1.59

(11.4) 1.40

(6.23) 1.86

[.239] .125

Urban Residence

Marmara

Note: It/-statistics are in parentheses; individuals aged 25 to 64 are included in the sample. Source: DIE 2002 HBS.

UnempIoyment by SchooIing Attainment The evolution of unemployment rates over time i s 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 (original data source i s the 1990-2003 Labor Force Surveys). Figure IV.12 shows that (a) since 2000, unemployment rates have been on the rise; (b) youth unemployment rates are significantly higher than overall unemployment rates; and (c) the probability of unemployment i s higher for males in recent years, especially among youths. While each of the remaining figures could be discussed separately, for brevity, three issues will be mentioned here. First, the female unemployment rate i s higher among (vocational and general) secondary school graduates. Second, the unemployment rate among vocational secondary school graduates tends to be at about the same level as the unemployment rate among general 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 i s a clear distinction between their overall unemployment rates (low compared with others in the sample) and youth unemployment rates (very high at around 35 percent). Since this trend has persisted since 1990, one explanation i s the possibility that reservation wages are higher for recent higher-education graduates, such that they are more selective when it comes to accepting jobs, compared to less-educated individuals. But overall unemployment rates are lower, as one would expect-skilled individuals have more opportunities in the labor market, and this effect seems to dominate other factors (such as differences in reservation wage) eventually when it comes to determining observed unemployment rates. Finally, among individuals who have completed higher education, the female unemployment rate i s much higher, despite the lack of a gender difference among the younger cohorts. This i s a finding for which we have no explanation at this stage. Figure IV.12. UnemploymentRates Over Time 25 0-

20 0-

m 1501 8

0 -E, n

E 100C

Figure IV.13. Illiterate UnemploymentRates Over Time 25.0-

200

P m

z C

150

k l

-Q 0

E 100C

50-

001

1990

-+Me

1931

1992

1993

1994

1995

1996

1937

Lk"@o,nmRae -%Fenale LhE@qmert Rae +Fn;leYu&

1998

1999

2ooo

2M1

'XJlZ

2003

Rae ~ _ _

UlnplOymrtFate uFenaleYadh LhE@u,m-d

Figure IV.14. Literate Without Diploma Unemployment Rates Over Time

0.0-

1990

1931

1992

1933

1994

19%

1996

1937

1998

1999

2Mx)

2M1

2002

xo3

Figure IV.15. Primary School Graduates’ UnemploymentRates Over Time

I I

20 0.

5 150. S

E 100-

I 50

Figure IV.16. Junior Secondary Graduates’ Unemployment Rates Over Time

250-

200.

2 C

E 150

10 0

Figure IV. 17. Vocational Junior Secondary Graduates’ Unemployment Rates Over Time

€00.

500

:400 0

:330 0

3 200

I 1004

Figure IV.18. Secondary School Graduates’ Unemployment Rates Over Time

400

350

i; 330 ?

;EO

i 200 )

Figure IV.19. Vocational Secondary School Graduates' Unemployment Rates Over Time

Figure IV.20. Higher Education Graduates' UnemploymentRates Over Time

Lao?

150

CHAPTER V: 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. I t i s largely based o n 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 in the framework o f the World Bank’s Turkey Health Sector Report (World Bank 2003~).

A. THE HEALTH CARESYSTEM 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 Ministry o f Healthy (MOH), the Social Insurance Organization (SSK), and the university hospitals. The MOH i s 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 reform program, the distinction in access to SSK and MOH hospitals and health centers has been removed, and patients are n o w free t o visit the facility o f their choice. Although still small compared t o 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 in companies with more than 50 employees, private hospitals, and foundation hospitals.

Given the age o f the available D H S data and the fact that the 2003 D H S data will become available during the second h a l f 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 W o r l d Bank using 1993 D H S data, was not undertaken. 25

Table V.l. Distribution of Primary Care Health Staff, by Region, 2002 Population/ Doctor

Population/ Midwife

Population/ Nurse

Marmara Aegean Mediterranean Central Anatolia Black Sea Eastem Anatolia Southeastern Anatolia

7,65 1 3,565 3,595 3,985 3,747 5,223 7,304

5,569 2,333 2,341 4,339 2,952 4,511 6,960

9,382 4,597 5,440 5,409 4,770 5,923 10,477

Turkey

4,708

3,672

6,196

Source: MOH, General Directorate of Primary Health Care web page.

Table V.2. MOH Health Centers and Health Posts Lacking Key Staff, by Region, 2002 Region

Number of Health % of All Number of Village YOo f 'YOof Births Centers without Health CentersHealth Posts without Health Posts Unattended Doctors Midwives b y Health Staff

Marmara

89 1

Aegean

129

814

Mediterranean

808

Central Anatolia

151

1,353

Black Sea

130

2,326

Eastern Anatolia Southeastern Anatolia

116 84

20 20

1,660 984

90 90

19% 20%

Turkey

785

8,836

Source: MOH, General Directorate o f Primary Health Care web page.

Although significant efforts have been made over the past t w o decades to expand the health care network and assure adequate physical access to health facilities across the country, the service delivery network remains highly uneven. Health facilities and health care professionals remain concentrated in urban areas, particularly in the three largest cities, Istanbul, Ankara, and Izmir, while many rural areas, especially in Eastem and Southeastem Anatolia, suffer f i o m severe shortages o f medical staff (Table V.l). The skewed distribution o f facilities, and particularly o f staff, has resulted in significant 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 o f operating budgets in many primary care facilities. This, in tum, has led to the perceived l o w quality o f care provided at the primary care level, as demonstrated by patients' frequently circumventing primary care and self-referring directly t o higherlevel 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.

Table V.3. Regional Distribution o f Hospital Beds, 2002 Total Number of Beds

Marmara Aegean Mediterranean Central Anatolia Black Sea Eastern Anatolia Southeastem Anatolia

49,655 21,541 17,306 3 1,679 22,303 13,823 5,928

Beds/10,000 Population

Admissions/ 1,000 Population

27.5 23.5 19.2 26.6 26.4 18.0 10.9

Turkey 162,235 23.3 Note: Admission figures are for 2000, beds for 2002. Source: MOH, In-patient Treatment Institutions, Statisical Yearbook (2003).

76.6 84.4 71.4 87.6 83.1 64.6 47.0 76.0

B. HEALTH INSURANCE Several public health insurance schemes currently provide financial protection to various target groups.

O f those, the health insurance o f the SSK i s the most important, and provides coverage t o those employed

in the formal sector and their dependents. Bag-Kur i s the health insurance for the self-employed, including in principle the rural population and informal sector workers. E m e k l i Sandigi provides health insurance to retired c i v i l servants, while those in active c i v i l 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. T o increase operational efficiency and improve health insurance coverage, the Government plans to shift from the currently multiple public insurance schemes t o universal health insurance, which would operate o n the principles o f solidarity and risk pooling, and provide coverage to the entire population. The Government would make contributions o n behalf o f those who cannot afford t o 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 201 1. B o t h the 2001 HCIS and the 2002 H B S provide information about the populationâ&#x20AC;&#x2122;s health insurance coverage (Figure V.1). B o t h surveys suggest that over one-third (36 to 37 percent) o f the population does not have access to health insurance, including the green card program.26 The data suggest that in rural areas, almost h a l f the population has n o health insurance, and in urban areas, about one-third has none. Because a larger share o f the lower-income groups i s employed in the informal sector, the share o f those in lower-income groups that i s covered by health insurance (including green card) i s significantly smaller

W h i l e 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 o f coverage. For example, the Draft National Health Accounts Report suggests that about one-third o f the population remains without health insurance coverage, compared to 36 percent implied by the 2002 HBS and 37 percent implied by 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). I t 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 f people without coverage. 26

than among upper-income groups (42 percent in the bottom quintile are covered, compared to 79 percent in the top quintile).

Figure V.1. Share of Population with Health Insurance, by Quintile

90% 80%

70%

6 0% 50% 40%

=Rural

30% 20% 10% 0% 1

Total

Quintile

Source: 2001 HCIS.

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 in paver$' (Table V.4). HBS data suggest that more than h a l f (58 percent) o f 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 o f green card holders are not poor.28 Thus, an important share o f 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. Health Insurance Coverage and Poverty Insurance Status

Compulsory Voluntary Compulsory and Voluntary Green Card None Total

Extremely Poor Poor % of poor with insurance 13% 0%

TOTAL

28% 3% 3% 9% 58%

18% 68%

54% 4% 2% 4% 36%

100%

Source: 2002 HBS.

This analysis i s based on the food poverty and poverty lines used throughout this report. 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 minimum wage, i s eligible for green card coverage. In 2002, the net minimum wage 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 in this report.

27 28

C. HEALTH OUTCOMES Health outcomes in Turkey are poorer than would be expected in a country with Turkeyâ&#x20AC;&#x2122;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. L i f e expectancy i s 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 in terms o f vaccination coverage.

Table V.5. Health Status Comparisons, 2002

Life Expectancy Male Female Infant Mortality per 1,000 life births Maternal Mortality

Turkey

Europe

Southeast and Central Europe

66.2 70.9 36.0 100.0

69.7 77.8 9.7 18.1

69.4 77.2 10.5 13.0

Sources: WHO-HFA, OECD Health Indicators, Turkey MOH. Note: Data are for 2002 or most recent available prior to 2002.

OECD

73.8 79.9 7.3

Key health outcome indicators vary markedly across urban and rural Turkey and across regions, reflecting the uneven supply o f 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 i s significantly lower in the poorer Eastern and Southeastern Provinces than in the rest o f the country, although the gap has begun to narrow over the past five years (Table V.6).

Figure V.2. Regional Variation in Infant and Child Mortality Rate; and Regional Variation in Prenatal Care and Home Births Regional Variation in Infant and Child Mortality

Regional Variation in Prenatal Care and Home Births 80

60 40

0%without pre-natal care

0 %home births

North South Center W e n

East

Rural Urban

Source: DHS (1998).

Health 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.

Table V.6. Vaccination Coverage of Infants and Pregnant Women (TT-2T), by Region, 1998-2002 BCG DBT-3/OPV-3 Measles 1998 2002 1998 2002 1998 2002

HBV-3 1998 2002

TT-2T 1998 2002

Marmara Aegean Mediterranean Central Anatolia Black Sea Eastem Anatolia SoutheastemAnatolia

95 86 86 87 85 42 46

88 88 90 87 88 57 61

90 85 86 88 88 63 54

79 80 85 83 82 68 63

88 85 83 86 89 67 52

86 85 89 87 85 71 67

69 87 64 84 83 42 23

73 82 83 84 78 55 49

27 57 55 47 45 19 12

25 58 53 48 47 24 18

Turkey

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 o f the population i s affected by noncommunicable diseases prevalent in developed countries. Morbidity and mortality associated with heart and cerebrovascular diseases have increased sharply over the last two decades as, for example, reflected by a fourfold increase in hospital discharges for ischemic heart disease and cerebrovascular problems between 1988 and 2002, and an increase o f over 150 percent in registered cancer cases (WHO HFA 2004).

D. SELF-REPORTED MORBIDITY Information o n 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 ill during the 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 o f people reporting an illness or injury during the past month, and the share o f people reporting an illness requiring hospitalization, are higher among the lower-income groups, particularly in rural areas (Figure V.3 and Table V.7).

Figure V.3. Incidence o f Illness Income Quintiles and Location

Ulllness/in]u!y past 30 days urban

20 0

MIllness/injury past 30 days rural

15 0 10 0

Oillness requiring hospitalization past 6 months urban

50 00

Total Per Capita Income Quintile

Oillness requiring hosoltalization Dast 6 months rural ' ~~

~~~~~

~~~

Source 2001 HCIS

Table V.7. Self-Reported Morbidity by Income Quintile Quintiles of Per Capita ircome % reporting being sick/injured in past 30 days % reporting illness requiring hospitalization in past 6 months

Severity of ///ness (% of those reporting Illnesshjury last Month) Lif e-threatening Very Serious Serious Not Serious

13.0% 5.6%

9.7% 3.6%

10.0% 2.9%

9.1% 3.3%

10% 15% 54% 21%

5% 21% 45% 29%

8% 20% 41% 31%

5% 16% 53% 26%

Total

11.0%10.6% 2.8% 3.6%

4% 16% 41% 39%

6% 17% 47% 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 during the month prior to the survey was l i f e threatening. The share o f 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 in the lowest quintile. The overall higher reported incidence and severity o f illness among the lower income groups, particularly in rural areas, may suggest poorer quality o f care and more limited access to and use o f health care among these groups, resulting in delayed 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 t o health care, which then lead those without insurance to less readily admit to being sick. International evidence has shown that selfreported morbidity tends to be associated with access to health care. Figure V.4. Incidence o f Morbidity by Health Insurance Status

% reporting being sick/injured

114%

I 0%

10%

I N o Insurance OOther I G r e e n t Card 0 Private OBagkhur ISSK DES

15%

Source: 2001 HCIS.

E. ACCESS T O AND USE OF HEALTH CARE O f those reporting an illness over the past month, 72 percent report seeking some form o f health care for the reported illness, with the share being lower in rural (66 percent) than in urban areas (74 percent) (2001 HCIS). The likelihood o f seeking care when ill i s significantly lower in the bottom two quintiles than in the upper quintiles, with the difference being particularly marked in rural areas (Figure V.5). On a

regional basis, the share o f those seeking care in Southeastern and Eastern Anatolia i s significantly lower than in the rest o f the country (Figure V.6).

Figure V.5. Share of Those ReportingIllness Who Sought Care by Quintile and Location

100% 1 ~

80% 60%

U%sought care rural

40%

20%

0% 1

Per Capita Income Quintile

Source: 2001 HCIS

Figure V.6. Propensity to Seek Treatment (by region and location)

130% 80% 30%

Region

0 Rural 0 Urban .

Source: 2001 HCIS.

Even among those who report that their illness was l i f e threatening or very serious, the share o f those seeking care remains a low 55 percent in the lowest-income group, while it i s 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 ill than those without insurance, underliningthe importance of insurance to improvingaccess to health care. Even among those who report that their illness was l i f e threatening or very serious, the share o f those seeking care remains a low 55 percent in the lowest-income group, while it i s 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, underlining the importance o f insurance to improving access to health care (Figure V.7).

Figure V.7. Propensity to Seek Care When I11 by Insurance Status Figure V.7. Propensityto Seek Care 'VVhen 111 by Insurance status Private Emekli Sandigi Eagkur SSK

Green Card None

0O h

20%

40 %

60 Yo

anyo

100%

Source; 2001 HCIS

Table V.8. Health Care Utilization by Severity o f Illness LifeThreatening Very Serious Quintile

Total

3 4 5

55% 87% 80% 83% 83% 72%

66% 58% 84% 86% 89% 76%

Serious

Not Serious

YOSeeking Care

70% 72% 79% 84% 84% 77%

54% 54% 72% 77% 74% 67%

Total

63% 63% 79% 79% 78% 72%

Source: 2001 HCIS

Hospital Care. O f the 552 individuals (3.6 percent o f population) who reported having suffered from 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 and to insurance status (Figure V.8). The data seem to suggest that possession o f a green card has a significant impact on access to hospital care, because there i s no marked difference in the hospitalization rate between those with insurance and those with a green card, but the share among those without any insurance protection i s 26 percentage points lower than among those with insurance or a green card.

Figure V.8. HospitalizationAmong Those Requiring Hospitalization,by Quintile

Hospitalization among Those Requiring Hospitalization,by Insurance Status

Hospitalization among Those Requiring Hospitalization, by Quintile 5 4 3 2 1 Quintile

no insurance

greencard

-c

insuranw

0.2

0.4

0.6

0.8

â&#x20AC;&#x2122;~

~~~~~

72t 20%

60%

40%

80%

Source: 2001 HCIS.

Determinants o f Care Seeking. The 2001 HCIS showed that the most important reason for not seeking outpatient care when sick, and for not seeking hospital admission when required, i s the lack of affordability (Table V.9). Close to three-quarters o f those who did not seek outpatient care when ill,and a similar share of those who did not seek hospital care when needed, reported that they could not afford to do so, with the share of those not being able to afford it expectedly falling rapidly with rising income levels. Table V.9. Reasons for Not Seeking Outpatient Care When I11 ~~

Reason for Not Seeking Care Quintile

1 2 3 4 5

Total Insurance Status

Emekli Sandigi SSK Bag-kur Private Green Card No insurance

Unaffordable Too Far

Poor Quality

Bad Time No Doctor N o Drugs

Other

95% 79% 80% 32% 25%

0% 0% 0% 0% 4%

2% 0% 5% 16% 14%

0% 6% 5% 5yo 14%

0% 2% 0% 5% 7%

2% 0% 0% 5% 0%

2% 13% 10% 37% 36%

72%

5yo

2Yo

1Yo

14%

17% 5 1% 40% 0% 90%

0% 2% 0% 0% 0% 0%

5 0% 6% 10% 0% 0% 3 yo

17% 10% 0% 0% 0% 3yo

0% 4% 10% 0% 0% 0%

0% 2% 0% 0% 10% 0%

17% 24% 40% 100% 0% 8%

85%

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,

regardless o f 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 i s 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 perceived to be o f low quality (Table V.10).

Table V.10. Reasons for Not Being Hospitalized When Needed ~~

Quintile Cannot Afford I t Too far Poor Quality Other

88% 4% 5% 3%

83% 8% 4% 5%

75% 17% 0% 8%

________~

Total

40% 0% 10% 50%

30% 0% 17% 53%

78% 5% 5% 12%

Source: 2001 HCIS.

Information pertaining to problems that those who sought care encountered, confirm that affordability o f health care i s 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 t o 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 o f that among the bottom quintile (Table V. 11).

Table V.11. Problems Encountered When Seeking Outpatient Care, by Quintile

Nature o f Problem N o problem N o t clean Long wait N o specialists Too expensive N o drugs Treatment did not work Staff was rude Other

46% 2% 12% 0% 20% 9 yo

50% 3% 20% 2% 10% 4% 3% 6% 1%

5% 4% 1%

Quintile (per capita income) 3 4 5

54% 3% 18% 2% 14% 3y o 7% 1% 0%

61YO 3% 18% 1% 10% 1Yo 4% 2% 0%

68% 1Yo 13% 2% 4% 2% 6% 3% 2%

Total

55% 2% 16% 1% 12% 4% 5% 3% 1%

Source: 2001 HCIS.

The share o f the population that had to pay for outpatient treatment, drugs, and hospitalization i s consistently higher among the lowest-income quintile than among the upper-income groups-a reflection o f 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 lowestincome group, and dropped with rising income and associated increased insurance coverage (Table V. 12). The situation i s less marked in the case of hospital treatments; while the share o f those 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).

Table V.12. Average Amount Paid During Last Outpatient Treatment by Those Who Paid Quintile Consultation

1 2 3 4 5 Total

34 28 27 35 30 31 Source: 2001 HCIS.

Drugs T L Million 24 30 19 20 17 21

Gift

Total

% Who Paid

4 1 3 3 1 2

62 59 48 57 48 54

60% 40% 35% 46% 37% 43%

Table V.13. Share and Average Amount of Payments for Last Hospital Admission Quintile

Total

TL Million

% Who Paid for Treatment

93 170 139 133 161 136

68% 58% 67% 50% 66% 62%

Note: Average paid i s average for those who paid only. Source: 2001 HCIS.

Multivariate analysis o f the determinants o f health-care-seeking behavior confirms that income, insurance coverage (including green card), household size, gender o f the household head, and severity o f illness are the most important determinants o f an individualâ&#x20AC;&#x2122;s seeking health care (see Table V.14). Together with the above information, these findings thus confirm that financial constraints constitute a significant access barrier to care for low-income groups, particularly those who have n o financial protection. They underscore the importance o f access to health insurance in the decision to seek health care.

Table V.14. Determinants o f Health Care-Seeking Behavior, Probit Estimates (based on 2001 HCIS data)

Employed head o f household Female head o f household Insurance: Emekli Sandigi

-0.363 1.076

-2.890 7.960

Variable

Insurance: SSK Insurance: Bag-Kur Insurance: Private Green card Severely ill Household size Constant

Coefficient

0.671 0.683 1.261 0.618 0.598 -0.062 0.024

t-ratio

7.570 5.590 2.110 3.670 6.520 -3.160 0.120

Bankâ&#x20AC;&#x2122;s Health Sector Report (2003c), particularly Chapter 2. This report also carried out 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, household size, gender of household head, and severity o f illness as the most important determinants o f an individualâ&#x20AC;&#x2122;s decision to seek health care.

Location o f Care. Lower-income groups are most likely to seek care at MOH facilities because they can gain access t o them even if they have n o insurance (or with green card coverage), and because the average cost o f treatment there i s lower than elsewhere, with the exception o f SSK and military facilities, both o 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 f r o m 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 t o seek care i s taken, perceived quality o f care i s an important determinant o f the location o f care seeking (Tables V.15 and V.16).30 The data also confirm that 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 Governmentâ&#x20AC;&#x2122;s recently introduced policy changes, which provide patients free choice between MOH and SSK facilities regardless o f insurance affiliation, are likely t o change this careseelung pattern somewhat. I t i s unlikely, however, that the distribution between private and public providers will substantially change as long as waiting times and perceived quality o f care at public facilities are not improved.

The World Bank Turkey Health Sector Report (2003~)found that the demand for health care i s very price inelastic, even among low-income groups. Thus, once a decision to seek care i s taken, particularly b y those who have to pay for the entire cost o f care themselves, quality considerations are likely to drive care seekers to seek care from private providers that are perceived as providing higher quality o f care. 30

Table V.15. Location o f Outpatient Care by Quintile

Location o f Care

M O H Hospital M O H Clinic SSK Clinic SSK Hospital University Hospital Military Hospital Private Clinic Private Hospital Consulting Room Pharmacy Other

43% 5% 6% 13% 3% 2% 6% 11% 8% 0% 4%

Quintile (per capita income) 3 4 5

2 38% 3% 4% 30% 4% 0% 6% 6% 7% 0% 3%

29% 4% 5% 38% 5% 1% 6% 5% 2% 0% 5%

18% 5% 9% 29% 11% 1% 8% 7% 8% 0% 2%

Total

26% 3% 3yo 29% 8% 1% 13% 8% 7% 1% 2%

31% 4% 5 yo 27% 6% 1% 1% 8% 6% 0% 3%

Source: 2001 HCIS.

Table V.16. Average Amount Paid for Outpatient Treatment by Facility Type M O H Hospital M O H Clinic SSK Clinic SSK Hospital University Hospital Military Hospital Private Hospital Private Clinic Doctor's Residence Pharmacists Other

Note: Payments include consultation, drugs, and gifts. Source: 2001 HCIS.

TL million 30.5 30.3 24.6 26.2 51.9 3.9 91.5 55.2 50.4 22.5 61.9

Preventive Health Care. Given access barriers to health care, preventive health care is, not surprisingly, a luxury item in Turkey. The 2001 HCIS data show that only about 1 percent 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. N o t surprisingly, the vast majority o f those in the bottom two quintiles indicated that they could not afford t o have a checkup, while the main reason for not having a checkup in the top t w o quintiles was that there was n o need (Table V.17).

Table V.17. Utilizationof Preventive Care % o f Population Who Had a Check-Up in Past 6 Months Reasons for Not Having a Check-Up Unaffordable Too Far N o t Time N o t Aware o f Need N o t Necessary Other

Total

1Yo

85% 1% 0% 2% 10% 1%

80% 1% 0% 3% 15% 0%

68% 1Yo 1% 2% 27% 1%

53% 1% 2% 2% 42% 1%

31% 1Yo 4% 2% 59% 3%

64% 1Yo 1% 2% 30% 1%

Source: 2001 HCIS.

F. HEALTH CAREEXPENDITURE Out-of-Pocket Payments. Household expenditure o n 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 f o r m o f out-of-pocket expenditures, but this share increases with income, suggesting that health care i s considered a luxury good. The top quintile spends about twice as high a share o f total household expenditure o n health care than the lowest quintile. More important, the top quintile spends over 12 times more per capita on health care than the bottom quintile (Table V. 18). Table V.18. Composition o f Health Care Expenditure, by Expenditure Quintiles 1 Per Capita Health Expenditure/month (mil TL) Share o f Total Expenditure

1.2 1.7% Composition of Health Care Expenditure (YOo f total OOP) Outpatient Consultation 9% Inpatient care 1% Dental Care 2% Diagnostics 1% Drugs 84% Other Medical supplies 4%

Total 2002

2.9 2.5%

3.4 2.2%

6.1 2.7%

14.8 3.5%

3.3 2.2%

16% 2Yo 4% 2Yo 73% 4%

16% 1% 4% 3% 71% 6%

20% 2% 4% 6% 65% 5%

19% 5% 9% 5% 58% 6%

14% 2% 4% 3% 74% 5%

1. Total may not add up to 100 due to rounding. 2. Health care expenditures adjusted for spatial price variations. Average unadjusted out-of-pocket (OOP) amounts to about TL2.7 million per capita per month. Source: DIE 2002 HBS.

The largest share o f out-of-pocket expenditures o n 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 o n 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

when ill,rather than seek professional care. There are larger differences in the average absolute amounts spent o n all types o f care across income quintiles: the top quintile spends almost 12 times more o n outpatient care per capita than the bottom quintile, and almost four times more o n drugs.

Catastrophic Effect of Health Care Expenditures. 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 o n 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 householdâ&#x20AC;&#x2122;s non-food expenditure were considered for this analysis. Table V.19 shows that there i s o n average 10 percent o f the population that spends more than 10 percent o f non-food expenditure o n health care, with the share dropping to 0.7 percent for a threshold o f 40 percent. This compares favorably t o 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 i t 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 i s also confirmed by the fact that the share o f those with catastrophic expenditure shares i s not significantly higher among those without health insurance. Table V.19. Proportiono f People with Catastrophic Health Care Expenditure 10% or M o r e o f Non-Food Expenditure

20% or M o r e of Non-Food Expenditure

40% or M o r e of Non-Food Expenditure

Quintile

1 2 3 4 5

9.3% 12.2% 11.0% 10.0% 11.0%

4.2% 5.6% 3.5% 3.6% 5.0%

0.5% 1.O% 0.1% 1.4% 1.9%

Total

10.0%

4.0%

0.7%

Poor Non-Poor

9.5% 10.8%

4.3% 4.4%

0.6% 0.8%

8.7% 15.6% 12.9% 11.6%

3.2% 6.4% 7.4% 5.1%

0.7% 0.8% 0.7% 0.9%

Insurance Status

Compulsory Voluntary Green Card None Source: DIE 2002 HBS.

Impoverishing Effects of Health Care Expenditures. T o see t o what extent health care expenditures can throw people into poverty by preventing them f r o m using the money spent o n 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 lead to a substantial increase in poverty. 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 s t i l l relatively small, the poverty impact o f health care expenditure i s bigger among those with green cards and n o insurance, than among those with insurance. The relatively modest overall impact o f health care expenditures o n 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 o f their extended families. Table V.20. Impoverishing Effects of Health Care Expenditures Pre-Payment

Post-Payment

Difference

Poor Extremely Poor

26.9% 1.3%

27.9% 1.4%

1.O% 0.1%

By Region Marmara Aegean Mediterranean Central Anatolia Black Sea Eastern Anatolia Southeastern Anatolia

18.8% 13.4% 41.6% 26.7% 26.7% 48.3% 37.9%

17.6% 12.9% 40.4% 25.7% 26.3% 47.4% 37.5%

1.2% 0.5% 1.2% 1.1% 0.4% 0.9% 0.4%

By Insurance Status Compulsory Voluntary Green Card None

13.7% 19.9% 65.9% 43.6%

14.4% 20.0% 67.4% 44.7%

0.7% 0.1% 1.5% 1.2%

Note: The post-payment poverty line was adjusted by the amount o f health care expenditure. o f those living just at the poverty line to account for lower overall spending Source: DIE 2002 HBS.

Public Expenditure on Health Care. Public expenditure o n health care grew at an average annual rate o f 7.3 percent between 1999 and 2003 (Figure V.9). The Governmentâ&#x20AC;&#x2122;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 i s distributed across four major areas, including health care programs and service delivery through MOH, the green card scheme, c i v i l servants health benefits, and government financing o f social health insurance schemes, when the latter run deficits. Central Government funding i s 3â&#x20AC;&#x2122;

This follows the methodology developed by Wagstaff and van Doorslaer (2002).

supplemented by limited funding from the provincial administrations and municipal governments, with the latter operating their own facilities in some o f the larger cities.

Figure V.9. Evolutiono f Public Sector Spending on Health

6.00

15,000,000

b 4.00

10,000,000

2.00 0.00

5,000,000 1999 2000

2001

2002

2003

.-B

'+Health I ExpenditureslGNP +Real Health Expenditure

Source: World Bank Staff Calculations, based on data provided by MOF and SPO.

Figure V.10. Composition o f Public Sector Spending on Health, 2003

Local Govern men

Social Securi Funds 50%

State Economic

Centra I Governmen 48%

Source: World Bank Staff Calculations based on M O F , 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).~~

Turkey Ministry o f Health School o f Public Health, National Health Accounts 2000, Draft Report, 2003.

Figure V.ll. Composition o f Total Health Sector Spending (2000)

Composition of of Total Health Sector Spending [2000]

Central GOV

Local GOY 0 Social Security Funds

0 Private Social Insurnace

Private Health Insurance Household OOP

Corporations

Other

Benefits incidence analysis carried out within the framework o f the National Health Accounts found that public sector spending o n 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.

Table V.21. Distributionof Public Sector Spending on Health Across Income Quintiles ~

Quintile

Inpatient

Outpatient

Total

15% 16% 19% 26% 24%

15% 17% 19% 26% 23%

% o f spending accruing to quintiles

Q1 Q2 43 44 Q5

16% 19% 20% 25% 20%

Source: Turkey M O H School o f Public Health: National Health Accounts, Draft Report,

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 t o health care, while over one-fifth goes toward providing civil servants with health care benefits, a population which i s traditionally not among the lowest-income groups (Figure V.12). Another fifth goes in subsidies t o 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.

Figure V.12. Destination of Central Government Funding for Health Care

Destination of Central Government Funding For Health Care

w e a l t h Services Delivery reen Card Rogram F I V I I ServantHealh Benefib

m o v Subsidies to SSI

Source:Ministy of Health ,Turkey Nabnal Health Pccounts, Drat.Report.2003

CHAPTER VI: LABOR In Turkey, as in most countries, poverty i s closely correlated with employment status and type o f job, whether formal or informal. Informally employed or casual workers have a noticeably higher rate o f poverty. In Turkey, unemployment o f the household head i s particularly associated with poverty. Education i s a key factor in explaining employment, and therefore poverty outcomes. Certain sectors o f 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 o f information: the Labor Force Surveys (LFS) conducted by the State Institute o f 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. In general, the H B S data confirm the overall trends o f the LFS, but the LFS should be viewed as definitive for measuring unemployment in Turkey, because this i s what they are designed for, as opposed to the HBS, which i s not.

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 increased after the 2001 economic crisis (Table VI.l) and remained at about 10 percent in 2003. However, unemployment i s affected to a large extent by l o w levels o f labor force participation, whereby those who do not have work, typically drop out o f 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 o f males aged 15 and above are not in the labor force, representing many discouraged workers.33

Table VI.1. Turkey: Unemployment Rate and Labor Force Participation Rate 11990 11991 11992 11993 11994 11995 11996 11997 11998 11999 PO00 PO01 PO02 nemployment

L a b o r 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.

For DIE labor 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.

There i s n o 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 i s primarily driven by the labor force participation rates. Unemployment among women i s lower since so few women are in the labor force. Under 30 percent o f women are labor force participants, and the female rate o f labor force participation in 2002 was significantly lower than in 1990. Figure VI.2 shows the l o w rate o f female labor force participation compared t o the male participation rate.

Figure VI.1. Turkey: Unemployment Rate

12 0 100

* e

0 Total

8 Male 17 Female

20 00

Figure VI.2. Turkey: Labor Force Participation Rate

+Total

+Male

Female _.

UNEMPLOYMENT AND INACTIVITY

In the 2002 HBS, 35 percent o f those aged 12 and above reported that they had worked in a paid j o b in the survey month. Of those reporting a paid job, 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 i s that 22 percent o f the sample are children, and that poverty i s concentrated among families with children, as demonstrated in the poverty profile chapter.

Turkeyâ&#x20AC;&#x2122;s l o w rate o f labor force participation, particularly o f women, i s detailed in the World Bankâ&#x20AC;&#x2122;s, Labor Market Study (forthcoming), and was summarized above. The 2002 HBS results confirm these findings (Figures VI.3 and VI.4). In the 2002 H B S sample, 22 percent were children and 10 percent were adults aged 60 and above, leaving an adult population o f 68 percent. O f the adult population, about h a l f (51.2 percent) reported that they had a paid j o b in the survey month, but women accounted for only 32 percent o f all reporting paid jobs (adults and a few elderly).

Figure VI.3. Turkey: Adults Employed

Figure VI.4. Turkey: Adults N o t Employed

0.4

Male Female

Unemployment i s measured in the HBS as those who report that they did not work for income or in kind in the survey month, and that they were seelung a job. Only 7.2 percent o f those aged 12 and above reported that they were looking for a job, which again underscores the low labor force participation rate in Turkey, because those not looking for a job are not in the definition o f the labor force. Households where the head was unemployed had a poverty rate o f 35 percent compared to 26 percent poor o f households whose heads were employed. While households with unemployed heads are poorer, this relatively l o w difference relates mainly to the fact that so few people are loolung for work, and are thus able to be defined as unemployed, and also to the fact that only 8 percent o f households reported that the head was unemployed (by meeting the definition o f being in the labor force and looking for work). For example, 39 percent o f those aged 12 and above reported that they were not seelung a j o b (Table VI.2) for a variety o f 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 Reason for Not Seeking Job

Found a j o b but waiting Student Housewife

Poor

Non-Poor

16.8 24.3 22.7

83.2 75.7 77.3

Percent o f Valid*

0.1 25.7 42.2

Population Subtotals

Percent o i Total

37,568 6,879,196 11.373.103

0.1 10.1 16.6

* The percent o f those who answered the question or for whom we have data.

Notes: Poverty in percentages. Population Subtotals: Number o f observations. Source: DIE 2002 HBS.

Some o f these categories were associated with an elevated risk o f 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. O f those, however, the poverty rate was 37 percent compared to the average o f 27 percent. Disability and illness accounted for 1 and 2 percent o f the responses, but with a poverty rate o f 36 and 33 percent, respectively. However, for those very few households headed by a disabled person, almost all were poor (86 percent).34

However, these households were only 0.3 percent of those sampled, so this number i s too l o w for robust estimation. It i s presented as indicative. 34

Figure VI.5. Turkey: Poverty and Inactivity

c. QUALITY

OF EMPLOYMENT

In Turkey, there i s a strong association between the type o f employment and the poverty status of the individual o r household. F o r example, the poverty rate o f those aged 12 and above who reported that their j o b was permanent employment was only 12 percent, but for those who reported their j o b as casual, the poverty rate was 44 percent (and i t was 30 percent for the very few people who described their employment as temporary) (Figure VI.6). The relative r i s k o f poverty for casual w o r k 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 30.0 25.0 20.0 15.0 10.0 5.0 0.0

Poor _ +Total _ _ _ _Poor _ ~

Permanent employment

Temporary Casual type of employment I work under contract for a fixed period ~~

In a different question o n the same theme, virtually the same percentages o f poor and non-poor were observed for casual employees and regular employees (corresponding to permanent employment) (Figure

VI.7).

Figure VI.7. Turkey: Poverty and Employment Status I 1

50.0 45.0 40.0 35.0 30.0

' '

2 25.0

20.0 15.0 10.0 5.0 0.0

In addition, poverty rates were higher for the self-employed and unpaid family workers. Reflecting the finding o n the l o w rate o f poverty for apprentices, i t i s 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 reported membership), while the poverty rate o f nonmembers was 20 percent. However, only 16 percent 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 f registration at a social security institution. Conversely, formal employment as measured by enrollment in social security i s a strong bulwark against poverty in Turkey. I t i s important to emphasize that reliable figures o n 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 twothirds o f the population. In the 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).

Figure VI.8. Turkey: Poverty and Social Security

40.0 35.0

; %:8 20.0

15.0 10.0 5.0 0.0

O f these formal jobs, the most significant in terms o f employees facing a lower r i s k 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), i t was only 5 percent (Figure VI.9). However, i t 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 o f Workplace 30.0

25.0 U

20.0

Poor +Total Poor-

2 15.0

10.0 5.0 0.0 Private

Public I Government

State-Owned Enterprise

Table M.3. Turkey: Poverty Rate of Status of Workplace

Notes: Poverty in percentages.

dation Subtotals: Number o f observations.

Poverty i s associated with the size o f the enterprise. People employed in enterprises 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 i t 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

25.0

1 2 1

-_____

Poor +Total Poor

20.0

Q 15.0 n 10.0

5.0 0.0

119 persons

10124 persons

25/49 persons

50 and over

Another way to gauge informality i s t o look at the legal status o f the respondentâ&#x20AC;&#x2122;s workplace, assuming that more individually owned workplaces would be likely to be informal, while few incorporated companies would be. Using these criteria, i t i s clear that the highest poverty rate o f those aged 12 and above (3 1 percent) was found in those who reported that their workplace status was individual ownership, and the lowest poverty rate o f 6 percent pertained to incorporated companies (Table VI.4).

Table VI.4. Turkey: Poverty Rate o f Legal Status o f Workplace

In Turkey, poverty i s associated more with lack o f work than with work. For example, the mean number 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 in the same low-paying jobs 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 o f 27 percent, suggesting that mobility i s slightly associated with lower risk o f poverty.

D. SECTOR OF EMPLOYMENT The largest sector in terms o f employment in Turkey i s agriculture (including forestry and hunting). O f the 35 percent aged 12 and above, fully 40 percent are engaged in agriculture (Table VIS). Agriculture i s 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 o f employment, as reported in the HBS.35 T h i s latter finding i s 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 i s lowest for mining and quarrying, where it i s under 3 percent poor, but less than 1 percent o f employees are employed in that 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. In terms o f poverty, â&#x20AC;&#x153;other social, public, and personal servicesâ&#x20AC;? have a higher poverty rate (30 percent) than average, but account for only 2 percent o f employment.

Note that according to the LFS, construction accounted for 4.5 percent o f employment in 2002 and 6.1 percent in

2003.

Table VIS. Turkey: Poverty Rate o f Basic Code of Main Activity o f Workplace

Agriculture, Hunting, Forestry, Fishing Mining and Quarrying Manufacturing Electricitv. Gas and Water Construction Wholesale and retail trade and hotels and restaurants Transportation communication and storage services Financial services and real estate lComunitv and Dersonal services Total N o t reDortin2 emdovment r l

Poor 36.5 2.6 16.8 16.6 35.9 19.2

14.1 15.3

N o n Poor 63.5 97.4 83.2 83.4 64.1 80.8

I I I

I I

85.9 84.7

Percent o f Valid*

40.5 0.7 15.3 0.5 5.3 17.0

I I

* The percent o f those who answered the question or for whom we have data.

I Population 1

Subtotals 9,786,383 168,655 3,708,013 121,490 1,280,230 4,101,936

Percent of Total 14.3 0.2 5.4 0.2 1.9

6.0

5.3

1,285,5531

13.0

3,176:20914.6

44,233,8571

1.9

64.7

Notes: Number of work hours per week 4 0 .

Poverty in percentages. Population Subtotals: Number of observations. Source: DIE 2002 HBS.

CHAPTER VII: SOCIAL PROTECTION Social protection in Turkey consists primarily o f limited formal systems o f 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, i s documented in the W o r l d 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. T h i s 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. In rural 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 t o extend t o the urban gecekondu (slum) areas, through networks o f people from villages o f origin, called hemseri. Additionally, religious charity plays an important role. In qualitative interviews conducted for the Social Risk Mitigation Project, and in general poverty monitoring, some urban very poor reported charity, including rent-free housing in gecekondu 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 i t s affiliated 93 1 Social Assistance and Solidarity Foundations (SYDVs).

A. TURKISHPENSION SYSTEM Turkey’s current social security system i s highly fragmented, with benefits and contributions dependent o n 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 c i v i l servants. C i v i l 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 i s a small noncontributory pension scheme available to those who reach age 65 and have n o 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 o f the labor force i s contributing to one or the other o f the schemes, with the bulk o f the coverage in SSK. O f the 42 percent o f the labor force covered, 48 percent are covered in SSK, 22 percent each in ES and in the self-employed scheme under BK, and an additional 8 percent in the farmers’ scheme under BK. The total number o f contributors to a l l schemes i s around 11 million.

100

Box VII.1. Parameters o f the Current Pension System SSK Retirement age is 58160 for full and 58160 for partial retirement 7,000 days o f contribution (19.4 years) are required for full retirement Replacement rate i s 54 percent for full retirement Average earnings in whole working life are considered in calculating pensions Present value o f earnings is calculated through inflating each yearâ&#x20AC;&#x2122;s earnings by associated Consumer Price Index (CPI) and gross domestic product (GDP) for the period Pensions are indexed t o inflation Contribution rate is 20 percent Insurable earnings are indexed t o C P I and GDP. Bag-Kur (self-employed) Retirement age i s 58160 for full and 60/62 for partial retirement 25 years o f contribution are required for 111 retirement Replacement rate is 65 percent for full retirement Average earnings in whole working life are considered in calculating pensions Present value o f earnings i s calculated through inflating each yearâ&#x20AC;&#x2122;s earnings by associated C P I an( GDP for the period Pensions are indexed t o inflation Contribution rate is 20 percent Insurable earnings are indexed t o C P I and GDP. Bag-Kur (farmers) Retirement age is 58160 for full and 60162 for partial retirement 25 years of contribution i s required for full retirement Replacement rate is 70 percent for full retirement Last income step where contributions are p a i d i s considered in calculating pensions Pensions are indexed t o c i v i l servant salary increases Contribution rate i s 20 percent Insurable earnings are indexed to increase in c i v i l servant salaries.

ES Retirement age is 58160 for full and 60162 for partial retirement 25 years o f contribution i s required for full retirement Replacement rate is 75 percent for full retirement Last salary i s considered in calculating pensions Pensions are indexed to c i v i l servant salary increases Contribution rate i s 20 percent Insurable earnings are indexed to increase in c i v i l servant salaries.

101

On the beneficiary side, only 29 percent o f the population over age 65 i s receiving an old-age pension from any o f the public schemes. O f 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 in Bag-Kur. A total o f 1.2 million 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 until just before retirement, and then make a large lumpsum payment that i s supposed t o represent the cumulative value o f a l l past-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 o f contributors i s far larger than the pensions going to the elderly. However, in 1991, the minimum retirement age was abolished, which, coupled with short minimum contributory periods in SSK o f less than 15 years, and only slightly longer periods in ES 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 in SSK, and as early as age 40 in E S 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 collecting pensions. The 1999 reform sought t o remedy this problem by reestablishing a minimum retirement age. However, the reform i s to be phased in extremely slowly, with the retirement age as l o w 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. B o t h the 1991 law and the retirement age aspect o f the 1999 reform applied to all three pension systems, although the impact o n SSK i s far greater than on the other systems, since its shorter minimum contributory period had resulted in such l o w 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 1 percent 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, 1 percent for maternity, 1.5 t o 7 percent for work injury, and 2 percent for unemployment insurance on top o f what i s 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. Contribution Rates to SSK (including unemployment insurance premiums) Employee Health Insurance Matemity Work Injury and Occupational Diseases Pension Unemployment Insurance2 Total

5% 0 0 9% 1% 15%

1. The rate i s subject to risks o f injury and disease in a particular sector. 2. There i s also a 1 percent state contribution for unemployment insurance.

Employer 6% 1% 1.5%-7%' 11% 2% 21.5%

102

The 1999 l a w 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 prorated by years of service under each system, it i s necessary t o describe the benefit structure both pre-reform and postreform. 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. In addition, workers were to have been members o f SSK for 25 years i f male, 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 o f contributions, or 10 years. Contribution ceilings and floors and maximum and minimum pensions were adjusted by Parliament in an ad hoc fashion, with some periods where the contribution ceiling actually fell below the minimum wage. Pensions post-retirement were adjusted by growth in c i v i l servant wages. During periods 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 a l l pensioners. The nominal amount o f this flat payment was decided annually by Parliament. By 1996, this flat payment had become so large that i t exceeded the average pension level, completely unraveling any relationship between pensions and contributions. Since 1996, this flat payment was frozen in nominal terms, and i s now only about 1 percent o f pensions, but i t i s still gwen t o 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 i s also being phased in slowly for current workers. Full retirement benefits are 54 percent o f lifetime career average, although the average i s computed by revaluing the nominal wages by growth in nominal GDP. Workers are s t i l l required to belong to S S K 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 s t i l l 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 up to 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 o n which contributions were t o be paid was indexed to nominal GDP growth. The maximum salary on which contributions were t o 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 indexed to inflation o n a monthly basis. One issue that has emerged from the automatic indexation i s that a discrepancy has arisen between minimum wage and the minimum insurable earnings. Particularly since the 2001 fiscal crisis, minimum wage, which i s s t i l l adjusted o n an ad hoc basis, has not risen as fast as nominal gross national product (GNP) growth. Thus, minimum insurable earnings are almost 50 percent higher than minimum wage. Employers are currently required by l a w to pay the full contribution, both employer and employee portion, for the difference between the two. T h i s 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 minimum wage, to leave i t at i t s previous 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 a l l systems that cover the selfemployed, there i s the perennial problem o f evaluating the income earned. Turkey initiated a system of minimum earnings 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. T h i s occurs for the f i r s t 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 farmersâ&#x20AC;&#x2122; income substantially below that o f 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 workerâ&#x20AC;&#x2122;s discretion beyond that. The nominal earnings level associated with each step i s n o w also automatically indexed to nominal GDP growth. However, i t should be noted that Bag-Kur has very l o w collection rates for its contribution revenue. Typically, workers pay very little i f anything during their working years. Just prior t o 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 f the 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 o n 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 o n the full earnings history (as defined by the steps). Pensions were also n o w automatically indexed t o inflation o n 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.

Emekli Sandigi. Emekli Sandigi i s 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 i s not covered by the pension fund. Rather, i t i s covered directly by the line ministries with which the c i v i l servants are employed. The pension fund covers only retirement age benefits, pensions, and health services during retirement. For this, a 36 percent contribution i s collected, of which 20 percent comes from the employer and 16 percent from the employee. N o 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 o n 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 workerâ&#x20AC;&#x2122;s total cash remuneration. When pension benefits are paid, they are paid o n the full remuneration, including all the bonuses, and so forth. Thus, from the initial design, there i s both a financing gap in EmeMi Sandigi and an equity issue, whereby lower-grade workers pay contributions o n a larger share o f their salary than higher-grade workers. While the retirement age change includes c i v i l servants, the 1999 l a w otherwise left c i v i l servants untouched. Their benefit as always i s 75 percent o f last salary, based o n all remuneration, with increases after retirement indexed to growth in c i v i l servant wages. In fact, 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 n o impact yet because civil servants had to work a minimum o 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 i s affected by the change yet.

104

Noncontributory pension benefits. 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, o f 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 in the same group. Finally, 136,949 men over age 18 with disability o f 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 m i l l i o n (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 n o w about one-sixth. In addition, the person receives an identification card that provides outpatient care at state hospitals, but n o medicines. Inpatient care i s provided by the hospital itself. The pension i s exempt from all taxes.

Eligibility i s o n the basis o f need as defined by L a w 2022 passed in 1976 and enacted in 1977. People have to apply through the district administrative council where their birth record i s registered. They can do so by mail. A six-person board decides whether the applicant i s needy. F o r the disabled, a health board of three physicians determines whether the person i s truly disabled. B o t h property 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 i s taken into consideration. If the 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 t o qualify under SSK or Bag-Kur if they 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 birth office. There i s some crosschecking with SSK and BagKur databases t o verify that the individual i s not receiving a pension from one o f those systems. Budget issues do not impact h o w many people are paid; if they qualify, they are paid. Issues in Turkish Pensions. The obvious main issue in the Turkish pension system i s its lack o f fiscal sustainability. Figure VII.l shows the fiscal situation o f the pension system, with overall deficits projected to be 4.5 percent o f GNP in 2004, after having hit a l o w o f 2.57 percent o f GNP in 2000. Figure VII.1. Budgetary Transfers to Social Security Institutions

% of GNP

1994

&SSK

1996

+Bag-Kur

1998

-&--

2000

Emekli Sandigt

2002

--rc-Total

105

While S S K i s still slightly below its pre-reform deficit, the large growth in deficits comes from Emekli Sandigi, which as noted, had made few changes in i t s policy since the 1999 reform. While the picture improves somewhat as the slow retirement age i s phased in, the increases in life expectancy in Turkey overcome the pace o f reform and, in the long term, the deficits increase t o 7 percent o f GDP.

But along with the unsustainability o f the pension system i s 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 i s provided to many people, rather than focusing o n 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 o f the work income when they truly retire. L i f e 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 o f contributions, can be spending almost twice as much time collecting pensions as they do contributing t o them. Obviously, a system cannot work this way. Usually pension systems count on contributions from a large group o f contributors t o 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 Turkish society. While the retirement age issue i s being addressed in the 1999 reform, although at a very slow pace, the number o f years required to collect a pension i s rising f r o m 15 years t o only 19.4 years. Even at that rate, people will spend as much time or more in retirement as in working and contributing. In most countries, people contribute twice or three times as long as they collect benefits, so the Turkish situation i s quite abnormal. On top o f that, the benefit i s front-loaded. Workers in SSK receive 35 percent for the f i r s t 10 years o f service-3.5 percent per year. This drops t o 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 o f their average salary. If they work an additional year, they get only 55 percent. It i s 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 t o 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 t o avoid the fairly high employee contributions (21 percent), and the culture o f workers all contributing never gets established. Benefit systems usually pay 1 t o 2 percent, closer t o 1 percent per year o f service. Even if people are allowed to retire with 20 years o f service, if their benefit i s 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 o f 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 in addition 15 percent o f salary for income taxes, with higher rates on higher-income workers. Thus, the take-home salary o f a worker i s 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. During working 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 both Turlush and intemational standards.

106

The one issue that i s consistent with both the perception o f l o w 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. In Bag-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 l o w pensions. The current system appears to reward this behavior by awarding large pensions for minimum years o f service. In addition, periodically, the Government grants additional social support payments t o 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 t o workers that it i s acceptable to under-declare earnings. Pensions will be provided at a reasonable level through the political process in any case. Related to this i s the level o f contributions. The overall labor charges are 36.5 percent o f the wage bill. I t i s 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 i t s primary role, which i s to replace a workerâ&#x20AC;&#x2122;s income during retirement. Workers then look to the political process t o 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 o n 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 o l d age to a limited extent, the above discussion should make clear that the system has significant problems that prevent i t from playing an even more constructive role. The Government i s in the process o f unifying a l l 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 i t s 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 93 1 regional affiliate Foundations (under L a w No. 3294, which entered into effect o n M a y 14, 1986). The Foundations are called vakzjZar in Turkish and in acronym, SYDVs. The purpose o f the S Y D T F was defined as, â&#x20AC;&#x153;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 t o improve social justice and to encourage social assistance and solidarity.â&#x20AC;? The Committee o f the S Y D T F 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 o f 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 i t s 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) t o 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 m i l l i o n from July 1, 1997 through M a r c h 26, 1999).36 The type o f assistance provided by the S Y D T F and i t s regional affiliates i s shown in Table VII.2. Project assistance refers to employment-generating projects, shown in Table VII.3. General revenues and expenditures are presented in Tables VII.4 and VIIS.

Table V11.2. Types o f SYDTF Assistance, July 1,1997-March 26,1999 (in trillion TL)

Other Total

--- = Not available.

1.3 128.4

1 100

Source: Cumhuriyet hukumetleri doneminde soysal hizmetler, ozurluler ve sosyal yardimlar.

4,232,022 Ankara: Nisan, 1999.

36 Turkey also has a program o f 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 â&#x20AC;&#x153;40 percentâ&#x20AC;? disabled, discussed in the pension section. In September 1998, there were over 900,000 beneficiaries, 80 percent o f whom were elderly and 20 percent o f whom were disabled (World B a n k 1999 Living Standards Assessment).

108

Table VII.3. SYDTF Employment-Generating Project Assistance, July 1, 1997-March 26, 1999 (million TL) Type of Assistance

Greenhouse Poultry Carpedtapestry Confection Small handicrafts Milk Fishery Beekeeping Growing h i t s Cow breeding Sheep breeding Miscellaneous

Disabled people Starting own business Growing vineyard Growing plants Culture mushrooms TOTAL

Amount

2,467,5 17 3 15,03 1 61 1,328 469,350 55,865 1,171,004 206,175 1,436,048 898,051 3,634,746 237 1,015 512,043 121,480 69,740 175,039 239,190 151,865 15,105,487

Share (percent) 16.34 2.09 4.05 3.1 1 0.37 7.75 1.36 9.5 1 5.95 24.06 17.02 3.39 0.80 0.46 1.16 1.58 1.01 100

No. of People

9,857 27,656 16,598 2,500 449 11,345 587 4,957 36,943 13,434 11,873 12,688 979 185 1,434 9,602 672 161,759

Central Social Solidarity Fund revenues are comprised o f 0 A 10 percent sum to be transferred from funds that exist or are to be established by statute or decree with the force o f law under the Decision o f the Council o f Ministers 0 Payments inserted into the budget 0 H a l f the revenue o f fines for motoring offenses A 2.8 percent share o f the total payments made by income and corporation taxpayers 0 A 15 percent share to be transferred from Turlush Radio and Television's advertising income 0 All types of donations and assistance from outside 0 Other income.

109

Table VII.4. SYDTF Revenues

0) 0 0 t230,OOO 1,8381 1,303

Fuel consumption tax F r o m national budget Incomes o f previous years

)229,000[ 01 01 01 I505,000[ 7,7951 2,0941

Loan Account o f World Bank Other Income Total income

0 0 0 3,313 7,803 17,114 102,050 2,090 4,380 9,155 454,7241519,627 841,698 890,217 826,583

Source: SYDTF

Table VIIS. SYDTF Expenditures 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 in provinces by the Provincial Governor, and in sub-provinces by the Sub-Provincial Govemor, i s composed in provinces o f the Municipal Mayor, the Security Director, the Provincial Head of Finance, the Provincial Director o f National Education, the Provincial Health Director, the Provincial Director o f Social Services and Child Care, and the Provincial Director o f Religious Affairs. In the sub-provinces, i t i s composed o f 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 o f the Board o f Trustees. The staff working within the Foundations are appointed by the Board o f Trustees in accordance with Labor L a w N o . 1475. Each Foundation has a separate legal entity and no hierarchical connection with the central Fund Administration. Foundations are independent in their decision-making. Taking into consideration the population structure o f the province and sub-province, the socioeconomic development index, and other social factors, each month resources 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 deprived people and families. Support i s 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 in which they can do so.

The medical expenses (for medicine) o f outpatients with green cards who are unable to meet health expenditures from their own means, and the costs o f medicine and treatment o f patients not covered by social insurance, are met by the SYDTF. In addition, all types o f equipment needs o f physically handicapped people (for hearing aids, prostheses, wheelchairs, and so forth) fall within the scope o f 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 o f 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 in areas o f intensive migration, and where there i s 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 o f and food for needy students who travel to the centers where schools are located in the implementation o f education requiring transportation, funds are transferred for the purpose o f providing these students with a mid-day meal. The assistance i s passed on through the SYDVs.

111

For the purpose o f solving the accommodation problems o f 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 o f hostels, thus assuming an important responsibility within the education system. The scholarship program for young people in universities has continued from 1989 to the present. Contributions are also made by the SYDTF to orphanages, old peopleâ&#x20AC;&#x2122;s homes, and rehabilitation centers aimed at the accommodation and rehabilitation o f the weak and destitute.

C. CONDITIONALCASHTRANSFERS Turkeyâ&#x20AC;&#x2122;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 Mitigation Project, SRMP), Turkey began a national program o f conditional cash transfers (CCTs). CCTs are payments made to the mothers o f 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 o f 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 i s 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 in the arsenal o f poverty-fighting interventions in Turkey. They are targeted to the poorest o f 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 children to school. The State Minister for the SYDTF and SYDVs, and the Ministries o f Education and Health, view CCTs as an important intervention that will help Turkey achieve multiple objectives: amelioration o f extreme poverty; secondary school enrollment o f girls; improved health care access for poor children, including immunization in urban 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 births and marriages.

As of M a y 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 o f the CCT program will be undertaken.

112

Table VII.6. Quarterly Figures for CCT PAYMENTS

NUMBER OF BENEFICIARIES EDUCATION HEALTH

TOTAL

EDUCATION

HEALTH

TOTAL

11,233

6,625

17,858

135,05 3

79,65 1

214,704

2003 4 3

25,697

14,647

40,344

233,934

180,228

4 14,162

59,206

28,027

87,233

73 1,472

294,2 18

1,025,690

225,324

87,919

3 13,243

3,194,2 14

847,457

4,041,671

305,666

113,103

418,769

3,979,778

1,389,752

5,369,530

2004

Figures for number o f beneficiaries are cumulative.

Payment amounts are not cumulative.

Q1 includes January and March figures. 42 includes May figures. 43 includes July and September figures. 44 includes November figures. Source: SRMP Project CoordinatingUnit.

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 f r o m 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 t o provide these people with sustainable livelihoods, thereby lifting them permanently out o f poverty. Incidence of Government Programs

Unfortunately, the 2002 HBS questionnaire was not well designed t o capture information about government social protection program receipt. For example, there i s a question asking, â&#x20AC;&#x153;Do you receive any transfers from persons (neighbor, relative, etc.) or foundations (SYDVs)?â&#x20AC;? but this question i s not further subdivided, so government programs cannot be analyzed separately from private charity or transfers. While there are detailed questions about transfers in kind in which government programs can be separated from private programs, this i s not the case for cash, which i s 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 i s presented in Table VII.7. All government transfers except old-age income and transfers in kind are regressive. This analysis i s based o n nominal transfers reported on an annual basis. I t i s not clear h o w 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: Incidence o f Nominal Transfers (annual, undeflated)

institutions in the last 12 months

14,024,605

11,869,626

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 i s presented here in greater detail. The source o f data i s the 2002 Household Budget Survey (HBS) data o f the State Institute o f Statistics (Devlet Istatistik Enstitusu, DIE). The methodology i s well established in the poverty literature and in World Bank research and 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 o f Well-Being T w o measures o f economic welfare are constructed in this report4onsumption expenditure and income-although our preferred measure o f 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 using consumption as the welfare indicator, as i s standard World Bank practice. For Turkey, it should be noted that an extensive informal sector and flows o f private remittances, coupled with a l o w level o f tax declaration compliance and enforcement, make money income an inferior measurement o f household welfare. In addition, a very important source o f food for not only rural households, but also many urban households in the gecekondu areas, i s food produced in garden plots, including both 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 i s used. In particular, the previous methodology included expenditures on consumer durables which, following Deaton and Zaidi (2002), have been excluded from the new methodology. T h e previous methodology did not include the imputed value o f owner-occupied housing in the 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 o f 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 rendered by 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 household was not available, nor was the age or purchase price o f the durable item. That i s why the imputed or reported value o f durable goods i s not included in the consumption aggregate. The housing rental market i s well developed in Turkey (according to the 2002 HBS data set, 20.2 percent of households were renting their housing), malung i t 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 consumption received by the household f r o m occupying i t s own dwelling. The significant share o f households paying rent makes possible the imputation o f the value o f owner-occupied housing. B o t h 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 i s done by a semi-log regression o f monthly rent on housing attributes, and then the regression coefficients were used to impute rent based o n the characteristics o f owneroccupied housing.

Income Aggregate The household income aggregate was constructed using the income section o f the HBS. The information on income i s collected o n 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 in country and remittances from abroad.

Methodology of Making Household Welfare 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 onetwelfth o f the total sample o f 9,555 interviewed in each month. Inflation was significant in Turkey, averaging 29.7 percent during 2002. 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 using three-dimensional price indexes by regions, type o f settlement, and months i s one of the possible solutions t o deflate the nominal welfare measures for the different times and areas, and to make households comparable to each other and t o any single national poverty line. Thus, real consumption can be compared to a single poverty line.37

The alternative approach to the solution o f the problem o f nominal values needing to be deflated i s 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 in this 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 i s not used in this report. 37

116

Household consumption and income, as well as other monetary measures, were deflated using threedimensional 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 C P I varies across a l l three dimensions: regions, urbadrural location, and months. The total number o f CPI i s equal t o 168 (7 regions times 2 settlement types times 12 months). However, it i s more preferable to use a price index from inside the survey database, based on survey prices, because a l l other information used for the poverty analysis i s based o n survey estimates, and introducing another data source could lead to bias (Glewwe and H a l l 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 o n the basis o f survey information, because o f the insufficient number o f purchases o f such items in terms o f 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 i s n o significant difference between the poverty figures based o n deflation by these two different price indexes. Appendix 3 contains the findings o f the sensitivity analysis.

Using Equivalent Measures T o make welfare comparisons across households with different demographic compositions, some way i s 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 t o 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. T o 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 i t s income-based poverty comparisons, and one from the United Nations Food and Agricultural Office (FAO) based o n dietary information.

I17

Various equivalent measures for income and consumption are considered in this appendix. Different scales can have major impacts on poverty measurement and the profile of the poor (Lanjouw, Milanovic, and Patemostro 1998).

Adult Equivalent Size General Formula T o measure the effects o f 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:

where Ai i s the number o f adults in the household, Ci i s the number o f children, and a and 6’ are parameters. Children are individuals aged 14 and below. In this appendix a value o f e= 0.6 was adopted to ensure comparability with other World Bank regional poverty findings. Two different values o f a were considered in this analysis (a = 0.9 and a = 1). The batch mode programs developed for the analysis allowed examination of the impact o f other possible values o f 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 f the household using the formula shown below. Adjusted adult equivalent size o f the household i (AE-ADJ;) i s defined as

AE-ADJ, =

+co AEi (A0 + a cole A0

where Auand Cuare the number o f adults and children in the “pivotal” household, respectively, and Aiand Ci are the number o f adults and children in the i” household.

The modal or pivotal household in Turkey i s a 4-member household with 2 adults and 2 children (Au = 2 and C, = 2). AEi here i s from the general formula above. OECD Equivalence Scale AE-OECD = 1+ [number of children under age 141 x 0.5 + ([number o f adults]

- 1) x 0.75.

F A 0 Equivalence Scale

AE FA0 = (n-ch5 x 0.64 + nch5-11 + nm12-17 nmZ0- x 0.88 + nf40- x 0.76) x 0.75

+ nfl2-17

x 0.84

+ nm18-39 +

nfl8-39 x 0.84 +

where n-ch5 i s the number o f children under age 5, nch5-11 i s the number o f children aged 5 to 11, nm12-17, nf12-17 are number of male and female children, respectively, aged 12 to 17, and nm40- and nmf40- are the number o f males and females aged 40 and above, respectively.

118

Table A.I. 1 describes the relation between 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 b y the Set-Type of Settlement rype of ettlement rurkey

Rural

Urban

Household size

4.30

4.07

4.16

Adult equiv. size general formula

2.29

2.23

2.26

AE adjusted by modal HH (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 o n equivalence scale type and parameters a and 8 are presented in Appendix 6. Since there i s n o 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 t o which to use. This analysis has chosen the adjusted adult equivalence scale based o n the modal household (Deaton and Zaidi) approach for its basic approach, and compared i t to the main poverty line (explained below) using a = 0.9 and 8 = 0.6.

Using Day-BasedMeasures T o make welfare comparisons across households surveyed during months 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 o f 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 30.41â&#x20AC;&#x2122;7 (the average number o f days in the month).

Structure o f Income and Consumption Income The average household disposable income according to the survey data was around TL672 million per month for 2002, o f which TL639 m i l l i o n i s the average cash income and the other TL33 million i s the average household income in kind.

Structure o f 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 o f cash income.

Table A.I.2. The Structure o f Cash Income ttlement Urban Share

1 0.4340 0.2206 0.0144

Total net income in cash from a l l previous jobs in the last 12 months Rent from real estate income in cash in the survey month Interest income in cash from movable property in the survey month Transfer income in cash in the survey month

0.0198 133,371,432

76,957,193

168,881,759

0.2086

0.1416

0.2414

26,197,746

10,929,818

35,808,248

0.0410

0.0201

0.0512

31,794,266

15,809,179

41,856,189

0.0497

0.0291

0.0598

116,345,155

87,212,289

134,683,036

0.1820

0.1605

0.1925

The Structure o f Consumption T h e main i t e m o f consumption o f T u r k i s h households i s the category o f housing a n d water consumption (28.7 percent). T h i s estimate includes the value o f real housing rents for households renting the apartment or house, and the value o f imputed rents for house and apartment owners. T h e second important i t e m i s food consumption (25.45 percent). For urban and rural households there i s a slight difference in the importance o f consumption items. T h e m a i n i t e m for rural households i s food consumption (26.96 percent), and the second i t e m i s housing (22.73 percent), w h i l e for the urban households the m a i n i t e m o f consumption i s housing (3 1.53 percent), and the second i t e m i s food consumption (24.73 percent).

120

Table A.I.3 describes the structure o f consumption.

Food received as income in kind from activities and real estate Non-food goods and services received as income in kind from activities and real estate F o o d received as a present Non-food goods and services received as a present

2,954,25 1

2,272,086

3,383,645

0.0053

0.0049

0.0055

23,219,048 5,384,374

13,191,239 5,513,449

29,531,122 5,303,127

0.0418 0.0097

0.0284 0.01 19

0.0481 0.0086

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 wellbeing of the population. One o f the common indicators o f inequality i s the decile 90/10. In this analysis, the decile 90/10 i s the adjusted (by the modal household) per adult equivalent deflated consumption (PEC) o f the 90th percentile divided by the PEC o f the 10th percentile, that is, the PEC o f the poorest person in the richest decile over the PEC o f the richest person in the f i r s t or poorest decile. This indicator i s easy to interpret, but i t 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 in Table A.I.4.

Table A.I.4. Average Monthly Consumption Per Capita and Per Adult Equivalent, by Deciles TL Per Month

Adult Equivalent Adjusted with Modal HH

55,348,090

100,4363 17 115.135.441

Turkey

132,137,395

Source: DIE, Turkey 2002 HBS.

96,184,245 112.996.782

133,701,603

Another common inequality measure i s the Gini coefficient, which i s sensitive to a l l the parts o f distribution. The Gini coefficient i s bounded between 0 and 1; i t i s 0 in the case o f absolute equality, when the per equivalent consumption o f each person i s the same, and i t i s 1 in the case o f absolute inequality, when one person consumes everything and others consume nothing. The Gini coefficient i s g v e n 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 householdâ&#x20AC;&#x2122;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. 1 describes the inequality based on PEC.

Figure A.I.l. Cumulative Distributiono f PEC

0.9 0.8 0.7 0.6 0.5 0.4 1 0.3 0.2 , 0.1 0

The Gini concentration coefficient i s calculated on the basis o f individual information o f the H B S database, which i s more accurate than if we use consumption or income by deciles. The Gini coefficient i s not an additive or decomposable measure, meaning that i t 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 t o inequality in the top o f the distribution, while the mean l o g 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 t o other countries or the same country at different points in time. B o t h 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 in Table A.I.5 are the Theil T index and Theil L index.

123

Table A.1.5. Inequality Measures

Both indexes indicate that inequality i s higher in urban areas than in rural areas. Inequality based o n income i s greater than inequality based on consumption.

Poverty Poverty Lines and Poverty Rates

Where the poverty line i s drawn depends at least o n h o w 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 â&#x20AC;&#x153;food poverty lineâ&#x20AC;? i s chosen. I t should be emphasized that there i s n o perfect indicator identifying the living standard. World Bank practice i s to set at least two poverty lines, one for extreme or food poverty, the other for overall poverty. However, the W o r l d Bank also uses several absolute poverty lines for international comparisons, and relative poverty i s used in the OECD countries for comparison. This appendix lays out several possible lines, and decides o n 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 i s estimated as TL663,575 for 2002. This i s DIEâ&#x20AC;&#x2122;Scurrent estimate o f PPP. In World Bank publications, such as the World Development Indicators, PPPs are often estimated by taking a base year and extending i t by the CPI. In the comparisons chapter, two different kinds o f PPPs are used, one the extended-by-CPI version, and the other the current DIE PPPs. Using the current DIE PPP, 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 i s drawn as 60 percent o f median consumption per capita. I t i s estimated as TL2,377,139 per day per adult. The relative approach i s 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 o f the population, priced out by using the country average survey prices for 2002. The calorie intake information from the 2002 H B S survey was calculated using the United States Department o f Agriculture (USDA) nutritional database. The composition o f the minimum food basket was calculated on this base to reach 2,100 Kilocalories per day (Kcal) per average person 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, i t i s estimated that the amount necessary for attaining the minimum food consumption i s TL1,082,359 per person per day in an average Turkish family. Table A.I.6 describes the structure o f a minimum food basket based on the consumption patterns o f the reference population, which are population deciles 3 and 4 by food consumption.

Table A.I.6. The Structure o f a Minimum Food Basket Estimated Based on Survey Data

Detailed information about the structure and cost o f a minimum food basket i s found in Appendix 5.

Complete Poverty 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 method o f Ravallion (1994) i s used here to determine the value o f the general poverty line. To determine the allowance for non-food consumption, using the survey data itself, first those individuals whose total consumption i s just above the value o f the food poverty l i n e are selected. This part o f 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 calculated for this reference group. This share i s the â&#x20AC;&#x153;allowanceâ&#x20AC;? for non-food consumption that 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 i s just above the value o f the food poverty line i s 57 percent, and food consumption represents 43 percent. The value o f the complete poverty line i s thus:

Complete Poverty Line (CPL) = Value o f Food Consumption + Value o f Non-Food Consumption where: Food Consumption = Value o f Food Poverty Line = TL1.08Mlm = 43% o f CPL Non-food Consumption = 57% o f CPL CPL = TL1.08Mln divided by 0.43

= TL2.5Mln = TL1.08

Mln + TL1.42 Mln.

The CPL i s 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 i s based on the consumption pattems o f population whose adjusted PEC i s just above the food poverty line. These households are in the first decile o f PEC. This method o f deriving the complete poverty line i s the simplest way to assess the value o f the minimum consistent with the actual consumption pattern o f the population. We have chosen the simplest method described above, because i t i s the most transparent, most easily replicable, and most intuitive. It could be argued that other lines would be more accurate. However, if a certain way to set the line i s not commonly understood, i t s use will not help the national poverty diagnostics. Given the fact that any poverty line i s 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 o f poverty in Turkey. The value o f the CPL i s presented in Table A. 1.7.

I I

Household Size

Average Poverty Line for Household Deflated

1 2

137.055.990 207,561,617

309,55 1,477

7 8 9 10

1 I

431,521,387 466,737,285 500.4 14.569 532,742,054

I 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. I t i s the most frequently used poverty measure. The main advantage o f this statistic i s i t s simplicity. If q i s the number o f poor people in the population o f size n, then the headcount i s given by:

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 Gap Index. Let Q be the subgroup o f poor; the poverty gap i s then given by:

The poverty gap also allows an interpretation in terms o f the potential fiscal cost for eliminating poverty by targeting transfers to the poor. Summing a l l the poverty gaps in the 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 t o the consumption (income) gap o f 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 a i m to reach the poorest, but i t i s 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 in the main text, but are in levels. T o obtain percentages, multiply by 100.

Table A.I.8. Poverty Headcount, Gap, and Severity Index, Based on Adjusted and Deflated PEC and Complete Various Poverty Lines

Food Poverty Line 1.082.359 TLlday

Complete Poverty Line 2.510.930 TL/day

Re'ative Poverty Line 1.980.949 TLlday

Poverty L i n e s 1

0.0135

0.2696

0.1474

Poverty Line=$2.15 PPP=1426686 TLlday

Poverty Line=%4.3 PPP=2853373 TLlday

0.0021

0.041 1

0.3414

ppp=663575 TLlday

0.0026

0.0688

0.03 13

0.0004

0.0083

0.0973

0.0008

0.0253

0.0104

0.0001

0.0027

0.0384

127

Other poverty figures based o n other welfare indicators, CPI, parameter o f economies o f scale, and adult equivalence are presented in Appendix 3.

Sensitivity of Poverty Statistics The poverty indexes are sensitive t o 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 o n several key parameters o f how they were calculated:

CPI DIE

Imputed Reported Rent by Rent by HH Regression Food Poverty 0.0438 I 0.0462 0.0083 0.0090 0.0026 0.0028 0.1894 0.1943 Complete Poverty 0.2896 0.2810 0.0764 0.0749 0.0288 0.0287

PO PI P2 Shorkfall PO PI P2

CPI Survey

Imputed Rent by Regression

Reported Rent by HH

0.0379 0.0067 0.0021 0.1775

0.0388 0.0073 0.0022 0.1877

0.2696 0.0688 0.0253

0.2618 0.0691 0.0259

More detailed information about the results o f the sensitivity analysis can be found in Appendix 3.

Comparison with Previous World Bank Estimates There are significant differences between the main poverty line used in this report and the poverty line and consumption aggregate o f the W o r l d Bankâ&#x20AC;&#x2122;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. T h i s poverty line i s much higher than the main one used herein. The data collected in both 1987 and 1994 seem to adequately address most o f the problems in measuring well-being. Unfortunately, it was impossible to follow the new methodology to recalculate basic results for 1994, because the 1994 data are not in a 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 t o 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 Turkish household data includes:

All monetary non-business and non-investment expenditures the (2002 version would exclude all durables; the 1994 version would exclude only expenditures o n selected items) Gifts, earnings, and transfers in kind Consumption from stocks Consumption from own production

128

Imputed rents from owner-occupied housing (imputed by regression in 2002 methodology and self-reported in 1994).

In addition, the methodology developed for the analysis o f 1994 data relied on spatial and time indexes o f prices from the official CPI statistics, while the new methodology i s based on survey unit values.

T o obtain comparable figures, the 1994 methodology was applied to the 2002 HBS, including the use o f the CPI instead o f the survey price index. Data are broadly comparable, but some important differences are noticeable in the detailed structure o f both aggregates. Table A.I.10 shows the consumption o f food per household. While overall the change o f total perhousehold consumption i s within the plausible range, some important changes in the structure reflect the comparability problem. In the 1994 consumption aggregate, both in-kind consumption o f own food and consumption from stock were included. The DIE was concerned that this might lead to double counting. In the new version o f the questionnaire, much more detailed information on consumption from stocks i s included, and only own production i s included in the food consumption.

Table A.I.lO. Food Consumption and Expenditures, 1994 and 2002 Consumed I n

Total Food per Household, TL/Month

Rural Urban

203.759.737 200,831,846 201.962.890 I

3.344.540 , , 3,988,774 3,676,828 4-2002. Times 60.9 50.3 54.9 ,

Urban

73.1% 87.6% 82.0%

62.8% 79.1% 72.6%

71.0 55.7 62.0

Earnings o r Gifts (aid)

Production

0.7% 6.5%

30.4% 7.8% 17.5%

31.8 4.8 20.5

Outside

4.9% 4.6% 4.8%

3.1% 3.9% 3.6%

96.5 59.9 73.1

I I

6.1% 7.0% 6.7% 3.8% 9.1% 6.3% 99.6 38.6 58.4

This factor, with some economic changes, explains the very small change in the consumption o f own food for urban households. On the other hand, the consumption o f meals outside o f the home has clearly not kept pace with the overall consumption o f food, especially in urban areas, suggesting that the current questionnaire forms for collecting these types o f expenditures may need revisions. Creating the “previous” methodology consumption aggregate included these adjustments: 0 Household consumption includes the purchase o f all durable goods, excluding cars. 0 The FA0 adult equivalence scale was used. 0 The official CPI was used for deflation. 0 The value o f the minimum food basket o f the P A was used as a food poverty line. 0 The quantities from the Hacateppe basket were priced out by using the 2002 survey prices. 0 The same non-food share i s added to the poverty line for different regions and settlement types.38 38 N o t e that

the PA generated these constants by regression. W e simply use t h e m “as is.”

129

The 2002 poverty line comparable with the P A i s TL2,361,383 = US$3.60 PPP at 2002 prices. Poverty rates according to the poverty line based on the minimal food basket used for P A 1994 and using comparable methodology are given in Table A.I. 11. Table A.I.ll. Comparable Poverty Measures, 1994 and 2002

2002 HBS, using 1994 food basket and previous algorithm for poverty line and welfare measure PA

Complete pl P2 Poverty Complete Complete

2002

0.352

0.108

0.045

1994

0.363

0.109

0.046

Thus, there was a very slight decrease in poverty between 1994 and 2002. Comparison o f 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 used in this analysis o f the 2002 HBS compared to what was done for the previous P A using the 1994 data. Thus, the poverty figures presented above are not comparable with poverty figures o f the 1994 PA. In the previous PA, the food basket was developed by a Turkish academic institution, and i t 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 P A (1994 data) i s around 3,000 Kcal/day. I t i s important to note that the new survey-based basket gives a much more realistic structure o f consumption o f the poor population than the one used for the 1994 data analysis. Table A.I. 12 lists the approximate comparison. Table A.I.12. Food Basket for Equivalent Adult: 2002 Survey-Based and HacettepeUniversity (kilograms per day per adult)

... = Not available.

n.e.c.

= Not elsewhere

classified.

The food poverty line based on the food basket developed by Hacettepe University i s 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 consumption o 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 o f the third and fourth deciles o f population (TL1,082,359 per person per day).

If we price out the minimum food 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 o n this poverty line i s 75 percent, which i s far too high to be usable for either public policy applications or research.

131

Appendix 1: Income and Consumption by Poverty Status and UrbadRural Dimension 1

Household's Average Monthly Consumption (TL) Turkey

Type o f Settlement

Consumption

02,387,828 510,357,187 660,317,142

700,347,775

Food and nonalcoholic beverages

41,54 1,355 125,083,823 15 1,900,660

157,345,897

r I

I ~

24.158.598

Clothing and footwear

36,289,947

Housing, water supply

59,589,070 105,441,687 193,672,510

Health

Alcohol and tobacco

Furniture HH appliances and home care services

Poverty Status Non-Poor Households

Urban

Rural I

43,2 16,912

23.370.600 30,474,680

40,37 1,957

24.654.608 39,950,407

27,146,480

14.236.1781

16.921.274

45,007,688

52,530,573

Poor Households

264,094,536 86,962,205

I I

68,670,110

1 I

11,053,266'

4.480.470 !

Transportation

46,917,368

33,023,943

55,662,679

58,045,979

8,486,002

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

Education Restaurants and hotels

8,387,213

4,497,613

10,835,548

10,675,340

485,429

15,037,810

13,358,012

16,095,170

18,077,677

4,539,983,

Various goods and services

25,330,134

22,389,654

27,18 1,039

30,953,323

5,911,102

Own produced food for s e l f consumption

14,465,225

35,028,585

1,521,478

13,837,979

16,631,345

Own produced non-food for selfconsumption

3,646,997

9,041,038

251,681

3,466,240

4,271,221

Food received as income in kind from activities and real estate

2,954,25 1

2,272,086

3,383,645

3,328,750

1,660,964

I ~

Non-food goods and services received as income in kind from activities and i real estate

23,219,048

Food received as a present

5,384,374

5,513,449

5,303,127

5,418,279

5,267,284

Non-food goods and services received as a present

6,797,208

5,596,940

7,552,724

7,798,547

3,339,198

13,840,292: 9,740,489

43,977,902 185,916,570

14.127.7121 13.955.395

13,191,239

2933 1,122

28,219,806

5,949,515

Source: DIE. Turkev 2002 HBS.

132

Household's Average Monthly Food Consumption (TL) Poverty Status

Turkey

Yon-Poor HH Poor HH

Food and nonalcoholic beverages

1413 4 1,355

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

1,893,908

2,198,002

843,758

Milk, cheese, and eggs

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

8,388,184

9,5 11,660

4,508,388

Fish

Nonalcoholic beverages Source: DIE, Turkey 2002 HBS.

Household's Average Monthly Alcohol and Tobacco Consumption (TL) Turkey Alcohol and tobacco Tobacco

Poverty Status Non-Poor HH

Poor HH

24,158,598 23,370,600 24,654,608

27,146,480

13,840,292

1,920,985

2,362,809

58 1,949

22,195,678 2 1,34 1,060 22,733,623

24,783,671

13,258,343

1,962,920

Alcohol

Type of Settlement ' Rural

2,029,541

Urban

Source: DIE, Turkey 2002 HBS.

Household's Average Monthly Clothing and footwear Consumption (TL) Turkey Clothing and footwear Fabrics Clothing (man, woman, children) Accessories and other clothing

Dry cleaning, mending, and hiring Footwear Repair and hiring o f footwear

Type of Settlement

Non-Poor HH

Poor HH

43,977,902

9,740,489

1,004,577

1,278,976

331,337

20,230,937 27,382,252

30,215,310

5,295,952

Rural

Urban

36,289,947 30,474,680 39,950,407 1,066,185 24,619,709

Poverty Status

1,164,060

1,426,808

1,253,045

1,536,184

1,64 1,346

685,926

416,676

286,582

498,565

526,456

37,565

8,636,495

7,460,496

9,376,736

10,17 1,666

3,334,960

124,073

79,560

152,092

144,148

54,747

Source. DIE, Turkey 2002 HBS.

133

Household's Average Monthly Housing Consumption (TL) Type of Settlement Turkey

Rural

Urban

Poverty Status Non-poor HH

Poor HH

Housing, water supply

159,589,070

105,44 1,687 193,672,5 10 185,9 16,570

68,670,110

Housing rents (real for owners and imputed for non-owners

77,976,501

44,322,664

99,160,139

87,538,065

44,956,754

Maintenance and repair

8,355,128

8,859,876

8,037,411

10,603,146

591,859

Various services related to 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, Turkey 2002 HBS.

Household's Average Monthly Expenditures on Furniture HH Appliances and Home Care Services (TL) Type of Settlement Turkey

Poverty Status

Rural

Urban

Non-Poor HH

Poor HH

$3,216,912

40,371,957

45,007,688

52,530,573

11,053,266

15,451,131

16,994,088

14,479,905

19,270,358

2,261,871

Household textile

3,568,869

3,792,127

3,42 8,3 3 8

4,536,687

226,622

Household appliances

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

Furniture HH appliances and home care services Furniture, household textile, carpet and other

floor coverings

House and garden appliances Nondurable household goods and services

972,847

1,2 19,101

817,840

1,108,330

504,97 3

10,155,452

7,42 1,129

11,876,590

12,015,775

3,731,041

Source. DIE, Turkey 2002 HBS.

Household's Average Monthly Expenditures on Health (TL Turkey

Type of Settlement Rural

Urban

Poverty Status Non-Poor HH Poor HH

Health

.4,127,712 13,955,395 14,236,178

16,92 1,274

4,480,470

Medical products, appliances, and materials

6,155,298

6,276,510

6,079,001

7,036,015

3,113,846

Treatment without keeping in hospital

6,366,685

6,560,875

6,244,45 1

7,832,358

1,305,156

Treatment in hospital

1,605,728

1,118,010

1,912,726

2,052,902

61,468

Source DIE, Turkey 2002 HBS.

134

Household's Average M o n t h l y Expenditures o n Transportation (TL) Type o f Settlement Turkey

Rural

Poverty Status

Urban

Non-Poor Households

Poor 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

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

9,338,287

Source: DIE, Turkey 2002 HBS.

Household's Average M o n t h l y Expenditures o n Communication (TL)

Type of Settlement Turkey Communication

Rural

27,362,109

Poverty Status

Urban

Non-Poor Households

Poor Households

32,599,264

9,276,204

20,820,121 3 1,480,008

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

23,94 1,048 17,027,500 28,292,828

28,355,165

8,697,408

Postal services

Source: DIE, Turkey 2002 HBS.

Household's Average M o n t h l y Expenditu es o n Entertainment and Culture(TL) Type of Settlement Turkey

Rural

Urban

Poverty Status Non-Poor Households

Poor Households

18,169,140

3,442,372 1,961,602

Entertainment and culture

14,862,268

12,101,713 16,599,919

Visual, auditory, and photography equipment

5,858,250

7,012,750

5,131,541

6,986,606

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

5 10,499

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

35 1,999

Newspaper, book, and stationery

3,878,372

2,617,93 1 4,67 1,764

4,715,980

985,787

Tours

1,065,695

1,373,599

2,384

8,831

1,730,945

Source: DIE, Turkey 2002 HBS.

135

Household's Average Monthly Expenditures on Education (TL) Type o f Settlement

Turkey

Rural

Urban

8,387,2 13 4,497,613 10,835,548

Education

Pre-primary and primary education 1,478,850

Poverty Status Yon-Poor Households Poor Households

10,675,340

485,429

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 1,2 18

4,607,833

162,840

University education

1,822,2 12

883,182

2,4 13,29 1

2,349,872

419,198

77,488

634,289

526,248

49,512

Education not definable by level Source: DIE, Turkey 2002 HBS.

Household'sAverage Monthly Expenditures on Education (TL) Turkev

Type of Settlement Rural

Urban

Poverty Status 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

2,018,640

59,545

Accomodation services

1,578,728

866,819

2,026,844

Source: DIE, Turkey 2002 HBS

Household's Average Monthly Expenditures on Education (TL) Turkey

Type o f Settlement Rural

Urban

Poverty Status 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

5,373,555

7,150,584

4,254,992

6,608,523

1,108,737

30,364

49,477

39,156

1,467,226

586,990

2,02 1,297

1,874,328

6 1,349

207,427

235,172

189,962

266,150

4,635

8,230,839

7,735,537

8,542,610

10,185,620

1,480,230

Other personal care Materials Social services Insurance Financial services

Other services Source: DIE, Turkey 2002 HBS.

136

Household's Average Monthly Income (TL) Type of Settlement

Turkey

Rural

Urban

Total monthly disposable income

672,103,224 564,545,784 739,805,987

Total monthly income in cash

639,211,405 543,466,796 699,478,501 32,891,819

Total monthly income in kind

21,078,988

40,327,486

Source: DIE, Turkey 2002 HBS.

Household's Average Monthly CASH Income (TL) Turkey

Type o f Settlement

Poverty Status

Rural

Urban

Non-Poor Households

Poor Households

j43,466,796

j99,478,50 1

738,542,973

296,181,381

Total monthly income in cash

639,211,405

Total net cash income as wage or salary in the survey month

24 1,134,556 141,984,036 303,545,527

281,879,866

100,425,364

N e t disposable income in cash in the survey month

125,712,674

80,282,906

154,308,750

148,329,645

47,607,591

Total net agricultural income in cash in the survey month

74,346,597

176,424,673

10,092,856

74,172,091

74,949,230

Total net income in cash from additional job in the survey month

19,295,668

27,982,891

13,827,436

21,735,790

10,868,989

133,371,432

76,957,193

168,88 1,759

127,474,292

153,736,516

Rent from real estate income in cash in the survey month

26,197,746

10,929,818

35,808,248

32,685,823

3,79 1,927

Interest income in cash from movable property in the survey month

3 1,794,266

15,809,179

4 1,856,189

39,973,071

3,549,714

87,212,289 134,683,036

136,385,222

47,139,116

Total net income in cash from all previous jobs in the last 12 months

Transfer income in cash in the survey 116,345,155 month Source. DIE, Turkey 2002 HBS

137

Household's Average Monthly Income I n Kind

L) Poverty Status

Type of Settlement Turkey

Rural

Urban

Non-Poor Households

Poor Households

32,891,819

:1,078,988

,0,327,486

38,174,736

14,647,880

Total income in kind as wage or salary in 22,482,203 the survey month

2,43 1,164

18,808,898

27,322,675

5,766,196

Total income in kind as entrepreneurship income in the survey month

3,500,954

2,78 1,567

3,953,778

4,013,223

1,73 1,892

7,585

14,474

3,249

9,782

182,557

236,119

148,842

202,875

112,392

4,823,847

2,042,993

6,574,275

4,856,705

4,710,377

2,053

3,345

2,647

6,483,009

5,298,640

7,2283 19

6,413,460

6,723,191

Total monthly income in kind

Total net agricultural income in kind in the survey month Total income in kind from additional job in the survey month Total income in kind from all previous jobs in the last 12 months Rent from real estate income in kind in the survey month Imputed rent for the survey month Transfer income in kind in the survey month Source: DIE, Turkey 2002 HBS.

Household's Average Annual Income (TL) Type of Settlement Turkey

Poverty Status

Rural

Urban

Non-Poor Households

Poor 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 cash

6,964,504,409

5,879,567,064

7,647,425,6 17

8,06 1,387,046

3,176,547,7 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's Avera

Annual Cash Income (TL) Type of Settlement

Turkey

Rural

Poverty Status Non-Poor HH

Urban

Poor HH

Total annual income in cash

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 in the 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 in the 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., in the last 12 months

35,172,034

12,356,576

49,533,379

44,111,001

4.302.351

Total expertise, counseling income in the last 12 months

2,337,690

213,510

3,674,769

3,013,178

4,969

1,656,452,040

478,033,898

Total per diem (travel allowance) income in the last 12 months

Total net disposable entrepreneurship income in cash . ,391,840,185 in the last 12 months

842,269,3 85 1,737,771,271

Total net agricultural income in cash in the last 12 months

892,143,203

!,117,055,050

121,114,091

890,05 1,084

899,368,096

Total net income in cash from additional j o b in the last 12 months

182,262,486

267,491,791

128,614,3 19

205,369,279

102,465,863

Total net income in cash from all previous jobs in the last 12 months

133,37 1,432

76,957,193

168,881,759

127,474,292

153,736,516

Rent from real estate income in cash in the last 12 months

296,235,392

121,396,133

406,289,156

370,04 1,275

41,355,353

Profit income from bank 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 in the last 12 months

15,921,198

22,954,920

11,493,774

19,978,704

1,909,075

Interest income from bond, debenture in the last 12 months

62,046,915

31,452,401

8 1,304,841

76,204,O 16

13,157,012

Interest income from capital association in the last 12 months

44,757,623

1,504,454

7 1,983,626

57,702,776

53,043

Interest income from private finance council in the last 12 months

10,975,697

582,618

17,517,691

14,151,698

7,744

Pension from government in the last 12 months

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

69,285,744

17,038,255

73,508,7 10

139

Household's Averar

Annual Cash Income (TL) Type of Settlement

Turkey

Rural

Urban

Poverty Status Non-Poor HH

Poor HH

Scholarship from government in the last 12 months

3,206,961

2,708,054

3,521,002

3,721,242

1,430,950

Old-age income from government in the last 12 months

43,412,524

96,626,467

9,9 16,645

40,071,717

54,949,613

1,425,121

2,805,205

556,417

1,811,657

90,264

Orphan and widower pension from government in the last 12 months

93,558,512

65,086,869

111,480,182

110,877,439

33,749,612

Disability (war veteran) pension from government in the last 12 months

10,035,59 1

4,762,177

13,354,978

12,057,892

3,051,809

Pension from abroad in the last 12 months

35,115,146

52,714,058

24,037,391

42,914,302

8,181,668

8,254,063

10,332,200

6,945,965

10,400,923

840,131

21,380,827

16,066,068

24,726,239

22,7 15,410

16,772,002

8,099,720

5,992,960

1,440,936

149,871,294

135,932,282

Unemploymentpension from government in the last 12 months

Foreign currency from abroad in the last 12 months Scholarship, alimony, etc., from abroad in the last 12 months Alimony income from other private persons or institutions in the last 12 months

4,970,810

Scholarship, alimony, etc., from other private persons or institutions in the last 12 months

128,668,333

94,983,796

103,583,13 1

Source: DIE, Turkey 2002 HBS.

140

Household's Average Annual Income I n Kind (T Poverty Status

Type of Settlement Turkey

Rural

Urban

Non-Poor Households

Poor Households

Total annual income in kind

300,756,381

177,267,287

378,487,433

344,929,771

148,208,714

Total income in kind as wage or salary in the last 12 months

!04,588,895

96,291,576

!72,757,380

246,762,2 17

58,948,229

Total net entrepreneurship income in kind in the last 12 months

39,046,736

29,247,897

45,2 14,68 1

44,308,530

20,875,740

414,796

589,529

304,809

458,947

262,326

Total income in kind from additional job in the last 12 months

3,205,989

5,337,639

1,864,208

2,909,242

4,230,773

Total income in kind from all previous jobs in the last 12 months

4,823,847

2,042,993

6,574,275

4,856,705

4,710,377

Rent from real estate income in kind in the last 12 months

854,378

1,753 8,184

285,470

820,816

970,279

Total net agricultural income in kind in the last 12 months

Imputed rent for the last 12 months Income in kind from government in the last 12 months

6,021,880

9,452,001

3,862,767

5,565,361

7,598,415

Transfer income in kind from abroad in the last 12 months

491,855

102,777

736,763

569,959

222,132

Transfer income in kind from other private persons or institutions in the last 12 months

40,002,97 1

29,925,532

46,346,284

37,145,3 10

49,871,573

Source: DIE, Turkey 2002 HBS.

141

Appendix 2: Poverty Figures f o r Various Poverty L i n e s

The Poverty Headcount According to Various Poverty Lines, Based on Consumption Per Capita Food Poverty Line TL

Zomplete Poverty Line TL

Relative Poverty i n e 50% of Median TL

?raction o f Poor

Fraction of Poor

.0379

.3 167

.2030

.0038

.09 18

.3893

.lo35

S902

.4701

.0158

.2380

.6639

.0324

.3240

.1887

.002 1

.Of303

.4 105

,0054

.lo83

.0541

,0209

.1466

University

.0012

,0228

.0111

.oooo .oooo

.0012

.0284

Rural

.0556

.3989

.2700

.0056

.1289

.4822

Urban

.0261

.26 19

,1583

.0026

.067 1

,3275

.0365

.2832

.1781

.0044

,0820

.3500

Yes

,0407

.3842

.2532

.0027

.1117

.4687

‘No

.0487

.3850

.2537

.0050

.1179

.4665

,Yes

.0013

.087 1

.0323

.oooo

.0042

.1298

Zero

.0019

.1283

.0594

.0008

.0126

,1792

.0107

.2304

,1167

.oooo

.032 1

.3044

.O 179

.3291

.1909

.0067

.0554

.4254

.0906

S789

.4151

.0040

.2440

.6705

4 and more

.1958

.7670

.6277

.0151

.3827

3276

Zero

.0342

.2989

.1891

.0038

.0874

.3708

One

.0298

.37 18

.2266

.oo 18

.0700

.4550

Two and more

.0820

.3569

.2697

.0078

.1677

.4 160

Male

.0380

.3 185

.2030

.0036

.0925

.3913

Female

.0352

.2905

.2028

.0071

.0823

.3613

,0543

.3802

.2545

.0057

.1204

.46 16

.0117

,2273

,1332

,0007

.05 10

,2887

,0016

.1144

.0243

.oooo

.0035

.1532

rurkey Lower than primary Education o f HH lead

rype o f settlement Access to land

Own automobile

’ primary

’ Secondarylvocational

One Children under age Two 14 Three

Number o f HH members age >60

Gender o f HH head

Dependency ratio n 0-0.25 o f employed 0.25-0.5 (EXCL unpaid family 0.5 and more workers)/HHsize

Absolute Poverty Line $1 PPP D a y

Absolute Poverty Line i2.15 PPP Day

4bsolute Poverty Line $4.3 PPP D a y

Source: DIE, Turkey 2002 HBS.

142

The Poverty Gap Index According to Various Poverty Lines, Based on Consumption Per Capita Complete Food Poverty Poverty ,ine TL line TL

Relative Poverty Line 50% o f Median TL

Absolute Poverty Line $1 PPP D a y

Absolute Poverty Line $2.15 PPP Day

4bsolute Poverty ,ine $4.3 PPP Day

Poverty Gap Index

.0067

.0982

.OS48

.0007

.0200

.1288

Lower than primary

.0188

.2223

.1395

,0028

.OS58

.2717

Primary

.0057

.0925

.0483

.0004

.O 169

.1254

Secondarylvocationa

.0006

.0264

.0123

,0030

.0384

University

.oooo

.005 1

,0024

.oooo .oooo

.0003

.0076

Rural

.0096

.1304

.0745

.OO 13

.0284

.1676

Urban

,0048

.0767

.04 16

.0003

.0144

.1029

,0075

,0882

,0502

.0007

.0195

.1156

Yes

.0052

.1184

.0640

.0006

,0211

,1554

.0086

.1225

.0696

.0009

,0257

.1589

Yes

.0003

.0165

.0049

.0009

.0275

Zero

,0006

.0286

.0110

.OO 17

.0433

One

,0015

.OS65

.0248

.oooo .oooo .oooo

.0062

,0817

Two

.005 1

.0883

.0425

.0017

.0116

.1228

Three

.0127

.2044

.1244

.0010

.0485

.2559

4 and more

.0343

.3092

.2066

.0016

.0945

.3684

Zero

.0068

,0917

.os 10

.0006

.O 187

.1210

One

.0048

.lo36

.os01

.0002

.0162

.1406

T w o and more

.0094

.1401

.093 1

.0022

.0375

.1693

Male

.0065

.0983

.OS47

.0007

.0200

.1290

Female

.0094

.0972

,0558

.0005

.0202

.1253

G0.25

.0096

.1236

.0715

.0011

,0280

.1593

0.25-0.5

.002 1

.0623

.0308

.OOOO

.0076

.0857

0.5 and more

,0002

.0186

.0037

.OOOO

.0006

,0322

Turkey Education o f HH head

Type o f settlement Access t o land Own automobile

Children under age 14

Number o f HH members age >60

Gender o f hh head Dependency ratio n o f employed (EXCL unpaid family workers)/HHsize

Source: DIE, Turkey 2002 HBS.

143

The Poverty Severity Index According to Various Poverty Lines, Based on Consumption Per Capita, Based on HH Consumption per Capita Food Poverty Line T L

Complete Poverty Line T L

Relative Poverty Ane 50% of Median TL

Poverty Severity Index

,002 1

.0423

.0214

.0002

.0068

.0584

Lower than primary

,0065

.lo48

.OS72

.0008

.o 195

.1371

Primary

.OO 16

.0381

.0185

.0001

.0056

.0541

Secondarylvocational

,0098

.0041

.0149

.0018

.0007

.oooo .oooo

.0007

University

.ooo 1 .oooo

.0001

,0028

Rural

.0032

.0575

.0299

.0004

.0098

.0783

Urban

.0013

.0322

.0158

.oooo

.0048

.0452

.0023

.0389

.0203

.0002

.0070

.053 1

Yes

.OO 16

.0493

,0237

.0002

.0063

.0692

.0026

.0535

.0274

.0002

.0087

.0733

Yes

.0001

.0047

.OO 13

.0003

.0086

Zero

.0002

.0092

.003 1

,0006

.0153

One

.0003

.0202

.008 1

.oooo .oooo .oooo

.0017

.0312

Two

.0023

.0339

.0148

.0005

.0047

.0501

Three

,0036

.0937

.0499

.0004

.0146

.1246

4 and more

.0095

.1555

.09 13

.0003

,0331

.1971

Zero

.002 1

,0396

.0201

.0002

.0065

.0547

One

.oo 12

,0402

.0179

.oooo

.005 1

.0588

Two and more

.0033

.0683

.0385

.0006

.0119

.0879

Male

.0020

.0423

,0214

.0002

.0067

.0584

Female

.0030

.0430

.0222

.0001

.0080

.0585

,0030

.055 1

.0289

.0003

.0096

,0746

.0005

.0239

.0102

.oooo

,0022

.0352

.oooo

,0044

.0011

.oooo

.0002

.0092

rurkey Education o f HH head

Type o f settlement Access to land

Own automobile

Children under age 14

Number o f HH members age >60

Gender o f HH head

Dependency ratio n 0-0.25 o f Employed 0.25-0.5 (EXCL unpaid family 0.5 and more workers)/HHsize

Absolute Poverty L i n e $1 PPP Day

Absolute Absolute Poverty Poverty Line Line $4.3 i2.15 PPP PPP Day Day

Source: DIE, Turkey 2002 HBS.

144

The Relative Poverty R i s k s According to Various Poverty Lines, Based on Consurr t i o i Per Capita Relative Risk

Relative Risk

Re1ative Risk

Relative Risk

.oo

Lower than primary

1.73

-86

1.32

3.13

1.59

.7 1

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

Urban

-.3 1

-.17

-.22

-.3 1

-.27

-.16

-.04

-.11

-.12

.14

-.11

-.lo

Yes

.07

.2 1

.25

-.28

.22

.20

.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

Two

-.53

.04

-.06

.75

-.40

.09

Three

1.39

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 '

One

-.2 1

.17

.12

-.52

-.24

.17

Two and more

1.17

.13

.33

1.03

.83

.O?

.oo

.o 1

-.06

.01

.oo

-.07

-.08

.oo .oo

.86

-.lo

-.07

.43

.20

.25

.3 1

.19

-.69

-.28

-.34

-.82

-.44

-.26

-.96

-.64

-38

-1.oo

-.96

-.61

rurkey

Education o f HH head

rype o f settlement Access to land

Own automobile

Children under age 14

Number o f HH members age >60

Gender o f HH head Dependency ratio n o f workers)/ HH size

Male Female

0-0.25 0.5 and more

Source: DIE, Turkey 2002 HBS.

145

The Relative Poverty R i s k s According to Various Poverty Lines, Based on Consumption Per Capita Relative Risk Index

Relative Risk Index

1.oo

2.73

1.86

2.32

4.13

2.59

1.71

Primary

.86

1.02

.93

.55

.87

1.05

Secondarylvocational

.14

.34

.27

.23

.38

University

.03

.07

.05

.oo .oo

.o 1

.07

Rural

1.47

1.26

1.33

1.47

1.40

1.24

Urban

.69

.83

.78

.69

.73

.84

.96

.89

.88

1.14

.89

.90

Yes

1.07

1.21

1.25

.72

1.22

1.20

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

Two

.47

1.04

.94

1.75

.60

1.09

Three

2.39

1.83

2.05

1.03

2.66

1.72

4 and more

5.17

2.42

3.09

3.95

4.17

2.13

Zero

.90

.94

.93

.99

.95

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

Male

1.oo

1.01

1.oo

.94

1.01

.93

.92

1.oo

1.86

.90

.93

1.43

1.20

1.25

1.50

1.31

1.19

0.25-0.5

.3 1

.72

.66

.18

.56

.74

0.5 and more

.04

.36

.12

.oo

.04

.39

Turkey Lower than primary Education o f HH head

Type o f settlement Access to land Own automobile

Children under age 14

Number o f HH members age >60

Gender o f HH head Dependency ratio n o f employed (EXCL unpaid family workers)/HH size

Female

0-0.25

Source: DIE, Turkey 2002 HBS.

146

The Poverty Headcount According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted PO, Food Poverty Line, Adjusted Ad. Eq. Measures

PO, Complete Poverty Line, Adjusted Ad. Eq. Measures

PO, Relative Poverty Line, Adjusted Ad. Eq. Measures

PO, Poverty Line=$l PPP=663515 TLlday, Adjusted Ad. Eq. Measures

PO, Poverty Line=$2.15 PPF1426686 TL/day, Adjusted Ad. Eq. Measures

PO, Poverty Line=$4.3 PPP=2853313 TLIday, Adjusted Ad. Eq. Measures

Fraction of Poor

.0135

,2696

,1474

,0021

,041 1

,3414

.0304

.5245

,3238

.0084

.0807

.5985

Education o f Primary HH head Secondary /vocational

.0133

,2709

,1406

,0013

,0429

,3568

,0019

.0915

.0423

.oooo

.0060

.I243

University

.oooo

.O 160

.0095

.oooo

,0012

,0267

Turkey

Lower than primary

Rural

,020 1

,3448

,1986

,0046

,0533

.43 13

settlement

Urban

.0092

,2195

,1133

,0005

,0330

,2815

Access to land

,0145

,2466

.I391

,0018

.0456

,3102

Yes

.01 I 6

.3 I62

.I641

.0027

,0320

.4044

OWn automobi1e

,0174

,3313

.I852

.0028

.0525

.4 129

Yes

.0007

,0623

.0202

.oooo

,0028

.IO1 1

Zero

,0043

.I594

,0735

,0008

,0131

,2201

One Children under age 14 Two

,0077

,2010

,0925

.oooo

.0178

.2741

,0125

.2702

,1277

,0066

.0347

,3526

Three

,0237

,4678

,2946

.0027

.0869

S458

4 and more

,0460

,5395

.3781

.0003

,1426

.6112

Zero

,0150

,2588

,1404

,0018

.0436

,3244

.0075

,2800

,1291

.0005

,0315

.3867

,0129

,3366

.2367

.0078

,0386

,3939

Gender o f hh Male head Female

,0128

,2662

,1458

,0022

,0389

.3388

,0235

.3197

,1699

.0008

,0729

,3792

Dependency C0.25 ratio n o f 0.25-0.5 employed (EXCL unpaid 0.5 and more family workers)/HH size

,0186

,3163

,1746

,003 I

.0535

,3880

,0059

,2034

,1122

,0006

,0212

,2823

.oooo

,1241

,0450

.oooo

,0135

,1604

Type o f

Number o f HH One members age Two and >60 more

Source: DIE, Turkey 2002 HBS.

147

The Poverty Gap Index According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted P1, Food Poverty Line, Adjusted Ad. Eq. Measures

P1, Complete Poverty Line, Adjusted Ad. Eq. Measures

P1, Relative Poverty Line, Adjusted Ad. Eq. Measures

P1, Poverty Line=$l PPP=663575 TLlday, Adjusted Ad. Eq. Measures

P1, Poverty Line=$2.15 PPF1426686 TLlday , Adjusted Ad. Eq. Measures

P1, Poverty Line=$4.3 PPP=2853373 TLlday, Adjusted Ad. Eq. Measures

Poverty Gap Index

,0026

.0688

.03 13

.0004

.0083

.0973

.008 1

.1475

,0696

,0012

.0200

.1975

,002 1

.0667

,0301

.0003

,0078

,0964

Secondary1 vocational

,0001

,0191

,0073

.oooo

.0009

,0298

University

.oooo

.0036

.0010

.oooo

.ooo 1

,0059

Type o f settlement

Rural

,0043

,0923

,043 1

.0009

.0117

.1282

Urban

,0015

.053 1

.0234

.oooo

.0060

,0767

Access to land

,0029

,0649

.03 10

,0004

,0092

.0907

Yes

,002 1

.0765

.03 18

,0003

,0064

,1107

OWn automobile

,0034

.0862

.0398

,0005

,0106

.1207

Yes

,0001

.0102

,0029

.oooo

,0004

.0185

I'urkey

Lower than primary Education of Primary HH head

,0009

,0345

,0119

,0001

,0024

,0530

.0009

,0444

.0184

.oooo

.0037

.0679

.0041

,0640

.0273

,001 1

.0086

,0939

,0046

,1315

,0665

,0007

.0162

,1763

,0062

.1695

,0897

.0001

,0268

.2191

,0028

,0667

,0312

,0004

,0087

,0935

,001 1

,0635

,0249

.oooo

,0054

.0970

.0041

.0953

,0442

,0009

,0095

,1280

Male

.0025

.0673

,0304

,0004

,0078

,0957

Female

,0043

.0897

,0446

.0001

,0148

,1212

Zero One Children under age 14 Two Three 4 and more Zero Number o f HH One members age >60 Two and more Gender of HH head

Source: DIE, Turkey 2002 HBS.

148

The Poverty Severity Index According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted P2, Complete Poverty Line, Adjusted Ad. Eq. Measures

P2, Relative ?overtyLine, Adjusted Ad. Eq. Measures

P2, Poverty Line=$l PPP=663575 TLlday, Adjusted Ad. Eq. Measures

P2, Poverty Line=$2.15 PPP=1426686 TLlday, Adjusted Ad. Eq. Measures

P2, Poverty Line=$4.3 'PP=2853373 TLIday, Adjusted Ad. Eq. Measures

Poverty Poverty ieverity Index Severity Index

Poverty ieverity Index

Poverty Poverty everity Index Severity Index

Poverty everity Index

P2, Food Poverty Line, Adjusted Ad. Eq. Measures

.0008

.0253

.0104

.0001

,0027

,0384

,0028

,0562

.0242

,0002

.0075

.0826

Education o f Primary HH head Secondary /vocational

.0006

.0243

.0099

.0001

.0023

,0373

.oooo

,0061

,0018

.oooo

.0002

.0102

University

.oooo

.0010

,0002

.oooo

.oo 19

Type o f settlement

rural

.0016

.0347

.0147

.0002

,0042

.05 19

urban

,0003

,0190

.0076

.oooo

.0017

,0294

Access to land

,0009

,0249

,0109

.0001

,0030

.0368

Yes

,0007

.0262

,0095

.oooo

,0021

,0415

,001 1

,0320

,0133

.0001

.0035

.0482

Yes

.oooo

,0028

,0007

.0001

,0054

Zero

,0003

.0109

,0035

,0008

,0183

.0001

,0151

.0056

.oooo .oooo .oooo

,0010

,0243

.0018

,0232

.0100

,0002

.0037

,0358

.OO 14

,0508

,0215

.0002

,0050

.0742

,0012

,0690

.03 13

.oooo

,0074

,0975

Zero Number o f HH One members age Two and >60 more

,0008

,0250

.O 107

.0001

,0029

,0375

,0003

,0212

,0076

.oooo

,0014

,0345

.0017

,0352

.0136

.0001

,0038

,0525

Gender o f

male

,0008

,0246

.o 100

.0001

,0026

,0375

HH head

female

,001 1

,0356

,0165

.oooo

.0045

,0514

,0012

.03 13

,0136

,000 1

.0038

.0466

,0002

,0169

.0057

.oooo

,0010

,0270

.oooo

,0063

,0019

.oooo

.0002

,0116

Turkey Lower than primary

Own automobile

One Children under age 14 Two Three 4 and more

Dependency 0 - 0.25 ratio n o f 0.25 - 0.5 employed (EXCL unpaid 0.5 and more family workers)/HH size Source: DIE, Turkey 2002 HBS.

149

The Relative Poverty R i s k s According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted Relative Risk

Access to land

Own automobile

Children under age 14

telative Risk

belative Risk

.oo

Lower than primary

1.24

.95

1.20

2.93

.96

.75

Primary

-.02

.oo

-.05

-.40

.04

Secondary /vocational

-.86

-.66

-.71

-1.00

-.85

-.64

-1.00

-.94

-1.00

-.97

-.92

Rural

.49

.28

.35

1.16

.30

.26

Urban

-.32

-.I9

-.23

-.78

-.20

-.18

.07

-.09

-.06

-.14

.I1

-.09

Yes

-.I4

.I7

.I1

.29

-.22

.18

.28

.23

.26

.30

.28

.2 1

Yes

-.95

-.77

-.86

-1.00

-.93

-.70

Zero

-.68

-.41

-.50

-.64

-.68

-.36

One

-.43

-.25

-.37

-1.oo

-.57

-.20

Two

-.08

.oo

-.I3

2.10

-.I5

.03

Three

.75

.73

1.oo

.25

1.11

.60

2.40

1.oo

1.57

-.87

2.47

.79

-.04

-.05

-.16

.06

-.05

-.45

.04

-.I2

-.76

-.23

.13

-.05

.25

.61

2.65

-.06

.15

-.05

-.01

.04

-.05

-.01

.74

.19

.15

-.61

.77

.11

.37

.I7

.18

.45

.30

.14

-.57

-.25

-.24

-.71

-.48

-.17

-1.00

-.54

-.69

-1.oo

-.67

-.53

4 and more Zero Number of HH members One age >60 Two and more Gender of HH head

Relative Risk

.oo

University Type o f settlement

Relative Risk

.oo

Turkey

Educationof HH head

telative Risk

Male Female

0-0.25 Dependency ratio n o f employed (EXCL unpaid 0.25-0.5 family workers)/HH size 0.5 and more

.ll

Source: DIE, Turkey 2002 HBS.

150

The Relative Poverty R i s k s According to Various Poverty Lines, Based on Consumption Per Aduk Equivalent Adjusted Xelative ielative Zelative Relative ielative telativc Risk Risk Risk Risk Risk Risk Index Index Index index Index Index

1.oo

2.24

1.95

2.20

3.93

1.96

1.75

Primary

.98

1.oo

.95

.60

1.04

Secondary1 vocational

.14

.34

.29

.oo

.15

.36

University

.oo

.06

.oo

.03

.08

Rural

1.49

1.28

1.35

2.16

1.30

1.26

Urban

.68

.81

.77

.22

.80

.82

1.07

.9 1

.94

.86

1.11

.91

Yes

.86

1.17

1.11

1.29

.78

1.18

1.28

1.23

1.26

1.30

1.28

1.21

Yes

.05

.23

.I4

.07

.30

Zero

.32

.59

.oo .36

.32

.64

One

.57

.75

.63

.oo

.43

.80

Two

.92

1.oo

.87

3.10

.85

1.03

Three

1.75

1.73

2.00

1.25

2.1 1

1.60

4 and more

3.40

2.00

2.57

.13

3.47

1.79

Zero

1.11

.96

.95

.84

1.06

.95

One

.55

1.04

.88

.24

.77

1.13

Two and more

.95

1.25

1.61

3.65

.94

1.15

Male

.95

.99

1.04

.95

.99

Female

1.74

1.19

1.15

.39

1.77

1.11

0-0.25

1.37

1.17

1.18

1.45

1.30

1.14

0.2 5-0.5

.43

.75

.76

.29

.52

.83

0.5 and more

.oo

.46

.3 1

.oo

.33

.47

Turkey Lower than primary Education o f HH head

Type o f settlement Access to land Own automobile

Children under age 14

' Number

age >60

o f HH members

Gender o f HH head Dependency ratio n o f employed (EXCL unpaid family workers)/HH size

Source: DIE, Turkey 2002 HBS.

151

The Poverty Headcount According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted PO, Food Poverty Line, OECD Ad. Eq. Measures

PO, Complete Poverty Line, OECD Ad. Eq. Measures

'0, Relative Poverty Line, OECD Ad. Eq. Measures

PO, Poverty Line=$l PPP=66357 5 TLIday, OECD Ad. Eq. Measures

PO, Poverty Line =$2.15 PPP=14266 86 TLIday, OECD Ad. Eq. Measures

PO, Poverty Line=$4.3 PPP=2853373 TLIday, OECD Ad. Eq. Measures

Fraction of Poor

Fraction o f Poor

Fraction of Poor

,0044

.I597

.0814

,0002

.0204

.2151

,0189

,4082

.2 186

,0015

.05 18

.49 10

Primary Education o f HH head Secondary /vocational

.0023

,1403

.0696

.oooo

.0188

,2026

.oooo

,0385

,0170

.oooo

.0013

.0534

University

.oooo

.0059

,0012

.oooo

,0116

Rural Urban

.0065

,2201

.I214

.0006

,0264

,2926

,0030

.I195

.0548

.oooo

.O 163

,1634

Access to land

,0047

,1382

,0700

,0004

,0235

.I850

Yes

,0038

,2032

,1045

.oooo

,0141

,2758

Own automobile

No Yes

.0057

,2010

,1044

.0003

,026 1

,2687

.oooo

.0208

.0039

,0012

,0349

Zero

,0019

,0658

.0253

,003 1

.1103

.0005

,0956

.0430

,0089

,1473

,007 1

,1400

,0506

.oooo .oooo .oooo .oooo

.0136

,1940

.0043

,3114

,1992

,0021

.0398

.3898

,0151

,4679

,2786

.oooo

,0926

,5423

.0046

.1445

,0719

,0003

,0205

,1983

.0019

.I708

,0755

,0173

.2394

,0078

.2612

,1685

.oooo .oooo

,0248

.3 044

Turkey

Lower than primary

Type of settlement

One Children under age 14 Two Three 4 and more Number o f Zero HH One members age >60 Two and more Gender o f

Male

,0042

,1587

,0824

.0003

,0202

.2 147

HH head

Female

,0078

.1754

,0673

.oooo

,0233

,2206

.0062

,1965

.I062

,0004

.0285

,2572

,0017

.I129

,0460

.oooo

.0074

,1610

.oooo

.O 176

,0049

.oooo

,0016

,0545

Dependency 0-0.25 ratio n o f 0.25-0.5 employed (EXCL unpaid family 0.5 and more workers)/HH size Source DIE, Turkey 2002 HBS.

152

The Poverty Gap Index According to Various Poverty Lines, Based on ConsumptionPer Adult Equivalent Adjusted PI, P1, Food Zomplete Poverty Poverty Line, Line, OECD OECD Ad. Eq. Ad. Eq. Measures Measures

PI, Relative Poverty Line, OECD Ad. Eq. vleasures

P1, Poverty Line=$l PP=663575 TLIday, 3ECD Ad. EqMeasures

P1, Poverty Line=$2.15 'PP=1426686 TLIday, OECD Ad. :q. Measures

P1, Poverty Line=$4.3 PP=2853373 TLIday, OECD Ad. :q. Measures

Poverty Gap Index

'overty Gap Index

Poverty Gap Index

'overty Gap Index

.0010

,0384

,0166

.oooo

.0039

.0563

Lower than primary

,0045

.lo24

.0467

.0002

.0119

.1440

Primary

.0005

.033 1

.0139

.003 1

.0497

Secondaryivocational

.0078

,0022

.0001

.0124

University

.oooo .oooo

,0010

.0003

.oooo .oooo .oooo

.oooo

.002 1

Type of settlement

Rural

,0019

.0553

.0246

.0001

.0061

,0793

Urban

.0005

.0272

.0112

.0024

.0410

Access to land

.0010

.0345

.0157

.004 1

.0496

Yes

.0011

.0465

.0183

.0034

.0698

No OWn automobile Yes

.OO 13

,0489

.0213

.0050

.07 13

.oooo

,003 1

.0007

.0002

.0059

Zero

.0003

.0124

,0036

.0007

.02 12

One

.0001

.0206

.0075

.0014

.0330

Two

.0024

,0279

.0107

.oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo

.0042

.0446

Three

,0013

.OS57

.0389

.0002

.0073

,1174

4 and more

.002 1

.1297

.0653

.0143

.1747

Number of Zero HH One members age >60 Two and more

.0010

.0349

.0152

.0038

.05 11

,0003

.0368

.0144

.0028

.0577

.0030

,0700

.03 16

.006 1

.0953

Gender o f

Male

.0010

.0386

.0166

.0038

.OS64

HH head

Female

,0010

.0363

.0165

.oooo .oooo .oooo .oooo .oooo .oooo

.0053

,0554

Turkey Education o f HH head

Children under age 14

Source: DIE, Turkey 2002 HBS.

153

The Poverty Severity Index According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted P2, Food Poverty Line, OECD Ad. Eq. Measures

P2 Relative P2, Complete Poverty Poverty Line, Line, OECD OECD Ad. Ad. Eq. Eq. Measures Measures

P2, Poverty Line=%l PPP=663575 TLlday, OECD Ad. Eq. Measures

P2, Poverty Line=$2.15 ?PP=1426686 TLIday, OECD Ad. Eq. Measures

P2, Poverty Line=%4.3 PPP=2853373 TL/day, OECD Ad. Eq. Measures

Poverty Severity Index

,0003

,0135

.0052

.oooo

,0012

,021 1

,0014

.0374

,0153

.oooo

,0042

.0568

Education o f Primary HH head Secondary /vocational

,0002

,0114

.0043

.oooo

,0008

,0182

.oooo

,0022

,0005

.oooo

.0040

University

.oooo

.0002

,0001

.oooo

.0005

Type of settlement

Rural

.0006

,0199

,0079

.0020

,0306

Urban

.0001

,0093

,0034

.OOM

,0149

Access to land

,0003

.0126

,0052

,0012

.O 193

Yes

,0004

,0153

.0052

,001 1

,0249

,0004

.O 173

.0067

Yes

.oooo .ooo 1 .oooo

,0008

.0002

,0035

,0010

,0065

.002 1

,0008

.0096

,0044

,0005

.0309

.0115

,0004

,0494

,0200

,0003

.0124

,0049

,0001

,0119

,004 I

,0011

,0253

,0096

Gender of hh Male head Female

,0003

,0136

,0052

,0002

,0131

,0058

Dependency Cb0.25 ratio n o f 0.254.5 employed (EXCL unpaid family 0.5 and more workers)/HH size

.0005

.0180

,0073

.0001

,0068

,0020

.oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo .oooo

.oooo

,0012

,0003

.oooo

Turkey

Lower than primary

OW automobile

Zero One

Children under age 14 Two Three

4 and more Zero Number o f HH One members age >60 Two and more

Poverty Poverty Severity ieverity Index Index

.0015

,0270

.oooo

,0016

,0003

,0065

,0003

.0110

.0019

,0156

.0020

.0469

,0033

,0727

.001 I

.0193

.0006

.o 199

.0025

.0383

.0011

.0212

.0014

,0205

,0016

,0275

,0004

.0120

.oooo

,0021

Source: DIE, Turkey 2002 HBS.

154

The Relative Poverty R i s k s According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted Relative Risk

Type o f settlement Access to land

Own automobile

Children under age 14

Number o f HH members age >60

Gender o f hh head Dependency ratio n o f employed (EXCL unpaid family workers)iHH size

Relative Risk

.oo

.00

.oo

Lower than primary

3.28

1.56

1.69

5.36

1.54

1.28

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

Rural

.47

.38

.49

1.50

.30

.36

Urban

-.3 1

-.25

-.33

-1.00

-.20

-.24

.07

-.13

-.14

.I5

-.14

Yes

-.14

.27

.28

-1.00

-.3 1

.28

.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

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

One

-.56

.07

-.07

-1.oo

-.15

.I1

.76

.63

1.07

-1.00

.22

.42

-.05

-.01

.01

.07

-.01

.oo

Female

.76

.10

-.17

-1 .oo

.14

.03

0-0.25

.40

.23

.30

.59

.40

.20

-.61

-.29

-.43

-1.00

-.64

-.25

-1.00

-.89

-.94

-1.00

-.92

-.75

Turkey

Education o f HH head

Relative Risk

Two and more

Male

0.25-0.5 0.5 and more

Source DIE, Turkey 2002 HBS.

155

The Relative Poverty R i s k s According to Various Poverty Lines, Based on Consumption Per Adult Equivalent Adjusted Relative Risk Index

Access to land Own automobile

Children under age 14

Relative Risk Index

1.oo

4.28

2.56

2.69

6.36

2.54

2.28

Primary

.5 1

.88

.86

.oo

.92

.94

Secondary1 vocational

.oo

.24

.2 1

.oo

.07

.25

University

.oo

.04

.01

.oo

.05

Rural

1.47

1.38

1.49

2.50

1.30

1.36

Urban

.69

.75

.67

.oo

.80

.76

1.07

.87

.86

1.50

1.15

.86

Yes

.86

1.27

1.28

.oo

.69

1.28

1.30

1.26

1.28

1.30

1.28

1.25

Yes

.oo

.13

.05

.06

.16

Zero

.42

.4 1

.3 1

.15

.5 1

One

.11

.60

.53

.44

.68

Two

1.61

.88

.62

.oo .oo .oo .oo

.67

.90

Three

.98

1.95

2.45

8.72

1.95

1.81

4 and more

3.41

2.93

3.42

.oo

4.55

2.52

Zero

1.04

.90

.88

1.36

1.01

.92

.44

1.07

.93

.oo

.85

1.11

1.76

1.63

2.07

.oo

1.22

1.42

.95

.99

1.01

1.07

.99

.83

.oo

1.oo

1.14

1.03

Lower than Primary

Type o f settlement

Relative Risk Index

1.oo

Turkey

Education o f HH head

Relative Risk Index

Number of HH members One age >60 Two and more Male

Gender o f hh head

Female

1.76

Dependency ratio n o f employed (EXCL unpaid family workers)/ HH size

C0.25

1.40

1.23

1.30

1.59

1.40

1.20

0.25-0.5

.39

.71

.57

.36

.75

0.5 and more

.oo

.ll

.06

.oo .oo

.08

.25

1,lO

Source: DIE, Turkey 2002 HBS.

156

Appendix 3: Sensitivity of Poverty to the MinimumFood Basket Caloric Requirements, CPI Source, Size Economiesâ&#x20AC;&#x2122; and Equivalence Scales Parameters

C P I DIE

kPoverty ciden.

P1 Poverty Gap Index

Poverty Shortfall Index

157

Alpha = 0.9 Theta = 0.6 Kcal = 2,100

PO Poverty Incident

Poverty Index

P2 Poverty Severity Index

Poverty Shortfall Index

Food Poverty Line Complete Poverty Line Food Poverty Line Complete Poverty Line

Adult equivalent Adjusted adult equivalent OECD Per capita Adult equivalent Adjusted adult equivalent OECD Per capita Adult equivalent Adjusted adult eauivalent OECD Per capita Adult equivalent Adjusted adult equivalent OECD

0.001 1

0.0013

0.0008

0.0010

0.0005 0.0468 0.0288

0.0006 0.0461 0.0287

0.0003 0.0423 0.0253

0.0004 0.0426 0.0259

0.0155 0.1894 0.1870

0.0155 0.1943 0.2 125

0.0135 0.1775 0.1939

0.0139 0.1877 0.2037

0.2171 0.3 148 0.2638

0.2367 0.3 180 0.2664

0.2337 0.3101 0.255 1

0.25 10 0.3123 0.2638

0.2430

0.2409

0.2406

0.2438

158

I Food

~Poverty n ~ i dte n

Poverty Line

Complete Povertv

Alpha = 1 Theta = 0.6 Kcal = 2450 CPI DIE CPI Survey Imputed Rent Reported Imputed Rent Reported by Regression Rent by HH by Regression Rent by HH Per capita 0.0714 0.0752 0.0596 0.0653 Adult equivalent 0.0306 0.0342 0.0252 0.0274 Adjusted adult 0.0153 0.0147 0.0131 0.0143 equivalent OECD Per capita Adult equivalent

0.4324

0.4251

0.4 184

0.4083

0.3906

0.3788

0.3652

0.3570

P1 Poverty Gap Index

P2 Poverty Severity Index

159

Alpha = 1 Theta = 0.6 Kcal = 2100 CPI DIE CPI Survey Reported Imputed Rent Reported Imputed Rent by Regression Rent by HH by Regression Rent by HH Per capita 0.0438 0.0462 0.0379 0.0388

PO Poverty Incident

Poverty Gap Index I

Poverty Severity ~ n ~

Line

equivalent OECD Per capita Food Adult equivalent 'Over@ Adjusted adult Line Poverty eauivalent OECD Shortfall Per capita Index Complete Adult equivalent Poverty Adjusted adult Line equivalent OECD I

0.0155 0.1894 0.1870

0.0158 0.1943 0.2080

0.0136 0.1775 0.1881

0.0139 0.1877 0.1961

0.2171 0.3 146 0.2624

0.2367 0.3 153 0.2658

0.2337 0.3097 0.2570

0.25 10 0.3119 0.2613

0.2430

0.24 15

0.2415

0.2435

160

Appendix 4: Regression Analysis Descriptive Model Variables Enteredmemoved (b,c) Variables Removed

Model Variables Entered Do you have in your dwelling-Closed garage, DHEAT-4, DDWEL-5, DDWEL-4, Do you have in your dwelling-Electricity, secondary education o f HH head, DDWEL-7, DDWEL-1, DDWEL-9, Do you have in your dwelling-Heating from ground, South eastern Anatolia, HH head i s female, DDWAGE-3, hh head i s 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, D o 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-Natural gas, AE adjusted by modal hh (2ad + 2 ch), How many square meters i s the utilized area (d), DDWEL-2, Do you have in your dwelling-Centralheating(a)

Method

Enter

4. A l l requested variables entered. 3. Dependent Variable: LNF.

Z. Weighted least squares regression-weighted by population weight.

Model Summary Model

R Square

Adjusted R Square

Std. E r r o r of the Estimate

.681(a)

.464

57.89849

A Predictors: (Constant), Do you have in your dwelling-Closed garage, 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 eastem Anatolia, HH head i s female, DDWAGE-3, hh head i s 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 o f 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 have in your dwelling-Piped water system, children under age 14 , D o 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 dwellingBathroom, D o you have in your dwelling-Natural gas, AE adjusted by modal hh (2ad + 2 ch), How many square DDWEL-2, Do you have in your dwelling-Central heating meters i s the utilized area (d),

161

ANOVA(b,c) Model Regression

Sum o f Squares

198,585,252,725,405

Mean Square

44 4,5 13,301,198,305

1 Residual

229,269,350,347,546

68,392,987

Total

427,854,603,072,951

68,393,03 1

F 1,346,355,937

3,352,235

a. Predictors: (Constant), Do you have in your dwelling-Closed garage, DHEAT-4, DDWEL-5, DDWEL-4, Do you have in your dwelling-Electricity, secondary education o f HH head, DDWEL-7, DDWEL-1, DDWEL-9, Do you have in your dwelling-Heating from ground, South eastern Anatolia, HH head i s female, DDWAGE-3, hh head i s unemployed, DDWEL-8, Do you have in your dwelling-Satellite antenna, D o you have in your dwelling-Waste disposal, Eastern Anatolia, DDWEL-10, higher education o f HH head, Central Anatolia, DDWAGE-1, University, master, doctorate education o f 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, D o you have in your dwelling-Hot water, Do you have in your dwelling-Toilet (indoors), Do you have in your dwelling-Bathroom, D o you have in your dwelling-Natural gas, AE adjusted by modal hh (2ad + 2 ch), How many square meters i s the utilized area (m2), DDWEL-2, Do you have in your dwelling-Central heating.

b. Dependent variable: LNF. c. Weighted least squares regression-weighted by population weight.

Coefficients (a,b) Unstandardized Coefficients Std. Error

Standardize d Coefficients

Sig.

Beta

95% Confidence Interval for B Lower Bound

Upper Bound

14.440

14.448

-.055

-.054

.144

.145

-.323

-.322

-.003

-.002

-.135

-.134

-.057

-.056

.025

,027

-1,322.171

.ooo .ooo .ooo .ooo .ooo .ooo .ooo .ooo .ooo

-.OS

-.057

Model

1 (Constant)

14.444

.002

-5.424E-02

-.045

-3 62.625

.084

747.964

-.200

-1,553.420

-.001

-12.147

-.099

-846.849

-.026

-227.808

.011

90.725

-.181

9,064.796

South eastern Anatolia

2.584E-02

Children under age 14

-5.752E-02

.ooo .ooo .ooo .ooo .ooo .ooo .ooo .ooo

Secondary education of HH head

.161

.ooo

,082

886.785

.ooo

.160

.161

Higher education of HH head

,264

,000

.136

1,391.592

.ooo

.263

.264

University, master, doctorate education of HH head

.399

.ooo

.I30

1,322.195

.ooo

,399

.400

-1.487E-02

.ooo

-.006

-61.519

.ooo

-.015

-.014

Rural Household Aegean Mediterranean Central Anatolia Black sea Eastern Anatolia

HH head i s female

.145 -.323 -2.22 1E-03 -.135 -5.622E-02

162

Coefficients (a,b) Unstandardized Coefficients lode1

Std. Error

Standardize d Coefficients

Sig.

Beta

95% Confidence Interval for B Lower Bound

Upper Bound

-.143 .ooo

-.073

-757.275

.ooo

-.144

-.143

1.059E-02

.ooo

-.027

- 178.748

.ooo

-.011

-.010

3DWEL-1

.359

.002

.02 1

233.354

.ooo

.355

.362

3DWEL-2

8.49 1E-02

-.073

-444.759

,000

-.085

-.084

D DWEL-3

-.176

.ooo .ooo

-.098

-753.416

-.176

-.175

3DWEL-4

8.633E-02

,001

-.009

-104.229

.ooo .ooo

-.088

-.084

DDWEL-5

5.205E-02

.ooo

.014

152.366

,000

.05 1

.053

DDWEL-7

.164

.001

.028

300.844

.163

,165

DDWEL-8

.140

.011

118.521

.137

.143

DDWEL-9

-.190

-.049

-499.557

.ooo .ooo .ooo

-.191

-.189

DDWEL- 10

-.35 1

-.063

-640.854

,000

-.352

-.349

.005

45.498

.010

,011

-.010

-91.800

-.017

-.016

.068

643.960

.080

.05 1

497.510

.090

.091

,251

,256

.258

.263

.456

.462

I H head i s unemployed 9E adjusted by modal HH 2ad + 2 ch)

DDWAGE- 1

1.053E-02

DDWAGE-2

1.670E-02

DDWAGE-3

7.986E-02

DDWAGE-5

9.034E-02

.oo 1 .ooo .oo 1 .ooo .ooo .ooo .ooo

DHEAT-2

.254

.001

.lo8

244.697

DHEAT-3

.260

.001

.073

242.991

DHE AT-4

.459

.001

.039

3 99.724

.ooo .ooo .ooo .ooo .ooo .ooo .ooo

How many rooms are there in your dwelling?

2.697E-02

.ooo

.054

360.789

.ooo

.027

How many square meters i s the utilized area (m2)?

1.737E-03

.ooo

.lo7

7 17.584

,000

.002

Do you have in your dwellingBathroom?

1.557E-02

.ooo

-.010

-78.179

,000

-.016

-.015

Do you have in your dwellingToilet (indoors)?

,145

,000

.125

1,003.096

.ooo

.147

Do you have in your dwellingKitchen?

.7.218E-02

.ooo

-.041

-348.0 14

.ooo

-.073

-.072

Do you have in your dwellingWaste disposal?

.14t

.oo1

.010

109.885

.ooo

.143

.149

Do you have in your dwellingCentral heating?

3.139E-02

.001

-.04 1

-79.875

.ooo

-.OS4

-.079

Do you have in your dwellingHeating from ground?

.350

.001

.027

298.03 1

.ooo

.347

.353

163

Coefficients (a,b) Unstandardized Coefficients

Standardize d Coefficients

Std. Error

Beta

D o you have in your dwellingPiped water system?

j.122E-02

.ooo

.035

312.081

D o you have in your dwellingElectricity?

.326

.002

,019

D o you have in your dwellingNatural gas?

.I25

.ooo

Do you have in your dwellingHot water?

,126

D o you have in your dwellingSatellite antenna? D o you have in your dwellingClosed garage?

lode1

Sig.

95% Confidence Interval for B Lower Bound

Upper Bound

.ooo

.061

.062

207.743

.ooo

.322

,330

,039

306.671

,000

.I24

.I26

.ooo

.I13

936.297

,000

.125

.I26

.126

.ooo

.035

346.933

.ooo

.125

.I27

.287

,000

.09 1

977.939

.ooo

.287

.288

a. Dependent Variable: LNF. b. Weighted least squares regression-weighted

b y population weight.

164

Appendix 5: Food Basket

165

Item Name

166

ItemCode

Item Name

ICalories in1

Price

Q2450

C2450

I VAL2450 I

167

168

Margarine (packaged)

719

0115302010

Sunflower oil

0115303010

Maize oil

0116101020

Orange

0 1 16102010

Grape (having seeds)

01 15201010

2,209,605

0.0099

71.1736

2 1,873

884

1,812,505

0.0393

347.1712

71,182

884

2,072,678

0.0058

51.2739

12,022

592,783

0.0496

23.2889

29,373

952,293

0.0062

4.1285

5,868

169

ItemCode

Item Name

/Calories in 100 grl

Price

43100

C3100 [VAL31001

170

Appendix 6: Estimation o f Average Household Adult Equivalent Size, Depending on the Formula and Parameters The Average Household Adult Equivalent Size

Alpha Theta Average adult equivalent size o f household Average adjusted adult equivalent size o f household Alpha Theta Average adult equivalent size o f household Average adjusted adult equivalent size o f household Alpha Theta Average adult equivalent size o f household Average adjusted adult equivalent size o f household

0.8 0.6

0.8 0.7

0.8 0.8

0.8 0.9

0.8 1

2.22

2.55

2.94

3.39

3.92

4.11

4.16

4.22

4.28

4.36

0.7 0.6

0.7 0.7

0.7 0.8

0.7 0.9

0.7 1

2.18

2.50

2.87

3.31

3.81

4.19

4.25

4.32

4.40

4.49

0.6 0.6

0.6

0.7

0.6 0.8

0.6 0.9

0.6 1

2.14

2.45

2.81

3.22

3.70

4.27

4.34

4.43

4.52

4.63

171

Relationship between HH Size, Composition, and Adult Equivalent Size for P ?ha= 0.9 and Theta = 0.6

HH Size 1 2

Adult Equivalent Size

Adult Equivalent Adjusted by Modal hh (2ad + 2 ch)

Adult Equivalent Size OECD

Adult Equivalent Size F A 0

Adult

Child

1.oo

1.80

1.oo

.85

1.47

2.64

1s o

1.50

1.52

2.72

1.75

1.47

1.85

3.33

2.00

1.97

1.89

3.40

2.25

2.05

1.93

3.47

2.50

2.03

2.19

3.94

2.50

2.55

2,23

4.00

2.75

2.59

2.26

4.06

3.00

2.57

2.30

4.13

3.25

2.56

2.50

4.49

3.00

3.14

2.53

4.54

3.25

3.06

2.56

4.60

3.50

3.07

2.59

4.66

3.75

3.04

2.63

4.72

4.00

3.03

2.78

4.99

3.50

2.81

5.05

3.75

3.52

2.84

5.10

4.00

3.53

2.87

5.16

4.25

3.47

2.90

5.21

4.50

3.46

2.93

5.26

4.75

3.49

3.07

5.52

4.25

3.95

3.10

5.57

4.50

3.98

3.13

5.62

4.75

3.92

3.16

5.67

5.00

3.87

3.19

5.72

5.25

3.91

3.21

5.77

5.50

3.93

172

3.32

5.97

4.75

4.23

3.35

6.01

5.00

4.35

3.38

6.06

5.25

4.42

3.40

6.11

5.50

4.32

3.43

6.16

5.75

4.36

3.46

6.21

6.00

4.28

3.48

6.25

4.32

3.56

6.39

5.25

4.73

3.59

6.44

5.50

4.63

3.61

6.48

5.75

4.69

3.64

6.53

6.00

4.75

3.66

6.58

6.25

4.75

3.69

6.62

6.50

4.75

3.71

6.67

6.75

4.73

3.74

6.71

7.00

4.74

3.79

6.80

5.75

5.13

3.81

6.84

6.00

5.12

3.84

6.89

6.25

5.15

3.86

6.93

6.50

5.16

3.88

6.98

6.75

5.14

3.91

7.02

7.00

5.09

3.93

7.06

7.25

5.03

3.96

7.1 1

7.50

5.18

3.98

7.15

7.75

5.30

4.00

7.19

6.25

5.54

4.03

7.23

6.50

5.31

4.05

7.28

6.75

5.54

4.08

7.32

7.00

5.40

4.10

7.36

7.25

5.41

4.12

7.40

7.50

5.40

4.15

7.44

7.75

5.43

4.17

7.49

8.00

5.37

4.19

7.53

8.25

5.62

173

4.24

7.61

7.00

5.67

4.26

7.65

7.25

5.86

4.31

7.73

7.75

5.75

4.33

7.77

8.00

5.93

4.35

7.81

8.25

5.85

4.37

7.85

8.50

5.83

4.40

7.89

8.75

5.81

4.5 1

8.09

8.25

6.43

4.53

8.13

8.50

6.23

4.55

8.17

8.75

6.05

4.57

8.21

9.00

6.17

4.59

8.25

9.25

6.56

4.72

8.48

9.00

6.48

4.75

8.52

9.25

6.56

4.77

8.56

9.50

6.74

4.79

8.60

9.75

6.56

4.81

8.63

10.00

6.57

4.83

8.67

10.25

6.50

4.85

8.71

8.75

7.30

4.91

8.82

9.50

6.94

4.93

8.86

9.75

6.74

4.95

8.90

10.00

7.02

5.00

8.97

10.50

6.85

5.08

9.12

9.75

7.33

5.10

9.15

10.00

7.13

5. I 4

9.23

10.50

7.16

5.28

9.48

10.50

7.58

5.32

9.55

11.00

7.83

5.34

9.58

11.25

7.62

5.42

9.72

12.25

7.83

5.65

10.14

11.75

8.28

5.66

10.17

12.00

8.12

5.68

10.20

12.25

8.07

5.83

10.47

12.50

8.64

5.89

10.57

13.25

8.59

5.91

10.6 1

13.50

8.99

5.92

10.64

13.75

8.87

174

Appendix 7: General Descriptive Statistics

Population Structure by Sex Type of Settlement

Turkey Gender

Rural

Urban

Male

48.7%

48.5%

48.7%

Female

51.3%

51.5%

5 1.3%

100.0%

Turkey Source: DIE, Turkey 2002 HBS.

Population Structure by Age Type of Settlement Rural

Age Group

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%

35-39

7.0%

7.7%

7.4%

4 w 4

6.6%

6.7%

45-49

5.8%

5.9%

50-54

4.9%

55-59

3.8%

3.1%

3.4%

60-64

3.7%

2.5%

2.9%

10.5%

4.8%

7.0%

100.0%

65+ Turkey

Urban

Source: DIE, Turkey 2002 HBS.

175

Population Structure b y Age and Sex Turkey

Gender

Age Group

Male

Female

0-5

52.2%

47.8%

10.0%

100.0%

6-14

50.1%

49.9%

18.4%

100.0%

15-19

48.6%

5 1.4%

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%

35-39

49.7%

50.3%

7.4%

100.0%

4w4

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%

48.7%

5 1.3%

100.0%

Turkey

Source DIE, Turkey 2002 HBS

Population (age 6 and over) Structure b y Education by the Type of Settlement Turk

Type of Settlement Urban

Rural

Educational Institutions Graduated From

Count

10.7%

7,709,053

12.5%

7,721,572

2 1.1%

3,212,854

2 1.5%

42.8% 12,291,816

33.6%

'2,958,546

37.3%

1,920,310

5.2%

3,215,992

5.2%

2,791,097

7.6%

4,005,484

6.5%

115,567

.3%

174,674

.3%

6.1%

4,193,597

11.5%

5,7 19,7 15

9.3%

1.9%

1,470,172

4.0%

1,934,897

3.1%

Count

Illiterate

3,798,846

15.3%

3,910,208

Literate without diploma

5,491,283

22.0%

Primary school (5 years)

10,666,730

Primary education(8 years)

1,295,683

5.2%

Junior high school (8=5+3)

1,2 14,386

4.9%

59,108

.2%

1,526,118 464,725

Vocational school at Jr. high (8-9 = 5+ 3-4) H i g h school 11-12 = 8-9 + 3-4 Vocational school at high

176

Population (age 6 and over) Structure by Education by the Type o f Settlement Turkey

Type o f Settlement Rural Count

Urban

Count

OO /

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%

8,012

.O%

121,175

.3%

129,186

.2%

Master and doctorate

14,908,885 100.0% 36,617,206 100.0% 61,526,091 100.0%

rurkey Yource: DIE, Turkey 2002 HBS.

Population (age 14-65) Employment Status by Type of Settlement Type of Settlement Turkey

Employed in some way Unemployed Economically inactive

Turkey

Rural

Urban

Col Yo

57.7%

37.2%

45.5%

1.6%

3.4%

2.7%

40.8%

59.4%

51.8%

100.0%

Source: Turkey 2002 HBS.

Population (age 12+) Employment Status by Sex Gender

Employed in some way Unemployed Economically inactive

rurkey

Turkey

Male

Female

Col OO/

Col Yo

64.2%

28.4%

45.5%

4.0%

1.4%

2.7%

31.8%

70.2%

5 1.8%

100.0%

100.0% 100.0%

iource: DIE, Turkey 2002 HBS.

177

Population (age 12 +) Employment Status by Education Education Lower than Primary

Primary

Col Yo

Employed in some way

Turkey

Secondary University Vocational

Col Yo

Col %

COl%

27.5%

49.9%

49.3%

70.2%

45.5%

.4%

2.3%

6.3%

6.4%

2.7%

72.1%

47.8%

44.4%

23.4%

51.8%

100.0%

Unemployed Economically inactive

Turkey

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

Regular employee

17.0%

60.5%

38.0%

Casual employee

5.8%

9.2%

7.4%

.2%

.3%

.2%

2.0%

5.8%

3.8%

Self-employed

36.5%

16.6%

26.9%

Unpaid family worker

38.6%

7.6%

23.6%

100.0%

Employment Status in

Apprentice

Workplace

Employer

Turkey

100.0%

Source. Turkey 2002 HBS.

Population (age 12+) Emp

iment by Sex Gender Male

Female

Col Yo

45.8%

21.9%

38.0%

7.8%

6.7%

7.4%

Apprentice

.3%

.1%

.2%

Employer

5.5%

.4%

3.8%

3 1.4%

17.4%

26.9%

9.2%

53.5%

23.6%

100.0%

Regular employee Casual employee Employment Status in Workplace

Turkey

Self-employed Unpaid family worker Turkey Source: DIE, Turkey 2002 HBS.

178

Population (age 12 +) Employment by Education Education

status workplace

Lower than Primary

Primary

Col Yo

rurkey Secondary Jniversity Vocational Col Yo

Col Yo

Regular employee

6.8%

32.2%

67.7%

81.8%

38.0%

Casual employee

8.9%

9.1%

2.6%

.6%

7.4%

Apprentice

.3%

.l%

Employer

.?yo

3.8%

5 .o%

7.3%

3.8%

Self-employed

40.0%

30.3%

11.5%

6.5%

26.9%

Unpaid family worker

43.3%

24.2%

13.0%

3.8%

23.6%

100.0%

Turkey

.2%

100.0% 100.0%

Source: DIE, Turkey 2002 HBS.

HH Head's Gender by the Type o f Settlement Type o f Settlement Rural Col Yo

Gender Turkey

Male

Turkey

Urban

ROW Yo

ColYo

ROW%

Col Yo

&ow Yo

92.0%

39.5%

88.8%

60.5%

90.0% 100.0%

8.0%

31.1%

11.2%

68.9%

10.0% 100.0%

100.0%

38.6%

100.0%

Female

61.4% 100.0% 100.0%

Source: DIE, Turkey 2002 HBS.

179

HH Head's Education by the Type of Settlement Type of Settlement Rural

Turkey

Urban

Col Yo ROWYo Col % ROWYo

10.5%

49.7%

6.7%

50.3%

8.2% 100.0%

9.3%

53.2%

5.2%

46.8%

6.8% 100.0%

44.3%

48.2%

55.7%

53.1% 100.0%

.l%

48.0%

.l%

52.0%

. l % 100.0%

7.4%

30.9%

10.3%

69.1%

9.2% 100.0%

.l%

13.7%

.5%

86.3%

.4% 100.0%

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%

.l%

6.0%

.9%

94.0%

.6% 100.0%

Illiterate Literate without diploma Primary school (5 years)

6 1.O%

Primary education (8 years)

Educational Institutions Graduated From

Col Yo ROWYo

Junior high school (8=5+3) Vocational school at Jr. high (8-9 = 5+ 3-4)

Master and doctorate

100.0%

Turkey

38.6% 100.0%

61.4% 100.0% 100.0%

Source: DIE, Turkey 2002 HBS.

Dwelling Type by the Type of Settlement Type o f Settlement Urban

.l%

.2%

Separated

74.1%

22.9%

42.7%

Semi detached house

12.2%

.O%

1.4%

Ground floor

1.9%

6.1%

4.5%

Regular floor

9.0%

51.3%

35.0%

Attic

.6%

1.5%

1.1%

Double floor

.l%

.6%

.4%

3.7%

2.7%

Basement

Shanty Other (Indicate)

Turkey

Rural

.2%

Luxurious house

What i s the type o f your dwelling?

Turkey

1.O% 3% 100.0%

.l% 100.0%

.4% 100.0%

Source. Turkey 2002 HBS.

180

Housing Rent Type o f Settlement Rural

Actual Rents

Turkey

Urban

Mean

60,287,195

100,553,2 13

94,554,897

Median

50,000,000

80,000,000

75,000,000

Source: Turkey 2002 HBS.

Number o f Rooms (frequency distribution )

Type o f Settlement

Turkey

Rural

Urban

eo1 Yo

Col Yo

e o 1 Yo

.92

.7 1

.79

11.80

9.86

10.61

38.90

43.48

41.71

35.84

40.79

38.88

8.26

4.19

5.76

2.63

.58

1.37

1.07

.16

.5 1

.so

.18

.30

.04

.03

.07

.01

.03

100.00

H o w many rooms are there in your dwelling?

9 10 Turkey

Source Turkey 2002 HBS.

Number o f Rooms Type o f Settlement

H o w many rooms are there in your dwelling?

Turkey

Rural

Urban

Mean

3.5

3.4

3.5

Median

3.0

Source Turkey 2002 HBS.

181

PersonlRoomRatio Statistics Type of Settlement Rural

Urban

1.301

1.277

Ratio number of members/ number of rooms

Turkey

1.286

Source: Turkey 2002 HBS.

Amenities Type of Settlement

hrkey

Rural

Urban

Do you have in your dwelling: Bathroom?

.836

.966

.916

Do you have in your dwelling: Toilet (indoors)?

.62 1

.943

.819

Do you have in your dwelling: Kitchen?

.883

.980

.942

Do you have in your dwelling: Waste disposal?

.001

.004

.003

Do you have in your dwelling: Central heating?

.OS4

.235

.165

Do you have in your dwelling: Heating from ground?

.ooo

.006

.004

Do you have in your dwelling: Piped water system?

388

.994

,953

Do you have in your dwelling: Electricity?

.999

Do you have in your dwelling: Natural gas?

,002

.149

.092

Do you have in your dwelling: Hot water?

SO9

.781

.676

Do you have in your dwelling: Satellite antenna?

.002

.093

.058

Do you have in your dwelling: Closed garage?

.033

,036

,035

Source: Turkey 2002 HBS.

182

Durable Goods: Average Number Per One Household rype of Settlement

rurkey

Rural

Urban

H o w many telephones do you have?

,869

,900

.888

H o w many cellular phones do you have?

,446

,781

.652

H o w many computers do you have?

.027

.lo9

.077

H o w many emets do you have?

.008

.043

.030

1.009

1.207

1.130

H o w many videos do you have?

,044

.096

.076

H o w many DVDs or VCRs do you have?

.066

.151

.118

H o w many video cameras do you have?

.009

,025

.019

H o w many satellite antennas do you have?

.296

.160

.212

H o w many HI-FI systems do you have?

,232

,434

.356

H o w many CD players do you have?

.033

.092

.070

H o w many refrigerators do you have?

.954

.98 1

.97 1

H o w many deep freezers do you have?

.062

.06 1

.061

H o w many dishwashers do you have?

.073

.281

.201

H o w many gas stoves with oven do you have?

.480

.629

,572

H o w many electric ovens do you have?

.379

.437

.415

H o w many microwave ovens do you have?

.017

.053

.039

H o w many automatic washing machines do you have?

.595

,842

.746

H o w many dryers do you have?

.006

,011

.009

H o w many presses (irons) do you have?

.005

,026

,018

H o w many vacuum cleaners do you have?

.592

,815

.729

H o w many carpet washing machines do you have?

.05 1

.161

.119

H o w many air-conditioners do you have?

.008

,044

.030

H o w many water heaters do you have?

.391

.625

.535

H o w many Jacuzzis do you have?

.003

.005

.004

H o w many aspirators do you have?

.143

.3 18

.250

H o w many televisions do you have?

Source: Turkey 2002 HBS.

183

Vehicles : Average Number P e r O n e Household rype o f Settlement

Turkey

Rural

Urban

.182

.275

.239

.004

.015

,011

Number o f Jeeps

.ooo

.001

Number o f minibus-caravans

.016

.013

,014

Number o f motorcycles

.05 1

.038

.043

Number o f motorized-motorized sea vehicle

.001

Number o f Yachts

.ooo

.ooo .ooo

.oo 1 .ooo

Number o f automobiles-owned Number o f automobile-provided

b y employer

Source: Turkey 2002 HBS.

184

ANNEX 11: POVERTY IN TURKEY: A LITERATURE REVIEW This literature review summarizes the methodology and results o f studies that focus on poverty in Turkey. Most o f these studies use 1987 and 1994 Household Budget Survey (HBS) data. One o f them i s a sociological study that examines poverty in Turkey o n the basis o f qualitative data that come f r o m 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 t o the type o f method used. Therefore, a brief review o f the methodology used should be an essential part o f any poverty assessment report. In this review, i t i s found that one problem with some Turkish poverty studies i s that they fail t o 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 o n the comparison o f the methodology and poverty measures used in the studies. In general, if there i s information o n the method used in the study, this i s reviewed in detail herein.

Studies Based on 1994 HBS A l i c i (2000) assesses poverty and examines its determinants using the 1994 Household Budget Survey (HBS) conducted by DIE.

An absolute poverty line based on a minimum required 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 i t 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 o f 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 in this reference group i s taken as the minimum required level o f calories. The resulting minimum required 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 contentare 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 nonfood goods and services-is constructed using the share o f non-food consumption among individuals whose total consumption i s just above the value o f the food poverty line. I t 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 minimum food basket in urban areas, and 1.5 times in rural 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 oneh a l f 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.

In this study, household sizes are adjusted using an appropriate economies-of-scale parameter for a l l the calculations. The results for different values o f this parameter are reported separately. Table A.II.l presents results using different poverty measures and different economies-of-scale parameters.

185

Table A.II.l. Different Poverty Measures and Results Yo

e = 0.75

Poverty) Basic Needs, CPL (Food + Non-Food)

e = 0.75

OECD scale

Consumption-Based Measures

Income-Based Measures

Basic Needs*

I33

economies o f scale 8 =

Medium economies o f scale

e = 0.75

0.90

High economies o f scale 8 = 0.5

OECD measure** (Relative Poverty)

5.09

15.69

** Comparison o f income for equivalent adult with one-half of the monthly median income. Note: Based on 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 in Turkey. East Anatolia i s the region facing highest poverty r i s k (1 1.25 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 i s different from the general findings in other studies that conclude that rural poverty rates are higher than urban poverty rates.

Regions

Share of Total Population ("A)

Basic Needs (CPL) (YO)

Extreme poverty (Food Poverty) (YO)

Marmara Aegean Mediterranean Central Rlack Sea

Southeast TOTAL

24.7 13.6 12.8 17.2

3.86 1.88 3.78 6.38

13 5

9.6 100.00

0.36 0.14 0.30 1.14 O 16

2 74

I 1

11.05 5.20

1 I

1.36 0.69

186

Table A.II.3. Urban and Rural Poverty Rates

Regions

Share of Total Population (”/)

Rural Urban

Basic

Needs (CPL)

(”/.I

Extreme Poverty (Food

Poverty) (YO)

46.46 53.54

5.21 5.20

0.80 0.59

While A l i c i (2000) i s a study that mainly focuses on poverty across various dimensions in Turkey in general, Erdogan (2000) i s a study that puts more emphasis on the structure o f 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 in Turkey, and poverty results based on the 1994 HBS. The main feature o f the study i s i t s emphasis on the detailed examination o f the structure o f poverty in Turkey. Erdogan’s (2000) choice o f poverty measures i s similar to A l i c i (2000). A Food Poverty Line based on a minimum required level o f calories, and the Complete Poverty Line based on consumption o f both food and non-food goods and services, are constructed. The results are reported for both o f the measures used. The important distinction between the poverty measures used by A l i a (2000) and Erdogan (2000) i s the difference in the methods they use in setting the minimum required level o f calories. A l i c i (2000) sets the minimum requirement o f calories using the actual survey data,39 whereas Erdogan calculates the calorie requirements for a four-person family according to Table A.II.4.

Table A.II.4. Calorie Requirements by Age and Gender

Source: Erdogan (2000).

In the Erdogan (2000) study, the Food Poverty Line is set as the estimated local cost o f the minimum food basket that meets the food energy requirements, and households with monthly monetary incomes below the cost o f the minimum food basket are considered poor. As the Complete Poverty Line (Cost o f Basic Needs Approach), cost o f a consumption bundle deemed to be adequate for basic needs (which include food and non-food components) i s estimated, and households with monthly monetary incomes below the cost o f 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 in needs in both approaches, according to location o f residence and household size.

For more details on Alici’s (2000) method, see previous explanation.

187

According to the first method, 8.37 percent o f the individuals and 5.66 percent o f the households are poor. In general, the higher poverty rates o f individuals relative to households can be explained by the fact that households with more members face more poverty compared to households with fewer members. In the rural areas, 11.82 percent o f individuals are poor, and in the urban areas 4.60 percent are poor. There i s a significant difference between urban and rural areas in terms o f poverty rates, and this difference holds after disaggregating according to certain demographic and socioeconomic characteristics. The highest percentage o f poor i s in Southeast Anatolia rural areas. Also, household size i s a significant determinant factor in poverty. As the household size gets bigger, poverty risk increases. In Turkey, 32.19 percent o f the households with 13 members are poor, while 27.32 percent o f households with 14 members, and 24.89 percent for households with 15 or more members, are poor. According to the second method, 24.30 percent o f individuals and 19.3 1 percent o f households are poor. Although the cost o f basic needs in rural areas i s lower than in urban areas, the poverty rates in the rural areas tend to be higher. The highest percentage o f poor i s again in the Southeast Anatolia rural areas.

The Erdogan (2000) study places strong emphasis on the structure o f poverty. Dimensions and characteristics o f poverty are examined in detail. Distribution o f age, gender, education, marital status, employment status, and economic activities among the poor in Turkey are important to understand who i s poor in the country, and this study thoroughly analyzes this issue. According to the first method, among the poor in Turkey, 5 1.49 percent are women and 48.5 1 percent 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 o f those aged 0 to 14 are poor, and 4.96 percent o f those 65+ are poor. Tables A.II.5 and A.II.6 present Erdoganâ&#x20AC;&#x2122;s findings regarding the structure o f poverty by age and gender. Table A.II.5. Structure o f Poverty b y Age and Gender (Food Poverty)

65+

2.96

2.94

2.98

188

Table A.II.6. Structure of Poverty by Age and Gender (Basic Needs)

Note: Based on 1994 HBS. Source: Erdogan (2000).

In Turkey, the population aged 12 and above constitutes 76 percent o f the total population. Tables A.II.7 and A.II.8 present the poverty o f 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 t o the second method, among the poor, 46.71 percent are employed and 53.29 percent are unemployed. There i s 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 o f the poor are employed. Within the group o f individuals who are not working, housewives, students, and the elderly face higher poverty r i s k relative to other groups. Table A.II.7. Structure o f Poverty by Employment Status (Food Poverty)

Sick Retired Elderly Other Working

2.05 1.08 9.54 0.43 53.67

2.70 0.55 12.09 0.57 62.49

1.10 1.86 5.83 0.22 29.69

Casual worker

16.56

11.50

45.52

Family worker w l

24.67 49.86

25.57 57.50

19.50 6.15

Source; Erdogan (2000).

189

Table A.II.8. Structure of Poverty by Employment Status (Basic Needs) Employment Status

Total Out o f work Unemployed Student Hoiisew ife

Handicapped Sick

Retired

YOAmong Poor 100.00 46.71 14.95 22.56 46 99

1.36 2.18 2.93

YOAmong Rural Poor 62.89 35.34 17.11 19.49 42 69 1.48 3.24 2.62

YOAmong Urban Poor 37.11 65.98 13.00 25.33 50.89

1.07 1.23 3.21

nopay

Source: Erdogan (2000).

Within the group o f 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 o f 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 A X . 10 summarize the distribution according to economic activities among the poor.

Table A.II.9. Structure o f Poverty by Economic Activity (Food Poverty)

--- =Negligible.

Source: Erdogan (2000).

190

Table A.II.lO. Structure o f Poverty by Economic Activity (Basic Needs)

trade Transportation Financial institutions

3.03 0.4 1

1.62 0.18

7.57 1.16

Source: Erdogan (2000).

Despite the recent shift o f poverty from rural to urban areas, rural poverty remains an important dimension o f poverty in Turkey. 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, i s the arithmetic mean o f the difference between the income o f individuals in poor households and the poverty line; the third measure, P2, i s the arithmetic mean o f the square o f this difference, and it captures the severity o f poverty. 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 a l l the tables in Pamukâ&#x20AC;&#x2122;s study i s that the rural population i s the group being examined. Therefore, a l l the numbers are indexes that compare certain subgroups within the rural population only, not the total population in Turkey.

Table A.II.ll. Regional Poverty Regions

Marmara Aegean Mediterranean Central Anatolia Black Sea East Southeast

Head Count Index

6.50 7.67 17.42 15.79 16.44 12.97 30.46

Poverty Severity o f Gap Index Poverty

2 1.76 23.36 27.77 26.73 29.44 26.17 26.93

0.53 0.68 1.94 1.84 2.05 1.27 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 o f the household head. Table A.II.12 summarizes Pamukâ&#x20AC;&#x2122;s findings for poverty by gender o f the household head. In general, households with female heads face higher poverty risk, and the severity o f poverty i s higher for such households.

Table A.II.12. Poverty and Gender of Household Head

Gender o f Household Head

Head Count Index

Poverty Severity of Gap Poverty Index

IMale Female

14.46 22.11

26.91 27.15

1.58 2.41

Total

14.80

26.92

1.62

Source: Pamuk (2000).

In the rural areas, just like in urban areas, household size and years o f education o f the household head are determinant factors o f poverty. As household size increases, the poverty risk faced by the household members increases, and as years o f education o f the household head increases, the poverty r i s k decreases. Some other socioeconomic variables she uses to disaggregate poverty are by employment status o f household head, number o f working individuals in the household, source o f income, and types o f economic activities the household head i s engaged in (the study also looks at some o f 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 i s that poverty i s examined according to certain factors that play an important role in determining the relative standing o f households in rural places. Its analysis o f poverty profiles that are geared solely toward the analysis o f the rural areas i s what gives this study i t s unique place in the poverty literature in Turkey. For example, Pamuk looks at poverty according to the agricultural activity the household i s engaged in, and ownership o f agricultural land and equipment. Table A.II.13 summarizes Pamukâ&#x20AC;&#x2122;s findings on poverty rates according to the type o f agricultural activity rural households engage in. In rural areas, households not engaged in agricultural activities face aboveaverage poverty risk. Table A.II.13. Poverty and Agricultural Activity of the Household Type of Agricultural Activity o f the Household D o not engage in agricultural activity Farmer Stockbreeder Both farmer and stockbreeder Total

Head Count Index

(Yo)

Total

Poor Population

(YO)

17.07

29.42

33.94

15.88 15.79 13.07

15.03 4.99 50.56

16.12 5.32 44.62

14.80

100.00

192

Table A.II.14 summarizes poverty rates according to the size and ownership o f agricultural land in rural areas. I t i s observed that the highest level o f poverty risk i s faced by households with agricultural land o f 6 to 10 decares (1 decare = 0.247 acres). As the size o f the agncultural land that i s 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. Poverty and Agricultural Land Size Agricultural Land Size (decares)

Head Count Index

("/.I

Total Population (YO)

Poor Population (YO)

Do not have land

16.72 17.88 21.36 16.80 13.80 5.64 2.54 0.00

36.26 7.77 8.97 13.24 18.27 9.71 5.60 0.18

40.95 9.38 12.95 15.03 17.04 3.70 0.96 0.00

Total

14.80

100.00

Ck5 6-10 11-20 21-50 51-100 101-500 501 +

Note: Based on 1994 HBS. *1 decare = 0.247 acres. Source: Pamuk (2000).

Table A.II.15 shows the poverty rates according to the size o f 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 in households with agncultural land o f 0 to 5 decares i s 48.23 percent, in this region. The regional breakdown of the poverty rates reveals that those households engaging in agricultural activity but that o w n relatively smaller amounts o f land face higher poverty r i s k than households not engaging in agricultural activity in most of the regions, except for Central Anatolia. This observation contradicts the results f r o m the previous table, that is, that households engaging in agricultural activity face lower poverty risk relative to those that do not (Table A.II.14). This i s a result o f the fact that lower poverty risk of those households engaging in agricultural activity i s mainly driven by those who o w n more land.

193

Table A.II.15. Poverty and Agricultural Land Size Across Regions Agricultural Land Size

O* 0** 0-5 6-10 11-20 2 1-50 51-100 10 1-500 501 +

Total

Marmara

Aegean

Mediterranean

Central Anatolia

Black Sea

East

Southeast

Total

8.23 5.36 8.20 15.37 4.67 4.33 0.52 0.00 0.00

9.47 13.84 4.61 2 1.84 8.97 1.75 0.00 0.00 0.00

20.39 18.20 19.88 23.22 18.05 12.93 3.56 0.00 0.00

17.04 15.06 27.18 11.47 19.56 23.74 6.22 3.44 0.00

10.71 22.68 21.84 22.24 19.67 17.57 0.00 0.00 0.00

18.90 15.36 19.52 22.50 20.37 5.40 3.16 0.00 0.00

39.45 20.27 48.23 42.65 34.86 22.35 15.79 5.79 0.00

17.07 15.79 18.60 22.40 17.36 12.05 4.17 2.09 0.00

6.50

7.67

17.42

15.79

16.44

12.97

30.46

14.80

** Households that engage only in stockbreedkg as an agricultural activity. *** 1 decare = 0.247 acres. Another important factor in rural places i s the ownership o f 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 o f poverty by tractor ownership (Table A.II.17) reveals that while the proportion o f rural households without a tractor i s 84.42 percent, their proportion within the poor population i s 92.32 percent.

Table A.II.16. Poverty and Number of Tractors Owned Aegean

Mediterranean

Central Anatolia

Black Sea

East

Southeast

Total

8.73

10.94

16.97

17.48

16.86

13.12

30.74

15.47

1 2

0.88 0.00

1.26 0.00

9.04 0.00

12.68 0.00

8.35 0.00

8.71 0.00

15.76 0.00

7.03 0.00

0.00

7.05

9.42

16.02

16.45

15.73

12.59

29.19

14.15

Number of Tractors

Do not own one

Total

Marmara

194

Table A.II.17. Structure of Poverty According to Number o f Tractors Owned

Mediter r-anean

Number of Tractors

Marmara

Aegean

Do not own

97.36

97.94

93.25

84.59

92.98

91.67

94.39

92.32

1 2 3

2.64 0.00 0.00

2.06 0.00 0.00

6.75 0.00 0.00

15.41 0.00 0.00

7.02 0.00 0.00

8.33 0.00 0.00

5.61 0.00 0.00

7.68 0.00 0.00

100.00

one

Total

ource: Pamuk (2000).

Central Anatolia

Black Sea

East

Southeast

Total

Studies Based on 1987 HBS Dumanli (1996) looks at dimensions o f poverty in Turkey using the 1987 Household Income and Expenditures Survey (HBS). The poverty lines are computed based on a minimum required level o f calories taken as 2,450 calories per day. The cost o f a minimum food basket that meets the daily calorieintake requirement i s taken as the food poverty line. Table A.II.18 presents the poverty lines used in this study.

Table A.II.18. Poverty Line Per Capita (Annual) Years 1987 1988 1989 1990 1991 1992 1993

TL 284,700 473,405 777,085 1,388,433 2,135,740 3.830.809 , , 14,696,360

TL = Turkish Lira. Source: Dumanlr (1996).

US$ 332.2 332.2 365.0 532.4 513.4 556.2 541.4

Table A.II.19 summarizes Dumanli's findings on poverty. Here, he takes an approach o f 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.

Table A.II.19. Comparison o f Income and Poverty Lines by 5 Percent of Population Income L

5% 5 yo 5 yo 5 yo 5 yo

1.156 1,176 1,264 1,293 1,361 1,361

915 1,127 1,298 1,483 1,591 1,677

195

Household Percentage 5% 5% 5% 5% 5 ?o' 5%

Poverty Line

1.330 1,347 1,361 1.341 1.463 -1,426 1,415 1,449 1,469 1,572 1,495 1,515 1,461

Household Income 1.863 2,249 2,454 2.692 2.960 -3,259 3,623 4,063 4,627 5,332 6,377 8,102 16,947

- 7

5% 5% 5% 5% 5% 5 yo 5 yo

-3-

*Values in 1,000 TL. Source: Dumanli (1996).

Comparative Studies Due to the difficulty o f comparing poverty across time, there are n o t many comparative studies that examine how poverty has changed over the years in Turkey. Dagdemir (1999) is unique in this respect, and analyzes data from the 1987 and 1994 Household Income and Consumption Expenditures Surveys (HBS) to assess changes in poverty during 1987-1994. For the 1994 poverty lines, Dagdemir uses the poverty lines that Erdogan (2000) calculates using the minimum food cost (MFC) and cost-of-basic needs (CBN) methods. The 1987 poverty 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 i s not addressed in detail in this study. I t i s 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 in poverty in Turkey during this period are summarized in Tables A.II.20 and A.II.2 1.

Table A.II.20. Poverty in Turkey, 1987-1 994 (Minimum-Food-Cost Approach Poverty (YO)

Marmara Mediterranean

I Central

Black Sea East Southeast Urban Rural Turkey

85 109 141 129 95 126 105

127 108 143 144 131 136 135 138

13.3 10.1 14.2

----_

6.9 21.2 11.5

13.6 10.1 18.8 14.7 19.6 8.7 20.2 11.5

196

Table A.II.21. Poverty in Turkey, 1987-1994, (Cost-of-Basic-Needs Approach) Regions Marmara

Aegean

Mediterranean

Central

Black Sea

Poverty Line (US$) 1987 1994 220 165 208 168 177 254 134 170 148 194

I I

Poverty (%)

1994 23.4 20.8

1987

---

I I

41 5 23.2 23.0

46 7

27.2 32.2

Dagdemir finds that the cost o f a minimum food basket increased from US$l05 to US$138 (monthly) between 1987 and 1994. For the C B N approach, the minimum cost o f meeting basic food and non-food requirements increased from U S 1 5 2 to US$198. I t i s observed that the household income o f the poor has followed an increasing trend parallel to the increase in the minimum food cost, and therefore the percentage o f households that cannot meet the minimum food cost has not changed significantly during this period. On the other hand, the percentage o f households whose income i s below the C B N poverty line has increased from 27 to 29 percent. According to the M F C method, the poverty rate in urban 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 C B N method, the poverty rate in urban areas increased from 14.3 to 20 percent, and in rural areas i t increased from 41.5 to 42.5 percent. One conclusion from these observations i s that there i s a sharp contrast in the change in poverty rates in rural areas compared to change in poverty rates in urban areas according to the C B N method. On the other hand, according to the M F C method, in both urban and rural areas, the change in poverty rates during this period i s not as significant. According to the C B N method, i t i s observed that the poverty rates in urban areas have increased significantly during this period. A closer look into the structure o f poverty reveals the share o f the urban poor in the total population o f poor has increased from 27.5 to 37.6 percent, while the share o f the rural poor has decreased from 72.5 to 63.4 percent. Based on these observations, i t can be concluded that poverty slowly shifted from urban to rural areas between 1987 and 1994. However, rural poverty still remains significantly higher than urban poverty in Turkey. 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. In general, i t i s observed that poverty rates have increased most significantly in the Aegean, Marmara, and Black Sea regions.

Perceptions o f Urban Poverty Perceptions of Urban Poverty in Turkey (Erdogan 2002) i s a sociological study on poverty. I t s focus i s the urban poor, and i t i s based on 160 interviews conducted with extremely poor households. The study consists o f articles by various authors, and a selection o f the 160 interviews. I t s main purpose i s to

197

identify h o w 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 i s achieved through using the interviews with the poor as the main source. The following are some o f the issues o n which the authors focus: H o w the poor feel about their place in society; what their daily experiences are and h o w they choose to express and define themselves; the conditions that the poor live in; h o w the poor perceive class and cultural hierarchy in the society; h o w 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 i s 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 r i c h 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 i s the conflict between “wanting t o speak up” while “never speaking up.” In other words, while the poor have the desire t o speak up about their painful experiences, they are also reluctant to speak at all, especially because they do not see themselves in a position t o speak. Erdogan emphasizes the fact that the poor are ashamed o f their way o f speaking. Therefore, listening to the poor should consist not only o f listening t o their words-it should also be listening t o their silence and interpreting it correctly. Another feature of the self-image o f the poor i s the reflection o f the poverty experience in the “physical body.” One example i s the general belief o f the poor that the r i c h see them as “animals.” In the interviews, it i s observed that this self-image i s also reflected in their body posture. For example, the respondents usually refuse to sit at the same level as the interviewer; they want to sit o n the floor or o n a lower chair, instead. It i s observed in the interviews that the extent o f informal assistance that i s in general believed t o exist among relatives, fiiends, neighbors, and hemseri i s rarely mentioned by the poor as a source o f help. Most o f the interviewees say that they never get support f i o m 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 j o b or gave them food and money, the general consensus i s that such sources o f aid are rare and unreliable. I t i s concluded that the social assistance provided by 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 i s 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. In most interviews, the women say that their husband i s too proud to ask for help. Therefore, i t i s usually the woman who tries these altematives .

The interviews provide a different perspective o n 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 i t s implications. This study i s unique in the way it sheds light on these cultural schemas by measuring or quantifying the perceptions o f the poor o f their daily life.

198

Conclusion The only way to combat poverty i s to eliminate the factors that create it. This can be done only if policymakers know what those factors are. The studies reviewed here aim to provide such information by examining various dimensions o f poverty. It i s important to bring this information together and compare the studies in terms o f the different aspects o f poverty they look at, and this review i s an attempt to achieve that goal. I t should be noted that there are some poverty studies that have not been reviewed here mainly due to lack o f access.

199

REFERENCES Alici, Sema. 2000. “Turkiye de Yoksullugun Sosyo Ekonomik Analizi” (“A Socioeconomic Analysis o f Poverty in Turkey”). In Yoksullukla Mucadele Stratejileri. Ankara: Hak-19 Konfederasyonu Yayini 2002. Behrman, Jere R., and Nancy Birdsall. 1983. “The Quality o f 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 and Time 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 Sourcebook for Poverty Reduction Strategies. Washington, D.C.: World Bank. Council o f Higher Education. 1998. “Parental Income, Educational Expenditures, Financial Aid and Job Expectations o f University Students.” YOK (Turkish) Council o f Higher Education. Dagdemir, Ozcan. 1999. “Turkiye Ekonomisinde Yoksulluk Sorunu ve Yoksulluk Analizi 1987-1 994” (Poverty Problem and the Analysis of Poverty in the Turkish Economy 1987-1994). H.U. Iktisadi ve Idari Bilimler Fakiiltesi Dergisi, Ciltl7, Say1 1. Datt, Guarav, and Martin Ravallion. 1992. “Growth and Redistribution Components o f Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s.” Journal of Development Economics, 38:275-95. Dayioglu, Meltem, and Insan Tunali. 2004. “Falling Behind W h i l e Catching Up: Changes in the Female-Male Wage Differential in Urban Turkey.” Unpublished manuscript, Koc University. Deaton, Angus. 1997. The Analysis of Household Surveys: Microeconomic Analysis for Development Policy. Baltimore: Johns Hopkins University Press for the World Bank. Deaton, Angus, and C h s t i n a Paxon. 1998. “Economies o f 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 University Press.

. 1986. “On Measuring Child Costs: With Applications to Poor Countries.” Journal of Political Economy, 94:72&44.

Deaton, Angus, and Salman Zaidi. 2002. “Guidelines for Constructing Consumption Aggregates for Welfare Analysis.” LSMS Working Paper 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 i t s Dimensions in Turkey”). Erdogan, Guzin. 2000. “Tiirkiye’de ve Diinya’da Yoksulluk Olqumleri Uzerine Degerlendirmeler” (“Poverty Assessment in Turkey and in the World”). In DIE, Iygiicii Piyasalan Analizleri. Ankara: Devlet Istatistik Enstitiisii Matbaasi. Erdogan, Necmi (ed). 2002. Yoksulluk Halleri (Perceptions of Urban Poverty). Istanbul: Demokrasi Kitapligi Yayinevi. Foster, James, J. Greer, and Eric Thorbecke. 1984. “A Class o f Decomposable Poverty Measures.” Econometrica, 52:761-66. Glewwe, Paul, and Gilette Hall. 1992. “Poverty and Inequality during Unorthodox Adjustment: The Case o f Peru, 1985-90.” World Bank Living Standards and Measurement Study Working Paper No. 86. Washington, D.C. Heckman, James J. 2003. “China’s Investment in Human Capital.” Economic Development and Cultural Change, 5 1(4):795-804. Lanjouw, Peter, Branko Milanovic, and Stefan0 Paternostro. 1998. “Poverty and the Economic Transition: H o w D o Changes in Economies o f Scale Affect Poverty Rates for Different Households?” Policy Research Working Papers 2009. Washington, D.C.: World Bank. Lanjouw, Peter, and Martin Ravallion. 1995. “Poverty and Household Size.” Economic Journal, 105: 1415-34. Liviatan, Nissan. 2002. “Lessons for Turkey from International Experience with Financial Crisis and Disinflation.” Turkey Country Economic Memorandum (2003) Background Paper. Washington, D.C.: World Bank. Mete, Cem. 2004. “The Inequality Implications o f Highly Selective Promotion Practices.” Economics of Education Review (forthcoming). Milanovic, Branko, and Sholomo Yitzhaki. 200 1. “Decomposing World Income Distribution: Does the World Have a Middle Class?” Policy Research Working Paper No. 2562. Washington, D.C.: World Bank. Mincer, J. 1974. “Schooling, Experience and Earnings.” National Bureau o f Economic Research. New York: Columbia University Press.

Mingat, Alain. 1998. “The Strategy Used by High-performing Asian Economies in Education: Some Lessons for Developing Countries.” World Development, 26(4):695-7 11. Ministry o f Health. 2003. National Health Accounts 2000. Draft Report. Turkey. Ministry o f National Education. 2003. “Milli Egitim Sayisal Veriler 2002-2003 .” Ankara. Ogut, Menevis, and R. Funda Barbaros. “Regional Development Inequalities in Turkey: An Assessment on the Distribution o f Investment Incentives.” Draft. Undated.

20 1

Pamuk, Muzeyyen. 2000. “Kirsal Yerlerde Yoksulluk” (“Poverty in Rural Places”). In DIE, Iggucu Piyasalari Analizleri. Ankara: Devlet Istatistik Enstitiisti Matbaasi. Psacharopoulos, George. 1994. “Returns to Investment in Education: 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 Lines in Theory and Practice.” LSMS Working Paper 133. Washington, D.C.: World Bank.

1999. “Poverty Lines in Theory and Practice.” Living Standards Measurement Study Worlung Paper No. 133. Washington, D.C.

. 2000.

“Poverty in Theory and Practice.” Living Standards Measurement Study Working Paper No. 133. Washington, DC: World Bank.

Schultz, T. Paul. 2002. “Why Governments Should Invest More to Educate Girls.” World Development, 30(2):207-25. Shorrocks, A. F. 1999. “Decomposition Procedures for Distributional Analysis: A Unified Framework Based on the Shapley Value.” Mimeo, University o f Essex. T C Milli Egitim Bakanligi (Republic o f Turkey Ministry o f Education). 2001. “2002 Yili Basinda Milli Egitim” (“National Education at the Beginning o f 2002”). Ankara, Turkey, December. Turlush Government. 2003. “Preliminary National Development Plan.” Ankara, December. Turkiye I s Bankasi. 1999. “75 Yilda Egitim” (“Education during the Last 75 Years”). Istanbul, Turkey, 1999.

UNDP (United Nations Development Programme). 200 1. “Human Development Report.” 2003. “New Poverty and the Changing Welfare Regime o f Turkey.” Ankara. Wagstaff, A., and E. Doorslaer. 2002. “Catastrophe and Impoverishment in Paying for Health Care: With Applications to Vietnam, 1993-1 998.” Washington, D.C.: World Bank. World Bank. 1990. World Development Report 1990: Poverty. Washington, D.C.

. 2000. “Making Transition Work for Everyone: Poverty and Inequality in Europe and Central Asia.” Washington, D.C.

. 2000. “Turkey: Country Economic Memorandum-Structural Growth.” Report No. 20657-TU, September 15.

Reforms for Sustainable

202

. 2000.

“Turkey: Economic Reforms, Living Standards, and Social Welfare Study.” Report No. 20029-TU. Washington, D.C., January 27.

. 2001.

World Development Report 2000/2001: Attacking Poverty. Washington, D.C.: World Bank and Oxford University Press.

. 2003a. “Turkey: Country Economic Memorandum-Towards Sustained Growth.” Report No. 26301-TU, July 28.

Macroeconomic Stability and

. 2003b. “Turkey: Poverty and Coping After Crises.” Report No. 24185-TR. Human Development Unit, Europe and Central Asia Region. Washington, D.C., July 28.

. 2003c.

“Turkey: Reforming the Health Sector for Improved Access and Efficiency.” Report No. 24358-TU. Washington, D.C., March.

------_-. 2004.

“Turkey: A Review o f the Impact o f The Reform o f Agricultural Sector Subsidization.” Washington, D.C. ,March.

. Forthcoming. “Labor Market Study.” Washington, D.C. W H O (World Health Organization). HFA. 2004. K e Xu, David, B. Evans, K e i Kawabata, 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 shown on this map do not imply, on the part of The World Bank Group, any judgment on the legal status of any territory, or any endorsement or acceptance of such boundaries.

RAILROADS

28°

*Province names are the same as their capitals.

32°

36°

44°

BULGARIA

Edirne Kirklareli

Sinop Bosporus

Kocaeli (Izmit)

Bursa

Aydin

Tokat

Aci Gölü

Denizli Burdur Burdur Gölü

Kirikkale Sivas

Erzurum

Erzincan

Tuz Gölü

Keban Reservoir

Nevsehir

Agri

Diyarbakir Tigris

Adiyaman

Osmaniye

Antalya

36°

Icel (Mersin)

Adana

Hakkari

Mardin Sanliurfa To Tall Birak

To Zakhu To Tall Kushik

Kilis

IRAQ

To Aleppo

Hatay (Antakya) 32°

Gaziantep

To Reza’iyeh

To Latakia

To Al Atarib and Aleppo

SYRIAN ARAB REPUBLIC 40°

100

200

KILOMETERS 44°

36°

IBRD 33581

SEPTEMBER 2004

ksu

To Sharafkhaneh

Siirt

Batman

Sirnak

Kahraman Maras

Beysehir Gölü

Gö

Van

Bitlis Malatya

Isparta

Gulf of Antalya

Lake Van

Mus

Elazig

Nigde

Konya

ISLAMIC REP. OF IRAN

Bingöl

Kayseri

Aksaray Egridir Gölü

40°

To Maku and Reza’iyeh

Tunceli

Kirsehir

Karaman

Sea

To Yerevan

Eup

Mugla

28° Mediterranean

Çayl

te hra

Usak Hoyran Gölü

ARMENIA

Gümüshane Bayburt

Igdir

Yozgat

Afyon

di z Nehri

Kelkit

Neh

Izmir

To Leninakan

Kars

r Ço

Amasya

a ry ka Sa

Sea

ANKARA

Sakarya Nehri

Irm

Ardahan

Giresun

Çorum

Nehri

Manisa

lice

Eskisehir Kütahya

Aegean

Irm

Bilecik

L. Apolyont

Balikesir

Artvin Trabzon Rize

Ordu

ra tes

Çanakkale

Çankiri

Bolu

Sim

40°

Sakarya (Adapazari)

Yalova Nilü fer

L. Manyas

Samsun

Eu ph

elles dan Dar

Kizil k Irma

Kastamonou

Sea of Marmara (Marmara Denizi)

To Akhaltsikhe

Bartin Karabük

Istanbul

Tekirdag

To Batumi and T’bilisi

Cey han

Kipoi

Zonguldak

e Ergen

ara

GREECE To

Sea GEORGIA

Ank

To Kapitan Andreevo and Kharmanli

Black

To Malko Turnovo

TURKEY

JOINT POVERTY ASSESSMENT REPORT (JPAR) GROSS DOMESTIC PRODUCT BY STATISTICAL REGION STATISTICAL REGION BOUNDARIES

PROVINCE CAPITALS*

PER CAPITAL GROSS DOMESTIC PRODUCT, BY STATISTICAL REGION, 2001 AT CURRENT PRICES

PROVINCE BOUNDARIES

2,500 .00 US$ 2,000 .00 US$ 1,500 .00 US$

This map was produced by the Map Design Unit of The World Bank. The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of The World Bank Group, any judgment on the legal status of any territory, or any endorsement or acceptance of such boundaries.

INTERNATIONAL BOUNDARIES *Province names are the same as their capitals.

1,000 .00 US$

28°

TURKEY

NATIONAL CAPITAL

32°

36°

44°

BULGARIA

Black

Sea GEORGIA

Edirne

Sinop

Kirklareli Bartin

Bosporus

Zonguldak

GREECE

Karabük

Istanbul

Tekirdag

Sea of Marmara (Marmara Denizi)

Kocaeli (Izmit)

Yalova

Bursa Çanakkale

ANKARA

Kirikkale

Sivas

Yozgat

Sea Aegean

40° Igdir Erzurum

Erzincan ate phr

Agri

Kütahya Kirsehir

Tunceli Keban Reservoir

Manisa

ARMENIA

Gümüshane Bayburt

Tokat

Çorum

Eskisehir

Kars

Giresun

Amasya

Çankiri

Ardahan

Trabzon Rize

Ordu

Bilecik

Balikesir

Artvin

Sakarya (Adapazari) Bolu

40°

Samsun Kastamonou

Afyon

Tuz Gölü

Usak

Nevsehir

Izmir

Eup

Kayseri Malatya

ISLAMIC REP. OF IRAN

Bingöl Mus

Elazig

Lake Van

Van

Bitlis

hrates

Aksaray Denizli

Burdur

Diyarbakir

Nigde

Konya

Aydin Isparta Beysehir Gölü

Hakkari Sirnak

Sanliurfa

Osmaniye

Karaman

Gaziantep

Antalya

36°

Adana

Hatay (Antakya) 32°

Kilis

IRAQ SYRIAN ARAB REPUBLIC 40°

100

200

KILOMETERS 44°

36°

IBRD 33582

SEPTEMBER 2004

Icel (Mersin)

Gulf of Antalya

Sea

Siirt

Mardin

Mugla

28° Mediterranean

Batman

Tigris

Adiyaman

Kahraman Maras